Research

Research Health burden of hip and other fractures in Australia beyond 2000 Projections based on the Geelong Osteoporosis Study Kerrie M Sanders, Geoffrey C Nicholson, Antony M Ugoni, Julie A Pasco, Ego Seeman, and Mark A Kotowicz MJA 1999; 170: 467-470 For editorial comment, see Morris et al; see also Pocock et al. Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Author's Detials - - More articles on Public and environmental health Abstract Objective: To calculate the expected increase in the number of fractures in adults attributable to the predicted increase in the number of elderly Australians. Data sources: All fractures in adult residents (> 35 years) of the Barwon Statistical Division (total population, 218 000) were identified from radiological reports from February 1994 to February 1996. The Australian Bureau of Statistics supplied predictions of Australia's population (1996 to 2051). Main outcome measure: The projected annual number of fractures in Australian adults up to 2051 (based on stable rates of fracture in each age group). Results: The number of fractures per year is projected to increase 25% from 1996 to 2006 (from 83 000 fractures to 104 000). Hip fractures are projected to increase 36% (from 15 000 to 21 000) because of a substantial rise in the number of elderly aged 85 years and over. Hip fractures are expected to double by 2026 and increase fourfold by 2051. Conclusions: In contrast to Europe and North America, where numbers of hip fractures are expected to double by 2026 and then stabilise, in Australia hip fractures will continue to place a growing demand on healthcare resources for many decades. These projections can be used for setting goals and evaluating the costs and benefits of interventions in Australia. Introduction Australia's population is ageing. Thus, diseases such as osteoporosis are affecting a greater proportion of the population. Cost-effective prevention and treatment of fractures will become increasingly important, as the proportion of the "prime working" population (aged 15 to 64 years) is expected to decrease by 7% between 1996 and 2051.1 Estimations of the annual number of fractures will facilitate effective allocation of limited healthcare resources. Previously, fracture rates in Australia have been extrapolated from rates in Dubbo and Busselton,2,3 populations that may not be representative of the national population as they are both fairly isolated towns with small numbers of men and women in the oldest age groups. Lord et al investigated changes in hip fracture admissions in New South Wales between 1979 and 1990 using International Classification of Diseases hospital discharge codes.4 Although frequently used for ascertaining hip fractures, the limitations of this method are well documented.5,6 The Geelong Osteoporosis Study is a population-based study of age-, sex- and site-specific fracture rate(s) within a well-defined geographic region sufficiently large to establish reliable rates of fracture. It is the largest Australian epidemiological study of adult fractures, identifying 2184 fractures over two years in people aged 35 years and over. Data from the 1996 national census confirm the findings of an earlier study7 that the regional population is typical of the nation in age distribution and socioeconomic range. The Australian Bureau of Statistics publishes projections of the size, structure and distribution of Australia's population into the next century.1 Assuming stable age- and sex-specific rates of fracture, we have projected annual numbers of fractures by anatomical site in adults over the next 50 years. Methods Population The Barwon Statistical Division is a region of southern Victoria7 defined by the Australian Bureau of Statistics which includes urban, semi-urban and rural residents and has a population of 218 000 (109 923 aged 35 years and over). Fracture ascertainment Using radiological reports from the two medical imaging practices in the region, we identified all fractures in adults (aged 35 years and over) residing within the Barwon Statistical Division over the two-year ascertainment period (17 February 1994 to 16 February 1996). Fractures were identified by a weekly computerised keyword search of all radiological reports. Although radiological services are available at several sites throughout the region, these sites are under the auspices of two medical imaging practices. Only reports of definite fractures were included as "cases". Reports referring to "likely" or "possible" fracture were not included unless radiologically confirmed at a later date. Vertebral fractures were included in the study on the basis of the radiological report, provided there was no previous record of a fracture at that level. The initial x-ray demonstrating each fracture was sought and the date of fracture was the date of the radiological diagnosis. As the region is well provided with medical facilities, few fractures would not have been assessed or followed up by services within the region. The standing orders for the ambulance service are to transport patients from residences within the study region to The Geelong Hospital. The study was well advertised and adults who may have sustained fracture(s) while outside the region were asked to contact study coordinators. Non-residents and patients with pathological fractures (metastatic cancer, Paget's disease and multiple myeloma) were excluded. We have previously reported that the exclusion of high trauma fractures may underestimate the prevalence of bone-fragility fractures in the community,8 so we did not exclude fractures on the basis of trauma classification. The method of fracture ascertainment has been previously described9 and validated, using hip fracture as a model.10 The study was approved by The Geelong Hospital Human Research and Ethics Advisory Committee. Projections We calculated the predicted number of fractures in each five-year age group for men and women as: Predicted number of fractures = Number of fractures in the study region Study region population X Projected Australian population X 0.5 (2 year ascertainment) The study region population was determined from the 1996 Australian Bureau of Statistics census. Annual population projections were also provided by the Australian Bureau of Statistics.1 This method of projection assumes that the fracture rate in each age group will remain unchanged over time. Confidence intervals around the projections were calculated by the bootstrap method.11 Results The number of fractures identified in the study population is shown in Table 1. Projections for all fractures During the ascertainment period, 2184 adults sustained fractures, producing an age- and sex-adjusted incidence of 102 per 10 000 person-years. In 1996, we estimate that about 83 000 Australians aged 35 years and over sustained fractures. We predict that by 2006 the total number of fractures per year will increase by 25%, to 104 000 (Figure 1). Projections for hip fractures Over the two-year ascertainment period, 428 people sustained hip fractures, producing an age- and sex-adjusted incidence in people aged 35 years and over of 17.2. The number of hip fractures in Australian women is projected to increase from 11 300 per year in 1996 to 44 700 in 2051. In men, the number is projected to rise from 4 000 to 15 300. Figure 2 shows the projection for all people aged 35 years or over. The number of Australians sustaining hip fractures each year is projected to increase by 15% every five years until 2036, then by about 10% every five years until 2051. A fourfold increase in hip fractures is expected by 2051, when about 23% of Australia's projected population will be aged 65 years and over (compared with 12% in 1996) and over 8% of the population will be aged 85 years and over (compared with 2% in 1996). Hip fractures are likely to increase more than fractures at other sites because the greatest population growth is expected in the oldest age groups, where the hip is the most common site of fracture.9 Projections for fractures at sites other than the hip During the ascertainment period, 1756 adults aged 35 years and over sustained fractures at sites other than the hip, producing an age- and sex-adjusted incidence of 77 per 10 000 person-years. We estimate that in 1996 about 68 000 Australians aged 35 years and over sustained fractures at sites other than the hip, and that the number of non-hip fractures is likely to increase by 9% every five years until 2036, and then by 4% every five years until 2051 (to 147 645 non-hip fractures per year). Projections for vertebral, Colles, humeral and pelvic fractures (the most common sites of fracture after the hip) are shown in Figure 3. Fractures at these sites are likely to increase more than fractures at other sites where fracture rates do not increase substantially with age.9 Population projections suggest that vertebral, humeral and pelvic fractures will increase by 12% every five years to 2036, then by 6% every five years to 2051. Colles fracture rates increase with age in women, but not in men,8 and the overall number of adults sustaining a Colles fracture will increase by 10% every five years until 2036, then by 5% every five years until 2051. Fractures other than hip, vertebral, Colles, humeral and pelvic fractures are predicted to increase at about half the rate predicted for these sites (6% every five years until 2036, then 3% every five years until 2051). Projection of fracture numbers by age group The number of fractures in adults aged 35 to 59 years is not predicted to change substantially over the projection period, as only a small population increase in this age group is anticipated. By contrast, among those aged 60 years and over, the number of fractures is predicted to increase by at least 10% every five years, and by almost 20% every five years among those aged 85 years and over (Table 2). Figure 4 shows the projected distribution of hip fractures across age groups for selected years from 1996 to 2051. Discussion The ageing of the Australian population is increasing the demand for health resources. Health expenditure per person aged 65 years and over is nearly four times higher than for younger individuals ($4900 v $1300).12It is likely that hip fracture, which becomes increasingly common with advancing age, contributes significantly to this higher expenditure. Almost all people with a fractured hip are hospitalised, with an average length of stay of 13 days.13 Studies in other white populations suggest that 50% of patients who survive hip fracture are discharged to nursing homes, and 25% remain institutionalised one year later.14,15 Comparable Australian data are not available, although the increased mortality after hip fracture has been confirmed.16 We estimate that hip fractures accounted for 0.9% of total government health services expenditure for 1995/96.12 Based on the number of Australian salary earners in 1998 (Australian Bureau of Statistics) and an average cost of $16 000 per hip fracture,17 treatment alone costs salary earners an average of $28 per year. In current dollar terms, this may escalate to $120 per salary earner per year by 2051 if the projected increase is realised. The number of adults sustaining a hip fracture is likely to more than double from 15 000 in 1996 to 34 000 in 2026, then almost double again by 2051. Fractures at other sites are expected to increase by 70% from 1996 to 2026, then by a further 26% to 2051. These rates of increase are far above the expected growth in total healthcare costs due to the ageing of the Australian population, which is estimated to be 4% every five years for the next 30 to 40 years.18 For the number of hip fractures to remain stable, their incidence (per 10 000 population per year) would need to decline from 17.2 in 19969 to 15.6 in 2001, 14.3 in 2006 and 11.3 in 2026. In the United States, an 18% reduction in hip fracture rates between 1988 to 2000 has been targeted by public health strategists.19 Current therapies (such as hormone replacement therapy and bisphosphonates) may reduce fracture rates by 50%, but their cost-effectiveness remains controversial.20,21 In most white populations the number of hip fractures is predicted to double between 1990 and 2025, and then plateau as the growth in the aged population ceases.22 However, in contrast with Europe and North America, the aged population and the number of hip fractures in Australia and New Zealand are likely to continue to increase from 2025 to 2050.22 Current trends suggest that almost three times as many women as men will sustain a hip fracture (Table 1). Largely for this reason, efforts to prevent hip fracture have focused on osteoporosis in women. However, mortality and rate of institutionalisation after hip fracture is higher in men,23 and increasing hip fracture rates among men, but not women, have been reported.24,25 The projected increased longevity in men suggests that osteoporosis will affect a growing number of Australian men, yet there are no data concerning efficacy of any drug treatment to prevent hip or vertebral fractures in men.26 Our projected number of vertebral fractures is likely to be an underestimate because our ascertainment relied on clinical indications for medical imaging. Between 50% and 75% of vertebral fractures do not come to medical attention.27,28 However, our data are likely to include some previously undiagnosed vertebral fractures that occurred before the ascertainment period. Nevertheless, the predicted number of vertebral fractures represents the number likely to come to medical attention each year. Excluding vertebral fractures does not alter the projected rate increase of all fractures. The validity of these projections is dependent upon reliable and stable incidence rates and accurate population projections. The major strength of our study was the comprehensive ascertainment of all fractures among adult residents of a defined region representative of Australia.9 A decline in hip fracture rates has recently been reported,19 but other studies have pointed to stable age-specific rates in women and younger men and increasing rates in older men.24,25,29,30 These projections highlight the need to decrease fracture rates among the elderly and can be used for setting goals and evaluating the costs and benefits of interventions in Australia. Acknowledgements This study was supported by the Victorian Health Promotion Foundation. We acknowledge the invaluable contributions of the radiologists (D Barry, JM Cameron, PJ Carman, WP Holloway, V Mercuri, PM Motterdam and DB Robertson) and staff at the Geelong Radiological Clinic, as well as the radiologists (NJ Ferris, DO Lun, CB Styles) and staff of the Medical Imaging Department at the Geelong and Colac hospitals. We also wish to thank Biljana Skoric and Soheila Panahi. References McLennan W. Projections of the populations of Australia, States and Territories, 1995-2051. Canberra: Australian Bureau of Statistics, 1996: 128. (Catalogue no. 3222.0.) Jones G, Nguyen T, Sambrook PN, Kelly PJ, Gilbert C, Eisman JA. Symptomatic fracture incidence in elderly men and women: The Dubbo osteoporosis study (DOES). Osteoporos Int 1994; 4: 277-282. Prince RL, Knuiman MW, Gulland L. Fracture prevalence in an Australian population. Aust J Public Health 1993; 17: 124-128. Lord SR. Hip fractures: changing patterns in hospital bed use in NSW between 1979 and 1990. Aust N Z Surg 1993; 63: 352-355. Bacon W, Maggi S, Looker A, et al. International comparison of hip fracture rates in 1988-89. Osteoporos Int 1996; 6: 69-75. Maggi S, Kelsey JL, Litvak J, Heyse SP. Incidence of hip fractures in the elderly: a cross national analysis. Osteoporos Int 1991; 1: 232-241. Ross KN. The development of the 1984 "Indicator of Disadvantage" and its application to resource allocation decisions for the "Disadvantaged Schools Program" in Australia. Geelong: School of Education, Deakin University, 1984. Sanders K, Pasco J, Ugoni A, et al. The exclusion of high trauma fractures may underestimate the prevalence of bone fragility fractures in the community: The Geelong Osteoporosis Study. J Bone Mineral Res 1998; 13: 1337-1342. Sanders K, Seeman E, Ugoni A, et al. The age- and gender-specific rate of fractures in Australia: a population based study. Osteoporos Int 1999. In press. Pasco JA, Henry MJ, Gaudry TM, et al. Identification of incident fractures: Geelong Osteoporosis Study. Aust N Z J Med 1999; 29: 203-206. Efron B, Tibshirani R. The introduction to the bootstrap. New York: Chapman & Hall, 1993. Australia's health 1998: the sixth biennial health report of the Australian Institute of Health and Welfare. Canberra: AGPS, 1998. Abraham B, d'Espaignet E, Stevenson C. Australian health trends 1995. Canberra: Australian Institute of Health and Welfare, 1995. Orwoll ES. The special problem of hip fracture. In: Favus MJ, editor. Primer on the metabolic bone diseases and disorders of mineral metabolism. 3rd ed. Philadelphia: Lippincott-Raven, 1996: 272-282. Keene G, Parker M, Pryor G. Mortality and morbidity after hip fractures. BMJ 1993; 307: 1248-1250. Center J, Nguyen TV, Schneider D, et al. Mortality after all major types of osteoporotic fracture in men and women: an observational study. Lancet 1999; 353: 878-882. Randell A, Sambrook P, Nguyen T, et al. Direct clinical and welfare costs of osteoporotic fractures in elderly men and women. Osteoporos Int 1995; 5: 427-432. Gray R. Insurance: the long term funding of aged care. National Healthcare 1998; 8(4): 32-33. Melton L, Atkinson EJ, Madhok R. Downturn in hip fracture incidence. Public Health Rep 1996; 111 (March/April): 146-150. Black DM, Cummings SR, Karpf DB, et al. Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures. Fracture Intervention Trial Research Group. Lancet 1996; 348: 1535-1541. Jonsson B, Christiansen C, Johnell O, Hedbrandt J. Cost-effectiveness of fracture prevention in established osteoporosis. Osteoporos Int 1995; 5: 136-142. Cooper C, Campion G, Melton LJ III. Hip fractures in the elderly: A world-wide projection. Osteoporos Int 1992; 2: 285-289. Sernbo I, Johnell O. Consequences of a hip fracture: A prospective study over 1 year. Osteoporos Int 1993; 3: 148-153. Melton LJI, O'Fallon WM, Riggs BL. Secular trends in the incidence of hip fractures. Calcif Tissue Int 1987; 41: 57-64. Bacon W. Secular trends in hip fracture occurrence and survival: Age and sex differences. J Aging Health 1996; 8: 538-553. Seeman E. Osteoporosis: trials and tribulations. Am J Med 1997; 103 (2A): 74S-87S. Kanis JA, McCloskey EV. Epidemiology of vertebral osteoporosis. Bone 1992; 13 Suppl 2: S1-S10. Cooper C, Atkinson EJ, O'Fallon M, Melton L. Incidence of clinically diagnosed vertebral fractures: a population-based study in Rochester, Minnesota, 1985-1989. J Bone Min Res 1992; 7: 221-227. Falch JA, Kaastad TS, Bohler G, et al. Secular increase and geograpical differences in hip fracture incidence in Norway. Bone 1993; 14: 643-645. Agnusdei D, Camporeale A, Gerardi D, et al. Trends in the incidence of hip fracture in Siena, Italy, from 1980 to 1991. Bone 1993; 14: S31-S34. (Received 13 Nov 1998, accepted 13 Apr 1999) Author's Detials University of Melbourne Department of Medicine, Barwon Health-The Geelong Hospital, Geelong, VIC Kerrie M Sanders, MNutrition, PhD, Research Fellow; Geoffrey C Nicholson, PhD, FRACP, Professor of Medicine; Julie A Pasco, PhD, Study Coordinator; Mark A Kotowicz, FRACP, Senior Lecturer in Medicine. Department of General Practice and Public Health, The University of Melbourne, Melbourne, VIC. Antony M Ugoni, BSc(Hons), Lecturer in Biostatistics. Austin and Repatriation Medical Centre, Melbourne, VIC. Ego Seeman, MD, FRACP, Associate Professor of Medicine, The University of Melbourne. Reprints will not be available from the authors. Correspondence: Dr M A Kotowicz, University Department of Medicine, Barwon Health-Geelong Hospital, PO Box 281, Geelong, VIC 3220. Email: m.kotowiczATmedicine.unimelb.edu.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> Back to textBack to textBack to textBack to textBack to textBack to text

Kerrie M Sanders · Geoffrey C Nicholson · Antony M Ugoni · Julie A Pasco · Ego Seeman · Mark A Kotowicz

Research An analysis of the causes of adverse events from the Quality in Australian Health Care Study Ross McL Wilson, Bernadette T Harrison, Robert W Gibberd and John D Hamilton MJA 1999; 170: 411-415 For editorial comment, see Vincent Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Administration and health services The Quality in Australian Health Care Study (QAHCS), published in the Journal in 1995,1 reported that 16.6% of hospital admissions were associated with an iatrogenic patient injury, termed an adverse event (AE) (see Box 1). This compares with the rate of 3.7% for AEs in the Harvard Medical Practice Study (HMPS),2 and a rate of 17% in a more recent study with an alternative observational method of determining AEs.3 Fifty per cent of the AEs in the QAHCS were judged to have a high preventability score (4 or more on a scale of 1-6 of increasing likelihood of preventability).1 The disability caused by these adverse events ranged from temporary disability (fully resolved in one month) in 46.6% of AEs, to death in 4.9% of AEs. Although recording AEs emphasises only the "complications" of rather than the benefits derived from healthcare, AEs are of great significance to individual patients as well as to the whole healthcare system. Abstract Objective: To examine the causes of adverse events (AEs) resulting from healthcare to assist in developing strategies to minimise preventable patient injury. Design: Descriptions of the 2353 AEs previously reported by the Quality in Australian Health Care Study (QAHCS) were reviewed. A qualitative approach was used to develop categories for human error and for prevention strategies to minimise these errors. These categories were then used to classify the AEs identified in the QAHCS, and the results were analysed with previously reported preventability and outcome data. Results: 34.6% of the causes of AEs were categorised as "a complication of, or the failure in, the technical performance of an indicated procedure or operation", 15.8% as "the failure to synthesise, decide and/or act on available information", 11.8% as "the failure to request or arrange an investigation, procedure or consultation", and 10.9% as "a lack of care and attention or failure to attend the patient". AEs in which the cause was cognitive failure were associated with higher preventability scores than those involving technical performance. The main prevention strategies identified were "new, better, or better implemented policies or protocols" (23.7% of strategies), "more or better formal quality monitoring or assurance processes" (21.2%), "better education and training" (19.2%), and "more consultation with other specialists or peers" (10.2%). Conclusion: The causes of AEs or errors leading to AEs can be characterised, and human error is a prominent cause. Our study emphasises the need for designing safer systems for care which protect the patient from the inevitability of human error. These systems should provide new policies and protocols and technological support to aid the cognitive activities of clinicians. Introduction An additional analysis of data from the Quality in Australian Health Care Study (QAHCS)1 was undertaken in order to understand more fully the causes of the adverse events (AEs) identified and to assist in developing prevention strategies. Here, we describe the error or errors in the delivery of healthcare which led to the AEs. This contrasts with our previous report,1 which focused on the patient characteristics associated with AEs and the nature and consequences of the AEs. Methods The method of determining AEs in the QAHCS has been described previously.1 The AEs were re-examined with the specific goals of determining the causes for, or the underlying errors leading to, each AE. In addition, strategies that were judged to have the potential to prevent AEs were recorded. To obtain this information the first and subsequent review forms (RF1 and RF2 forms1) collected during the QAHCS were re-examined. The source material for these forms had been the hospital medical records, but neither the hospitals nor the medical records were revisited in this analysis. Categories for the causes of the AEs were devised by an iterative process during a three-day workshop. For this, we sought additional expertise in clinical epidemiology and qualitative research methods. Using these categories, the AEs recorded on the review forms were assessed by three of the senior medical specialists who had originally reviewed the medical records in the QAHCS. All the material from each AE was reviewed by only one reviewer, as the agreement between the medical reviewers in determining the presence or absence of an AE during the QAHCS was 80% (kappa, 0.55). A proforma was completed which asked the reviewer to identify the error and then classify it by human cause and preventive strategy. All AEs were also categorised into some of the processes of clinical care. Results for the "delay", "treatment" and "investigation" categories are presented. The categories were not mutually exclusive. These data were then entered into a database, merged with the original data from the QAHCS for each case and analysed. Two of the original total of 2353 AEs were missed in this review; thus, results are given for 2351 AEs. Our analysis provides the frequency of occurrence of each of the categories of causes of AEs, together with the proportion in each category resulting in permanent disability (including death) and the proportion with high preventability. Results Human error categories Box 2 shows the frequency of occurrence of each of the human error categories, and the proportion of the AEs in each category judged to have permanent disability and high preventability. Of the 2351 AEs, 1922 (81.8%) were associated with one or more human error categories. As the error categories were not mutually exclusive, the 1922 AEs were associated with 2940 causes. "Complication of, or failure in, the technical performance of an indicated procedure/operation" was the most frequent cause of AEs; examples of this category are shown in Box 3A. Human errors associated with categories of failure of cognitive function were the next most frequent cause of AEs (Box 2). These included "Failure to synthesise, decide and/or act on available information", "Failure to request or arrange investigation, procedure or consultation", and "Misapplication of, or failure to apply, a rule; or use of a bad or inadequate rule". The most frequent error category, "complication of, or failure in, the technical performance of an indicated procedure/operation", had a lower proportion of AEs with permanent disability (14.2%). The next five most frequent human error categories all had a high proportion of AEs with permanent disability (25% or more) (Box 2). This pattern was also seen in the proportions of AEs with death as the outcome: 2.2% in the first category, and 8% or more in each of the next five categories. Of the 1201 AEs having high preventability, 9 (0.7%) were not associated with a human error category; for the remaining 1192 AEs, 2051 causes were identified (Box 2). Delay categories The importance of timeliness to the quality of healthcare led to further analysis of all AEs to ascertain the nature and role of delay in their causation (Box 4A). Delays contributed to 20.0% of AEs: of these, delays in diagnosis accounted for 56.8% and treatment delays for 40.6%. Diagnostic delay was usually the failure to make, or attempt to make, a diagnosis of a patient's condition rather than just providing symptomatic or even no treatment. Treatment delay was when the diagnosis had been made but there was a delay in initiating specific therapy. Examples of AEs in the delay category are included in Box 3B. The AEs with delay categories were judged to have very high preventability (86%-90%) compared with the average (51.2%) for all AEs (Box 4A). Treatment categories AEs categorised as caused by a treatment error were also analysed (Box 4B). In 19.6% of all AEs, treatment error contributed to the cause. The majority of AEs in this group fell into the categories of "no or inadequate treatment" (51.5%), or "wrong or inappropriate treatment" (27.4%). As with AEs caused by delay, these AEs were judged to have much higher preventability than the average for all AEs. Examples of AEs involving treatment errors are shown in Box 3C. Investigation categories Analysis of the AEs caused by patient investigation issues is shown in Box 4C, and examples are given in Box 3D. There was a problem with clinical investigation in 10.7% of AEs. Paralleling the results in the treatment category, most (78.6%) of these AEs were in this category because an investigation was not done, rather than the investigation being inappropriate (3.6%), or not acted upon (15.5%). Consistent with other AEs that are attributed to cognitive failure, there was a very high percentage of these AEs rated as high preventability. Strategies for preventing AEs When describing AEs, preventability refers to the identification of an avoidable error that led to the adverse event. This is not to say that the error could be avoided on every occasion, and that the adverse event would not occur. Rather, it implies that, with the current state of knowledge and technology, it is possible to identify and avoid that particular error, and hence reduce the probability of an AE. The reviewers were making a judgement, having identified the error, on the particular strategy for a change in the healthcare system that could have prevented the AE. The outcomes of these judgements are given in Box 5. Nineteen (1.6%) of the 1201 high preventability AEs did not have a prevention strategy category. Of the 2613 prevention strategies identified in the 1182 AEs with high preventability, 24.7% (646) were for "better education and training", 20.9% (545) were for "new or better implemented policies or protocols" and 18.6% (486) were for "more or better formal quality monitoring or assurance processes". Discussion AEs are important to patients, healthcare providers and to the custodians and funders of health services. One estimate of the national cost to the Australian healthcare system of just the additional hospital bed-days (as a result of the AEs identified in 19921) is in excess of $800 million dollars per year.4 This estimate ignores any subsequent hospital admissions and out-of-hospital healthcare expenses, loss of productivity of the patients involved, and long term community costs of permanent disability from AEs. It also ignores the benefits received from healthcare. Providing insights into how AEs occur can help in developing prevention strategies to reduce the frequency and severity of patient injuries during healthcare. Our review and analysis of the AE data from the QAHCS have shown that the causes of AEs or errors leading to AEs can be characterised, and that human error is a prominent cause. It is important to recognise that human error is inevitable for even the best-trained and best-qualified healthcare providers. Weed has recently pointed out that the unaided human mind is incapable of performing consistently at the necessary level to provide optimal healthcare.5 However, other studies6 have noted that the label "human error" is prejudicial and non-specific; it may retard rather than advance our understanding of how complex systems fail. It is postulated that within complex systems error is a symptom of organisational problems, and this is likely to apply to healthcare. Therefore, we need a healthcare-system response to error that moves the system towards being as "failsafe" as possible rather than one that blames the clinician who may have erred. Examples from the more frequently studied area of adverse drug events7 would be decision-support technology for antibiotic prescribing,8 with its demonstrated benefits, and electronic prescribing to reduce prescribing and transcription errors in hospital.9 Our analysis identified broad functional categories that are linked to the processes that make up the system of healthcare delivery and hence cut across specialties, diagnosis-related groups (DRGs) and particular patient groups. The sample size is large enough to provide useful information even when several AEs could not be classified into the categories chosen, or insufficient information was available to indicate cause. On the other hand, several factors bias the information available for assessing AEs because of an emphasis on procedures and short term outcomes and possible under-reporting of the contribution of the supporting systems to the cause of the AEs. Firstly, because the original data source was the hospital medical record, the information available about AEs is biased towards the patient involved and away from other potentially important contextual events at the time. Further, the medical record often focuses more on the actions of clinicians involved in direct or procedural patient intervention, and less on the actions of other staff or systems with a more supportive role. These and other factors will lead to an emphasis on procedures and short term outcomes, and a possible under-reporting of the contribution of supporting systems in causing AEs. Finally, information about subsequent or prior hospitalisations is usually only available if the patient attended the same hospital on all occasions. Having acknowledged these potential limitations, cognitive failure (Box 2) appears to have a role in 57% of all the causes of AEs, and most of the AEs involved were judged to be of high preventability and to have caused significant disability. These AEs were largely associated with errors of omission rather than commission. Does this represent a minimum "obligatory" error rate resulting from a combination of human error and our healthcare system, and hence which cannot be improved? Our data are not able to answer this question unequivocally, but we believe they show sufficient opportunities for moving the system towards a failsafe mode to suggest that the answer is no. Until recently there has been an under-recognition of the role and responsibilities of the healthcare system and its custodians in providing a "safe environment" using systems-improvement tools.10 One response to these data should be to look at the factors in healthcare delivery that may interfere with the cognitive or technical performance of healthcare providers. Insufficient use of information technology to assemble the necessary information at the time of decision-making may increase error. Another important factor is fatigue, which has already been shown to increase error in doctors.11 Sleep deprivation may have a much more significant role in human error in healthcare than the current work-load patterns pay heed to, but more research is needed. Other factors that may be important include the level of supervision provided to junior staff, and the pervasive effect of the culture of medical practice, which can unhelpfully portray error as individual failure or deviation from perfection.12 Our study method does not provide direct information about the role of these factors. The high proportion of causes of AEs involving cognitive failure must represent a manifestation of human error occurring in a system that is not patient protective, if one accepts that these practitioners are appropriately trained and competent by international standards. Our study provides clear guidance on methods for improvement, with "new, better, or better implemented policies or protocols" accounting for 24% of prevention strategies identified for the AEs, "quality monitoring and assurance processes" accounting for 21%, and "better education and training" for a further 19%. In summary, improvement is needed in the agreed processes of care, supported by information systems that allow general dissemination of current knowledge of diseases or treatments, and information on outcomes of care for each patient, through appropriate quality processes. Simple examples are the availability of practice guidelines and protocols at the point-of-care, and the use of automated reminders for patients and practitioners when a particular test or follow-up is required. In addition, having adequate patient "outcome" information in a form that can be benchmarked is a powerful tool in identifying unacceptable variation. Acknowledgements We acknowledge the contributions of Professor B Armstrong, Professor W R Runciman, Professor R Holland, Dr T Robertson and Dr A Hobbes. References Wilson RMcL, Runciman WB, Gibberd RW, et al. The Quality in Australian Health Care Study. Med J Aust 1995; 163: 458-471. <eMJA pdf> Brennan TA, Leape LL, Laird N, et al. Incidence of adverse events and negligence in hospitalised patients: results of the Harvard Medical Practice Study I. N Engl J Med 1991; 324: 377-384. Andrews LB, Stocking C, Krizek T, et al. An alternative strategy for studying adverse events in medical care. Lancet 1997; 349: 309-313. The Final Report of the Taskforce on Quality in Australian Health Care. Appendix 7. Canberra: AGPS, June 1996. <http://www.health.gov.au/pubs/hlthcare/toc.htm> Weed LL. New connections between medical knowledge and patient care. BMJ 1997; 315: 231-235. Cook RI, Woods DD. Operating at the sharp end: the complexity of human error. Human performance in anaesthesia: a corpus of cases. Report to the Anaesthesia Patient Safety Foundation, 1991. Columbus, Ohio: The Ohio State University: 255-307. Classen DC, Pestonick SL, Evans RS, et al. Adverse drug events in hospitalised patients: excess length of stay, extra costs and attributable mortality. JAMA 1997; 227: 301-306. Evans RS, Pestonick SL, Classen DC, et al. A computer-assisted management program for antibiotics and other anti-infective agents. N Engl J Med 1997; 338: 231-238. Bates DW, Boyle DL, Vander Vliet MB, et al. Relationship between medication errors and adverse drug events. J Gen Intern Med 1995; 10: 199-205. Leape LL. A systems analysis approach to medical error. J Eval Clin Pract 1997; 3: 213-222. Nocera A, Khursandi DS. Doctors' working hours: can the profession afford to let the courts decide what is reasonable. Med J Aust 1998; 168: 616-618. Leape LL. Error in medicine. JAMA 1994; 272: 1851-1857. (Received 4 May 1998, accepted 20 Jan 1999) Authors' details Royal North Shore Hospital, Sydney, NSW 2065. Ross McL Wilson, MB BS, FRACP, Senior Specialist in Intensive Care; Director of QARNS (Quality Assurance Royal North Shore); Bernadette T Harrison, RN RM, Manager QARNS. University of Newcastle, Newcastle, NSW 2308. Robert W Gibberd, PhD, Associate Professor, Department of Statistics; and Director of Health Services Research Group. John D Hamilton, MB BS, FRCP, Professor of Medicine, Faculty of Medicine and Health Sciences. Reprints: Dr R McL Wilson, Director of QARNS, Royal North Shore Hospital, Pacific Highway, St Leonards, NSW 2065. Email: rwilsonATdoh.health.nsw.gov.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> 1: Terms and definitions Adverse event (AE): An AE was defined as an injury or complication which resulted in disability or prolongation of hospital stay and was caused by the healthcare received rather than by the disease from which the patient suffered. The AE either occurred during the hospital admission, or during an earlier contact with healthcare services, and was responsible for all or part of the hospital admission. Error: An act of commission or omission that caused, or contributed to the cause of, the unintended injury. for errors of commission this will usually be the immediate morbid consequences of the error for errors of omission this will usually be the continuation of, and consequences of, an existing morbid state that could have been cut short, or had a better outcome, if the error had not occurred. Prevention strategy: Changes in the system in which an error occurred that mayreduce the probability of the error occurring increase the probability that the error would be remedied before an unintentional injury occurredPreventability: Preventability of an AE was assessed by the detection of "an error in management due to the failure to follow accepted practice at an individual or system level"; accepted practice was taken to be "the current level of expected performance for the average practitioner or system that manages the condition in question". Back to textBack to text 3. Examples of categories of causes of adverse events and preventability scoresA. Human error categories Example 1: A 50-year-old man sustained a bowel perforation from colonoscopy for investigation of abdominal pain. Laparotomy required. Category: "Technical performance". Preventability score: 3. Example 2: A 71-year-old man required six operations for femoral hernia repair. Category: "Technical performance". Preventability score: 2. Example 3: A 32-year-old woman had persisting severe back pain after two laminectomies, three myelograms, one decompression/fusion, and three thecal/epidural injections over 18 months. In pelvic traction at the time of review. Category: "Technical performance". Preventability score: 4. Example 4: A failed attempt at percutaneous endoscopic gastrostomy on a 32-year-old woman was followed by an open procedure. The patient died 9 days later. Autopsy revealed acute peritonitis, subphrenic abscess and bilateral pneumonia. There did not appear to have been an antemortem diagnosis of intra-abdominal sepsis, or any specific treatment for it. Category: "Technical performance". Preventability score: 6. B. Delay in diagnosis and/or treatment categories Example 1: Diagnosis of cancer of the colon was delayed until the patient, a 62-year-old woman, presented with a ruptured caecum and peritonitis from an obstructing tumour. In hospital 3 months earlier with a history suggestive of cancer of the colon and iron-deficiency anaemia, but no investigation performed. Categories: Diagnosis delay, violation of protocol or rule; failure to synthesise, decide or act on available information; lack of care/attention. Preventability score: 5.5. Example 2: A 28-year-old man with abdominal pain was treated with cholecystectomy. Gallbladder was macroscopically and histologically normal. Small bowel lymphoma was eventually diagnosed and treated, with resolution of the presenting symptoms. Categories: Diagnosis delay; failure to synthesise, decide or act on available information; failure to request or arrange an investigation, procedure or consultation. Preventability score: 5. C. Treatment categories Example 1: A 52-year-old man with known asthma was prescribed a beta-blocker for hypertension. This resulted in acute respiratory failure leading to artifical ventilation and tracheostomy. Categories: Wrong or inappropriate treatment; misapplication of or failure to apply a rule; failure to synthesise, decide or act on information. Preventability score: 6. Example 2: A 54-year-old man developed gastrointestinal bleeding (haemoglobin level, 45 g/L) while receiving non-steroidal anti-inflammatory drugs and steroids for rheumatoid arthritis. This required hospital admission and blood transfusion, at which time the correct diagnosis of osteoarthritis was made. Categories: Wrong or inappropriate treatment; acting on insufficient information; failure to request or arrange an investigation, procedure or consultation. Preventability score: 5.5. Example 3: Hospitalisation and surgical intervention for septic arthritis that followed steroid injection into a joint. Categories: Unclassified treatment; technical; lack of care/attention. Preventability score: 4.5. D. Investigation categories Example 1: A 75-year-old woman died from acute renal failure after developing gentamicin toxicity. Gentamicin was used to treat an infected pleural effusion, and drug levels were not measured. Categories: Investigation not performed; failure to request or arrange an investigation, procedure or consultation; lack of care/attention. Preventability score: 5. Example 2: A 58-year-old woman had recurrent hospital admissions for chest pain and impaired cardiac function without specific investigation, and hence a reduction in treatment options. Categories: Investigation; violation of protocol or rule; failure to synthesise, decide and/or act on available information. Preventability score: 5. Back to text 4: Contribution of delay, treatment and investigation categories to adverse events (AEs). Values are number (%) of AEsPermanentHighA. Delay categoryFrequencydisabilitypreventabilityDiagnostic delay267 (56.8%)93 (34.8%)231 (86.5%)Treatment delay191 (40.6%)53 (27.7%)172 (90.1%)Administrativedelay12 (2.6%)3 (25.0%)11 (91.7%)Total470 (100%)149 (31.7%)414 (87.9%) B. Treatment categoryNo or inadequatetreatment237 (51.5%)72 (30.4%)176 (74.3%)Wrong/inappropriatetreatment126 (27.4%)35 (27.8%)96 (76.2%)No or inadequateprophylaxis41 (8.9%)9 (22.0%)34 (82.9%)Treatmentunclassified36 (7.8%)7 (19.4%)28 (77.8%)Missed treatment20 (4.4%)4 (20.0%) 16 (80.0%)Total460 (100%) 127 (27.6%)350 (76.1%) C. Investigation categoryInvestigationnot performed198 (78.6%)81 (40.9%)171 (86.4%)Investigationnot acted on39 (15.5%)13 (33.3%)36 (92.3%)Investigationinappropriate9 (3.6%)2 (22.2%)9 (100.0%)Investigationunclassified6 (2.4%)2 (33.3%)5 (83.3%)Total252 (100%)98 (38.9%)221 (87.7%)Back to text 5: Frequency of occurrence of categories of prevention strategies and the proportion of adverse events (AEs) judged as causing permanent disability or having high preventability. Values are number (%) of AEsPermanentHighCategoryFrequencydisabilitypreventabilityNew, better, or better implementedpolicies or protocols884 (23.7%)206 (23.3%)545 (61.7%)More or better formal quality monitoringor assurance processes790 (21.2%)186 (23.5%)486 (61.5%)Better education and training715 (19.2%)160 (22.4%)646 (90.3%)Consultation with other specialistsor peers391 (10.5%)133 (34.0%)294 (75.2%)Don't know341 (9.2%)51 (15.0%)186 (54.5%)Better access to, or transfer of, information135 (3.6%)40 (29.6%)100 (74.1%)Discharge procedures and protocols122 (3.3%)27 (22.1%)100 (82.0%)Other89 (2.4%)22 (24.7%)46 (51.7%)Changes in organisation management88 (2.4%)22 (25.0%)75 (85.2%)Changes in organisation culture77 (2.1%)26 (33.8%)66 (85.7%)More or better personnel72 (1.9%)29 (40.3%)53 (73.6%)More or better equipment or otherphysical resources22 (0.6%)8 (36.4%) 16 (72.7%)Total3726 (100%)*910 (24.4%)2613 (70.1%) * Total is greater than the number of AEs (2351) as the categories were not mutually exclusive. Back to text

Bernadette T Harrison · Robert W Gibberd · John D Hamilton

Research Consultants in cases of intended euthanasia or assisted suicide in the Netherlands Bregje D Onwuteaka-Philipsen, Gerrit van der Wal, Piet J Kostense and Paul J van der Maas MJA 1999; 170: 360-363 For editorial comment, see Hendin Abstract - Introduction - Methods - Results - Discussion - Acknowledgement - References - Authors' details - - More articles on Ethics Abstract Objective: To investigate how often physicians act as a consultant in the review of intended euthanasia and assisted suicide (EAS), by whom physicians are asked to act as a consultant, and the consultant's reasons for not agreeing with the intended performance of EAS. Design: A retrospective descriptive study. Setting: The Netherlands. Participants: A stratified random sample of 405 Dutch physicians. Main outcome measures: Number of times the physician has been a consultant; how often a physician had previously been asked to be a consultant by the same treating physician; why consultants advised against EAS. Results: 42% of interviewed physicians had acted as a consultant for EAS and 11% had been a consultant more than three times. Half the physicians who acted as a consultant more than once were invited to do so by the same attending physician, and 41% of consultants had previously consulted the attending physician. The main reasons consultants advised against EAS were because treatment options were still available, the patient's request was not well-considered or persistent, and the patient's suffering was not unbearable and hopeless. Conclusions: Many physicians have at some time been a consultant in a case of intended EAS, but only very few have been able to gain experience in consultancy. To guarantee high standards of consultation, it may be advisable to appoint and train specific consultants for EAS. Introduction In the Netherlands, physicians can be involved in euthanasia or assisted suicide (EAS) either by performing or by being consulted by another physician who intends to perform EAS. Consultation is considered to be an important aspect in the review of cases of EAS,1 and is a requirement that must be met to avoid prosecution for performing EAS. Consultation takes place in about 99% of reported cases of EAS (41% of cases are reported) and in about 37% of unreported cases. In 88% of cases of EAS in which consultation took place the consultant had seen the patient.2,3 In a consultation, a physician formally confers with an independent colleague in considering whether to grant a request for EAS. The consultant determines whether the patient's physician has acted according to the requirements for prudent practice: the patient's request is voluntary, well-considered and persistent, the patient's suffering is unbearable and hopeless, and there are no alternatives for treatment remaining. A consultant should be competent and independent of both the attending physician and the patient, and should visit the patient.1,4 Consultation for EAS is a relatively new task for physicians, and is not part of general medical training. Physicians generally are not used to pronouncing formal judgement on the decision-making process of their colleagues and are not used to seeing patients in a consultancy role. Moreover, consultants need to have knowledge relevant to EAS, such as the possibilities of palliative care.5 Gaining experience as a consultant seems to be important for a physician to become comfortable in this role. As part of a nationwide study on EAS and other practices involving the end of life in the Netherlands,3,6 we set out to determine how often physicians act as a consultant, whether physicians frequently consult the same colleague (or vice versa), whether groups of physicians are asked to be a consultant more often than other groups, the reasons why consultants do not agree with the intended performance of EAS, and the extent to which the consultant feels responsible for the attending physician's final decision to grant or refuse a request for EAS. Methods Study population For this retrospective, descriptive study, we interviewed a stratified random sample of Dutch physicians.2 Participants were stratified according to specialty. The physicians in each stratum were ordered by postal code of their work address and every nth physician was selected. The stratum size was based on the frequencies of medical decisions concerning the end of life and the homogeneity of the patient population (morbidity and age) per stratum. To interview the desired number of physicians, 559 were included in the sample; 83 did not meet the selection criteria and 21 had a chronic illness or could not be located. Of the remaining 455, 50 (11%) did not respond. The final sample of 405 physicians included 124 general practitioners, 74 nursing home physicians (nursing homes in the Netherlands are multifunctional institutions which care for predominantly elderly patients with chronic diseases and physical and/or mental disorders and handicaps) and 207 specialists in cardiology, surgery, internal medicine, respiratory medicine and neurology (oncology and palliative care are not distinct specialties in the Netherlands, but are practised by specialists in other disciplines, such as internal medicine). Physicians in the above-listed specialties attend 87% of all deaths occurring in hospitals. Together with the general practitioners and nursing home physicians, they attend about 95% of all deaths in the Netherlands. Physicians who were not practising in their registered specialty in the same institution since 1 January 1994 were excluded. Measuring instruments The interviews were conducted from November 1995 to February 1996 by 30 experienced physicians who were specially trained for the study. The questionnaire consisted mainly of open-ended questions for the respondents, with prestructured response categories for the interviewers. In the interviews, the definitions of euthanasia, assisted suicide and consultation (Box 1) were explicitly described to the respondents. The interviews took about 2-3 hours. For this study we predominantly used data on how often the respondents had been consultant ever and in 1994 and 1995, and data on the most recent case (all 108 cases occurred between 1994 and 1996) in which the respondent had been a consultant. Analysis To make the data of the stratified samples representative for all deaths in the Netherlands, we weighted the data per stratum. We calculated weights based on the proportion of the various types of physicians in the sample. In addition, the weights of the five specialties were corrected for the 13% of in-hospital deaths that were attended by other medical specialists. Proportions and 95% confidence intervals for these proportions were obtained by direct standardisation7 to adjust for marked variation among the different types of physicians. The normal approximation to the binomial distribution was used. Estimates of the number of consultations in 1995 were based on the (weighted) data on how often each physician had acted as consultant in 1994 and 1995. Multiple logistic regression analyses were used to obtain insight into determinants of whether physicians had ever been a consultant. Because of the stratification according to specialty, the variable "specialty" was included in all analyses. To deal with this categorical variable we used indicator variables, choosing the general practitioners as the reference category. Results In 1995, in the Netherlands, we estimate that almost 4000 consultations took place in cases of intended EAS. In most of these, the consultant was a general practitioner; nursing home physicians rarely acted as a consultant (Box 2). Physicians who had been a consultant Of the 405 physicians in the sample, 42% at some time had been a consultant in intended EAS (Table 1). In 1994 (the year in which the notification procedure was legally enforced) or 1995, 32% of physicians had been a consultant -- general practitioners more often than medical specialists, and specialists more often than nursing home physicians (Box 2). Eight physicians had refused to act as a consultant, for various reasons: lack of time (2), not independent of the attending physician or the patient (2), doubt whether requirements for prudent practice had been met (2), disagreement with notification procedure (1), or the attending physician did not intend to perform euthanasia (1). Eleven per cent of the physicians had been a consultant more than three times (Box 3), and in 1995, 3% had been a consultant three or more times. Previous consultations between consultant and consulting physician Half of the physicians who had been a consultant more than once had previously been consulted by the same physician who consulted them in their most recent case. In 24% of these cases, the treating physician and the consultant had previously acted as consultants for each other (Box 4). Physicians who previously consulted or had been consulted by the same physician agreed more often with the intended EAS than physicians who did not (90% v. 80%), but this difference was not significant. Reasons for advising against EAS The 28 physicians who had at some time advised against the performance of EAS were each asked to describe up to three such cases. Together, they described 48 cases in which they had given this advice. In 42 cases EAS was not carried out, in three it was, and in three instances the consultant did not know the outcome. The consultants gave the following reasons for advising against EAS: there were still alternative treatment options (20 cases), the patient's request was not well-considered or persistent (12 cases), the patient's suffering was not unbearable and hopeless (nine cases), the request was made under pressure of the family (five cases), the patient was already dying (five cases), and the attending physician felt manipulated by the patient (one case). The consultant's responsibility Sixty-five per cent of consultants considered that they had joint responsibility only in those cases in which the attending physician acted according to their judgement, and 30% did not consider that they had any joint responsibility. Medical specialists felt that they had joint responsibility more often than general practitioners (80% [95% CI, 68.6%-91.3%] v. 61% [95% CI, 47.7%-74.6%]). Determinants of having been a consultant With univariate analyses corrected for specialty, physician's age, sex, religion, region in which the physician lives, belief that every case of EAS should be reviewed and belief that consultation should take place in every case of EAS were not significantly related to whether the physician had been a consultant. The results of multiple logistic regression analysis for those determinants that were predictive in the univariate analysis are shown in Box 5. Male physicians had more often been a consultant than female physicians. The strongest association was found for the variable "ever performed EAS". Physicians who had performed EAS had been a consultant more frequently than physicians who had never performed EAS. Discussion We estimate that consultation with another physician in cases of intended EAS took place almost 4000 times in the Netherlands in 1995 (see Box 6). In about 60% of consultations, the consultant was a general practitioner. Overall, 42% of Dutch physicians had been a consultant; 11% had been a consultant more than three times. The most common reason why consultants advised against the performance of EAS was the availability of alternative treatment options. Most consultants considered that they have joint responsibility for the final decision to grant or refuse a request for EAS. Male physicians, general practitioners and physicians who had performed EAS had more frequently been a consultant. The forming of "pairs" of consultants suggests that familiarity is very important in the choice of consultant. An earlier study found that an important reason for choosing a consultant is accessibility and that physicians mainly consult physicians of their own specialty.2 A problem with these consultations may be that the independence of the consultant with regard to the attending physician might be threatened. This is suggested by the fact that "consultants of a pair" more often agree with the intended performance of EAS than other consultants, although this difference is not statistically significant. A possible way of assuring independence of the consultant while safeguarding the consultant's accessibility would be to appoint independent trained consultants who could be contacted by all physicians in a region. Such a system was implemented for general practitioners in Amsterdam in 1997.8 It might also be useful in increasing the frequency of consultation, and possibly the reporting of EAS. The reasons given by consultants for advising against EAS all related to the requirements for prudent practice; the fact that physicians very rarely carry out EAS when the consultant advises against it suggests that consultation can have an important function in assuring the quality of this kind of medical practice. However, our results do not show how often consultants agreed with EAS in cases in which not all the requirements for prudent practice were met. Nursing home physicians, neurologists, surgeons and cardiologists are less likely than general practitioners to have been a consultant. A possible explanation is that, for being asked to act as a consultant, it is not only important to have carried out EAS, but also to have done so relatively frequently: general practitioners, respiratory specialists and specialists in internal medicine carry out EAS more frequently than other physicians.6 In general, Dutch physicians do not have much experience in acting as a consultant in cases of intended EAS. Of the 42% of physicians who have been a consultant, only 27% had been a consultant more than three times, and only 3% more than 10 times. Because acting as a consultant differs greatly from a physician's normal working relationship with colleagues and patients, and because the consultation concerns a matter of life and death, it is important that consultants are experienced and specifically trained. A training program for consultants, in which the skills needed, knowledge about the requirements for prudent practice, palliative care and medicotechnical aspects of EAS are addressed, has been developed by the Royal Dutch Medical Association. In the future it might be advisable to permit only specifically trained physicians to act as a consultant. Acknowledgement This study was funded by the Dutch Ministry of Health, Welfare and Sports and the Ministry of Justice. We are indebted to Professor J Th M van Eijk for his comments on previous versions of this manuscript. References Board of the Royal Dutch Medical Association. Vision on euthanasia. In: Euthanasia in the Netherlands. 5th ed. Utrecht 1996. 24-56. Van der Wal G, van der Maas PJ. Euthanasia and other medical decisions concerning the end of life. The Hague, the Netherlands: Staatsuitgeverij, 1996 (in Dutch). Van der Wal G, van der Maas PJ, Bosma JM, et al. Evaluation of the notification procedure for physician-assisted death in the Netherlands. N Engl J Med 1996; 335: 1706-1711. Van der Wal G, Dillmann RJM. Euthanasia in the Netherlands. BMJ 1994; 308: 1346-1349. Onwuteaka-Philipsen BD. The role of the consultant. In: Legemaate J, Dillmann RJM, editors. Physician-assisted death: between norm and practice. 105-114. Bohn Stafleu Van Loghum, Houten 1998 (in Dutch). Van der Maas PJ, van der Wal G, Haverkate I, et al. Euthanasia, physician-assisted suicide, and other medical practices involving the end of life in the Netherlands, 1990-1995. N Engl J Med 1996; 335: 1699-1705. Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford: Blackwell, 1994; 436-440. Dillman RJM, Krug CHM, Onwuteaka-Philipsen B, et al. Support and consultation in cases of euthanasia in Amsterdam. Med Contact 1997; 52: 743-745 (in Dutch). (Received 7 Jul, accepted 21 Dec, 1998) Authors' details Institute for Research in Extramural Medicine, Vrije Universiteit, Amsterdam, The Netherlands. Bregje D Onwuteaka-Philipsen, MSc, Researcher, and Department of General Practice, Nursing Home Medicine and Social Medicine; Gerrit van der Wal, MD, PhD, Professor, and Department of General Practice, Nursing Home Medicine and Social Medicine; Piet J Kostense, PhD, Epidemiologist/Statistician, and Department of Epidemiology and Biostatistics. Department of Public Health, Erasmus University, Rotterdam, The Netherlands. Paul J van der Maas, MD, PhD, Professor. Reprints will not be available from the authors. Correspondence: B D Onwuteaka-Philipsen, Vrije Universiteit, Institute for Research in Extramural Medicine, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands. Email: B. Philipsen. EMGOATmed.vu.nl Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> 1: Definitions Euthanasia was defined as the administration of drugs with the explicit intention of ending the patient's life, at the patient's explicit request. Assisted suicide was defined as the prescription or supply of drugs with the explicit intention of enabling the patient to end his or her own life. Consultation was defined as consultation of a colleague, as stipulated in the notification procedure for EAS. Back to text 2: Physicians consulted in cases of intended EAS, and number of consultations in 1995 in the NetherlandsGeneral practitioners Medical specialists (n=124)* (n=207)† Proportion§ (95% CI) Proportion§ (95% CI) Physicians consultedEver consulted49% (40.2%-58.1%) 30% (23.9%-36.6%) Consulted in 1994 or 199540% (31.3%-49.1%) 22% (16.5%-28.2%) Number (95% CI) Number (95% CI) Consultations in 19952502 (2004-3086) 1424 (1217-1631) Nursing home physicians Total (n=74)‡ (n=405) Proportion§ (95% CI) Proportion§ (95% CI) Physicians consultedEver consulted19% (10.9%-30.1%) 42% (35.6%-47.7%) Consulted in 1994 or 199514% (6.77%-23.7%) 32% (25.1%-40.4%) Number (95% CI) Number (95% CI) Consultations in 199559 (33-98) 3985 (3419-4551) * 4 missing observations; †1 missing observation; ‡ 6 missing observations; § Calculated by direct standardisation.7 Back to text 3: Number of consultations (ever and in 1995) by physiciansGeneral practitioners (n=124)* Specialists (n=207)* Total (n=405)* Proportion‡ (95% CI) Proportion‡ (95% CI) Proportion‡ (95% CI) Number of consultations everno consultation51% (41.9%-59.8%) 69% (62.3%-75.5%) 58% 1 consultation15% (8.61%-21.4%) 10% (5.9%-14.7%) 13% more than one34% (25.7%-42.7%) 21% (15.1%-26.4%) 29% 2 or 323% 7% 18% 4 or 5 5% 4% 5% 6 to 10 5% 7% 5% more than 101% 2% 1% Number of consultations in 1995no consultation54% (70.0%-85.0%) 83% (77.3%-88.1%) 80% one or more46% (36.9%-54.7%) 17% (11.9%-22.7%) 20% 1 consultation32% 9% 13% 2 consultations9% 3% 4% more than 25% 5% 3% * General practitioners: 4 missing observations; medical specialists: 2 missing observations; total: 7 missing observations. †Includes 74 nursing home physicians. ‡ Calculated by direct standardisation.7 Back to text4: Previous consultations between attending physician and consultant*General practitioners (n=40) Specialists (n=39) Total (n=80) Proportion‡ (95% CI) Proportion‡ Proportion‡ Attending physician previously consulted the consultant§53% (36.1%-68.5%) 40% 50% Consultant previously consultedthe attending physician¶46% (30.1%-62.8%) 22% 41% Attending physician and consultantpreviously consulted each other§28% (14.6%-43.9%) 11% 24% CI = confidence interval. * Analysis is restricted to those physicians who had been a consultant twice or more and described their most recent consultation for EAS. †Numbers are too small for calculating confidence intervals. ‡ Calculated by direct standardisation. § Medical specialists: 4 missing observations; total: 4 missing observations. ¶General practitioners: 1 missing observation; medical specialists: 3 missing observations; total: 4 missing observations. Back to text5: Determinants of having ever been a consultant (n = 405)*Number Odds ratio (95% CI) Specialtygeneral practice119 1 nursing homemedicine72 0.42 (0.20-0.87) neurology34 0.36 (0.15-0.89) respiratorymedicine35 0.96 (0.43-2.12) surgery35 0.36 (0.15-0.90) cardiology32 0.21 (0.07-0.65) internal medicine64 0.78 (0.41-1.49) Male327 2.14 (1.08-4.85) Ever carried outEAS166 3.04 (1.91-4.85) * 14 missing observations. CI = confidence interval. EAS = euthanasia or assisted suicide. Back to text6: Confirmation of our estimate of number of consultations From our data on how often physicians had acted as a consultant in 1994 and 1995, we estimated that almost 4000 consultations took place in 1995. This estimate is reasonably consistent with the estimate we can make based on other data. There were 3600 granted requests for EAS, and 44% of 6100 refused requests. Consultation occurs in 63% of acceded requests (in 10%, the attending physician consults two, and in 2% three or more, colleagues) and in 16% of refused requests.2,3 Number of consultations = 0.63 x 3600 + 0.63 x 0.1 x 3600 + 0.63 x 0.02 x 3600 x 2 + 0.16 x 0.44 x 6100 = 3015 consultations. This estimate does not take into account the (unknown) frequency of consultation in the approximately 2625 (0.43 x 6100) requests for EAS that were not carried out because the patient died.6 Back to text

Bregje D Onwuteaka-Philipsen · Piet J Kostense

Research Rainfall, evaporation and the risk of non-contact anterior cruciate ligament injury in the Australian Football League John Orchard, Hugh Seward, Jeanne McGivern and Simon Hood MJA 1999; 170: 304-306 Abstract - Introduction - Methods - Results - Discussion - Acknowledgement - References - Authors' details - - More articles on Emergency medicine Abstract Objective: To determine if weather conditions affect the risk of anterior cruciate ligament (ACL) tear in Australian Football. Design: Prospective observational analytic study of football matches. Setting: The Australian Football League (AFL), a professional competition. Participants: All players in 2280 matches from 1992-1998. Main outcome measures: Surgically-proven ACL injury, not involving a direct contact mechanism, during a match; rainfall; water evaporation. Results: 59 ACL injuries not involving direct contact occurred during the study period, more commonly in cities north of Melbourne (χ2 = 17.0; df = 1; P < 0.001). Senior grade matches (relative risk [RR], 3.03; 95% confidence interval [CI], 1.52-6.03), high water evaporation in the month before the match (RR, 2.80; 95% CI, 1.53-5.10) and low rainfall in the year before the match (RR, 1.93; 95% CI, 1.12-3.34) were significantly associated with these injuries. Conclusion: Low water evaporation and high rainfall significantly lower the risk of ACL injuries in AFL footballers. The likely mechanism is a softening of the ground, which lowers shoe-surface traction. Consistent extra watering and covering of grounds during periods of high water evaporation may lower the rate of ACL injuries. Introduction Anterior cruciate ligament (ACL) injuries of the knee are the most costly injuries in football at both professional and amateur levels. Of the commonly occurring injuries, they are the most devastating for the player, as they usually require reconstruction and long term rehabilitation in order to retain normal knee function. In the Australian Football League (AFL), ACL injuries account for 12% of all missed playing time,1 which costs the competition over $1 million annually in injury payments alone. The rate of ACL injury in the AFL competition has increased over recent years.2 The exact numbers and costs of ACL injuries in amateur football are unknown. One report that tried to estimate the cost of sports injuries in Australia3 suggested that the cost of ACL injuries in all grades and codes of football was at least 10% of the nation's entire bill for sports injuries. It estimated 5000 football-related ACL injuries in Australia annually, directly costing $5000 each (not including lost work time and the increased susceptibility to later knee problems caused by these injuries).4 Recent studies have noted that there have been relatively more ACL injuries in the early months of the AFL season, in non-Victorian games, and in recent seasons of dry weather.2,5 Our aim was to examine the strength of the relationship of weather conditions to ACL injury. Methods We studied 2280 matches (all AFL home and away season, finals, reserve grade and Ansett Cup matches which occurred from 1992-1998 in the cities of Adelaide, Brisbane, Canberra, Darwin, Geelong, Hobart, Melbourne, Perth and Sydney), involving 2239 individual players. Approximately 75% of the matches were held in Melbourne, 5% to the south (Geelong and Hobart) and the remaining 20% in the northern cities. The occurrence of ACL injuries was determined by an ongoing injury surveillance program, which has attempted to determine the cause of every game missed through injury of listed players. The accuracy of this survey in determining the cause of missed games reached 100% in 1997, and has been greater than 90% in every year.1 In addition, a separate register for ACL injuries is held by the AFL. ACL injuries in the AFL are managed surgically and have the diagnosis proven at the time of simultaneous arthroscopy and reconstruction. The mechanism of injury was recorded by team doctors and divided into the categories of direct contact (where the player's knee or leg was contacted by another player) and no direct contact (which was subdivided into indirect contact and no contact). Daily weather variables were measured prospectively (but obtained by us retrospectively) by the Bureau of Meteorology at central locations in each city studied. No differentiation was made between rainfall in the differing suburbs of any city, except for Waverley Park (30 km from the centre of Melbourne), for which rainfall data were taken from the nearest centre (Moorabbin). Maximum and minimum temperature of the day of the match, rainfall, water evaporation and maximum wind gust were the raw weather variables considered. Water evaporation is a meteorological variable measuring the change of surface water into water vapour -- it is affected by temperature, sunshine, humidity and wind.6 Composite rainfall and evaporation variables were created (measuring the totals for the previous 7, 14, 28, 90 and 365 days), which were considered to have had potentially more effect on ground conditions than the weather only on the day of the match. Statistical analysis was carried out using SPSS for Windows.7 Chi-square and t tests (after variances assessed using Levene's test) were performed in the initial stages of the analysis. Multivariate analysis was performed using a logistic regression forward stepwise technique, with a significance of 0.05 to enter the equation. Continuous variables in the logistic regression were redefined into binary variables based on group median values, to calculate risk ratios adjusted for confounding. Results There were 111 surgically-proven ACL tears recorded in AFL players during the study period. Excluded from analysis were 33 injuries that did not occur in AFL matches (13 occurred during practice matches, 14 during training sessions and six in other league matches). Of the 78 injuries in AFL matches, 19 occurred through direct contact, 15 involved indirect contact (to another part of the body than the leg) and there was no contact involved in 44. Overall, games played north of Melbourne had a significantly higher rate of all ACL injuries (Table). However, injuries involving contact were not significantly associated with being north of Melbourne, although indirect-contact injuries were more common north of Melbourne. For the 59 ACL injuries that did not involve direct contact, the most predictive rainfall and evaporation variables for ACL injury were 28-day evaporation (t = - 3.8; df = 59; P < 0.001) and 365-day rainfall (t = 1.34; df = 2278; P = 0.18). Other significant associations were higher grade of match (t = 4.8; df = 59; P < 0.001), lower minimum temperature (t = - 3.7; df = 59; P < 0.001) and month (fewer injuries in winter) (t = 3.1; df = 2278; P = 0.002). Non-significant associations were night games (t = - 1.7; df = 59; P = 0.09) and wind speed (t = - 0.18; df = 2278; P = 0.86). Although 365-day rainfall was not significant according to the t test, this variable was entered into the logistic regression equation as significant. This is because of the confounding effect of evaporation -- venues with high evaporation also tend to have high rainfall. In a multivariate analysis including evaporation, variance in rainfall is more significant than when rainfall is considered alone. In contrast, night games and minimum temperature had less significance in the logistic regression model. All variables were considered in a logistic regression model with match grade, 28-day evaporation and 365-day rainfall qualifying to be entered into the equation. Senior grade matches had a 3.03-times risk of injury compared with reserve grade matches (95% confidence interval [CI], 1.52-6.03). Matches with high evaporation in the previous 28 days (> 48 mm) had a 2.80-times greater risk of injury (95% CI, 1.52-6.03). Matches with low rainfall (< 449 mm) in the previous year had a 1.93-times greater risk of injury (95% CI, 1.12-6.03). Figures 1 and 2 illustrate a strong association between the rate of ACL injuries not involving direct contact and 28-day evaporation and 365-day rainfall, respectively, for the matches in Melbourne over the study period. Discussion Risk factors for injury can be divided into intrinsic (personal) and extrinsic (environmental). Two intrinsic risk factors for ACL injury have been established: female sex8-11 and narrowed intercondylar notch.12-14 In a recent study, both factors were prospectively followed, and it was concluded that narrow intercondylar notch accounts for the increased number of ACL tears seen in female athletes. At this stage, narrowed intercondylar notch is not a reversible risk factor, but it can be measured at the time of injury and used to counsel athletes on the risk of recurrence, particularly to the contralateral knee. The most promising extrinsic risk factors postulated relate to the shoe-surface interface. It has been hypothesised for many years that increasing traction between a football boot and the playing surface would cause an increase in the rate of knee injuries.15 A recent study showed that American football boots, with a greater number of cleats and higher torsional resistance, were prospectively associated with an increased number of ACL injuries.16 Our results show that high water evaporation in the month before and low rainfall in the year before an AFL match confer an increased risk of ACL injury. This relationship is strong for most injuries not involving contact, but may not apply when the mechanism of injury involves a direct blow to the knee. The mechanism of this association is almost certainly through ground-related factors (either soil moisture content or amount and/or quality of grass). Increased speed of the game on dry grounds may also be relevant. These results are in keeping with the theory that excessive shoe-surface traction is a risk factor for ACL injury. Friction and torsional resistance from football boots has been shown to be higher in dry conditions on natural grass compared with wet conditions.17 In a recent review of non-contact ACL injuries in American Football, it was noted that almost all injuries on natural grass occurred in dry conditions,18 but that review did not measure or control for conditions on days when injury did not occur. We found that the long term effects of rainfall and water evaporation (over a period of months) are more relevant than the amount of rainfall and evaporation on the day or in the days leading up to a match. Further studies are required to determine the exact mechanism by which this relationship occurs and what the effects of intervention would be. The AFL has embarked on a study, beginning with the 1998 season, where matches have ground hardness readings taken with a Penetrometer, a device used in horse racing to measure track hardness. Penetrometer results are reliable and correlate with the speed of races.19 The preliminary results of the AFL study confirm that low rainfall and high evaporation are associated with hard (low) Penetrometer readings (< 4.7 cm).20 We did not study differences between ground conditions in coastal and inland locations, as (except for Canberra) all the major AFL venues are currently in coastal cities. Further study could include inland cities, which may experience different ground conditions due to much colder and drier winters. Irrespective of their mechanism of action, the effects of high water evaporation and low rainfall could be reversed on football grounds by consistently watering grounds during times of lower rainfall and covering them during times of increased sunshine. If grounds were prepared in this way to simulate the ground conditions typically experienced in a Melbourne winter, the number of ACL injuries would almost certainly be reduced. This is unlikely to lead to an increase in injuries of other types, as overall injury rates are consistently higher outside Victoria than in Victoria.1,5 In rugby league in the United Kingdom, injury rates increased across the board when the season was changed from winter to summer,21,22 suggesting harder grounds may be a universal risk factor for football injuries. In AFL, the main competing concern would be whether games played on consistently softer grounds would be a lesser spectacle for the attending and television public. In amateur football, similar reductions in ACL injuries could probably be achieved, without the standard of spectacle being an issue. The cost of manipulating ground conditions might seem initially prohibitive in this environment, but when the massive cost of ACL injury is considered, great overall savings could be made. In conclusion, low water evaporation and high rainfall significantly lower the risk of non-contact ACL injuries in AFL footballers. The mechanism is likely to be a softer surface, with lower shoe-surface traction and consequently less force transferred to the knee in movements such as pivoting. Consistent extra watering and covering of grounds during times of high water evaporation (sunny, windy periods with no rain) is likely to lower the rate of these devastating and costly injuries. Acknowledgement The Australian Football League funds the AFL Injury Surveillance System from which data for this study were derived. References Orchard J, Wood T, Seward H. AFL injury report 1997. Football Record 1998; 87: 54-61. Seward H. Can ACL injuries be prevented? In: Australian Conference of Science and Medicine in Sport. Canberra: Sports Medicine Australia, 1997. Egger G. Sports injuries in Australia: causes, costs and prevention. Sydney: National Better Health Program, 1990. Deacon A, Bennell K, Kiss ZS, et al. Osteoarthritis of the knee in retired, elite Australian Rules footballers. Med J Aust 1997; 166: 187-190. Orchard J, Seward H, Garlick D. Ground conditions and AFL injuries. In: Australian Conference of Science and Medicine in Sport. Canberra: Sports Medicine Australia, 1997. Lewis R, editor. Meteorological glossary. 6th ed. London: HMSO Publications, 1991. SPSS for Windows [computer program]. Version 6.0. Chicago, Ill: SPSS Inc, 1992. Arendt E, Dick R. Knee injury patterns among men and women in collegiate basketball and soccer. NCAA data and review of the literature. Am J Sports Med 1995; 23: 694-701. Ferretti A, Papandrea P, Conteduca F. Knee ligament injuries in volleyball players. Am J Sports Med 1992; 20: 203-207. Gray J, Taunton J, McKenzie D. A survey of injuries to the anterior cruciate ligament of the knee in female basketball players. Int J Sports Med 1985; 6: 314-316. Bjordal J, Arnoy F, Hannestad B, Strand T. Epidemiology of anterior cruciate ligament injuries in soccer. Am J Sports Med 1997; 25: 341-345. Souryal T, Moore H, Evans J. Bilaterality in anterior cruciate ligament injuries: associated intercondylar notch stenosis. Am J Sports Med 1988; 16: 449-454. Shelbourne K, Facibene W, Hunt J. Radiographic and intraoperative intercondylar notch width measurements in men and women with unilateral and bilateral anterior cruciate ligament tears. Knee Surg Sports Traumatol Arthrosc 1997; 5: 229-233. Shelbourne K, Davis T, Klootwyk T. The relationship between intercondylar notch width of the femur and the incidence of anterior cruciate ligament tears: a prospective study. Am J Sports Med 1998; 26: 402-408. Torg J, Quendenfeld T, Landau B. The shoe-surface interface and its relationship to football knee injuries. J Sports Med 1974; 2: 261-269. Lambson R, Barnhill B, Higgins R. Football cleat design and its effect on anterior cruciate ligament injuries: a three year prospective study. Am J Sports Med 1996; 24: 155-159. Heidt R, Dormer S, Cawley P, et al. Differences in friction and torsional resistance in athletic shoe-turf surface interfaces. Am J Sports Med 1996; 24: 834-842. Scranton P, Whitesel J, Powell J, et al. A review of selected noncontact anterior cruciate ligament injuries in the National Football League. Foot Ankle Int 1997; 18: 772-776. Neylan J, Stubbs A. Assessing racetrack conditions: a review of available devices. Canberra: Rural Industries Research & Development Corporation, 1998. Orchard J. Measurement of football ground hardness using the racetrack Penetrometer [abstract]. Med Sci Sports Exer. In press, 1999. Hodgson Phillips L, Standen P, Batt M. Effects of seasonal change in rugby league on the incidence of injury. Br J Sports Med 1998; 32: 144-148. Gissane C, Jennings D, White J, Cumine A. Injury in summer rugby league football: the experiences of one club. Br J Sports Med 1998; 32: 149-152. (Received 30 Jul, accepted 28 Nov, 1998) Authors' details Sports Medicine Unit, University of New South Wales, Kensington, NSW. John Orchard, MB BS, FACSP, Visiting Fellow. Australian Football League Medical Officers Association, Melbourne, VIC. Hugh Seward, MB BS, FACSP, President; Jeanne McGivern, MB BS, FRCS, Club Medical Officer; Simon Hood, BAppSci(PE), Research Officer. Reprints: Dr J Orchard, South Sydney Sports Medicine, 111 Anzac Parade, Kensington, NSW 2033. Email: johnorchardATmsn.com.au Make a comment Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> Back to text Back to text Back to text

John Orchard · Hugh Seward · Jeanne McGivern · Simon Hood

Research Colorectal cancer after open-access colonoscopy: a community and case survey John Croese MJA 1999; 170: 251-254 For editorial comment, see Bolin & Korman Abstract - Introduction - Methods - Results - Discussion - References - Author's Details - - More articles on Gastroenterology Abstract Objective: To evaluate whether colonoscopy protects against subsequent colorectal cancer (CRC). Design: Case and population survey. Setting: Townsville region in northern Queensland, between mid 1985 and January 1998. Subjects: All 8430 patients who underwent 11 148 colonoscopies performed by the author at the Mater Private Hospital (a community-based open-access colonoscopy service) between July 1985 and December 1996; those who were subsequently diagnosed with CRC, to January 1998; and all 476 residents diagnosed with colorectal cancer between 1994 and 1997. Main outcome measures: Age-standardised CRC incidence for 1994-1997 for patients who had had a previous colonoscopy and for the remaining community; Dukes' grade of CRCs. Results: For people aged 50 years and over, the incidence of CRC was significantly lower among those who had had a previous colonoscopy than in the remaining community (1.14 versus 2.31 per 1000 patient-years; P = 0.0046). For people aged 35-49 years, the incidence was 0.35 versus 0.31 per 1000 patient-years (P = 0.86). Thirty-one CRCs developed in 29 people who had had previous colonoscopy; only five of these CRCs were graded Dukes C, with none graded Dukes D. In contrast, almost half the CRCs in the rest of the community were graded Dukes C or D (P < 0.001). All but one of those diagnosed with CRC on repeat colonoscopy had risk factors (personal or family history), and 23 of the CRCs were subclinical, with 20 being diagnosed during surveillance colonoscopy. Conclusions: Patients aged over 50 years who had previously undergone a colonoscopy and ensuing treatment were less likely to be diagnosed with CRC than otherwise expected. Surveillance colonoscopy led to diagnosis of CRCs with lower Dukes grades. Introduction Colorectal cancer (CRC) is the most common internal malignancy and the second most common cause of death from cancer in Australia. CRC incidence in Australia is similar to that in other developed countries,1 but in the US both incidence and mortality are now decreasing.2 This change has been attributed to removal of premalignant polyps, detection of proportionally more early lesions by colonoscopy, and more effective treatment.3 Strategies based on colonoscopic surveillance and targeting people at increased risk of developing CRC are being promoted.3-5 However, colonoscopy, particularly in community-based, open-access practice, has not been shown to reduce CRC mortality. Indeed, although the rate of colonoscopies in Australia has increased fourfold since 1984, the incidence of CRC in New South Wales between 1973 to 1992 continued to increase by 2% per year in men and 0.9% per year in women.6This study aimed to evaluate whether colonoscopy protects against subsequent CRC by comparing CRC incidence and pathological grading between people who have had a previous colonoscopy and the rest of the population in a geographically isolated area. Methods Setting The study was set in Townsville and the surrounding region (defined by the postcodes 4804-4822, 4849 and 4850; Figure 1). Townsville is a regional centre in northern Queensland that provides centralised health services, including colonoscopy, for a population of 198 000 dispersed over 200 000 km2. The nearest alternative colonoscopy services are located at Cairns and Mackay, 400 km distant. Colonoscopy patients Subjects were patients who underwent colonoscopy performed by myself at the Mater Private Hospital, Townsville. All patients who underwent colonoscopy between July 1985 and December 1996 were identified, most from the hospital's detailed computerised records, but some of those examined between 1985 and 1990 from a hospital work ledger which gave only surname and given name. In addition, detailed demographic and clinical information was collected prospectively in a procedural database for all colonoscopy patients from April 1994 to December 1997. Similar information was obtained retrospectively from the case records of 600 randomly chosen patients who underwent colonoscopy between 1986 and 1990. Colonoscopic procedures and surveillance Colonoscopies were performed on patients referred by a general practitioner or specialist either for surveillance or for investigation of symptoms. Through concessions available until mid 1996, the service was equally available to all patients irrespective of financial resources. Fibreoptic colonoscopes were used before 1990, and video colonoscopes after then. Patients were lightly sedated with fentanyl (100 µg) and midazolam (2.5-5 mg). Treatment (eg, polypectomy, CRC resection) was given as necessary. Surveillance recommendations were mostly included in procedure reports and passed to both the patient and the referring doctor. While these recommendations changed over time consistent with published guidelines,3,7 a general summary is: Annual surveillance for either active ulcerative pancolitis of seven years' duration or longer or a previous malignancy plus a family history of hereditary non-polyposis colorectal cancer (HNPCC);3,7 One- to two-yearly surveillance for a previous CRC before 50 years and for those older than 25-40 years with either CRC developing in a first-degree relative before 50 years or a family history of HNPCC; Two- to five-yearly surveillance for longstanding quiescent pancolitis or active limited colitis, CRC or polyps in a first-degree relative, previous CRC, or large (>1 cm diameter) or multiple colonic polyps; and Five- to 10-yearly surveillance, depending on age, for a small adenomatous polyp. Colorectal cancer diagnoses All patients diagnosed with CRC in Townsville between January 1994 and December 1997 were identified retrospectively by searching the computerised databases of all three pathology services, three hospitals, three endoscopy services, three colonoscopists (including myself) and one oncology service provider in Townsville, and from the CRC audits maintained by the six surgeons in Townsville. From 1995, patients diagnosed with CRC were also identified prospectively by clinicians and institutions. Patients diagnosed with CRC who had had a previous colonoscopy performed by myself at the Mater Private Hospital were identified to January 1998. Clinical records of all patients diagnosed with CRC were audited by myself. A modified Dukes classification (A, B, C or D) was used for staging cancer spread.8,9 A malignant polyp was classified separately if colonoscopic resection was regarded as the definitive treatment. In cases of synchronous lesions, the lesion with the most invasive grading was registered. Incidence of CRC Colonoscopy population: The incidence of CRC was calculated as the number of cases per thousand patient-years for the period 1994-1997 for patients who had had a previous colonoscopy and still lived in the region in 1997 (colonoscopy population). Residence was determined from the electoral register current in January 1997,10 which is considered reliable as voter registration is compulsory in Australia. To reduce mismatch errors caused by individuals with identical names, only patients with a known middle name (duplication rate, 0.3%) were cross-referenced against voters who also had a recorded middle name (duplication rate, 1.3%). The number of patients without a middle name who were still resident was estimated and added to the above on the assumption that the proportion still resident would be the same in the groups with and without a known middle name. CRC incidence in each year was calculated for patients who had undergone previous colonoscopy up until the previous calendar year. For example, the incidence of CRC in 1994 was calculated for patients who had undergone previous colonoscopy up to 1993. Age of colonoscopy patients was determined for the year of incidence. The number of patient-years was the total for all patients in a given age range who had previously had a colonoscopy up to 1993, 1994, 1995 and 1996. Community: The incidence of CRC in the remaining population (community) was determined from the number of cases that occurred between 1994 and 1997 in people not registered as a colonoscopy patient per the region's population less the colonoscopy population. Population data were obtained from the August 1996 census undertaken by the Australian Bureau of Statistics.9 Statistical analyses Binary data were compared in two by two contingency tables using chi-squared analyses.12 The age-standardised incidences of CRC in colonoscopy patients versus the community were tested for the hypothesis that the ratios were equal to one.13 Results Colonoscopies From mid 1985 to the end of 1996, I performed 11 148 colonoscopies on 8430 patients (Figure 2). The number increased steadily, from 590 in the 18 months 1985-1986 to 2708 in the two years 1995-1996. The number of repeat colonoscopies increased from 12 (2.0%) in 1985-1986 to 875 (32.3%) in 1995-1996. Clinical and procedural characteristics are summarised in Box 1. Slightly more women than men had colonoscopies. Patients undergoing repeat colonoscopies were an average six years older than those newly referred and were more likely to have had surveillance for increased CRC risk as the primary indication (50.7% of repeat colonoscopies versus 15.0% of first colonoscopies; P < 0.0001). Primary indication also varied with time. For example, an abnormal barium enema was a common indication before 1991 (7.0%), but accounted for few after 1994 (0.3%; P < 0.0001). In contrast, a family history of polyps or CRC accounted for a greater proportion of colonoscopies after 1994 (10.6% of first and 11.2% of repeat colonoscopies) than before 1991 (6.3%; P < 0.0001). From the outset, the caecal completion rate exceeded 95%, and from 1994 it was 98.9% overall and 99.5% in those without a malignant obstruction. Polyps were diagnosed (and removed) in a greater proportion of repeat than first colonoscopies (36.1% versus 29.8%; P < 0.0001). Both these rates were higher than for colonoscopies performed before 1991 (24.5%; P < 0.0001). However, CRC was diagnosed less often in repeat than in first colonoscopies (0.6% versus 2.2%; P = 0.001). Resident populations Complete details, including a middle name, were recorded for 5762 of the 8430 colonoscopy patients (68.4%), and 4200 of these (72.9%) were registered voters and residents of the Townsville region in 1997. A surname and one given name only were recorded for the remaining 2668 patients -- 1913 from the hospital's computer register and 755 from the work ledger. This gave an estimated total number of resident colonoscopy patients of 6195 in 1997. Colorectal cancers Between 1994 and 1997, 476 new CRCs were diagnosed in residents of the Townsville region, with 474 in people aged over 35 years. Eighteen were in patients who had had a previous colonoscopy; each of these was diagnosed per colonoscopy by myself, nine at the study hospital (registered in the procedural database and shown in Box 1), and the remainder elsewhere. Incidences of CRC between 1994 and 1997 are shown in Box 2. In people aged over 50 years, the annual incidence of CRC in the colonoscopy population (1.14) was just less than half that in the remaining community (2.31; P = 0.0046). Between July 1985 and January 1998, I diagnosed 31 CRCs in 29 patients who had had a previous colonoscopy (two patients had a second CRC diagnosed two years after the first in each case). All but one of these patients had a personal or family history that warranted surveillance, and 21 had been enrolled in surveillance programs, with 20 (65%) having had multiple previous colonoscopies (mean, 3.4; range, 2-9). For 20 of the CRCs, planned surveillance was the indication for the repeat colonoscopy. Among the 11 people with symptoms as the primary indication, these symptoms were considered unrelated to the CRC in at least three. The time between most recent previous colonoscopy and diagnosis averaged 37 months (range, 3-136 months). Staging of CRCs is shown in Box 3. Metastatic spread from CRCs was less common in patients who had had a previous colonoscopy than in the community; only five of the 31 cases in colonoscopy patients were graded C, and none were graded D, while 219 of the 458 community cases (48%) were graded C or D (P < 0.001). Among the five colonoscopy patients with metastatic spread (Dukes C), the time between most recent previous colonoscopy and diagnosis was 19, 24, 29, 32 and 70 months, respectively. Discussion The study evaluated a colonoscopic service that followed and promoted contemporary surveillance guidelines similar to those currently recommended by professional cancer and gastroenterological societies.3 The incidence data suggest, but do not prove, that colonoscopic surveillance confers a benefit. Age-standardised incidence of CRC among people aged 50 years and over in the Townsville region was lower among those who had had a previous colonoscopy, along with any treatment considered necessary (eg, polypectomy or bowel resection), than in the community. This was despite the fact that many of those who had had a previous colonoscopy had a personal or family history likely to increase their risk of developing CRC. Furthermore, the CRCs that occurred in those who had had a previous colonoscopy were of a lower Dukes grade than those occurring in the community. The difference in incidence suggests but does not prove that colonoscopy is protective against CRC. CRC incidence for 1994-1997 in those who had had a previous colonoscopy may have been reduced, at least partly, by detection of subclinical CRCs during their pre-1994 colonoscopies. On the other hand, selection bias suggests that these people would develop more CRCs than the general population. The impact of each of these factors could not be measured, and there is no historical benchmark or matched population for comparison of outcomes. However, the result does suggest that colonoscopy confers a benefit, possibly because of removal of polyps and certainly because of detection and treatment of subclinical CRCs. In people aged 35-49 years, CRC incidence in those who had had a previous colonoscopy was similar to that in the community. Without a control group, a benefit of colonoscopy cannot be dismissed, as the colonoscopy group was expected to have higher CRC incidence. However, it is evident that, because of the large number of people aged 35 to 49 years and the low incidence of CRC, surveillance must be targeted to be effective. Polypectomy rate was high, and higher in repeat than in first colonoscopies. This also implies that patients having repeat colonoscopies had increased risk of developing CRC.14 While the high polyp rate may have been due to their older average age,15 the latter would also be expected to increase the CRC rate, which did not occur. Given an expectation that all lesions seen at the previous colonoscopy had been dealt with, this outcome (high polyp versus low CRC rate) validates the selection criteria for surveillance. The higher polypectomy rate after 1994 compared with that before 1990 probably relates to other circumstances, such as a higher caecal completion rate, while both indices probably reflect improved instrument technology. Thirty-one new primary cancers developed in 29 colonoscopy patients, with two-thirds diagnosed by planned surveillance colonoscopy. Metastatic spread occurred in only five of these patients. These findings confirm, firstly, that new CRCs will develop and, secondly, that outcome can be improved through early (subclinical) diagnosis.3,16 The number of cancers diagnosed in patients who had had a previous colonoscopy was of concern and suggested lesions might have been missed in the earlier examination. Colonoscopy, even when performed by an expert, does not identify all small lesions, while adverse conditions sometimes obscure gross pathology.17 Colon morphology, quality of the bowel preparation, instrument capabilities and operator proficiency may also impose limitations.3,18 However, substandard colonoscopy is unlikely to have been responsible. The caecum was reached at a rate exceeding the accepted standard (95%),3 and CRCs were observed in all parts of the bowel, arguing against an operator-dependent blind spot. Possibly, the comparatively large number of CRCs found in people undergoing colonoscopic surveillance was simply the outcome of increasing enlistment of an appropriate, at-risk cohort. Although most sporadic CRCs evolve slowly through malignant transition in a polyp, this sequence is truncated or absent for some sporadic CRCs and for CRCs developing in patients with ulcerative colitis or a genetic predisposition.19 It is unrealistic to imagine that surveillance colonoscopy with polypectomy as necessary will much reduce CRC incidence in such at-risk populations. Indeed, it might conceivably increase apparent incidence by uncovering subclinical CRCs. The results support the current practice of targeting individuals with recognised risk factors for surveillance colonoscopy. However, it is important to explain to patients that surveillance does not provide complete protection and that new CRCs are inevitable. Early diagnosis through repeated testing is the essential component of surveillance-derived protection. References Parkin DM, Pisani P, Ferlay J. Estimates of the world-wide incidence of eighteen major cancers in 1985. Int J Cancer 1993; 54: 594-606. SEER Program (National Cancer Institute). Surveillance, epidemiology, and end results (SEER) program. Bethesda, Md: National Cancer Institute, 1973-1992. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997; 112: 594-642. Bolin TD, Korman MG. How can we reduce the incidence and mortality of colorectal cancer [editorial]? Med J Aust 1997; 166: 175-176. Macrae FA. Screening for colorectal cancer, 1996 [editorial]. Med J Aust 1996; 165: 102-105. Bell JC, McCredie M, Coates MS, Armstrong BK. Trends in colorectal cancer incidence and mortality in New South Wales, 1973-1992. Med J Aust 1997; 166: 178-181. Mecklin J-P, Jarvinen HJ, Peltokallio P. Cancer family syndrome. Genetic analysis of 22 Finnish kindreds. Gastroenterology 1986; 90: 328-333. Astler VB, Coller FA. The prognostic significance of direct extension of carcinoma of the colon and rectum. Ann Surg 1954; 139: 846-851. Dunlop MG. Polyps and carcinoma. In: Shearman DJC, Finlayson N, Camillieri, Carter D, editors. Diseases of the gastrointestinal tract and liver. 3rd ed. New York: Churchill Livingstone, 1997: 1399-1448. Australian Electoral Commission. Elector information access system. Electoral roll information for Queensland. Canberra: Australian Electoral Commission, 1997. Australian Bureau of Statistics. 1996 census of population and housing. Community profile, Canberra: ABS, 1996 (Cat. no. 2020.0). Approximate significance for contingency tables. In: Matthews DE, Farewell VT. Using and understanding medical statistics. 2nd ed. Basel: Karger, 1988: 20-66. The binomial distribution. In: Snedecor GW, Cochran WG. Statistical methods. 8th ed. Ames: Iowa State University Press, 1989: 107-134. Atkin WS, Morson BC, Cuzick J. Long-term risk of colorectal cancer after excision of rectosigmoid adenomas. N Engl J Med 1992; 326: 658-662. Williams AR, Balasooriya BAW, Day DW. Polyps and cancer of the large bowel: a necropsy study in Liverpool. Gut 1982; 123: 835-842. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. N Engl J Med 1993; 328: 1365-1371. (Published erratum appears in N Engl J Med 1993; 329: 672.) Rex RK, Cutler CS, Lemmel GT, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997; 112: 24-28. Baille J, Ravich WJ. On endoscopic training and procedural competence. Ann Intern Med 1993; 118: 73-74. Kuramoto S, Oohara T. Flat early cancers of the large intestine. Cancer 1989; 64: 950-955. (Received 1 May, accepted 21 Dec, 1998) Author's Details 42 Ross River Road, Townsville, QLD. John Croese, MD, FRACP, Gastroenterologist. Reprints will not be available from the author. Correspondence: Dr J Croese, 42 Ross River Road, Townsville, QLD 4812. Email: jcroeseATmedeserv.com.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> Back to text Back to text Back to text Back to text Back to text

John Croese

Research Hyperbaric or normobaric oxygen for acute carbon monoxide poisoning: a randomised controlled clinical trial Carlos D Scheinkestel, Michael Bailey, Paul S Myles, Kerry Jones, D James Cooper, Ian L Millar and David V Tuxen MJA 1999; 170: 203-210 For editorial comment, see Moon & DeLong Abstract - Introduction - Methods - Results - Discussion - Conclusion - References - Authors' details - - More articles on Anaesthesia and intensive care Abstract Objective: To assess neurological sequelae in patients with all grades of carbon monoxide (CO) poisoning after treatment with hyperbaric oxygen (HBO) and normobaric oxygen (NBO). Design: Randomised controlled double-blind trial, including an extended series of neuropsychological tests and sham treatments in a multiplace hyperbaric chamber for patients treated with NBO. Setting: The multiplace hyperbaric chamber at the Alfred Hospital, a university-attached quarternary referral centre in Melbourne providing the only hyperbaric service in the State of Victoria. Patients: All patients referred with CO poisoning between 1 September 1993 and 30 December 1995, irrespective of severity of poisoning. Pregnant women, children, burns victims and those refusing consent were excluded. Intervention: Daily 100-minute treatments with 100% oxygen in a hyperbaric chamber -- 60 minutes at 2.8 atmospheres absolute for the HBO group and at 1.0 atmosphere absolute for the NBO group -- for three days (or for six days for patients who were clinically abnormal or had poor neuropsychological outcome after three treatments). Both groups received continuous high flow oxygen between treatments. Main outcome measures: Neuropsychological performance at completion of treatment, and at one month where possible. Results: More patients in the HBO group required additional treatments (28% v. 15%, P = 0.01 for all patients; 35% v. 13%, P = 0.001 for severely poisoned patients). HBO patients had a worse outcome in the learning test at completion of treatment (P = 0.01 for all patients; P = 0.005 for severely poisoned patients) and a greater number of abnormal test results at completion of treatment (P = 0.02 for all patients; P = 0.008 for severely poisoned patients). A greater percentage of severely poisoned patients in the HBO group had a poor outcome at completion of treatment (P = 0.03). Delayed neurological sequelae were restricted to HBO patients (P = 0.03). No outcome measure was worse in the NBO group. Conclusion: In this trial, in which both groups received high doses of oxygen, HBO therapy did not benefit, and may have worsened, the outcome. We cannot recommend its use in CO poisoning. Introduction Carbon monoxide (CO) poisoning is one of the most common lethal poisonings,1 with neurological or psychiatric sequelae occurring in up to 67% of survivors.2Treatment with hyperbaric oxygen (HBO) is recommended because it reduces carboxyhaemoglobin (COHb) dissociation half-life from more than four hours at room air or 45 minutes on 100% oxygen to 23 minutes at 2.5 atmospheres absolute (ATA).3 Carbon monoxide also inhibits cellular respiration by binding to cytochrome oxidase, a component of the mitochondrial electron transport chain.4 Hyperbaric oxygen enhances the dissociation of CO from this enzyme.5 Despite these physiological effects, it has not been established in humans that HBO either improves survival or decreases neuropsychological deficits. Much of the evidence that HBO is more efficacious than normobaric oxygen (NBO) therapy in humans arises from isolated case reports,6-8 uncontrolled clinical observations,9-11 small,12 non-randomised13 and unblinded series9,11,13-20 and incomplete assessment of outcome (no neuropsychological testing).9,11,15-18 All reported non-randomised studies have suggested benefit from HBO. Of the four published randomised studies, two report benefit from HBO12,14 and two report no benefit15,17 (Box 1). Three restricted entry to mildly poisoned patients, while the fourth15 included severely poisoned patients, but did not allocate any to NBO treatment. None of the randomised studies blinded patients by using sham treatments for NBO, only one blinded outcome assessment12 and only one used neuropsychological tests to assess outcome.14 Hence, the benefit of HBO in CO poisoning has been questioned1,2,21-27 and remains unproven. We therefore performed a randomised double-blind trial in patients with all grades of CO poisoning, comparing HBO and NBO (with sham treatments for the NBO group), and using an extended series of neuropsychological tests to assess both persistent and delayed neurological sequelae (PNS and DNS). Methods The multiplace chamber at the Alfred Hospital, a university-attached quarternary referral centre, provides the only hyperbaric service in the State of Victoria (population, 4.5 million; area, 228 000 km2). Between 1 September 1993 and 30 December 1995, most CO-poisoned and all severely poisoned patients were referred for treatment. We included all referred patients, irrespective of severity of poisoning. Patients were excluded if they were pregnant, children, burns victims or if they did not consent. Informed consent to enter the trial was requested from patients with a Mini-mental score >2428 and from the next of kin for those obtunded or with a score ≤24. The Alfred Hospital's Ethics Committee approved the trial, conditional on an independent blinded interim analysis after recruitment of 50 patients (using a stopping rule of P < 0.001); this allowed continuation of enrolment to completion. Randomisation and blinding Patients were randomly allocated to HBO or NBO treatment. To ensure a similar distribution of causes and severity of poisoning in both groups, patients were first stratified into four groups (suicide versus accidental, then mechanically ventilated versus non-ventilated). A hyperbaric technician then allocated patients to treatment groups by opening envelopes chosen from random blocks, each with equal numbers of HBO and NBO selections. To minimise the impact of the trial on daily practice, we used cluster randomisation for patients who presented simultaneously from the same CO exposure, allocating them all to the same treatment group. Cluster randomisation accounted for the difference in numbers between HBO and NBO groups. As patients presenting simultaneously could be uniquely identified by having identical measurements for three continuous baseline severity measurements (exposure time, time to COHb measurement and time to treatment), any effects due to cluster randomisation could be controlled and adjusted for by including these variables in the generalised linear model. The hyperbaric technicians and nursing staff had knowledge of the treatment group but patients and outcome assessor did not. Interventions Before arrival at Alfred Hospital, non-intubated patients received high flow oxygen by non-occlusive facemask and intubated patients received 100% oxygen. All patients were admitted to hospital, received three treatments on a once-daily basis and continuous oxygen by non-occlusive facemask at 14 L/min (100% oxygen for ventilated patients) between treatments. Patients randomised to NBO therapy were treated for 100 minutes in the multiplace chamber with 100% oxygen at 1.0 atmosphere absolute (ATA). Non-ventilated patients used an occlusive facemask attached via a non-rebreathing valve to a Laerdal adult ventilation bag (1.6 L) with an oxygen reservoir (2.6 L; Armund S Laerdal, Stavanger, Norway). The chamber door was closed and the chamber flushed with air regularly to simulate pressurisation, but the chamber was not pressurised (sham treatment). HBO patients received 100% oxygen by hood, occlusive facemask or mechanical ventilator in the hyperbaric chamber for 100 minutes (60 minutes at 2.8 ATA). After the third treatment, patients were reassessed medically and underwent full neuropsychological assessment. Patients who were clinically abnormal or had poor neuropsychological outcome received three further treatments and received high flow oxygen between treatments. Outcome measures Patient assessment at entry included length of CO exposure, COHb level, time from end of exposure to COHb measurement and to treatment, Mini-mental score and clinical effects of poisoning (Box 2). We then assessed patients at completion of treatment (three or six treatments), and, wherever possible, at one month. We attempted to quantify deficits known to occur in CO poisoning by assessing attention, information processing, memory and learning. A clinical psychologist trained in neuropsychological assessment of brain-injured patients performed all tests at completion of treatment and at follow-up. Computerised testing was used to standardise administration and data-recording procedures and increase objectivity. The tests used were the digit span subtest of the Wechsler Adult Intelligence Scale -- Revised,29 comprising (i) Digit span forward and (ii) Digit span backwards (in which patients are asked to repeat a series of numbers read to them), which measures immediate auditory-verbal memory span, working memory and attention; computerised reaction-time tests,30 consisting of (iii) Simple reaction time (in which subjects are requested to press the space bar on a computer keyboard as soon as they see anything appear on the screen) to give a basal measure of alertness or arousal, and (iv) Choice reaction time (which requires subjects to ignore stimuli in a centre box and to respond selectively to the word "SEVEN" as it appears around the periphery of the computer screen) to test selective attention (reaction time was tested because it can show diffuse cerebral dysfunction, and because processing speed is considered to underlie attention deficits31); (v) a score on the Rey auditory verbal learning test, in which a 15-word list (List A) is presented over five learning trials, followed by an interference trial (List B), after which (vi) Short term free recall is tested without any further presentation of the word list, and (vii) Long term free recall is tested 20 minutes later (this provides a measure of learning across trials, and retention of information following short and long delay periods32). Raw scores of these seven neuropsychological tests were converted to z scores ([score - mean in normal population] / standard deviation), and then t scores (McCall's T; an adjusted z score so that the mean is 50 and the standard deviation is 10).33 Age-based and education-based norms were used where available to calculate t scores. A t score more than one standard deviation below the mean was considered abnormal, and two or more abnormal scores constituted a poor outcome. Patients with poor outcome at hospital discharge were considered to have persistent neurological sequelae (PNS). Delayed neurological sequelae (DNS) were defined as morbidity found at follow-up that was not obvious at hospital discharge, or deterioration of neuropsychological subtest scores by more than one standard deviation. Statistical analyses Statistical analyses were made using mixed linear models to adjust for all baseline covariants (age, sex, suicide attempt, COHb level, time to COHb level, duration of exposure, time to treatment, and presence of other drugs). Data were presented as mean and SD, median and interquartile range (IQR) or number and per cent. Continuous data were first assessed for normality and then analysed by unpaired two-tailed Student's t-test, or Wilcoxon rank sum test. Proportions were compared with chi-squared tests (with Yates' correction), or Fisher's exact test, as appropriate, with multiple logistic regression being used to adjust for confounding factors. We calculated odds ratios and 95% confidence intervals (95% CI) for the difference between proportions. The 95% CI for the difference between means and P values were calculated after adjustment for baseline covariants. All statistical analyses were performed using SAS.34 Results Two hundred and thirty patients with CO poisoning were referred for treatment. Thirty-nine were excluded (one child, eight burns victims, and 30 who refused consent) and treated with HBO. Thus, 191 patients entered the trial (Box 2). Based on the most sensitive neuropsychological test (Short reaction time), with 191 patients and a significance level of 0.05, we had greater than 99% power to detect a 10% difference between groups (ie, 408 seconds v. 450 seconds; SD, 63 seconds) (Clinical Trials Design Program, Biosoft, Cambridge, UK). The groups (104 HBO patients, 87 NBO patients) were comparable in age, sex, incidence of suicide attempt, mechanical ventilation, Mini-mental score, and other markers of severity, including loss of consciousness (coma). Forty-four per cent of patients who had attempted suicide (44% HBO and 44% NBO) also had evidence of self-administration of drugs or alcohol. Most of our patients (73%) had severe CO poisoning, defined by any of the following before or on arrival at Alfred Hospital: a Mini-mental score ≤24, COHb level >30%, confusion, focal neurological deficits, loss of consciousness, electrocardiogram abnormalities, arrhythmias, pulmonary oedema, metabolic acidosis, hypotension, convulsions, and cardiac arrest. All mechanically ventilated patients met the criteria of severe poisoning. Overall mortality was 3%, and the incidence of PNS was 71% at hospital discharge and 62% at follow-up, with no significant differences between the HBO and NBO groups (Box 3). A smaller proportion of NBO patients than HBO patients were considered to be medically or neuropsychologically impaired after three treatments and thus received additional treatments (all patients, 15% v. 28%, P = 0.01; severely poisoned patients, 13% v. 35%, P = 0.001). The only statistically significant difference between groups in neuropsychological performance was in the learning test at completion of treatment (Boxes 3 and 4); this was in favour of the NBO group for both all patients (P = 0.01) and severely poisoned patients (P = 0.005). NBO patients had a significantly lower number of abnormal test results at completion of treatment (all patients, 3.4 v. 2.7, P = 0.02; severely poisoned patients, 3.7 v. 2.6, P = 0.008) and, for those severely poisoned, there were fewer NBO patients with a poor outcome (85% v. 65%; P = 0.03). All five relapses (DNS) occurred in HBO patients (P = 0.03) at a median of 40 days (IQR, 29-81 days) after initial treatment; these patients then received a mean 4.5 (SD, 2.5) additional treatments. Although three of these patients improved with further treatments, all DNS patients had a poor outcome after re-treatment, with a mean 6.3 (SD, 1.2) abnormal test results. The evaluation at completion of treatment showed no difference in outcome between the HBO and NBO groups for patients: treated within four hours of exposure; with severe poisoning and treated within four hours of exposure; who required ventilation; and who were accidentally poisoned (as opposed to those who attempted suicide). Only 46% of patients attended the one-month follow-up. Thus, the numbers in subgroups of interest at one month were small, but showed no difference in any test between HBO and NBO groups. Ten patients had chamber-related complications; seven HBO patients experienced ear barotrauma, one HBO patient developed oxygen toxicity (convulsions) and two patients (one HBO and one NBO) developed severe claustrophobia. The incidence of such complications was thus 9% for HBO and 1% for NBO treatment. Discussion In patients with acute CO poisoning, we found no benefit and possible adverse effects of HBO therapy compared with three days of high-flow NBO. Our multiple comparisons between groups may have produced type 1 errors, and some differences may be spurious. However, differences were consistent and all suggested a more detrimental outcome in the HBO group. Despite multiple comparisons, we found no evidence to support HBO therapy. Our findings thus contrast with those of all other published studies, which have suggested benefit or lack of benefit, but never detriment, from HBO therapy (Box 1). To explain this, careful comparison with previous studies is required. Baseline severity No previous study has compared HBO with NBO in severely poisoned patients. Unlike the non-randomised studies (Box 1) which used HBO for all severely poisoned patients and NBO only for mildly poisoned patients,9,13,16,18-20 all patients in our study were randomised. Of the randomised studies (Box 1), three included patients with mild CO poisoning only12,14,17 (two showing benefit and one no benefit from HBO),17 while the fourth compared one versus two HBO treatments for severely poisoned patients.15Whereas most CO poisoning in the northern hemisphere occurs as a result of heating accidents, in Australia most results from suicide attempts. Not only had a high proportion of our patients (69%) attempted suicide, but, as in other studies,9,35 many (44%) had ingested other drugs. Stratified randomisation equalised this factor between groups and hence could not account for lack of benefit in the HBO group. Further, the six potential factors thought to influence baseline severity (listed in the Methods) were subsequently adjusted for in the generalised linear modelling process. This adjustment also accounted for the small imbalances in the data resulting from cluster randomisation. Carboxyhaemoglobin level Although COHb level is often used as an indicator of the severity of CO poisoning and to determine need for HBO,12-16,20 our findings and other reviews9,14,18,24,36-40 have shown no relationship between COHb level and outcome. COHb level depends on CO exposure, time elapsed to measurement and whether or not oxygen has been given. The low COHb levels in our study (HBO 20.5%; NBO 22%) reflect the delay to the measurement and use of high flow oxygen before measurement. Most previous studies did not report time to measurement and used isolated COHb levels to compare severity of poisoning between groups.12-16,20 Our multivariate analysis showed no correlation between outcome and COHb level even when taking the time to measurement into account. The low COHb levels do not explain the lack of benefit of HBO. Treatment delay Animal studies5 reporting beneficial effects of HBO given immediately after CO exposure cannot readily be extrapolated to clinical practice because treatment delay is to be expected in all clinical scenarios. Our study had a geometric mean treatment delay of 7.1 hours (95% CI, 1.9-26.5 hours), which is longer than in others,14,16,20 but well within entry criteria limits of most studies that report treatment delay15,17,39 (Box 1). Some authors have suggested that the benefits of HBO diminish with treatment delay,12,41 that more than six hours' delay increases DNS and mortality,9 and that treatment delay is associated with increased neuropsychological sequelae.13 Other studies have found treatment delay to be unimportant,42 while some case studies43 and small case series7 report HBO benefit regardless of treatment delays ranging from days to months. Most North American hyperbaric facilities surveyed in 1995 treated CO-poisoned patients who had neurological deficits despite presentation delays ranging from six hours to 56 days,44 and HBO has been advocated for DNS occurring weeks after initial exposure.20,44 In our study, analysis of patients commencing treatment within four hours (all patients or severely poisoned only) showed no differences in outcome between HBO and NBO. We also analysed time to treatment in quartiles (<3, 3-6, 6-12 and >12 hours) and found no difference in outcome between HBO and NBO. Further, multivariate regression analysis did not identify delay in treatment as a predictor of poor outcome. Thus, there was no evidence that delay to treatment might explain the lack of benefit from HBO. Oxygen dose There are no universally accepted recommendations for depth of pressurisation or duration of hyperbaric treatment for CO poisoning (Box 1). The only studies of the benefit of multiple treatments reached contradictory conclusions.13,15,38 Raphael et al15 found no difference in recovery at one month among 286 patients with transient loss of consciousness who received either one or two HBO treatments 12 hours apart. Gorman and Runciman reviewed 13 case series involving 3441 patients and concluded that HBO at 2-3 ATA for 1-2 hours on three or more occasions achieved the lowest mortality, PNS and DNS.38Based on the conclusions of Gorman and Runciman, our study was designed to provide maximum advantage for HBO, with a daily 60-minute treatment at 2.8 ATA on three consecutive days. Because the required dose of NBO for treating CO poisoning is unknown, to ensure we did not undertreat patients, and to maximise similarity of treatment in HBO and NBO groups, all our patients received a treatment on at least three consecutive days and continuous oxygen by non-occlusive facemask at 14 L/min between treatments. Compared with most previous studies, we performed more treatments, of longer duration, at higher ATA and in conjunction with prolonged high flow oxygen therapy between treatments (Box 1). Our HBO group received oxygen therapy equating to approximately 35.7 COHb-dissociation half-lives, while the NBO group received the equivalent of 28.5 COHb-dissociation half-lives. Most other studies have used total oxygen doses of less than 7.0 COHb-dissociation half-lives,9,12,14,15,17 with two series using up to 18 COHb-dissociation half-lives.13,16 It is possible that some of the reported beneficial effects of HBO are purely oxygen-dose related, and that adequate NBO, as given in our study, may achieve the same result. This is supported by an uncontrolled study in which a single HBO treatment had no benefit over NBO, but two or more HBO treatments of 60 minutes at 2.8 ATA resulted in significantly less PNS at hospital discharge and DNS at one month (P < 0.005).13 The apparent worse outcome in our HBO group may also be oxygen-dose related, with higher doses of oxygen adding no further benefit and possibly causing adverse effects. Hampson et al45 (discussing seizures rather than neuropsychological sequelae) have suggested that CO-poisoned patients are at greater risk of brain injury because of the higher ATA used in treatment, concomitant use of other drugs and toxins (particularly in patients who have attempted suicide), as well as the underlying CO poisoning. Assessment of outcome Abnormalities of the basal ganglia, subcortical white matter and hippocampus are the most consistent neuropathological findings in victims of CO poisoning42,46 and are associated with deficits of attention, information processing and memory.47 These deficits can be easily missed on casual assessment or simple neurological examination unless specifically targeted. Studies that did not use neuropsychological assessments9,12,15 reported a lower incidence of PNS than those that did,48 including our study. Appropriately targeted neuropsychological assessment provides the most objective, reliable and sensitive evaluation of outcome after CO poisoning,24,49,50 and in studies that did not report these data9,12,15-17 it is possible that significant adverse effects were missed, making resulting conclusions unreliable. The Carbon monoxide neuropsychological screening battery (CONSB)50 does not adequately measure memory, which may be impaired following CO poisoning.47 The neuropsychological tests we used were therefore more comprehensive than the CONSB, very sensitive to the deficits known to occur in CO poisoning and were computerised (thus increasing objectivity). Further, as one clinical psychologist performed all our testing, interviewer bias was eliminated. Because a full pretreatment neuropsychological assessment was not practical, a Mini-mental examination was used as our baseline neuropsychological assessment, as it gives a global assessment of severity of cerebral injury. Although not ideal, it had the advantages that (i) it could readily be performed by the assessing doctor and quantified, (ii) it enabled us to determine which patients were capable of giving informed consent, and (iii) as it was tested on presentation, completion of treatment and at follow-up, patients served as their own "controls". Our definition of an abnormal test result (>1 SD below the mean) would have included 16% of normal patients. This definition was deliberately chosen to be sensitive to small group differences. We defined a poor outcome as at least two test scores more than one standard deviation below the mean, which would have included less than 2.6% of normal patients. While it is possible that the true incidence of poor outcome or PNS may therefore be slightly lower than we report, the important analyses in making the group comparisons were based on the raw data. Non-randomised studies (Box 1) suggest beneficial effects of HBO, but only two13,20 used neuropsychological tests. One used extensive neuropsychometric evaluation at the one-month review,13 but no evaluation before or at completion of treatment. The other used the CONSB before treatment whenever possible,20 but not after treatment. Of the randomised studies (Box 1), one supplemented clinical assessments with electroencephalogram and cerebral blood flow reactivity to acetazolamide,12 but the clinical relevance of these is uncertain, as the abnormal results were found in patients who were clinically normal. Another used the CONSB, but only after completion of treatment and if patients were "fatigued", the tests were performed in the patients' homes within 12 hours, and a three-month review was only a telephone interview. The lack of baseline assessment, the variable circumstances of immediate outcome assessment and the restricted assessment of delayed outcome greatly limit the interpretation of the findings. Insensitive assessments and lack of blinding may also have missed important differences and created bias in these randomised studies. Study outcomes Our high rate of PNS (compared with previous studies) is probably attributable to our high proportion of severe poisonings (73%) and suicide attempts (69%, likely to be associated with depression and possibly a poor outcome on neuropsychological testing), as well as the comprehensive neuropsychological testing we employed. A type 1 error resulting from multiple testing may have contributed, as well as our cautious definition of PNS (≥2 test scores <1 SD below the mean). The rate of DNS in our study (2.6%) is lower than in most previous reports9,13,16,17,51 (especially given our higher proportion of severe poisonings) and possibly the result of effective treatment with higher doses of oxygen. Our findings differ from those of most published series (Box 1). Although all the non-randomised studies9,11,13,16,18-20 have reported benefit from HBO, none had a control group of matching severity, most had no neuropsychological assessment, and all had low doses of oxygen (brief treatment times) in the NBO groups. With these significant limitations, it is not possible to rely on the conclusions of these studies nor to compare them with adequately conducted randomised trials. None of the four randomised trials included pretreatment or follow-up neuropsychological assessments or sham treatment of the NBO group (our study is unique in doing this). Three did not have blinded outcome assessments. Both randomised trials that concluded there was benefit from HBO12,14 studied only patients with mild CO poisoning, thereby excluding the patients in whom the effects of HBO might be most important. The other two found no benefit in HBO and support our findings.15,17 Mathieu et al found a significantly different incidence of neurological sequelae between HBO and NBO at three-month review (9.5% v. 15%; P = 0.016), but not at completion of treatment, one month, six months or 12 months.17 Raphael et al randomised only mildly poisoned patients,15 and both treatment regimens used (HBO and NBO) have been criticised.38 Interim results (61 patients) of a United States randomised controlled trial enrolling all patients, irrespective of severity of poisoning and also using sham normobaric treatments, found no difference in PNS between NBO and HBO.52 Of the four published randomised studies, the two small ones12,14 reported a benefit from HBO, whereas the two larger studies15,17 did not. If our study is included, three studies involving 1395 patients have now shown no benefit for HBO, compared with two studies involving 91 patients which showed a benefit. Thus, it appears that much of the evidence supporting HBO for CO poisoning is flawed. Although our multiple tests increased the likelihood of a type 1 error, as the main outcome measures (the number of abnormal tests and a "poor outcome") were based on combining all tests we have minimised the chance of a spurious result. Delayed review Despite repeated efforts, only 46% of patients attended for follow-up. This low rate was probably affected by many of our patients having characteristics associated with suicide attempts and depression, many being referred from distant locations, and lack of incentive. However, the follow-up rate was equal in both groups, and evenly distributed across subgroups. Our follow-up assessment was more comprehensive than in all but one other study,13 but failed to show benefit for HBO. Others studies have had significant non-attendance rates at delayed review of 11%-47%.12-15,19,48 Most studies do not quote the "drop-out" rate.9,16-18,20 Conclusion Our prospective, randomised controlled trial of CO-poisoned patients of all severities attempted to address the shortcomings of previous studies by incorporating sham treatments for the NBO group, blinded outcome assessment and extensive neuropsychological assessment. Our HBO protocol was designed to provide the maximum potential advantage for HBO therapy based on currently available knowledge. When compared with three days of normobaric oxygen, we could find no evidence that treatment with HBO was beneficial to outcome and therefore do not recommend its use. References Meredith T, Vale A. Carbon monoxide poisoning. BMJ 1988; 2 96(6615): 77-79. Tibbles PM, Perrotta PL. Treatment of carbon monoxide poisoning: a critical review of human outcome studies comparing normobaric oxygen with hyperbaric oxygen. Ann Emerg Med 1994; 24: 269-276. Pace N, Strajman E, Walker EL. Acceleration of carbon monoxide elimination in man by high pressure oxygen. Science 1950; 111: 652-654. Tibbles PM, Edelsberg JS. Hyperbaric-oxygen therapy. N Engl J Med 1996; 334: 1642-1648. Brown SD, Piantadossi CA. In vivo binding of carbon monoxide to cytochrome c oxidase in rat brain. J App Physiol 1990; 68: 604-610. Dean BS, Verdile VP, Krenzelok EP. Coma reversal with cerebral dysfunction recovery after repetitive hyperbaric oxygen therapy for severe carbon monoxide poisoning. Am J Emerg Med 1993; 11: 616-618. Myers RAM, Snyder SK, Linberg S, Adams Cowley R. Value of hyperbaric oxygen in suspected carbon monoxide poisoning. JAMA 1981; 246: 2478-2480. Thomson LF, Mardel SN, Jack A, Shields TG. Management of the moribund carbon monoxide victim. Arch Emerg Med 1992; 9: 208-213. Goulon M, Barois A, Rapin M, et al. Carbon monoxide poisoning and acute anoxia due to breathing coal gas and hydrocarbons. J Hyperbar Med 1986; 1: 23-41. Yang ZD, Sun CQ, Cao XL. Clinical analysis of the effect with hyperbaric oxygen therapy on 672 cases with acute carbon monoxide poisoning. J Hyperbar Med 1986; 1: 188. Roche L, Bertoye A, Vincent P. Comparison de deux groupes de vingt intoxications oxycarbonees traitees par oxygenenormobare et hyperbare. Lyon Med 1968; 49: 1483-1499. Ducasse JL, Celsis P, Marc-Vergnes JP. Non-comatose patients with acute carbon monoxide poisoning: hyperbaric or normobaric oxygenation? Undersea Hyperb Med 1995; 22: 9-15. Gorman DF, Clayton D, Gilligan JE, Webb RK. A longitudinal study of 100 consecutive admissions for carbon monoxide poisoning to The Royal Adelaide Hospital. Anaes Intens Care 1992; 20: 311-316. Thom SR, Taber RL, Mendiguren II, et al. Delayed neuropsychologic sequelae after carbon monoxide poisoning: prevention by treatment with hyperbaric oxygen. Ann Emerg Med 1995; 25: 474-480. Raphael J, Elkharrat D, Jars-Guincestre M, et al. Trial of normobaric and hyperbaric oxygen for acute carbon monoxide intoxication. Lancet 1989; 2: 414-419. Mathieu D, Nolf M, Durocher A, et al. Acute carbon monoxide poisoning: risk of late sequelae and treatment by hyperbaric oxygen. J Toxicol Clin Toxicol 1985; 23: 315-324. Mathieu D, Wattel F, Mathieu-Nolf M, et al. Randomized prospective study comparing the effect of HBO versus 12 hours NBO in non-comatose CO poisoned patients: results of the interim analysis. Undersea Hyperb Med 1996; 23 Suppl: 7. Willms SJ, Turner F, Kerr J. Carbon monoxide or smoke inhalations treated with oxygen (hyperbaric vs normobaric): 118 reviewed. Undersea Biomed Res 1985; 12 Suppl: S56. Ely EW, Moorehead B, Haponik EF. Warehouse workers' headache: emergency evaluation and management of 30 patients with carbon monoxide poisoning. Am J Med 1995; 98: 145-155. Myers RAM, Snyder SK, Emhoff TA. Subacute sequelae of carbon monoxide poisoning. Ann Emerg Med 1985; 14: 1163-1167. Weaver LK, Hopkins RO, Larson-Lohr V. Hyperbaric oxygen and carbon monoxide poisoning. Ann Emerg Med 1995; 26: 390-391. Mitchell CA, Carroll PA. Acute toxicity of inhaled gases and particulates. Med J Aust 1989; 150: 717-720. Seger D. The science (or lack thereof) in the treatment of carbon monoxide poisoning. Am J Emerg Med 1994; 12: 389. Seger D, Welch L. Carbon monoxide controversies: neuropsychological testing, mechanism of toxicity, and hyperbaric oxygen. Ann Emerg Med 1994; 24: 242-248. Olson KR, Seger D. Hyperbaric oxygen for carbon monoxide poisoning: does it really work? Ann Emerg Med 1995; 25: 535-537. Van Meter KW, Weiss L, Harch PG, et al. Should the pressure be off or on in the use of oxygen in the treatment of carbon monoxide-poisoned patients. Ann Emerg Med 1994; 24: 283-288. Weaver LK. Randomized clinical trial in carbon monoxide poisoning needed. Am J Emerg Med 1994; 12: 685. Folstein MF, Folstein SE, McHugh PR. Minimental state. J Psychiatr Res 1975; 12: 189-198. Wechsler D. Wais-R manual. New York: Psychological Corporation, 1981. Miller EN, Satz P. The Californian computerised assessment package (Calcap). Los Angeles, CA: Miller EN & Satz P, 1987. Van Zomeren AH, Brouwer WB. Clinical neuropsychology of attention. New York: Oxford University Press, 1994. Lezak MD. Neuropsychological assessment. New York: Oxford University Press, 1995. Howell DC. Statistical methods for psychology. 2nd ed. Boston: Duxbury Press, 1987. Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS [computer program]. Version 6.12. Cary, NC: SAS Institute Inc, 1996. Power BM, Prentice DA. Carbon monoxide poisonings at The Royal Perth Hospital. Anaesth Intens Care 1991. 19: 149. Myers RAM. Planning an effective strategy for carbon monoxide poisoning. Emerg Med Rep 1987; 8: 193-201. Runciman WW, Gorman DF. Carbon monoxide poisoning: from old dogma to new uncertainties. Med J Aust 1993; 158: 439-440. Gorman DF, Runciman WB. Carbon monoxide poisoning. Anaesth Intens Care 1991; 19: 506-511. Norkool DM, Kirkpatrick JN. Treatment of acute carbon monoxide poisoning with hyperbaric oxygen: a review of 115 cases. Ann Emerg Med 1985; 14: 1168-1171. Choi IS. Delayed neurologic sequelae in carbon monoxide intoxication. Arch Neurol 1983; 40: 433-435. Origani G, Michael M, Tuscano R, Arghetti S. Outcome of patients suffering from CO poisoning. Undersea Hyperb Med 1996. 23 Suppl: 83. Pracyk JB, Stolp BW, Fife CE, et al. Brain computerized tomography after hyperbaric oxygen therapy for carbon monoxide poisoning. Undersea Hyperb Med 1996; 22: 1-7. Samuels AH, Vamos MJ, Taikato MR. Carbon monoxide, amnesia and hyperbaric oxygen therapy. Aust N Z J Psychiatry 1992; 26: 316-319. Hampson NB, Dunford RG, Kramer CC, Norkool DM. Selection criteria utilized for hyperbaric oxygen treatment of carbon monoxide poisoning. J Emerg Med 1995. 13: 227-231. Hampson NB, Simonson SG, Kramer CC, Piantadosi CA. Central nervous system oxygen toxicity during hyperbaric treatment of patients with carbon monoxide poisoning. Undersea Hyperb Med 1996; 23: 215-219. Miura T, Mitoma M, Kawai R, Harada K. CT of the brain in acute carbon monoxide intoxication: characteristic features and prognosis. Am J Neuroradiol 1985; 6: 739-742. La Plane D, Baulac M, Widlocher D, Dubois B. Pure psychic akinesia with bilateral lesions of basal ganglia. J Neurol Neurosurg Psychiatry 1984; 47: 377-385. Hopkins RO, Weaver LK. Long-term outcome in subjects with carbon monoxide poisoning. Undersea Hyperb Med 1994; 21 Suppl: 17. Starkstein SE, Berthier ML, Leigurada R. Psychic akinesia following bilateral pallidal lesions. Int J Psychiatry Med 1989; 19: 155-164. Messier LD, Myers RAM. A neuropsychological screening battery for emergency assessment of carbon monoxide-poisoned patients. J Clin Psychol 1991; 47: 675-684. Hopkins RO, Weaver LK. Does late repetitive hyperbaric oxygen improve delayed neurologic sequelae associated with carbon monoxide poisoning? Undersea Biomed Res 1991; 18 Suppl: 34. Weaver LK, Hopkins RO, Larson-Lohr V, et al. Double-blind, controlled, prospective, randomized clinical trial (rct) in patients with acute carbon monoxide (CO) poisoning: outcome of patients treated with normobaric oxygen or hyperbaric oxygen -- an interim report. Undersea Hyperb Med 1995; 22 Suppl: 14. (Received 25 Mar, accepted 1 Dec, 1998) Authors' details Alfred Hospital, Melbourne, VIC. Carlos D Scheinkestel, FRACP, DipDHM, Deputy Director, Department of Intensive Care and Hyperbaric Medicine, and Head, General Intensive Care Unit; Paul S Myles, MD, FANZCA, Head of Research, Department of Anaesthesia and Pain Management; D James Cooper, MD, FRACP, Head, Trauma Intensive Care Unit; Ian L Millar, FAFOM, DipDHM, Head, Hyperbaric Medicine; David V Tuxen, MD, FRACP, Director, Department of Intensive Care and Hyperbaric Medicine. Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, VIC. Michael Bailey, BSc, MSc(Stat), Statistical Consultant. School of Psychological Science, La Trobe University, Melbourne, VIC. Kerry Jones, BBSc(Hons), MPsych, Psychologist, and PhD student. Reprints: Dr C D Scheinkestel, Department of Intensive Care and Hyperbaric Medicine, Alfred Hospital, Commercial Road, Prahran, Melbourne, VIC 3181. Email: cdschATozemail.com.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> 1: Published trials of carbon monoxide poisoning treated with hyperbaric oxygen (HBO) and normobaric oxygen (NBO)StudyPatients includedNo. HBO patientsNo. NBO patientsBlindedNeuropsychological testsNon-randomised Roche et al11All2020NoNo Mathieu et al16All20327NoNo Myers et al20All13182NoYes Willms et al18All7246NoNo Goulon et al9All27329NoNo Gorman et al13All928NoYes Ely et al19All426NoNoRandomised Raphael et al15Non-comatose Severely poisoned173 145170 141No NoNo No Thom et al14Mildly poisoned3332NoYes Ducasse et al12Non-comatose1313YesNo Mathieu et al17Non-comatose299276NoNo This study*All10487YesYesStudyNumber of treatmentsMaximum ATATime at maximum ATA (min)Entry criteria time to treatment (h)Actual time to treatment (h)HBO benefit reportedNon-randomised Roche et al111-102.560??Yes Mathieu et al161-52.590?~4Yes Myers et al20?1.8-2.046?1-1.5Yes Willms et al1832.5???Yes Goulon et al92290??Yes Gorman et al131, 2, >22.860??Yes Ely et al1912.5120??YesRandomised Raphael et al151 1 v. 22 2120 12012 12? ?No No Thom et al141 2.8 230 90 <6 ~1 Ducasse et al121???~1Yes Mathieu et al1712.59012?No This study*≥32.860<247.1 (95% CI, 1.9-26.5)NoATA = atmospheres absolute; HBO = hyperbaric oxygen; NBO = normobaric oxygen; ? = not reported; ~ = approximately. * Our study was the only one listed in which there were sham treatments for NBO patients. 2: Patient characteristics and description of severity of carbon monoxide poisoning for patients treated with hyperbaric oxygen (HBO) and normobaric oxygen (NBO)HBO (n = 104)NBO (n = 87)PDemographic characteristics Age37.8 (35.1-40.5)34.8 (32.0-37.6)0.13+ Male89 (86%)67 (77%)0.13++ Suicide attempt68 (65%)63 (72%)0.3++Baseline severity Ventilated20 (19%)16 (18%)0.88++ Exposure time (h)2.6 (2.0-3.2)2.5 (1.9-3.1)0.87+ Time to carboxyhaemoglobin level (h)*3.2 (2.6-3.8)2.6 (2.1-3.1)0.11+ Carboxyhaemoglobin level (%)20.5 (18.0-23.0)22.0 (19.6-24.4)0.39+ Time to treatment (h)*7.5 (6.6-8.6)6.6 (5.7-7.5)0.16+ Mini-mental score27.0 (26.1-27.9)26.4 (25.4-27.4)0.27+ No. with criteria for severe poisoning72 (69%)67 (77%)0.23++Signs Coma53 (51%)49 (56%) Acidosis11 (11%)13 (15%) Focal neurological deficits9 (9%)6 (7%) Electroencephalogram changes7 (7%)9 (10%) Hypotension3 (3%)2 (2%) Arrhythmias2 (2%)7 (8%) Pulmonary oedema2 (2%)1 (1%) Convulsions1 (1%)3 (3%) Cardiac arrest1 (1%)1 (1%)Symptoms Headache55 (53%)38 (44%) Fatigue47 (45%)38 (44%) Difficulty in thinking46 (44%)37 (43%) Dizziness38 (37%)21 (24%) Nausea38 (37%)25 (29%) Acute confusional state20 (19%)10 (11%) Paraesthesiae11 (11%)6 (7%) Visual disturbance9 (9%)4 (5%) Palpitations8 (8%)3 (3%) Chest pain7 (7%)3 (3%) Tinnitus4 (4%)2 (2%) Abdominal pain3 (3%)0 Diarrhoea2 (2%)2 (2%)Figures are number (%) or mean (95% CI). HBO = hyperbaric oxygen; NOB = normobaric oxygen. *Geometric mean; +t test; ++chi-squared test. 3: Neuropsychological outcome for all patients treated with hyperbaric oxygen (HBO) and normobaric oxygen (NBO)HBO (n = 104)NBO (n = 87)Difference (95% CI) in favour of HBO*PAt end of treatment No. requiring >3 treatments29 (28%)13 (15%)OR 2.8 (1.3-6.2)++0.01++ Deaths3 (3%)3 (3.4%)0.96++Average neuropsychological test results+ Simple reaction time (s)385375-10 (-50 to 30)0.63s Choice reaction time (s)10.411.1-0.7 (-1.9 to 0.5)0.25s Digit span forward (no. digits recalled)8.28.10.1 (-9 to 1.1)0.87s Digit span backwards (no. digits recalled)5.35.6-0.3 (-1.1 to 0.5)0.55s Rey auditory verbal learning test (score)42.247.7-5.5 (-9.8 to -1.2)0.01s Short term free recall (no. objects)3.23-0.2 (-1.1 to 0.7)0.54s Long term free recall (no. objects)4.44.2-0.2 (-1.5 to 1.1)0.76s Improvement in Mini-mental score00.7-0.7 (-2.7 to 1.3)0.53sAverage number of abnormal tests 3.42.7-0.7 (-1.3 to -0.1)0.02sPoor outcome (PNS)0.740.68OR 1.7 (0.8-4.0)++0.19++Relapse (DNS)5 (4.8%)00.03**Figures are number (%) or mean (95% CI). HBO = hyperbaric oxygen; NBO = normobaric oxygen; OR = odds ratio; PNS = persistent neurological sequelae (>2 abnormal test results); DNS = delayed neurological sequelae. *After adjustment for age, sex, suicide attempt, carboxyhaemoglobin (COHb) level, time to measurement of COHb level, duration of exposure, time to treatment, presence of other drugs; +See Methods section for description of neuropsychological tests; ++chi-squared test; sF-test; **Fisher's exact test. 4: Characteristics of 139 patients with severe carbon monoxide poisoning, and the neurophysiological outcome for this subgroupHBO (n = 72)NBO (n = 67)Difference (95% CI) in favour of HBO*P*Demographic characteristics Age38.6 (35.2-42.0)36.7 (33.5-39.9)0.42** Male60 (83%)55 (82%)0.99s Suicide55 (76%)54 (81%)0.69++Baseline severity Ventilated20 (28%)16 (24%)0.74s Exposure time (h)2.5 (1.9-3.1)2.2 (1.6-2.8)0.45** Time to carboxyhaemoglobin level (h)+2.9 (2.3-3.7)2.2 (1.9-2.7)0.07** Carboxyhaemoglobin level (%)+23.0 (19.8-26.2)22.9 (22.0-25.8)0.94** Time to treatment (hours)7.3 (6.3-8.5)6.1 (5.4-6.9)0.08** Mini-mental score25.6 (24.3-26.9)25.8 (24.5-27.1)0.8**At end of treatment No. requiring >3 treatments25 (35%)9 (13%)OR 5.4 (2.0-14.8)0.001s Deaths3 (4.2%)3 (4.5%)OR 1.0 (0.2-6.0)0.97sAverage neuropsychological test results+ Simple reaction time (s)410377-33 (-72 to 6.0)0.01ss Choice reaction time (s)9.910.7-0.8 (-2.4 to 0.8)0.32ss Digit span forward (no. digits recalled)7.98-0.1 (-1.3 to 1.1)0.91ss Digit span backwards (no. digits recalled)55.5-0.5 (-1.6 to 0.6)0.35ss Rey auditory verbal learning test (score)4249.2-7.2 (-12.2 to -2.2)0.005ss Short term free recall (no. objects)43.5-0.5 (-1.7 to 0.7)0.43ss Long term free recall (no. objects)4.65.20.6 (-1.1 to 2.3)0.47ss Improvement in Mini-mental score21.60.4 (-0.2 to 2.8)0.71ssAverage number of abnormal tests3.72.6-1.1 (-1.9 to -0.3)0.008ssPoor outcome (PNS)0.850.65OR 3.6 (1.1-11.9)0.03sFigures are number (%) or mean (95% CI). HBO = hyperbaric oxygen; NBO = normobaric oxygen; OR = odds ratio; PNS = persistent neurological sequelae (>2 abnormal test results); DNS = delayed neurological sequelae. *After adjustment for age, sex, suicide attempt, carboxyhaemoglobin (COHb) level, time to measurement of COHb level, duration of exposure, time to treatment, presence of other drugs; +Geometric mean (95% CI); ++See Methods section for description of neuropsychological tests; schi-squared test; **t-test; ssF-test.

Carlos D Scheinkestel · Michael Bailey · Paul S Myles · Kerry Jones · Ian L Millar · David V Tuxen

Research Hospital in the home: a randomised controlled trial Gideon A Caplan, John A Ward, Nicholas J Brennan, Janis Coconis, Neville Board and Ann Brown MJA 1999; 170: 156-160 For editorial comment, see Montalto Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Administration and health services Abstract Objectives: To compare treatment of acute illness at home and in hospital, assessing safety, effect on geriatric complications, and patient/carer satisfaction. Design: Randomised controlled trial. Setting: A tertiary referral hospital affiliated with the University of New South Wales. Participants: 100 patients (69% older than 65 years) with a variety of acute conditions, who were assessed in the emergency department as requiring admission to hospital. Interventions: Patients were allocated at random to be treated by a hospital-in-the-home (HIH) service in their usual residence or to be admitted to hospital. Main outcome measures: Geriatric complications (confusion, falls, urinary incontinence or retention, faecal incontinence or constipation, phlebitis and pressure areas), patient/carer satisfaction, adverse events, and death. Results: There was a lower incidence of confusion (0 v. 20.4% [95% CI, 9.1%-31.7%]; P = 0.0005), urinary complications (incontinence or retention) (2.0% [95% CI, -1.8%, 5.8%] v. 16.3% [95% CI, 6.0%, 26.6%]; P = 0.01), and bowel complications (incontinence or constipation) (0 v. 22.5% [95% CI, 10.7%, 34.1%]; P =0.0003) among HIH-treated patients. No significant difference in number of adverse events and deaths (to 28 days after discharge) in the two groups was found (although numbers were small). Patient and carer satisfaction was significantly higher in the HIH group. Conclusions: Home treatment appears to provide a safe alternative to hospitalisation for selected patients, and may be preferable for some older patients. We found high levels of both patient and carer satisfaction with home treatment. Introduction Acute care of patients at home is one of the fastest-growing healthcare sectors in the United States1-3 and is gaining acceptance in many countries.4 Although there have been randomised controlled trials of patients receiving home versus hospital treatment for deep venous thrombosis,5 there are few data from trials involving other conditions, or assessing the safety of acute care at home, especially for older patients. It is these patients who occupy an increasing proportion of hospital beds and may derive most benefit from home treatment.6,7Recent studies characterising hospital-associated adverse events8,9 make home treatment more attractive,10 especially for elderly patients who are known to suffer a higher incidence of complications in hospital.11 When treated at home, patients do not have to change their environment or routine, they are not exposed to nosocomial infection, and they do not need to adapt to the sociological culture of the hospital. On the other hand, the incidence and severity of intrinsic complications in older patients can be minimised by round-the-clock supervision by expert staff in hospital. However, there is inadequate evidence to support the popular belief that programs which transfer care to the home result in a worse outcome for the patients.12,13 We have conducted a randomised controlled trial of a hospital-in-the-home (HIH) program targeting older patients. We selected conditions which were amenable to home treatment, but tried to include a variety of diagnostic groups. To accentuate the difference between the two groups, the HIH patients were taken home on the day of presentation to the emergency department, or on the following morning if they presented at night. We studied the safety, efficacy and patient/carer satisfaction of home treatment compared with hospital treatment. Methods Patients and randomisation One hundred patients requiring admission to hospital and meeting criteria for study inclusion were randomised to HIH treatment (intervention group) or treatment in hospital (control group). We targeted patients older than 65 years, including those living in nursing homes, but also accepted younger patients. Patients were accepted only if they had been assessed as requiring admission by the relevant medical or surgical team. The medical, surgical and emergency department staff were encouraged to refer patients with acute (pneumonia, urinary tract infections and cellulitis) and subacute (endocarditis and osteomyelitis) infections requiring treatment with intravenous antibiotics, deep venous thrombosis (DVT), minor cerebrovascular accidents (not affecting mobility or swallowing), and cardiac failure. Patients were excluded from the study if they had evidence of shock (systolic blood pressure < 100 mmHg); if they required oxygen (Pao2 < 60 mmHg); if they were judged too unwell by the study team; if they had no available carer; if they lived outside the local area; or if their home was unsuitable for home treatment (lack of running water, electricity or an inside toilet, concerns about safety, or dangerous pets). If any doubts about home suitability arose during the emergency department assessment, a study nurse visited the home before randomisation. After informed consent was obtained from the patient (written) and the patient's carer (verbal), randomisation (stratified according to whether the patient lived at home or in a nursing home, or had DVT) was achieved by computer-generated random numbers coded into sealed envelopes. The study protocol required that patients in the HIH group be discharged from hospital within 24 hours of diagnosis. HIH patients receiving intravenous antibiotics were given the first dose in hospital. The study protocol was approved by the hospital's Research Ethics Committee. Data collection For all study patients, study nurses completed a Barthel Index of Activities of Daily Living,14 a modified Instrumental Activities of Daily Living Index15 and a Mental Status Questionnaire16 on admission and discharge. Baseline demographic data on current medical diagnoses, a detailed social history, physical function, medications and allergies were also recorded. Hospital treatment Hospital patients (control group) were admitted under the appropriate physician or surgeon of the day and treated in accordance with standard regimens without the intervention of the study team. The hospital team treating the patient was notified that the patient had been included in the trial. HIH treatment HIH patients (intervention group) were treated according to the presenting diagnosis by the hospital community outreach team. The range of treatments in the study protocol included administration of parenteral antibiotics and other medications, and blood transfusions. Infections were generally treated with once-daily intravenous antibiotics such as ceftriaxone, gentamicin or vancomycin, as appropriate, according to the result of bacteriological tests (if positive) or by diagnosis. Intravenous access was usually via cannulas inserted by study nurses. Owing to safety concerns, some patients with dementia did not have cannulas left in situ, and were cannulated daily with butterfly cannulas. Patients requiring long term treatment or with difficult venous access were treated via a peripherally inserted central cannula. Patients with DVT were treated with daily subcutaneous enoxaparin injections (1.5 mg/kg) and oral warfarin until their international normalised ratio was 2.0 or above. GP involvement Before the study, we conducted an educational program consisting of an evening lecture and a question-and-answer session for the local division of general practice. Study nurses also visited general practitioners (GPs) and nursing homes to explain the program. Our preference was for the patient's own GP to be the primary medical manager, and we therefore sought the agreement of the patient's GP before entering the patient in the trial. If GPs declined to participate, the hospital provided medical support. Follow-up After discharge, an unmarked satisfaction survey based on the principles outlined by Draper and Hill,17 colour coded to differentiate between responses from HIH and hospital treatment, was sent to patients, carers and GPs. This was followed by a phone call to encourage response. Patients were asked "How would you rate your treatment overall?" and offered a four-point scale of answers: excellent (1), good (2), fair (3), and poor (4). A similar question about satisfaction with the patient's care was asked of the carer and GP. Patients were recontacted by telephone at one and six months after discharge to ascertain their health status. Complications Complications were assessed by two methods. Study personnel performed a systematic medical record review, focusing on the more common geriatric syndromes, as recorded in the patients' notes. The geriatric syndromes18 include confusion, falls, incontinence of bladder and bowel, constipation, urinary retention, and pressure ulcers. These are not specific to older patients, but occur more commonly in frail elderly patients (geriatric patients). The medical records were also assessed according to the validated method of counting adverse events developed for the Harvard Medical Practice Study8 and refined in the Quality in Australian Health Care Study (QAHCS). This was done by independent reviewers engaged from the team of the Australian study.9 This method records an adverse event -- unintended injury or complication -- only if it results in disability, death or prolonged hospital stay and is caused by healthcare management. The reviewers were a doctor and a nurse with extensive experience of QAHCS procedures. Reliability assessments were undertaken by a senior member of the QAHCS study team. The only difference from the standard technique was that the reviewers were instructed only to seek adverse events arising after randomisation, whereas the original study also sought those which had caused the admission. Statistical analyses The study had a power of 80%, assuming = 0.05, to find a difference of about 20% in the occurrence of complications. All analyses were performed on the basis of intention to treat, using the SPSS for Windows and Epi Info statistical packages.19,20 Continuous data are expressed as means and 95% confidence intervals (95% CI) and were compared by t tests. Satisfaction scores were compared by the Mann-Whitney U test. Fisher's exact test was used to compare proportions. All statistical tests were two-tailed. Results Between October 1995 and February 1997, 129 patients requiring admission to hospital were assessed for participation in the study. Reasons for exclusion before randomisation were: patient too unwell for home treatment (12 patients), patient declined (6), carer refused or no carer available (9), and patient lived too far from the hospital (2). One hundred patients were enrolled; 51 were allocated at random to HIH treatment and 49 to hospital treatment. The two groups were comparable in characteristics and diagnoses (Tables 1 and 2). A quarter of the patients lived in nursing homes and 69% were 65 years of age or older. For 92 of the patients their GP agreed before randomisation to assist in their care. Four patients had no GP, one GP was not interested and three patients' GPs could not be contacted before randomisation. Adverse events and complications The proportion of adverse events was similar in the HIH group and the control (hospital) group (11.8% [95% CI, 3.0%, 20.6%] v. 16.3% [95% CI, 6.0%, 26.6%]). However, there was a significantly lower occurrence in the HIH group of confusion, urinary (either incontinence or retention) complications, and bowel (either constipation or faecal incontinence) complications (Table 3). Home visits On average, patients in the HIH group were seen, at home, on 9.0 occasions by the study nurse, 0.8 times by their GP, 0.9 times by a doctor from the hospital, 0.2 times by a physiotherapist, and 0.1 times by an occupational therapist. This translates to one visit per day by the nurse, given that each patient was also seen by a study nurse in the emergency department. Outcomes Treatment failed in four of the 51 patients in the HIH group and they were admitted to hospital. Three and five patients in the HIH group and the hospital-treated group, respectively, were recorded as having an unplanned readmision within 28 days of discharge. There was one death in each of the HIH- and hospital-treated groups during the admission. Total deaths up to 28 days after discharge were three in the HIH group and four in the hospital-treated group, and at six months after discharge six and seven deaths, respectively, had occurred. There were no significant differences between the two groups in these outcomes. Patient, carer and GP satisfaction The response rates for the satisfaction surveys were (HIH v. control): patients 78% v. 40%, carers 55% v. 27%, and GPs 63% v. 37%. There was significantly greater satisfaction with overall HIH treatment among both patients (P < 0.0001) and their carers (P = 0.0001). Despite the increased workload and despite having to deal with more severely ill patients than usual, GPs were equally satisfied by home and hospital treatment (Table 4). Discussion In our study, treatment of a range of acute infections and other diseases at home was as safe as treatment in hospital. Previous studies which claimed to have demonstrated safety documented so few complications that risk appeared to have been eliminated by patient selection.1-3 Our study group's 7% mortality up to 28 days after discharge (compared with zero mortality in previous studies1-3) establishes our patients' illness severity. The most significant finding resulting from our inclusion of geriatric complications was the higher incidence of confusion in hospital-treated patients. The role of the hospital environment in the aetiology of delirium has long been recognised,21 although this is not universally accepted.22 For some of the patients in our study recorded as having confusion, all of the features of delirium according to the Diagnostic and statistical manual of mental disorders (DSM-IV)23 were not present. Our aim was to be as inclusive as possible of geriatric complications, so that if a problem was noted by ward staff or carers it was included even if study personnel had not observed it. These complications were not always serious clinically and often did not have a great impact, but we valued the perspective of patients, families and hospital staff. Similarly, the occurrence of urinary and bowel complications reflects the difficulty of caring for frail elderly patients24 (especially those from nursing homes) in an acute hospital ward. Patients with mobility or communication problems require carers attuned to their routines and signals to maintain continence. Because of small numbers our study was limited in its power to draw conclusions about the difference between HIH and hospital treatment as far as adverse events and deaths were concerned. As there have been no previous studies addressing this question, it was not possible to estimate the size of the expected difference. Another limitation of our study was that, although we did not directly inform the adverse event reviewers about which patients were in each group, we were not able to completely blind them to the status of each group because of stylistic differences between the patient records of the hospital- and the HIH-treated patients. Few similar comparative studies have been published. As reports and discussions entitled "home hospitalisation" or "hospital at home", for example, may include few or no acute patients, the home treatment of acute illness is obscured in the medical literature.25,26 Our study included only patients who met the criterion of requiring admission to hospital because of illness. By bringing the hospital care to the patient's home, we substituted completely for hospital facilities. There is a need for more studies of patients with illness of this severity treated at home. A number of factors may explain the greater satisfaction with treatment at home. In their own homes, patients are in their familiar environment, their privacy is protected, their sleep less interrupted, and they can eat their usual food.27,28 As an example, for an elderly patient with limited cognitive function and/or decreased visual acuity, finding the bathroom at night in hospital may be impossible. Moreover, elderly patients may be unable to get out of bed because of rails on the sides of the bed. These difficulties may have contributed to the higher incidence of incontinence in our hospital-treated group. At home, patients are more active participants, often partners, in their care. All these factors invariably translate into greater satisfaction with home treatment.29 Frail, older patients -- because of their atypical disease presentations, more rapid deterioration when ill and greater prevalence of disease -- need hospital-level services more frequently than younger patients. However, admission to hospital, like any medical intervention, has side effects as well as benefits.8 Elderly patients have more to gain from many treatments (eg, thrombolytic therapy) because of their higher underlying mortality rates,30 but they also suffer from a much greater occurrence of iatrogenic complications.10 Although many of these complications may be explained by an interaction between their underlying frailty and their disease process,8 we were able to detect a measurable effect attributable to a change of environment. The results of this study do not imply that older patients should never be admitted to hospital. On the contrary, the elderly as a group have more need for the technical expertise of the acute hospital than younger people. However, if such technical expertise can be delivered at home, this alternative should at least be evaluated. On the other hand, our study does not imply that treatment which can be delivered at home should always be delivered at home. Successful home treatment depends on careful patient and home selection, as the necessary minimum level of sanitation, facilities and support from carers is not universally available.4 However, where it is feasible, we have demonstrated that home treatment of acute illness, even for frail elderly patients, can be effective, safe and more satisfying for the patient and their families. Acknowledgements We would like to acknowledge the invaluable assistance of the staff of Post Acute Care Services and the Emergency Department at Prince of Wales Hospital and Ms Bernie Harrison and Dr Ross Wilson from QARNS at Royal North Shore Hospital. This study was supported by a grant from the Commonwealth Department of Health and Family Services - Hospital Access (National Projects) Program. References Antoniskis A, Anderson BC, Von Volkinburg EJ, et al. Feasibility of outpatient self-administration of parenteral antibiotics. West J Med 1978; 128: 203-206. Kind AC, Williams DN, Persons G, et al. Intravenous antibiotic therapy at home. Arch Intern Med 1979; 139: 413-415. Stiver HG, Telford GO, Mossey JM, et al. Intravenous antibiotic therapy at home. Ann Intern Med 1978; 89: 690-693. Tice AD. Outpatient parenteral antibiotic therapy in different countries. Int J Infect Dis 1996; 1: 102-106. Levine M, Gent M, Hirsh J, et al. A comparison of low-molecular-weight heparin administered primarily at home with unfractionated heparin administered in the hospital for proximal deep-vein thrombosis. N Engl J Med 1996; 334: 677-681. Balinsky W, Nesbitt S. Cost-effectiveness of outpatient parenteral antibiotics: a review of the literature. Am J Med 1989; 87: 301-305. Poretz DM, Eron LJ, Goldenberg RI, et al. Intravenous antibiotic therapy in an outpatient setting. JAMA 1982; 248: 336-339. Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalised patients: results of the Harvard Medical Practice Study II. N Engl J Med 1991; 324: 377-384. Wilson RM, Runciman WB, Gibberd RW, et al. The quality in Australian health care study. Med J Aust 1995; 163: 458-471. Caplan GA, Brown A. Post acute care: can hospitals do better with less? Aust Health Rev 1997; 20: 43-54. Fretwell MD. Acute hospital care for frail older patients. In: Hazzard WR, Andres R, Bierman EL, Blass JP, editors. Principles of geriatric medicine and gerontology, 2nd ed. New York: McGraw-Hill, 1990: 247-253. Maxwell RJ. Why rationing is on the agenda. Br Med Bull 1995; 51: 761-768. Larkins RG, Martin TJ, Johnston CI. The boundaryless hospital -- a commentary. Aust N Z J Med 1995; 25: 169-170. Mahoney FI, Barthel DW. Functional evaluation: the Barthel index. Md Med J 1965; 14: 61-65. Lawton MP, Brody EM. Assessment of older people: self-maintaining and instrumental activities of daily living. Gerontologist 1969; 9: 179-186. Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. J Am Geriatr Soc 1975; 23: 433-441. Draper M, Hill S. The role of patient satisfaction surveys in a national approach to hospital quality management. Canberra: AGPS, 1996. Hazzard WR, Andres R, Biermen EL, Blass JP, editors. Principles of geriatric medicine and gerontology. 2nd ed. Section 4 -- geriatric syndromes and special problems. New York: McGraw-Hill, 1990: 1055-1222. SPSS for Windows [computer program], version 6.0. Chicago: SPSS Inc, 1993. Dean AG, Dean JA, Burton AH, Dicker RC. Epi-Info [computer program], version 5. Stone Mountain, Ga: USD Inc, 1990. Wahl CW, Golden JS, Liston EH, et al. Toxic and functional psychoses: diagnosis and treatment in a medical setting. Ann Intern Med 1967; 66: 989-1007. Lipowski ZJ. Delirium and impaired consciousness. In: Evan JG, Williams TF, editors. Oxford textbook of geriatric medicine. Oxford: Oxford University Press, 1992: 490-495. American Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-IV). Washington, DC: APA, 1994. Caplan GA, Brown A, Croker WD, Doolan J. Risk of admission within four weeks of discharge of elderly from the emergency department -- the DEED study. Age Ageing. In press. Stressman J, Ginsberg G, Hammerman-Rozenberg R, et al. Decreased hospital utilization by older adults attributable to a home hospitalization program. J Am Geriatr Soc 1996; 44: 591-598. Donald IP, Baldwin RN, Bannerjee M. Gloucester hospital-at-home: a randomized controlled trial. Age Ageing 1995; 24: 434-439. Consumer reports of hospital experiences. In: Draper M, Hill S, editors. The role of patient satisfaction surveys in a national approach to hospital quality management. Canberra: AGPS, 1996: 12-20. Cruse PJE, Foord R. The epidemiology of wound infection: a 10-year prospective study of 62,939 wounds. Surg Clin North Am 1980; 60: 27-40. Brown RB. Selecting the patient. Hosp Pract 1993; 28(Suppl 1): 11-15. Fibrinolytic Therapy Trialists' (FTT) Collaborative Group. Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet 1994; 343: 311-322. (Received 17 Jun, accepted 22 Oct, 1998) Authors' details Post Acute Care Services, Prince of Wales Hospital, Sydney, NSW. Gideon A Caplan, MB BS, FRACP, Director; and Lecturer, University of New South Wales, Sydney, NSW; Nicholas J Brennan, MB BS, FRACP, Senior Registrar; Janis Coconis, RN, Clinical Nurse Specialist; Neville Board, RN, BA, Research Assistant; Ann Brown, RN, Nurse Manager. Community Health Services and Programs, South Eastern Sydney Area Health Service, Sydney, NSW. John A Ward, MB BS, MSc, FRACP, Community Geriatrician; and Lecturer, University of New South Wales, Sydney, NSW. Reprints: Dr G A Caplan, Post Acute Care Services, Prince of Wales Hospital, Randwick, NSW 2031. Email: g.caplanATunsw.edu.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> Back to text Back to text Back to text Back to text

Gideon A Caplan · John A Ward · Nicholas J Brennan · Janis Coconis · Neville Board · Ann Brown

Research Mortality associated with New South Wales methadone programs in 1994: lives lost and saved John R M Caplehorn and Olaf H Drummer MJA 1999; 170: 104-109 For editorial comment, see Ali & Quigley Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Drugs and alcohol Abstract Objectives: To estimate the effects of methadone programs in New South Wales on mortality. Design and cases: Retrospective, cross-sectional study of all 1994 New South Wales coronial cases in which methadone was detected in postmortem specimens taken from the deceased. Cases were people we identified as patients in NSW methadone maintenance programs or those whose deaths involved methadone syrup diverted from maintenance programs. Outcome measures: Relative risks of fatal, accidental drug toxicity in the first two weeks of treatment and later; the number of lives lost as a result of maintenance treatment; preadmission risks and the number of lives saved by maintenance programs, calculated from data from a previous study. Results: There was very close agreement between this study's classifications and official pathology reports of accidental drug toxicity. The relative risk (RR) of fatal accidental drug toxicity for patients in the first two weeks of methadone maintenance was 6.7 times that of heroin addicts not in treatment (95% CI RR, 3.3-13.9) and 97.8 times that of patients who had been in maintenance more than two weeks (95% CI RR, 36.7-260.5). Despite 10 people dying from iatrogenic methadone toxicity and diverted methadone syrup being involved in 26 fatalities, in 1994 NSW maintenance programs are estimated to have saved 68 lives (adjusted 95% CI, 29-128). Conclusions: In 1994, untoward events associated with NSW methadone programs cost 36 lives in NSW. To reduce this mortality, doctors should carefully assess and closely monitor patients being admitted to methadone maintenance and limit the use of takeaway doses of methadone. Introduction Methadone maintenance greatly reduces heroin addicts' risk of death.1 A 15-year follow-up of patients in New South Wales showed methadone maintenance saved lives by reducing addicts' risk of fatal heroin overdose.1 When combined in a meta-analysis with the results of overseas cohort studies, the relative risk of death in methadone maintenance was a quarter that of addicts not in treatment (95% CI, 0.19-0.33).1 However, methadone maintenance is also a cause of death. Patients are at risk of fatal iatrogenic toxicity and other drug users may die from taking methadone syrup diverted from maintenance programs.2-6Mortality associated with NSW maintenance programs was independently investigated. The first report from this project presented the case histories of the 13 patients who died in the first two weeks of treatment.7 It identified 10 probable cases of fatal iatrogenic methadone toxicity (ie, where prescribed doses of methadone either caused or contributed to fatal accidental drug toxicity).7 This, the second report, presents an estimate of the relative risk of fatal accidental drug toxicity in the first two weeks and later maintenance. It also presents estimates of the effect of admission to methadone maintenance on the risk of fatal accidental drug toxicity and of the number of lives saved by NSW maintenance programs in 1994. Methods This study was approved by the Human Research Ethics Committee of the Western Sydney Area Health Service and the NSW State Coroner. In late 1995 the database at the NSW Health Department's Division of Analytical Laboratories was searched to identify 1994 coronial cases in which methadone was detected in postmortem specimens. These analytical laboratories receive specimens for toxicological analysis in all cases of sudden death referred to the NSW State Coroner. Autopsy, toxicology and police reports and the statements of family and friends, prescribers and other witnesses were collected from coronial files. The methadone treatment histories of the deceased were extracted from data held by the NSW Health Department's Pharmaceutical Services Section. The Department also provided data on the number of people admitted to and treated with methadone maintenance in NSW in 1994. Cases were grouped according to the source of the methadone: methadone syrup given as maintenance treatment; methadone syrup diverted from the maintenance program; and methadone tablets (Physeptone; Glaxo Wellcome, Boronia, Vic.) prescribed for pain relief. As the Sydney black market consists almost entirely of methadone syrup diverted from maintenance programs,8 illicit drug users who obtained methadone from an unknown source were classified as having taken diverted syrup. We used two parallel classifications of cause of death -- that on the official pathologist's report, and our own. In our classification, we initially established cause of death independently of one another, with one of us (O H D) blind to the official cause of death. Cases were first categorised as "accidental drug toxicity" and "other". The "other" category included suicides, deaths from natural causes and trauma, and deaths in which drug toxicity was considered to have contributed to a death from natural causes. The "accidental drug toxicity" cases were further categorised into "methadone" and "other drug or drugs" on the basis of whether or not methadone was considered to have either caused or made a significant contribution to the death. There were no simple criteria for establishing the contribution of methadone to deaths involving other drugs. However, as deaths to which methadone contributed closely resembled cases of fatal methadone toxicity,2,3,7 a relatively confident decision could be made after a thorough examination of the documentary and toxicological evidence and the autopsy report.2,7,9-12 Police statements and photographs of the deceased at the scene of death provided some assistance. A brownish, frothy oedema fluid was often observed coming from the deceased's mouth or nose (see Box 1).7 Witnesses' statements provided a guide to likely tolerance and chronologies of ingestion and of the development of symptoms and signs of toxicity.2,7 These statements were particularly useful in cases involving methadone as death usually occurred some hours after the drug was taken,14 and some time after the development of coma (see Box 1).2,3,7,10,11 Postmortem blood methadone concentration was helpful but not definitive, as fatal concentration varies widely with tolerance11,12 and the blood concentration of methadone increases after death.15 Moreover, the postmortem increase in blood methadone concentration varies unpredictably from one part of a cadaver to another.16 The autopsy findings were remarkably consistent in cases of fatal drug toxicity involving methadone, with the immediate cause of death being pulmonary oedema secondary to hypoventilation.2,7,10,11 As methadone toxicity usually causes a gradually worsening hypoventilation, the hypoxia and resulting pulmonary hypertension are generally prolonged and severe, and significant quantities of water and electrolytes, large proteins and red blood cells leak from the pulmonary capillaries into the air spaces. Consequently, brownish oedema fluid was often observed in the large airways and the lungs were unusually heavy (see Box 1). Microscopic examination of lung specimens often showed areas of patchy bronchopneumonia and other evidence of prolonged hypoventilation and suppression of the cough reflex.2,7,10 Statistical analysis We used published estimates of NSW methadone patients' risks of death after leaving treatment as approximations of 1994 NSW methadone patients' risks before admission to treatment.1Rates were adjusted for age, as the risk of death was significantly higher for those aged 20-29 years compared with those aged 30-39 years.1 Weighted average risks were calculated in the knowledge that, in 1994, 68% of NSW maintenance patients were at least 30 years of age.17 We assumed half of those admitted to maintenance were aged 20-29 years and half 30-40 years. The 95% confidence intervals of mortality rates were calculated by dividing the estimates by significance factors taken from a published table.18 The standard errors of the relative risks were estimated using the binomial approximation of the Poisson distribution.19 Results Methadone was detected in postmortem material from 89 NSW coronial cases in 1994. These cases comprised 41 methadone maintenance patients (38 registered with the NSW Health Department and three with the Queensland Health Department), one neonate being breastfed by a NSW methadone maintenance patient, 29 cases considered to have involved methadone syrup diverted from the NSW methadone program, and 18 cases considered to have involved methadone tablets. In 18 of the 29 cases involving diverted methadone syrup, either a bottle used to dispense methadone syrup (5 cases), a statement from a witness (10 cases), or both (3 cases), indicated that the maintenance program was the source of the methadone. In the remaining 11 cases, it was assumed methadone syrup was obtained from the black market.8 In 16 of the 18 cases involving methadone tablets, either a statement from the prescribing doctor (8 cases), a tablet bottle (4 cases), or both (4 cases), indicated the source of the methadone. The remaining two people had professional access to methadone tablets and committed suicide. We excluded the three Queensland maintenance patients, the neonate and all cases involving methadone tablets, leaving 67 cases in the study. Methadone maintenance patients Box 2 shows that, of the 38 NSW maintenance patients, 13 died in the first two weeks after admission, and 25 died later in treatment. We and the official pathologists concluded that 12 of the 13 fatalities in the first two weeks of maintenance and six of the 25 deaths later in treatment were caused by accidental toxicity. Three of six deaths from accidental drug toxicity among established maintenance patients were caused by heroin, one by dextromoramide, one by the combined effects of heroin and dextroproxyphene, and one involved injected methadone syrup. Diverted methadone syrup Box 2 shows that, for the 29 cases involving diverted methadone, we concluded methadone contributed to 26 of 27 deaths from accidental drug toxicity compared with 24 of 26 on the official pathologists' reports. One death which we classified as accidental drug toxicity was officially attributed to bronchopneumonia with methadone intoxication as a contributing factor. In another case, we concluded injected, diverted methadone contributed to a death which was officially attributed to acute heroin poisoning. Witnesses' statements or autopsy reports indicated that methadone syrup was injected in 16 of the 26 cases we classified as accidental drug toxicity to which diverted methadone contributed. One of the 10 cases involving oral ingestion of diverted methadone was that of an infant who either took or was given some of his mother's syrup. Relative risks of accidental drug toxicity We concurred with official pathologists' conclusions that 12 patients died of accidental drug toxicity during the first two weeks of maintenance treatment in NSW in 1994 (see Box 2). To calculate the rate of fatal accidental drug toxicity, we estimated the total time patients spent in the first two weeks of maintenance treatment. In 1994, 4449 people were admitted to methadone maintenance in NSW. Assuming all new admissions stayed at least two weeks in treatment,20,21 patients spent approximately 170.5 person-years in the first two weeks of maintenance. Using this estimate as the denominator, the rate of fatal accidental drug toxicity in the first two weeks of maintenance was 70.4 deaths per thousand per year (Box 3). We also agreed with official pathologists' conclusions that six NSW methadone patients died from accidental drug toxicity after being in maintenance treatment for at least two weeks (Box 2). An approximation of the total time methadone patients spent in treatment in NSW in 1994 was derived from the average of the number in treatment at the beginning and end of the year (7975 and 9038, respectively).22 The 170.5 person-years spent in the first two weeks' maintenance were subtracted from the average of the totals, 8506.5, to estimate the total time spent in later maintenance -- 8336 person-years. When this was used as the denominator, the rate of fatal accidental drug toxicity in later maintenance was 0.72 deaths per thousand per year (Box 3). When combined with the previous estimate, the risk of fatal accidental drug toxicity in the first two weeks of treatment in NSW in 1994 was estimated to have been 97.8 times the risk later in maintenance (95% CI RR, 36.7-260.5 times). Based on the results of a previous study,1 the rate of fatal accidental drug toxicity for addicts on the street was estimated to be 10.4 per thousand per year. Using this estimate, the risk of fatal accidental drug toxicity in the first two weeks of methadone maintenance in NSW in 1994 was 6.7 times the risk before admission (95% CI RR, 3.3-13.9 times). Lives saved by NSW maintenance programs The age-adjusted approximation of the expected mortality from all causes among heroin addicts was 15.5 deaths per thousand per year (95% CI, 11.0-21.9 deaths).1 Using this estimate, 132 deaths would have been expected to occur in 8506.5 person-years (95% CI, 93-187 deaths). As 64 people either died while receiving maintenance (38) or from the toxic effects of diverted methadone (26), NSW methadone programs are estimated to have saved 68 lives in 1994 (95% CI, 29-123 lives saved). To save one life approximately 125 patients needed to be given methadone maintenance for a year (95% CI, 69-293 patients). To adjust for possible bias, we assumed that up to three of the 11 cases classified as involving diverted methadone syrup may have actually involved methadone tablets. When added to the two cases involving diverted methadone syrup in which there were differences in the official and study classifications of cause of death (Box 2), the number of lives saved may increase by up to five. Consequently, the upper limit of the confidence interval increased to give an adjusted 95% CI of 29 to 128 lives saved. If all 10 cases of fatal iatrogenic methadone toxicity7 and 26 deaths to which diverted syrup contributed had been avoided, NSW maintenance programs would have saved 104 lives in 1994 (adjusted 95% CI, 65-164 lives saved), making them up to 53% more effective at saving lives (adjusted 95% CI, 37%-124%). Discussion We found that, in NSW in 1994, the risk of fatal accidental drug toxicity in the first two weeks of methadone maintenance was nearly seven times the risk before admission to treatment. A previous report suggested that this excess mortality was primarily the result of iatrogenic methadone toxicity.7 However, the risk of fatal accidental drug toxicity later in maintenance was approximately one-hundredth the risk in the first two weeks of treatment and less than one-tenth the risk before admission. As there was complete agreement between our classification and that of official pathologists, our estimate of the relative risk of fatal accidental toxicity in the first two weeks and later maintenance is unlikely to have been significantly affected by misclassification of causes of death. Further, in estimating that NSW methadone programs saved 68 lives in 1994, we allowed for the difference between our opinion and that of the official report on the role of diverted methadone in two cases when calculating the upper limit of the adjusted 95% confidence interval (29-128) for the number of lives saved by NSW maintenance programs. Another consideration in estimating the number of lives saved is that mortality among patients discharged from maintenance is only an approximation of preadmission risk. If the real risk on the streets was higher than our estimate, NSW methadone programs would have saved more lives and admission to maintenance would not have caused such a dramatic increase in the risk of fatal accidental drug toxicity. Conversely, if the real risk was lower, the reverse applies. Our estimates of the number of lives saved and the increase in the risk of fatal accidental drug toxicity associated with admission to maintenance are approximations only. However, as our estimated 71% reduction in mortality is very similar to that observed in the US during the early 1970s, in Sweden during the 1980s, in Germany in the 1990s and Australia during the 1970s and 1980s,1 they are probably reasonably accurate. Previous Australian studies have also identified mortality associated with methadone programs. Eighteen people died from methadone toxicity in Western Australia in the years 1975 to 1980. However, there were virtually no such deaths after WA maintenance patients were required to take their methadone under supervision.23 In South Australia, nine maintenance patients died from drug toxicity in the years 1984 to 1994, while 12 other people died from the toxic effects of diverted methadone syrup.22 The number of deaths per 1000 SA maintenance patients was approximately 75% of that observed in our study. Our finding that diverted methadone syrup contributed to 26 deaths in NSW in 1994 is supported by the results of a previous investigation which suggested that diverted methadone syrup was involved in up to 100 deaths between July 1990 and December 1995.6 The WA experience23 suggests the number of deaths from diverted methadone syrup is related to the number of takeaway doses dispensed to maintenance patients for consumption on subsequent days. In 1994, two-thirds of private sector patients received four takeaway doses a week, with some programs giving five or six a week to newly admitted patients.24 Although the NSW Health Department argued against such practices, there was no policy enforcement.24 To minimise the diversion of methadone syrup from maintenance programs, the NSW Health Department should monitor and ensure compliance with its current policy which strictly limits the number of takeaway doses available to recent admissions while giving stable, long-term patients access to generous takeaway privileges. A serious problem with iatrogenic methadone toxicity was identified in Victoria, where 10 deaths occurred among newly admitted methadone patients in the last six months of 1989.2 As Victorian methadone programs treated fewer than 1200 maintenance patients in this period, the rate of iatrogenic methadone toxicity was many times that observed in our study. It is noteworthy that, during 1989, the number of Victorian maintenance patients and programs increased rapidly and a number of inexperienced and poorly trained prescribers entered the field.2,3 Persons with minimal or no tolerance were prescribed initial, daily methadone doses of 50-70 mg, with fatal results.2 Two recent British studies, from Sheffield and Manchester, have similarly identified significant numbers of deaths from iatrogenic methadone toxicity early in maintenance treatment.4,5 These problems also arose after the relaxation of admission criteria and during a period of rapid increase in the numbers of maintenance patients and the involvement of new, inexperienced prescribers.4,5 While the official criteria for admission to methadone maintenance in NSW have not changed since 1988,25 they were not being implemented in 1994.26 Statements made by its Chairman in 1996 indicate that the NSW Medical Committee had not been applying the official admission criteria for some time.26 This is significant because, under the NSW Poisons Act, the Medical Committee advises the NSW Health Department on applications from doctors to prescribe methadone maintenance to addicts. There were also problems with prescriber training. Since 1993, the NSW Methadone Prescribers' Accreditation Program has used the Methadone prescribers' manual as its course material.27 Contrary to NSW Health Department policy,25 the "Manual" states heroin users need not have a history of physiological dependence on opioids to be eligible for maintenance treatment.26,27 We urge the NSW Health Department to revise its Methadone prescribers' manual,27 review prescriber training and to ensure compliance with its current admission criteria for maintenance treatment.25,26 In 1994, Victorian and Queensland methadone prescribers were required to examine new patients during the first days of maintenance for signs of toxicity.28,29 Unfortunately, the NSW Health Department did not, and still does not, have a similar policy. Indeed, many private practitioners in NSW are only available to see maintenance patients one day a week (see Box 1), and the day-to-day supervision of patients attending public clinics is left to nurses working in busy dispensaries. The first two weeks of methadone maintenance will always be the "danger period" owing to the difficulty in determining a safe and effective starting dose. There is wide variation in opioid-na•ve individuals' response to and ability to metabolise and excrete methadone,30 and applicants' self-reports of recent drug use are an unreliable measure of tolerance.7 Given this uncertainty and variability, it is not possible to define safe, effective starting doses of methadone. We recommend prescribers be made aware of the risks, signs and symptoms of methadone toxicity and be required to examine newly admitted patients every day for the first one to two weeks of maintenance. People seeking methadone maintenance should be required to give written consent after being warned about the dangers of misleading their doctor and of the use of other drugs, particularly benzodiazepines.7,31 We believe that the forthcoming NSW methadone maintenance treatment clinical practice guidelines will address these issues. We strongly recommend the establishment of independent, expert committees to investigate methadone-related deaths in States and Territories with maintenance programs. These committees should be modelled on those used to monitor anaesthesia-related deaths. Acknowledgements For their advice and generous assistance, we thank the NSW State Coroner at Glebe, the Westmead Court, the Division of Analytical Laboratories, and the Pharmaceutical Services Branch of the NSW Health Department, and the Drugs of Dependence Unit, Queensland Health Department. We also thank Professor Geoffrey Berry of the Department of Public Health and Community Medicine, University of Sydney, who assisted with the statistics and presentation of results. References Caplehorn JRM, Dalton MSYN, Haldar F, et al. Methadone maintenance and addicts' risk of fatal heroin overdose. Substance Use Misuse 1996; 31: 177-196. Drummer OH, Opeskin K, Syrjanen SM, Cordner M. Methadone toxicity causing death in ten subjects starting on a methadone maintenance program. Am J Forensic Med Pathology 1992; 13: 346-350. McPherson CJ. Coronial inquiry into methadone related deaths. Melbourne: State Coroner Victoria, 1996. Clark JC, Milroy CM, Forrest ARW. Deaths from methadone use. J Clin Forensic Med 1995; 2: 143-144. Cairns A, Roberts ISD, Benbow EW. Characteristics of fatal methadone overdose in Manchester, 1985-94. BMJ 1996; 313: 264-265. Sunjic S, Zador D. Methadone-related deaths in New South Wales, Australia, 1990-1995. Euro-Methwork Newsletter 1998; issue 13: 11. Caplehorn JRM. Deaths in the first two weeks of maintenance treatment in NSW in 1994: identifying cases of iatrogenic methadone toxicity. Drug Alcohol Rev 1998; 17: 9-17. Darke S, Ross J, Hall W. The injection of methadone syrup in Sydney, Australia. Sydney: National Drug and Alcohol Research Centre, 1995. (Technical Report No. 23). Helpern M, Rho Y-M. Deaths from narcotism in New York City: incidence, circumstances and postmortem findings. NY State J Med 1966; 66: 2391-2408. Siegel H, Helpern M, Ehrenreich T. The diagnosis of death from intravenous narcotism. J Forensic Sci 1966; 11: 1-16. Segal RJ, Catherman RL. Methadone -- a cause of death. J Forensic Sci 1974; 19: 64-74. Worm K, Steentoft A, Kringsholm B. Methadone and drug addicts. Int J Leg Med 1993; 106: 119-123. Transcript of Proceedings, Case 011 0487/95. New South Wales Coroner's Court Westmead. Ruigrok M, Caplehorn J. A case of near-fatal methadone toxicity following repeated injections of methadone [letter]. Drug Alcohol Rev 1997; 16: 433. Pounder DJ, Jones GR. Post-mortem drug redistribution -- a toxicological nightmare. Forensic Sci Int 1990; 45: 253-263. Levine B, Wu SC, Dixon A, Smialek JE. Site dependence of postmortem blood methadone concentrations. Am J Forensic Med Path 1995; 16: 97-100. Drug and Alcohol Directorate. NSW methadone program 1993/94: annual statistical report. Sydney: NSW Health Department, 1995. Bailar JC, Ederer F. Significance factors for the ratio of a Poisson variable to its expectation. Biometrics 1964; 20: 639-643. Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford: Blackwell Scientific, 1994: 131. Caplehorn JRM, McNeil DR, Kleinbaum DG. Clinic policy and retention in methadone maintenance. Int J Addict 1993; 28: 73-89. Caplehorn JRM, Irwig L, Saunders JB. Physicians' attitudes and retention of patients in their methadone maintenance programs. Substance Use Misuse 1996; 31: 663-677. Williamson PA, Foreman KJ, White JM, Anderson G. Methadone-related overdose deaths in South Australia, 1984-1994. Med J Aust 1997; 166: 302-305. Swensen G. Opioid drug deaths in Western Australia: 1974-1984. Aust Drug Alcohol Rev 1988; 7: 181-185. Transcript of Proceedings, Case 010 1924/94. New South Wales Coroner's Court Glebe, Thursday 15th June 1995: 31-37. Directorate of the Drug Offensive. Policies and Procedures for the Methadone Treatment of Opioid Dependence in NSW. Sydney: New South Wales Department of Health, undated: 15, pars 4.3.1 and 4.3.2. Caplehorn JRM. Official and de facto admission criteria for methadone maintenance in New South Wales, Australia [letter]. Drug Alcohol Rev 1998; 17: 121-122. Bell J, O'Connor D. Methadone prescribers' manual. Sydney: NSW Health Department, 1993: 11, 23. Public Health Branch. Victorian methadone program guidelines for providers. Melbourne: Department of Health and Community Services, 1995: 7, par 3.5.2. (No. 93/0388). Bolton M, Reynolds A, Biggs L. Queensland methadone program: policy and procedures and treatment manual. Brisbane: Queensland Health, 1995: 35, par 4.14. Wolff K, Sanderson M, Hay AWM, Ralstrick D. Methadone concentrations in plasma and their relationship to drug dosage. Clin Chem 1991; 37: 205-209. Caplehorn JRM. More on iatrogenic methadone toxicity [letter]. Drug Alcohol Rev 1998; 17: 467-468. (Received 22 Sep 1997, accepted 18 Oct 1998) Authors' details Department of Public Health and Community Medicine, University of Sydney, NSW. John R M Caplehorn, MB BS, MPH, PhD Student. Victorian Institute of Forensic Medicine, Melbourne, VIC. Olaf H Drummer, PhD, Head of Scientific Services, and Associate Professor, Department of Forensic Medicine, Monash University. Reprints: Dr J R M Caplehorn, Department of Public Health and Community Medicine, Building A27, University of Sydney, Sydney, NSW 2006. Email: johnc@dph1.health.usyd.edu.au 1: A case of fatal iatrogenic methadone toxicity This 1995 NSW case highlights the danger of daily doses of 30-40mg methadone in non-tolerant individuals and presents a classic history of fatal iatrogenic toxicity. The deceased had clear, early signs of methadone toxicity: somnolence; unsteady gait; vomiting; and a general feeling of being unwell. The terminal events were also typical: prolonged coma following sleep; very slow, deep, irregular, noisy breathing; brown pulmonary oedema fluid coming from the mouth or nose. Six weeks before his death, the 19-year-old man was admitted to hospital with hypothermia, pneumonia, right brachial plexus neurapraxia, rhabdomyolysis and acute renal failure after a heroin overdose. He reported using amphetamines for six months and heroin for two weeks. Liver function test and echocardiogram findings were normal, and at discharge three days later his serum creatinine level had fallen from 0.18mmol/L to 0.10mmol/L (upper normal limit, 0.12mmol/L). He was referred to a short-stay, residential program and told the admitting officer he had had problems with alcohol for five years, cannabis for seven years and amphetamines for one year, but had only used heroin six times. While he was considered suitable for admission to a drug-free rehabilitation program, this was delayed pending full recovery of his arm. However, he was advised not to wait to enter this program as he was facing trial for a criminal offence. Soon after, the deceased apparently told a general practitioner and a methadone prescriber he had been using heroin daily for a year. He was prescribed 30mg methadone, with the dose to be increased by 5mg every day for six days and then reviewed. The deceased vomited several times after receiving his second dose (35mg). The next morning, he was difficult to rouse, had trouble walking and urinating and kept falling asleep. His father was unable to contact the methadone prescriber, who was on holiday, and the nurses at the private methadone clinic did not seem to have recognised the seriousness of the situation. By late afternoon he felt much better and travelled by public transport to receive his last dose of methadone (35mg) at 1830. That evening he seemed well, was in a very good mood and ate a large dinner. However, he was still having difficulty urinating. He went to bed at 2245. Around 0645 the next morning his father was unable to wake him, he was breathing deeply, noisily and irregularly and had brown fluid coming from his mouth. After about fifteen minutes he stopped breathing and died. At autopsy, the body weighed 72kg. No "track" or recent injection marks could be identified. The lungs weighed 960g (right) and 860g (left) and were described as "very oedematous and congested". The heart and liver were macroscopically and microscopically normal. The postmortem blood methadone concentration was 0.32mg/L. No other drugs were detected in blood, bile or urine samples. The investigating pathologist determined the cause of death was methadone toxicity. The Deputy NSW State Coroner determined the fatal toxicity was caused by the administration of three daily doses of methadone (30mg, 35mg and 35mg).13 At the inquest, the deceased's methadone prescriber said he did not physically examine methadone patients and had not rejected an applicant for maintenance in the past two years. He routinely saw methadone patients only one day a week at a private methadone clinic.13 Back to text 2: Causes of death determined in this study and in official pathologists' reports for 38 patients in New South Wales methadone maintenance programs and 29 people whose deaths involved methadone diverted from maintenance programs Accidental drug toxicityMethadone*Other drug(s)Other causes of deathMethadone maintenance patientsDeath in first two weeksThis study11 1 1 Official report10 2 1 Death after two weeksThis study1 5 19 Official report1 5 19 Diverted methadoneThis study26 1 2 Official report24 2 3 TotalsThis study38 7 22 Official report 35 9 23 *Methadone either caused or contributed to the death. Back to text 3: Rates of fatal accidental drug toxicity and relative risks of fatal accidental drug toxicity and sudden death from all causes for patients in New South Wales methadone maintenance programs in 1994Rate Relative (Deaths/1000/yr) risk 95%CI Deaths from accidental drug toxicityIn first two weeks' maintenance70.4 36.3-122.8 after two weeks' maintenance0.72 0.26-1.57 First two weeks' maintenance v.out of treatment* 6.7 3.3-13.9 First two weeks' maintenance v.after two weeks' maintenance 97.8 36.7-260.5 Out of treatment* v.after two weeks' maintenance 12.2 4.8-30.6 Deaths from all causesOut of treatment* v.all maintenance 3.5 2.2-5.6 Out of treatment* v.after two weeks' maintenance 5.2 3.1-8.7 *Calculated from approximations derived from previously published data. Back to text

Olaf H Drummer · John R M Caplehorn

Research Australian general practitioners' views and use of colorectal cancer screening tests Michael J Sladden and Jeanette E Ward MJA 1999; 170: 110-113 For editorial comment, see Young Abstract - Introduction - Method - Results - Discussion - Acknowledgements - References - Author's details - - More articles on Gastroenterology Abstract Objectives: To determine general practitioners' (GPs) current beliefs, knowledge and self-reported practices of screening for colorectal cancer. Design and setting: Postal survey of national random sample of 1271 GPs in 1996. Outcome measures: GP views on effectiveness of faecal occult blood testing (FOBT) and flexible sigmoidoscopy in reducing premature death from colorectal cancer in "average-risk" patients (asymptomatic with no family history); views on frequency of tests and target group; use of these tests; and independent predictors of views and use. Results: Response rate was 67%. FOBT and flexible sigmoidoscopy were said to be effective as screening tests by 38% and 61% of GPs, respectively, but 30% and 25% were unsure. Independent predictors of belief in screening effectiveness were State of practice (for FOBT), male sex and awareness of Gut Foundation guidelines (for flexible sigmoidoscopy) and increasing age (for both). Most often chosen screening frequencies were every year for FOBT (29%), and five-yearly for flexible sigmoidoscopy (24%), although 19% and 26%, respectively, were unsure of the appropriate screening interval. Most often cited target group was people aged over 40 years with first-degree relatives with colorectal cancer: 63% of GPs would offer FOBT and 74%, flexible sigmoidoscopy. Fewer than 3% of GPs were likely to adopt an opportunistic approach to screening, yet 15% would be highly likely to recommend FOBT during a dedicated health check-up for a 58-year-old male patient, and 9% for a female patient. Conclusion: The absence to date of a coherent national policy on colorectal cancer screening is associated with wide variations in views and practice that are inconsistent with the available evidence. If GPs are to be involved in implementing population screening, national policy must be widely and effectively promulgated. Introduction The past decade has seen a saga of unfinished policy initiatives on screening for colorectal cancer in Australia. As early as 1990, screening by faecal occult blood test (FOBT) was being promoted for people with "average risk" of colorectal cancer, in the absence of convincing evidence or national policy.1 Since then, a range of bodies have produced guidelines with quite different recommendations, particularly for this "average risk" group.2-8 Examples are shown in Box 1. More recent among these was the 1997 report of the Australian Health Technology Advisory Committee (AHTAC) Working Party on Colorectal Cancer Screening. This working party was formed after a 1994 call for a national symposium to develop "recommendations about early detection of colorectal cancer" as part of Australia's national health goals and targets,9 and a 1995 deputation to the Commonwealth from the Australian Cancer Network. The Network recommended pilot projects of screening and, if these were successful, implementation of national screening. AHTAC found that FOBT was effective in reducing mortality from colorectal cancer in the average-risk population (defined as the "well population aged over 50") and recommended studies to determine the effectiveness of different screening strategies, particularly how best to involve general practitioners (GPs). Meanwhile, in 1998, another set of draft guidelines, not yet nationally endorsed, was circulated by the Clinical Oncological Society of Australia and the Australian Cancer Network.8 Despite the unresolved status of national policy, research has continued to involve Australian GPs overtly or covertly in colorectal cancer screening.10-14 Most recently, a survey of Perth GPs reported high uptake of this screening.15 As no national studies have been conducted, we designed a postal survey to assess GPs' perceptions of the effectiveness of screening for colorectal cancer and self-reported levels of screening. Method GP sample and survey administration A covering letter, questionnaire and reply-paid envelope were mailed in May 1996 to a national random sample of 1271 GPs, as described elsewhere.16 Standardised response-aiding strategies were used to follow up non-respondents.16 Questionnaire Questions on colorectal cancer screening were part of a larger questionnaire on current status of cancer screening in general practice. Respondents were asked to tick the response that matched their views on: Effectiveness of FOBT and flexible sigmoidoscopy in reducing premature deaths from colorectal cancer in "average risk" patients (ie, asymptomatic patients with no family history); Frequency of screening; Who should be screened; Likelihood of their initiating a discussion about FOBT or flexible sigmoidoscopy with a well 58-year-old male, and next female, patient during non-urgent consultations for ear syringing as well as during scheduled health check-ups; and Usefulness of three guidelines available at the time of the surveys (first three items in Box 1). The final section of the questionnaire included eight sociodemographic questions. Data analysis Descriptive, cross-tabulation and logistic regression analyses were performed with SPSS for Windows 7.17 Logistic regression analyses were performed to determine independent predictors of knowledge and practice outcomes. Fourteen potential predictors were considered: GP's sex; age; State; full- or part-time practice; practice type (solo or group); membership of a Division of General Practice; Royal Australian College of General Practitioners (RACGP) affiliation; membership of the Australian Medical Association; personal history of cancer; family history of cancer; awareness of RACGP guidelines; Australian Gastroenterology Institute/Australian Cancer Society guidelines; and Gut Foundation of Australia (GFA) guidelines; and practice location (metropolitan or other18). All independent variables were included in the full model. A further variable -- belief that FOBT/flexible sigmoidoscopy is an effective screening test -- was included in the model to predict behaviour. A backwards stepwise modelling strategy was used to identify significant independent variables, whereby all dependent variables were initially included and non-significant terms were progressively eliminated. Significance was assessed using the Wald chi-squared statistic. Results We received 855 usable questionnaires from 1271 eligible GPs (67% response rate). The response rate for women (75%) was significantly higher than that for men (63%) (chi-squared = 15.4, 1 df, P < 0.001), but did not vary with age, vocational registration status, RACGP affiliation, or practice size and location. GP views about screening GP views about the effectiveness of screening for colorectal cancer in preventing premature deaths are shown in Box 2. More GPs believed flexible sigmoidoscopy to be effective (61%) than believed FOBT to be effective (38%), but there was substantial uncertainty about both screening tests, with 25% and 30% of GPs unsure. Independent predictors of believing either test to be effective are summarised in Figure 1. For FOBT, these comprised increasing GP age and State of practice (specifically, South Australia). For flexible sigmoidoscopy, they comprised increasing GP age, male sex, and awareness of the Gut Foundation guidelines (which favour screening). GPs' views on appropriate screening frequencies and target groups are shown in Box 2. Most often chosen frequencies were every year for FOBT (29%), and five-yearly for flexible sigmoidoscopy (24%). However, 19% and 26%, respectively, were unsure of the appropriate screening interval. The most often cited target group was people aged over 40 years with first-degree relatives with colorectal cancer: 63% of GPs would offer FOBT and 74%, flexible sigmoidoscopy. Self-reported practice Likelihood of suggesting colorectal cancer screening is shown in Box 3. Fewer than 3% of GPs were "highly likely" to adopt an opportunistic approach to colorectal cancer screening for a 58-year-old patient (by either test) or to recommend flexible sigmoidoscopy screening during a dedicated health check-up. More would be "highly likely" to recommend FOBT during a dedicated health check -- 15% if the patient was a man and 9% if a woman. This sex difference was significant (McNemar's chi-squared = 43.7, 1 df, P < 0.0001). Independent predictors of being "highly likely" to include FOBT in a health check-up are shown in Figure 2. They comprised increasing GP age, State of practice, and belief in the effectiveness of the tests. Usefulness ratings of guidelines available at the time of the survey are shown in Box 4. While RACGP, AGI/ACS and GFA guidelines were rated as "very" or "somewhat" useful by 33%, 50% and 45% of respondents, respectively, they were unable to be recalled by 52%, 35% and 41%. Discussion Our study suggests considerable confusion about colorectal cancer screening at the "front line" of general practice. About a third of respondents believed that FOBT was an effective screening test in average-risk individuals, while two-thirds believed this of flexible sigmoidoscopy. These beliefs are inconsistent with the evidence available at the time of our study: level II (randomised controlled trial) evidence19 that FOBT reduced colorectal cancer mortality, but only level III (case-control study) evidence20 supporting flexible sigmoidoscopy. A wide range of responses were given about appropriate screening frequency. This is not surprising, as the three guidelines available at the time offered different and contradictory advice, and none met the criteria for being systematically evidence-based. Further confusion is likely if yet more guidelines8 are published that contradict the evidence-based AHTAC guidelines.7Far fewer GPs reported using the screening tests than reported they were effective; most respondents did not use the tests. This discordance was greater for sigmoidoscopy than for FOBT. Again, it is likely that the lack of clarity and the variation between guidelines, as well as availability of conflicting information and educational material, have caused uncertainty among GPs, with concomitant inconsistency in their behaviour. Perhaps FOBT is used more often than sigmoidoscopy because it is easier to arrange. Increasing GP age and physician belief in screening effectiveness independently predicted self-reported provision of screening. South Australian GPs were more likely to advocate screening, perhaps because of research studies and community-based initiatives in that State.10 GPs also favoured screening men rather than women for colorectal cancer. Sex bias not been reported previously in colorectal cancer screening, and may reflect the higher incidence of colorectal cancer among men.21 Alternatively, perhaps GPs place a lower priority on colorectal cancer screening in women, for whom cervical and breast cancer screening are widely performed, but a higher priority in men, for whom there is, as yet, no "male cancer" screening of proven benefit. National levels of self-reported use of screening tests were less than previously reported,15,22,23 although the higher levels in South Australia were consistent with earlier data.10 GPs' wider endorsement of screening for "above-average-risk" individuals with first-degree relatives with colorectal cancer (63%-74%) was similar to the levels found in recent studies (80%-94% GP support),10,15 but considerably higher than in a similar 1982 study (23% GP support).22 Screening of relatives is apparently becoming more acceptable. Usefulness and recall of guidelines was low, similar to other recent findings.15 Our results suggest an urgent need for a national colorectal cancer screening policy. As GPs remain the most respected source of health information,24 it is vital they have access to timely and accurate information. The AHTAC report, released in 1998, was necessary but insufficient. If GPs are to be involved in implementation of colorectal cancer population screening (which, by definition, will target average-risk individuals), strategies are needed to inform them of the importance of screening and to facilitate appropriate changes in behaviour. When there has been a concerted effort to communicate agreed policy to Australian GPs (eg, about mammographic screening), uniformity of GP views has been achieved.25 However, when evidence is lacking (eg, for clinical breast examination),25 or State-based initiatives are patchy,26 it appears that GPs differ considerably in their views of screening effectiveness and self-reported behaviour. Our study concentrated on screening average-risk individuals, for which previous guidelines have been confusing, non-evidence based and at variance with each other. Most guidelines distinguish between screening for average-risk and above-average-risk individuals, generally recommending colonoscopic screening for above-average-risk groups (eg, those with familial adenomatous polyposis and hereditary non-polyposis colorectal cancer). Clearly, strategies for effective evidence-based screening of above-average-risk groups must also be promoted. Thus, our results reveal our failings to date as an organised healthcare system to provide GPs with timely, consistent and evidence-based health policy, as they and their patients deserve. Our study provides a baseline against which the effectiveness of future dissemination of colorectal cancer screening policy may be measured. Acknowledgements A Commonwealth General Practice Evaluation Program seeding grant funded this study. We thank the GPs who participated in our research with no financial incentive, Phoebe Holt for contributing to questionnaire design, and Tracey Bruce for diligent survey administration. The study was approved by the Ethics Committee of the Royal Prince Alfred Hospital, Sydney, NSW. References Woodward A, Weller D. Colorectal cancer: implications of mass screening for public health. Med J Aust 1990; 153: 81-88. Guidelines for screening for colorectal cancer. Sydney: Australian Gastroenterology Institute, 1991. Australian Cancer Society. National cancer prevention policy, 1993. Sydney: Australian Cancer Society, 1993. Bolin T, Collopy B, Cowen A, et al. Colorectal cancer: prevention, diagnosis and treatment. Sydney: The Gut Foundation and Colorectal Surgical Society of Australia, 1993. Goulston K, St John DJ, Bokey L, et al. Guidelines for early detection, screening and surveillance for colorectal cancer. 2nd ed. Sydney: Australian Gastroenterology Institute and Australian Cancer Society, 1994. Guidelines for preventive activities in general practice. 3rd ed. Sydney: Royal Australian College of General Practitioners, 1994. Australian Health Technology Advisory Committee (Standing Committee of the National Health and Medical Research Council). Colorectal cancer screening. Canberra: Commonwealth Department of Health and Family Services, 1997. Clinical Oncological Society of Australia and Australian Cancer Network. Guidelines for the prevention, early detection and management of colorectal cancer. Draft 3. June 1998. Better health outcomes for Australians. Canberra: National Health Goals and Targets Section, Department of Human Services and Health, 1994. Cockburn J, Thomas R, McLaughlin S, et al. Acceptance of screening for colorectal cancer by flexible sigmoidoscopy. J Med Screen 1995; 2: 79-83. Rae L. Community screening for colorectal cancer in north-eastern New South Wales, 1987-1996. Med J Aust 1998; 168: 382-385. King J, Fairbrother G, Thompson C, Morris D. Colorectal cancer screening: optimal compliance with postal faecal occult blood test. Aust N Z J Surg 1992; 62: 714-719. King J, Fairbrother G, Thompson C, Morris D. Influence of socioeconomic status, ethnicity and an educational brochure on compliance with a postal faecal occult blood test. Aust N Z J Public Health 1994; 18: 87-92. Olynyk J, Aquilia S, Fletcher D, Dickinson J. Flexible sigmoidoscopy screening for colorectal cancer in average-risk subjects: a community-based pilot project. Med J Aust 1996; 165: 74-76. Olynyk J, Aquilia S, Platell C, et al. Colorectal cancer screening by general practitioners: comparison with national guidelines. Med J Aust 1998; 168: 331-334. Ward J, Bruce T, Holt P, et al. Labour-saving strategies to increase response rates in general practice surveys. Aust N Z J Public Health 1998; 22: 394-396. Statistical Package for the Social Sciences. SPSS for Windows 7.5.1. Chicago (Ill): SPSS Inc, 1996. Commonwealth Department of Primary Industries and Energy and Commonwealth Department of Human Services and Health. Rural, Remote and Metropolitan Areas Classification: 1991 Census edition. AGPS, Canberra: 1994. Mandel J, Bond J, Church T, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. N Engl J Med 1993; 328: 1365-1371. Selby J, Friedman G, Quesenberry C, Weiss N. A case-control study of screening sigmoidoscopy and mortality from colorectal cancer. N Engl J Med 1992; 326: 653-657. Jelfs P, Coates M, Giles G, et al. 1996 Cancer in Australia 1989-1990 (with projections to 1995). Canberra: Australian Institute of Health and Welfare (Cancer Series No. 5). Macrae FA, Hill DJ, Dent O, et al. Colorectal cancer: knowledge and attitudes of doctors in Victoria. Aust N Z J Med 1982; 12: 278-283. Rolfe I, Pearson S. Screening recommendations in general practice: a survey of graduates from different medical schools. Med J Aust 1996; 165: 14-17. Cumming R, Barton G, Fahey P, et al. Medical practitioners and health promotion: results from a community survey in Sydney's western suburbs. Community Health Stud 1989; 13: 294-301. Young J, Ward J, Holt P. Breast cancer screening in Australian general practice: results of a national survey. Med J Aust 1998; 169: 364-368. Ward J, Donnelly N, Holt P. Impact in general practice of the policies of the organised approach to preventing cancer of the cervix. Aust N Z J Public Health 1998; 22: 336-341. (Received 25 Jun, accepted 28 Oct, 1998) Author's details Division of Community and Rural Health, University of Tasmania, Hobart, TAS. Michael J Sladden,FRACGP, MAppEpid, Honorary Senior Lecturer. Needs Assessment and Health Outcomes Unit, Central Sydney Area Health Service, Sydney, NSW. Jeanette E Ward, PhD, FAFPHM, Director. Reprints will not be available from the authors. Correspondence: Dr M J Sladden, Division of Community and Rural Health, PO Box 252-33, University of Tasmania, Hobart, TAS 7001. Email: M. Sladden@utas.edu.au Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> Figure 1 : Independent predictors of general practitioners (GPs) reporting that faecal occult blood testing or flexible sigmoidoscopy is effective in reducing premature deaths from colorectal cancer. Back to text Figure 2 : Independent predictors of general practioners (GPs) reporting they would be "highly likely" to discuss faecal occult blood testing during a deicated health check-up with a 58-year-old patient. (Adjusted odds ratios with 95% confidence intervals are available from the authors.)Back to text 1: Recommendations about colorectal cancer screening for average-risk* people Gut Foundation of Australia (1993)4 Screening by annual FOBT; consider 3-5-yearly flexible sigmoidoscopy Australian Gastroenterology Institute/Australian Cancer Society (1994) 5 Routine screening not recommended; can be performed at patient request Royal Australian College of General Practitioners (1994)6 Screening not recommended Australian Health Technology Advisory Committee (1997)7 Screening by FOBT; research required to determine method and frequency Clinical Oncological Society of Australia/Australian Cancer Network (draft 1998)8 Screening by annual FOBT; consider 5-yearly flexible sigmoidoscopy FOBT=faecal occult blood test. *Asymptomatic people aged over 50 years with no family history Back to text Back to text Back to text Back to text

Michael J Sladden · Jeanette E Ward

Research General practitioners' use of evidence databases Jane M Young and Jeanette E Ward MJA 1999; 170: 56-58 For editorial comment, see Van Der Weyden Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on General practice and primary care Abstract Objective: To determine the awareness and use of the Cochrane Library and access to the Internet by general practitioners in New South Wales. Design: Cross-sectional postal survey in September 1997. Participants: 311 of 428 (73% response rate) randomly selected general practitioners in New South Wales. Main outcome measures: Proportion of respondents with access to the Internet at home or at work; proportion of respondents aware of, with access to, and ever using the Cochrane Library; independent predictors of awareness of the Cochrane Library. Results: 134 respondents (43%) had access to the Internet either at home or at work; 42 (14%) were "on line" at their workplace. Seventy (22%) were aware of the Cochrane Library, although only 20 (6%) had access to it and 13 (4%) had ever used it. Those in group practice and members of Divisions were independently more likely to be aware of the Cochrane Library. Conclusions: As patient outcomes will improve with systematic implementation of evidence-based treatments, these low rates of access to useful evidence databases raise issues regarding the best ways to support general practitioners with information technology. Introduction There has been increasing interest in the use by clinicians of evidence databases and other resources, such as systematic reviews, meta-analyses and evidence-based guidelines, as aids for clinical decision-making. The first report of Australian clinicians' use of evidence databases was recently published in the Journal.1 In that study, 72% of neonatologists and 44% of obstetricians reported using evidence databases, with higher rates of use among those familiar with computers. Although lack of awareness of evidence databases does not preclude evidence-based practice,2 the inability of practitioners to access research findings readily at the time of decision-making is a major impediment to best practice.3Because of the breadth of their work, general practitioners have diverse needs for evidence to inform their practice.4 Accessible evidence databases potentially represent an essential resource to meet these needs. The Cochrane Library, which includes the Cochrane Database of Systematic Reviews and the Database of Abstracts of Reviews of Effectiveness (Box 1), is recognised as one of the best resources for evidence. General practitioners can use it on CD-ROM or through the Internet. Research from other countries suggests that general practitioners are reluctant to embrace information technology to support evidence-based clinical decision-making. Two recent surveys both reported that, at most, 40% of British general practitioners were aware of the Cochrane Database of Systematic Reviews.5,6 Furthermore, despite positive attitudes towards evidence-based medicine, general practitioners reported low levels of use of either printed or electronic summaries of evidence, even among those who were aware of these resources.6 In 1995, it was reported that a quarter of a national random sample of Australian general practitioners had access to a computer with a modem but less than 10% had access to the Internet.7 No reports have been published more recently to assess the uptake of information technology by general practitioners. The aim of our study was to determine New South Wales general practitioners' current awareness of, access to, and use of the Cochrane Library, and their access to the Internet both at home and at work. Methods Survey content and administration We added the following questions to a statewide random postal survey of general practitioners in NSW conducted in September 1997:Are you aware of the Cochrane Library? Do you have access to the Cochrane Library? Have you ever used the Cochrane Library? Do you have access to the Internet at your practice? Do you have access to the Internet at home? Respondents could indicate "Yes", "No" or "Unsure" to each of these questions. Respondents also completed eight standard sociodemographic questions. A copy of the questionnaire is available from the authors on request. Four hundred and twenty-eight eligible general practitioners in NSW, randomly selected from a commercial list, were contacted by telephone in advance of our survey. Two mail reminders and a telephone prompt were used to maximise the response rate. Data analysis Proportions and 95% confidence intervals were calculated for responses to questions about the Cochrane Library and Internet. The univariate association between awareness of and access to the Cochrane Library and personal and professional characteristics of respondents were assessed using c2 tests, or Fisher's exact test where expected cell frequencies were less than five. Logistic regression using a backwards stepwise modelling strategy was then carried out to identify factors that significantly and independently predicted positive responses to these questions. All analyses were conducted using SAS for Windows.8 Ethics approval This study was approved by the Central Sydney Area Health Service Ethics Review Committee and the Human Ethics Committee of Sydney University. Results We received completed questionnaires from 311 general practitioners (73% response rate). Although the response rate for women (80%) was significantly higher than for men (70%) (chi-squared = 4.5; df = 1; P = 0.03), respondent characteristics were similar to those of general practitioners in NSW.9 Respondents ranged in age from 24 to 72 years (mean, 45 years), 96 (31%) were women, 236 (76%) worked full time, and 202 (65%) were in group practice. Professional characteristics of respondents included RACGP affiliation (141; 45%), AMA membership (109; 35%), and membership of a Division of General Practice (242; 78%). A third of respondents (107; 34%) had trained with the Family Medicine Program. Responses to the questions about the Cochrane Library and Internet are shown in Box 2. Less than a quarter of respondents were aware of the Cochrane Library and only 13 (4%) had used it. Nearly one in five respondents were unsure if they had access to this resource. One hundred and thirty-four respondents (43%) had access to the Internet either at home or work, significantly higher than the 9% reported previously (chi-squared = 86.6; df = 1; P < 0.001).7 Awareness of the Cochrane Library was unrelated to age (t = -1.1; df = 298; P = 0.2) or sex (chi-squared = 0.6; df = 1; P = 0.4). The only significant associations were with general practice Divisional membership and working in group practice. These variables remained independently predictive of awareness of the Cochrane Library following logistic regression analysis (Box 3). The number of respondents who had actually used the Cochrane Library were too few for further analysis. Discussion Overall, 22% of respondents were aware of the Cochrane Library. As awareness was greater among those in group practice and members of their local Division, peer contact appears to be an important mechanism to promote evidence databases. Nonetheless, the level of awareness in our study was considerably lower than that reported in the United Kingdom,5,6 where the Cochrane Database of Systematic Reviews has been available since 1992.10 However, our finding that only 4% of respondents had ever used the Cochrane Library is comparable. Our finding of a marked uptake since 1995 of Internet access by general practitioners is reassuring. Nearly half had Internet access either at home or at their practice. However, only 14% were "on-line" at their practices, where clinical decisions are likely to be made. Evaluation of strategies to support the uptake of information technology for desktop Internet access will be an immediate challenge in ensuring evidence databases are used in general practice. Access to evidence databases is crucial to support the scientific paradigm now advocated in healthcare.11 Having accessed an evidence database, general practitioners can focus on treatments for which there is Level I (meta-analysis of randomised controlled trials) or Level II (randomised controlled trials) evidence of effectiveness. By ensuring treatments with such compelling evidence are used, GPs can confidently anticipate that their patient outcomes will positively and predictably improve. Less confidence can be placed on interventions for which only Level IV (descriptive case reports) evidence exists. Measurement and improvement of care based on Level I or II evidence of effectiveness should also be emphasised in quality assurance activities.12 Three years ago it was argued that "the health care system needs an infrastructure for the dissemination of evidence-based medicine into clinical practice".13 Subsequently, some people have suggested that general practitioners need mediated search services.14 Other problems to overcome include training general practitioners to appraise evidence4 and to incorporate research findings into their daily consultations with patients.15 Our findings suggest we have a long road ahead. Since June 1998, members of the Royal Australian College of General Practitioners (RACGP) have had access to the Cochrane Library through the RACGP Virtual Resource Centre. Evaluation of the impact of electronic evidence resources, including evidence databases or Web-based guidelines, on decision-making in general practice is the next step. Initiatives to encourage evidence-based decision-making in general practice are likely to generate dissatisfaction with the limitations of currently available evidence.16 We are optimistic this will accelerate the quality and quantity of research conducted in general practice. Syntheses of current knowledge prevent the reinvention of wheels or repetition of past mistakes, minimising expenditure on populist strategies without strong evidence of effectiveness. Gaps in current knowledge of effective interventions in clinical practice are tellingly revealed in evidence databases, inviting a responsive academic research agenda. General practitioners adopting an evidence-based approach may be more inclined to participate in research which is relevant, rigorous and responsive to gaps in evidence sorely felt in clinical decision-making. Acknowledgements The participation of general practitioners in our research, without financial incentive, is acknowledged gratefully. We thank Nancy Harding for organisational support and Leonie Cambage for data entry. J M Y is supported by an NHMRC research scholarship. References Jordens CFC, Hawe P, Irwig LM, et al. Use of systematic reviews of randomised trials by Australian neonatologists and obstetricians. Med J Aust 1998; 168: 267-270. Phillips PA. Disseminating and applying best evidence. Med J Aust 1998; 168: 260-261. Haines A, Jones R. Implementing findings of research. BMJ 1994; 308: 1488-1492. Ridsdale L. Evidence-based learning for general practice. Br J Gen Pract 1996; 46: 503-504. Prescott K, Lloyd M, Douglas HD, et al. Promoting clinically effective practice: general practitioners' awareness of sources of research evidence. Fam Pract 1997; 14: 320-323. McColl A, Smith H, White P, Field J. General practitioners' perceptions of the route to evidence based medicine: a questionnaire survey. BMJ 1998; 316: 361-365. Gupta L, Ward J, Hayward RSA. Future directions for clinical practice guidelines: needs, lead agencies and potential dissemination strategies identified by Australian general practitioners. Aust N Z J Public Health 1997; 21: 495-499. SAS for Windows [computer program]. Version 6.11. Cary, North Carolina: SAS Institute, 1995. Commonwealth Department of Health and Family Services. General practice in Australia: 1996. Canberra: Commonwealth of Australia, 1996. Silagy C. Randomised controlled trials: the challenge of Archie Cochrane. Med J Aust 1993; 158: 656-657. Risdale L. How do you know? The process of scientific reasoning. In: Evidence-based general practice: a critical reader. London: WB Saunders, 1995; 160-169. Ward J, Del Mar C, Colmer P, O'Connell D. Quality and outcomes in general practice. In: General practice in Australia: 1996. Canberra: Commonwealth of Australia, 1996; 169-199. Ahmed T, Silagy C. The move towards evidence-based medicine. Med J Aust 1995; 163: 60-61. Fowler C. Evidence-based learning in general practice. Br J Gen Pract 1996; 46: 754-755. Jacobson LD, Edwards AGK, Granier SK, Butler CC. Evidence-based medicine and general practice. Br J Gen Pract 1997; 47: 449-452. Campion-Smith C. Evidence-based general practice. Br J Gen Pract 1997; 47: 462. (Received 18 May, accepted 15 Sep, 1998) Authors' details Needs Assessment and Health Outcomes Unit, Central Sydney Area Health Service, Newtown, NSW. Jane M Young, MB BS, MPH, Postgraduate Fellow; Jeanette E Ward, PhD, FAFPHM, Director. Reprints: Associate Professor J E Ward, Needs Assessment and Health Outcomes Unit, Central Sydney Area Health Service, Locked Bag 8, Newtown, NSW 2042. Email: jwardATnah.rpa.cs.nsw.gov.au Make a comment Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Jane M Young · Jeanette E Ward

Abstract Objective: To ascertain the effectiveness of clinical pathways for improving patient outcomes and decreasing lengths of stay after hip and knee arthroplasty. Design and setting: Twelve-month randomised prospective trial comparing patients treated through a clinical pathway with those treated by an established standard of care at a single tertiary referral university hospital. Participants: 163 patients (56 men and 107 women; mean age, 66 years) undergoing primary hip or knee arthroplasty, and randomly allocated to the clinical pathway (92 patients) and the control group (71 patients). Main outcome measures: Time to sitting out of bed and walking; rates of complications and readmissions; match to planned discharge destination; and length of hospital stay. Results: Clinical pathway patients had a shorter mean length of stay (P = 0.011), earlier ambulation(P = 0.001), a lower readmission rate (P = 0.06) and closer matching of discharge destination. There were beneficial effects of attending patient seminars and preadmission clinics for both pathway and control patients. Conclusion: Clinical pathway is an effective method of improving patient outcomes and decreasing length of stay following hip and knee arthroplasty. Introduction The past two decades have seen an 85% rise in Australian health costs to 36.6 billion dollars, with the largest proportion of this expended in acute hospital care.1 Newer health policies now incorporate measures to rationalise and improve the efficiency of many services. Such policies, however, are economically driven and frequently fail to consider the optimum level of service required by the community.2 Treatment protocols, variously known as clinical pathways, critical pathways and care paths, that aim to streamline and standardise management through a systematic approach so that high quality care may be provided in a timely and cost effective manner3,4 have been developed. Clinical pathways describe the course of hospitalisation for patients with a specified illness and encompass a predetermined plan of treatment. The use of clinical pathways is now well established and their successes are widely reported.5-7 Joint arthroplasty is a common and costly procedure associated with high resource use that is frequently performed in the elderly who may have many coexisting morbidities. These characteristics suggest that joint arthroplasty may be a suitable procedure to incorporate into a clinical pathway.8 As part of a "best practice" initiative in line with quality assurance activities at St Vincent's Hospital, the hospital's Orthopaedic Service has developed clinical pathways for hip and knee joint arthroplasty for treating osteoarthritis which aim to maximise the use of all available resources and minimise negative patient outcomes, thereby improving patient care. To this end, we report the effects of introducing clinical pathways at our hospital on quality indicators such as mobilisation, complication rates, discharge planning and readmission rates while also exploring the impact on length of stay. Methods We used a prospective randomised control group design to compare the outcomes of patients who underwent hip or knee joint arthroplasty at St Vincent's Hospital, Melbourne (a tertiary referral hospital affiliated with the University of Melbourne), between 1 January 1996 and 30 December 1997. All such patients were randomly allocated to either the control or clinical pathway group by a clerical assistant who was blinded to their demographic and clinical profiles. Diagnostic category and comorbidities had no bearing on the allocation of patients to either the pathway or control groups, but patients were excluded from the study after randomisation if they were having revision arthroplasty, simultaneous bilateral joint arthroplasty, arthroplasty for acute trauma or complex tumour surgery. The management of patients undergoing joint arthroplasty at St Vincent's Hospital, Melbourne, is outlined in Box 1. Outcome measures Length of stay (calculated from the time of the patient's admission to the time of discharge and expressed in days); Time to sitting out of bed and ambulation (time between surgery and the patient's first day of sitting out of bed or walking with assistance); Complications (wound infections, including all wound erythema lasting more than 24 hours, chest infections, deep vein thrombosis [DVT] as diagnosed by clinical features and confirmed by ultrasonography, joint dislocation, decubitus pressure areas, failure to cope at home and a decreased range of motion after discharge); Readmission (for complications during a follow-up period of three months from discharge); and Discharge matching (between the presumptive discharge destination given at the preadmission clinic and the patient's postdischarge destination). Clinical pathway and control patients Patients randomly allocated to the clinical pathway received proactive treatment whereby specific goals were set each day for the patient and treating team. Their hospital records included a special written protocol which listed milestones to be achieved, identified tests that should be ordered, set daily tasks for patients and members of the treating team, and provided space for documenting any variation in treatment or patient response. Each intervention was signed by the treating health professional and the discharge plan was re-evaluated daily to ensure it remained realistic and appropriate to the patient's needs. The clinical pathway formalised in writing the participation of the various members of the treating team. Patients not allocated to the pathway received "reactive" treatment whereby the treating team responded to the will and condition of the patient in providing postoperative care. Statistical analysis Results were analysed with SigmaStat V2 software.9 Data were compared using t tests for independent groups and multiple linear regression where appropriate. We used the z test for comparisons of proportions between groups. As the data for length of stay (LOS), time to sitting out of bed and time to ambulation were not normally distributed, these data were transformed using a logarithmic transformation before analysis with t tests. We calculated the sample size for this study after reviewing all hip and knee arthroplasty patient data for 1995, which showed a mean LOS of 13 days (range, 5.8-43.3; SD, 5.3). We believed that a 20% reduction in LOS (2.6 days) would represent a clinically significant outcome. Therefore, we calculated that to detect a reduction of 2.6 days in LOS at a significance level of 0.05 with a power of 0.8 would require two groups with a minimum of 65 subjects in each group. Results During the study period 175 patients underwent hip or knee joint arthroplasty and were randomly allocated to the pathway (94 patients) and control (81 patients) groups. Twelve patients were then excluded by the crtiteria listed in the methods, leaving 163 patients -- 92 in the clinical pathway group and 71 in the control group. The sample comprised 56 men and 107 women, with a mean age of 66 years (range, 67-93 years). All patients were followed for a minimum of three months and none were lost to follow-up. Our findings are summarised in Box 2. There was no significant difference between control and pathway patients in terms of age or weight. Although the clinical pathway group included more patients with premorbid conditions than the control group, this difference was not statistically significant (95% CI, - 0.03 to 0.21). Length of stay (LOS) was significantly shorter for the pathway group than for the control group (t = 2.585; P = 0.011). When LOS was analysed for the subgroups of patients in each group with premorbid conditions, this was still significantly shorter for the pathway group than the control group (t = 3.152; P = 0.001) despite the larger number of patients with premorbid conditions in the pathway group. Patients in the clinical pathway group sat out of bed and walked earlier after surgery than control patients. Multiple linear regression for each group showed that time to ambulation was the only significant contributor to reduction in log LOS in the clinical pathway group (time to ambulation -- coeff = 19.6, standard error [SE] = 9.6, P = 0.04; time to sitting out of bed -- coeff = - 4.35, SE = 9.3, P = 0.64, R2 = 0.127). Neither time to ambulation nor time to sit out of bed was significantly associated with reduced log LOS in the control group (time to ambulation -- coeff = 21.05; SE = 26.28, P = 0.42; time to sit out of bed -- coeff = - 4.13, SE = 28.54, P = 0.88, R2 = 0.0251). Patients from both the clinical pathway and control groups who attended either the preadmission clinic (n = 122) or the patient information seminar (n = 61) had a shorter LOS (7.22 days and 6.84 days, respectively) than patients who attended neither (n = 36; LOS, 8.55 days). The 54 patients who attended both the clinic and seminar had the shortest LOS at 6.6 days, and t tests showed that the shorter LOS for these patients relative to those who attended neither the clinic nor seminar was significant (t = 2.66; P = 0.009). Post-hoc t tests showed that the shorter LOS for patients who had attended both preadmission clinics and information seminars relative to those who had attended neither was significant (t = 2.66; P = 0.009). Box 2 shows that a greater proportion of clinical pathway patients were discharged to their planned discharge destination than control patients (95% CI, - 0.05 to 0.23), and that there were fewer readmissions in clinical pathway patients (95% CI, 0.006-0.174), although neither result was statistically significant. However, there were significantly fewer complications in clinical pathway patients (95% CI, 0.036-0.27). Discussion We found that a clinical pathway for hip and knee joint arthroplasty had a beneficial impact on the duration of admission, with patients on the pathway having a 1.5-day shorter stay than control patients. The seven-day LOS for our pathway patients compared favourably with that of Gregor et al,10 who showed a reduction in LOS from 12 to nine days for pathway patients. Length of stay was significantly shorter for the pathway group than the control group despite the larger proportion of pathway patients with premorbid conditions. This result should be interpreted cautiously, as the small overall number of patients with premorbid conditions meant that the test had less than optimal power (0.45). However, we conclude that comorbidities per se should not exclude patients from clinical pathways. Patients with comorbid conditions may actually be better served because of the greater fastidiousness and vigilance imposed by the daily protocol. While our findings that there were fewer complications and readmissions in clinical pathway patients were not significant, we believe that given the appropriate number of subjects in future studies both of these areas may approach significance. We noted that reducing the length of stay did not increase the complication rate, a finding corroborated by others.11 In addition, the readmission rate for complications for pathway patients was one-third that of controls. This contrasts with some studies which have reported an inverse relationship between length of stay and readmission rates.11 We, like other authors,12 believe that it is a lower quality of care and not length of stay per se that increases the risk of unplanned readmission. Discharge planning is an important part of the clinical pathway which appears to be closely linked with the length of stay. Appropriate matching of predetermined discharge destinations is a correlate of shorter admissions. If we are able to improve on our destination matching rate of 70%, we may be able to further reduce our length of stay, thereby making more resources available for other patients. Education of patients and their relatives appeared to have a positive influence on the patients' recovery after joint arthroplasty, with earlier mobilisation and discharge from hospital. Attending information seminars and preadmission clinics assisted in reducing the length of stay by almost two days. Patients and their relatives who understand the disease and the necessary treatment may be in a better position to assist with care and rehabilitation. Attendances for our information seminar and preadmission clinic were 38% and 74%, respectively, and we are endeavouring to increase these. First introduced by the New England Medical Center, clinical pathways are now incorporated into the management philosophy of many hospitals worldwide.13,14 Pathways involve input from medical, nursing, paramedical and administrative staff, and reflect the expertise of all members of the healthcare team while highlighting the interdependent nature of these roles in achieving positive outcomes for patients.15 A valuable subsidiary purpose of pathways is in providing information from which the financial cost of care may also be derived.16 Accurate costing of treatment is fundamental to the operation of institutions where prospective payments are made in accordance with diagnosis-related groups (DRGs), standardised lengths of stay and fixed reimbursement for care. Clinical pathways thus provide an important tool for coordinating and managing clinical resources. However, the driving force behind clinical pathways must remain the need to improve the quality of care and patient outcomes, and not their utility as a tool to ensure that budgetary demands are met. We are encouraged by our findings, which indicate substantial improvements for patients on a clinical pathway. To our knowledge, no other study has investigated the effect of clinical pathways on joint arthroplasty using a contemporaneous control group. 1 Management of joint arthroplasty patients at St Vincent's Hospital, Melbourne Preadmission clinics Preoperative review for patients undergoing elective joint replacement involves a multidisciplinary approach and includes medical, nursing, physiotherapy and occupational therapy consultation and anaesthetic and social work screening. Preexisting conditions are identified and testing and treatment are undertaken to achieve an optimum level of preoperative health. A discharge destination is determined based on medical and projected rehabilitation needs. Appropriate referrals are initiated. Patient information seminars Groups of patients and their families are invited to attend an information seminar about the surgery. The surgeon explains the aetiology of the disease, principles of management, nature of potential risks and their prevention. The nursing staff discuss acute postoperative care, including pain relief, pressure and wound care, intravenous therapy, and prophylaxis for deep venous thrombosis. The physiotherapist discusses the regimen of postoperative exercises, cautions and mobilisation. The occupational therapist describes the availability and use of various personal aids which assist the patient in preventing complications such as falls, injury or dislocation. Patients are able to raise any questions related to their surgery. Patients and their families are encouraged to take an active role in the postoperative management, and are acquainted with their very important role in the postdischarge phase. All members of the team stress the philosophy that the primary intention is to return patients home in preference to a rehabilitation hospital after the surgery. Discharge Patients are discharged home or to a rehabilitation unit. For those discharged home, community nursing care is provided at regular intervals for the first three weeks after discharge. Community nurses pay special attention to the nature of the patient's wounds, their exercise regimen and general medical condition. Any concerns are immediately related to the medical staff for further attention. Patients are followed up on a regular basis in the outpatient department. 2 References MacIntyre CR, Brook CW, Chandraraj E, Plant AJ. Changes in bed resources and admission patterns in acute public hospitals in Victoria, 1987-95. Med J Aust 1997; 167: 186-189. Parry TG. Health expenditure in Australia -- the current dilemma. Med J Aust 1992; 156: 592-594. Wigfield A, Boon E. Critical care pathway development: the way forward. Br J Nursing 1996; 5: 732-735. Grudich G. The critical path system. AORN J 1991; 53: 705-714. Gouveia WA, Massaro FJ. Critical pathway experience at New England Medical Center. Am J Health-Syst Pharm 1995; 52: 1068-1070. Saltiel E. Critical pathway experience at Cedars-Sinai Medical Center. Am J Health-Syst Pharm 1995; 52: 1063-1068. Stevenson LL. Critical pathway experience at Saratosa Memorial Hospital. Am J Health-Syst Pharm 1995; 52: 1071-1073. Leininger SM. Tools for building a successful orthopaedic pathway. Orthop Nurs 1996; 15: 11-19. SigmaStat [computer program]. Version 2. San Rafael, CA: Jandel Scientific Software, 1995. Gregor C, Pope S, Werry D, Dodek P. Reduced length of stay and improved appropriateness of care with a clinical path for total knee or hip arthroplasty. Joint Commiss J Qual Improv 1996; 22: 617-628. Rushworth RL, Rob MI. Readmissions to hospital: the contribution of morbidity data to the evaluation of asthma management. Aust J Public Health 1995; 19: 363-367. Ashton CM, Kuykendall DH, Johnson ML, et al. The association between the quality of inpatient care and early readmission. Ann Intern Med 1995; 122: 415-421. Zander K. Managed care within acute care settings: design and implementation via nursing case management. Health Care Supervisor 1988; 6: 27-43. Bower KA. Managed care: controlling costs, guaranteeing outcomes. Definition 1988; 3: 14. Heacock D, Brobst RA. A multidisciplinary approach to critical path development: a valuable CQI tool. J Nurs Care Qual 1994; 8: 38-41. Weilitz PB, Potter PA. A managed care system. Financial and clinical evaluation. J Nurs Admin 1993; 23: 51-7. (Received 27 Jan, accepted 20 Aug, 1998) Authors' details Department of Orthopaedics, St Vincent's Hospital, Melbourne, VIC. Michelle M Dowsey, BN, GradCertOrth, Clinical Nurse Specialist; Meredith L Kilgour, BN, GradDipAdvClinPrac, Nurse Unit Manager; Nick M Santamaria, BAppSc, PhD, Director of Nursing Research; Peter F M Choong, MD, FRACS, Professor, and Director of Orthopaedics. Reprints: Professor P F M Choong, Department of Orthopaedics, St Vincent's Hospital, 41 Victoria Parade, Fitzroy, VIC 3065. Email: PeterChoongATc031.aone.net.au

Michelle M Dowsey · Meredith L Kilgour · Nick M Santamaria

MJA 1999; 170: 11-14 For editorial comment, see Conyers Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Author's details Abstract Objective: To assess the effect of a total quality management (TQM) approach on the appropriateness of clinical laboratory testing of patients with acute myocardial infarction. Design: Retrospective, control-group, quasi-experimental study. Setting: Two metropolitan teaching hospitals: one involved in the intervention (TQM experimental group); the other had no intervention (control group). Intervention: A multidisciplinary team involved in the development of improvement strategies, including laboratory testing guidelines, education programs, data collection, monitoring and feedback. Outcome measures: Total number of requested clinically indicated and non-clinically indicated tests. Results: The proportion of clinically indicated tests that were requested increased from 77.5% to 88.2% (P < 0.01) and the number of non-clinically indicated tests was reduced by 81.7% with the introduction of the TQM strategies (P < 0.01). Conclusions: The introduction of a TQM approach has improved the appropriateness of test ordering in acute myocardial infarction. The reduction in non-clinically indicated tests could produce substantial savings in hospital pathology costs, and the increase in clinically indicated tests provides better patient care. Introduction The determination of plasma creatine kinase (CK) and creatine kinase-MB fraction (CK-MB) is pivotal in the diagnosis of acute myocardial infarction (AMI).1 However, the inappropriate use of these and other laboratory tests has increased markedly during the past 10-15 years, contributing to escalating healthcare costs2 and undesirable clinical consequences because of false positive results in AMI, often leading to unnecessary further investigations.3A number of reports have indicated an excessive use of laboratory tests in patients admitted to hospital with AMI or other clinical disorders. One Australian study found that at one hospital up to 20% of all laboratory tests were unnecessary.4 Another found that there were three times as many clinical chemistry tests ordered as were appropriate for patients admitted with chest pain.5 Similarly, a recent study found inappropriate testing of calcium, phosphate and magnesium in the emergency department of a United States teaching hospital.6 The conventional approach to address inappropriate laboratory testing has been to target doctors through education,7 feedback8 and providing information on laboratory testing costs.9 Such measures have had limited success, and in most cases have had limited input from clinicians.10 I have examined the effect of a total quality management (TQM) approach (that was developed and introduced by clinicians and other healthcare professionals) on the appropriateness of clinical laboratory test use in the management of patients with AMI. Total quality management refers to the overall approach of managing the total aspects of an organisation's quality. In my study, this involves changing the way quality in test use is viewed, focusing on the customer and including everyone involved in its provision and use in the process of continuous improvement. The specific TQM model I used was the FOCUS-PDCA approach to quality improvement (Box 1).11 Methods Setting and patient groups The study was conducted from March 1993 to August 1995 at two teaching hospitals in Sydney. Patients at Bankstown-Lidcombe Hospital were designated the "experimental" group, and those at Nepean Hospital were the "control" group. Only the experimental group was subjected to the TQM intervention. At the time of the study, Bankstown-Lidcombe Hospital was a 454-bed hospital with 21 724 annual admissions from the Emergency Department and 2355 admissions to the Coronary Care Unit (CCU). Nepean Hospital was a 415-bed hospital with 20 485 annual admissions from the emergency department and 1891 admissions to the CCU. Within the study groups, two subgroups of AMI patients were identified: "suspected AMI" and "confirmed AMI". The suspected AMI group included all patients who were admitted to the CCU via the emergency department with chest pain and electrocardiographic (ECG) signs suggesting AMI: these were greater than 0.1 mV of ST-segment elevation in two or more limb leads or greater than 0.2 mV in two or more contiguous precordial leads, and the Minnesota code was used to classify Q-wave changes.12 Not all the cases of suspected AMI did in fact progress to AMI. Patients were included in the confirmed AMI group if they had a primary discharge diagnosis in accordance with the World Health Organization criteria.13 The diagnosis was confirmed if the patient had at least two of the following three findings: the presence of a typical history of characteristic chest pain, new ECG changes of pathological Q waves, and an elevation of serum CK to 390 U/L or above during the first 72 hours of admission. Improvement process A quasi-experimental design involving a pre-intervention and post-intervention phase with a concurrent control was used to test the effect of the TQM approach. The study was conducted in two 15-month stages. The first stage involved the collection of pre-intervention data. In the second stage the TQM intervention took place. Using the FOCUS-PDCA TQM model, a multidisciplinary team was empowered to make the necessary improvements, which were introduced during the intervention period as they were developed. The team included representatives from all clinical areas involved in the process of laboratory testing. These included the Emergency Department (Director of Emergency, Assistant Director of Nursing, Clinical Nurse Specialist and Senior Medical Officer), CCU (Director of Cardiology, Nurse Unit Manager, Clinical Nurse Specialist), Pathology Department (Director of Pathology, Senior Technologist), and Administration (Deputy Director of Medical Services). The hospital's Quality Assurance Officer was appointed the Quality Advisor for the TQM team. TQM team meetings were held weekly over the first two months of the 15-month intervention period, and then monthly. The TQM team established an overall review and improvement of the total systems and processes involved in test ordering. The team's mission statement was To ensure that the pathology services used are appropriate, effective and efficient for supporting clinical care in AMI. Practice guidelines were introduced for laboratory testing in the management of AMI. Clinicians in the team were actively encouraged to participate in the improvement process. Draft copies of the guidelines were circulated widely to other clinicians for comment. Suggestions for change were considered and incorporated into the guidelines as determined by the team. The guidelines provided details of all recommended laboratory tests during the 72-hour period following the admission of a patient with suspected AMI into hospital. Strategies were developed for education and training programs, feedback mechanisms and ongoing monitoring of performance through data collection and analysis. To implement the changes, the TQM team assigned responsibility to motivated individuals to accomplish specified tasks. Progress reports were provided at subsequent meetings. Medical staff requesting laboratory tests for suspected AMI patients were issued with pre-stamped pathology request forms that listed the recommended tests from the guidelines. Clinically indicated tests Requested laboratory tests were designated as "clinically indicated" if the ordered tests matched those listed in the AMI practice guidelines devised by the team, or if the tests were found to be justified when the patient's records were checked by one of the study's hospital medical officers. Conversely, "non-clinically indicated" tests were all requested tests found to be outside the recommendations of the AMI practice guidelines that could not be justified for inclusion when checked by one of the study's hospital medical officers. Data collection and statistical analysis Once the suspected AMI patients were transferred to the CCU of the experimental group hospital, an audit was undertaken of the laboratory tests requested and the time and date of each blood collection. The data were collected by me in collaboration with the Senior Technologist of the Pathology Department. The findings on the appropriateness of test use were reported to the Emergency Department and CCU staff, and at the team meetings. For each of the confirmed AMI groups, the number of clinically and non-clinically indicated tests were retrospectively determined. Group equivalency between the pre-TQM and post-TQM groups was determined using the t test. Only the results for the confirmed AMI groups are presented here. All data were coded, entered and analysed using SPSS for Windows.14 Ethical approval The study was approved by the Ethics Committees of both the South Western Sydney Area Health Service and the Nepean Health Service. Results The experimental and control groups did not differ in age, sex ratio or length of stay in hospital (Box 2). The proportion of clinically indicated tests that were requested for the experimental group increased from 77.5% before the TQM intervention to 88.2% in the intervention period (Box 3). For the control group, the proportion of clinically indicated tests that were requested did not change significantly (Box 4A). For the experimental group, the number of non-clinically indicated tests per admission was reduced by 81.7% in the intervention period (Box 4B), whereas for the control group this number did not change significantly. Overall, the total number of tests requested at the experimental hospital was approximately halved in the intervention period, whereas there was no change at the control hospital (Box 5). Discussion The introduction of TQM strategies achieved statistically highly significant improvements to the appropriateness of laboratory test ordering in AMI. The findings were sustained over the 15-month intervention period and were not replicated within the concurrently monitored control group hospital. The findings have a number of potentially important economic and patient care implications. Firstly, the 81.7% reduction of unnecessary tests may provide a substantial reduction in the overall cost of hospital inpatient pathology services if the TQM strategies were to be successfully introduced to more clinical situations. Such potential savings remain the challenge for future investigations. Secondly, there was an overall benefit to patient care through the increased use of clinically indicated laboratory tests. In addition, although not reported here, there was a statistically significant improvement to the clinically recommended timing of the blood collections for cardiac enzyme testing. I have not been able to find any study published in medical and health sciences journals during the past 15 years that has used TQM specifically to improve the appropriateness of clinical laboratory use in AMI. However, the FOCUS-PDCA model has had widespread and successful application within the healthcare system.15-16 As the model is process driven, it is likely to be adaptable to other test-ordering processes. My study supports the recent findings by Nardella et al of significant cost reductions and increased appropriateness of testing through the use of continuous quality improvement strategies.17 Although Nardella et al did not specifically use the FOCUS-PDCA model, they adopted a similar improvement approach. Methodologically, however, my study provided several strengths, with the inclusion of a pre-test and a concurrent control group to enable substantial control over threats to internal and external validity. The introduction of the TQM intervention was accompanied by additional staff time spent undertaking activities such as training, meetings, data collection and analysis. There was also medical, nursing and clerical staff time saved as a result of the accompanying improvements to the clerical systems used in test ordering. Although such costs and benefits were not quantified, I suspect that any savings made in staffing time were expended in establishing and maintaining the intervention. Following the 15-month intervention and the disbandment of the formal TQM team, the AMI laboratory testing guidelines remain in operation. However, recent discussions with key personnel from the original team reveal that the pre-stamped pathology request forms are no longer in use and that compliance with the testing guidelines has generally declined. These observations are consistent with those of other studies where improvements have failed to be sustained once improvement strategies had been completed.18-19 In view of the important economic and patient care implications, further investigations should be undertaken on the effects of the TQM approach in a variety of other clinical laboratory testing situations. Acknowledgements This research formed part of a major doctoral study undertaken by the author under the supervision of Professor Graeme Rawson, former Dean of the Faculty of Health, University of Western Sydney, Macarthur, NSW. Thanks also to Ms Margaret Gehrig, Quality Advisor to the study. I also acknowledge the contribution of the staff of the Bankstown-Lidcombe Hospital, particularly the Emergency Departments, Coronary Care Units, Pathology and Clinical Information Departments. Similarly, thanks also to the Clinical Information Department of the Nepean Hospital. References Hamm CW. New serum markers for acute myocardial infarction. N Engl J Med 1994; 331: 607-608. Deeble J, Lewis-Hughes P. Directions for pathology. Background Paper No. 6. Melbourne: National Health Strategy, 1991; 7. Gama R, Swain DG, Nightingale PG, et al. The effective use of cardiac enzymes and electrocardiograms in the diagnosis of acute myocardial infarction in the elderly. Postgrad Med J 1990; 66: 375-377. Grantham P, Weinstein S. Reducing pathology test misuse. Aust Health Rev 1993; 16: 16-23. Ratnaike D, Hunt D, Eilermann R, et al. The investigation of chest pain: audit and intervention. Med J Aust 1993; 159: 666-671. Rose WD, Martin JE, Abraham FM, et al. Calcium, magnesium and phosphorus: emergency department testing yield. Acad Emerg Med 1997; 4: 559-563. Hindmarsh JT, Lyon AW. Strategies to promote rational clinical chemistry test utilization. Clin Biochem 1996; 29: 291-299. Mugford M, Banfield P, O'Hanlon M. Effects of feedback of information on clinical practice: a review. BMJ 1991; 303: 398-402. Winkens RA, Ament AJ, Pop P, et al. Routine individual feedback on requests for diagnostic tests: an economic evaluation. Med Decis Making 1996; 16: 309-314. Goldman L. Changing physicians' behavior -- the pot and the kettle. N Engl J Med 1990; 322: 1524-1525. Hospital Corporation of America. Hospitalwide quality improvement process, strategy for improvement: FOCUS-PDCA. Nashville: Hospital Corporation of America, 1989. Rose GA, Blackburn H. In: Cardiovascular survey methods. WHO Monograph Series No. 56. Geneva: World Health Organization, 1968; 137-153. World Health Organization criteria for the diagnosis of acute myocardial infarction. Geneva: World Health Organization Cardiovascular Disease Unit, 1981. SPSS Inc. SPSS for Windows [computer program]. Version 6.0. Chicago Ill: SPSS Inc, 1992. Rhew E. Quality improvement project reviews: a tool to accelerate the transformation. Jt Comm J Qual Improv 1994; 20: 79-89. Schneider PD. FOCUS-PDCA ensures continuous quality improvement in the outpatient setting. Oncol Nurs Forum 1997; 24: 966-969. Nardella A, Farrell M, Pechet L, et al. Continuous improvement, quality control, and cost containment in clinical laboratory testing. Arch Pathol Lab Med 1994; 118: 965-968. Reeves CA, Bednar DA. What prevents TQM implementation in health care organisation? Qual Prog 1993; 26: 41-44. Chan YCL, Ho SJK. Continuous quality improvement -- a survey of American and Canadian healthcare executives. Hosp Health Serv Adm 1997; 42: 525-544. (Received 24 Oct 1997, accepted 3 Aug 1998) Author's details Division of Public Health, University of Western Sydney, Macarthur, NSW. Godfrey Isouard, BSc, PhD, Lecturer. Reprints will not be available from the author. Correspondence: Dr G Isouard, Division of Public Health, University of Western Sydney, Macarthur, PO Box 555, Campbelltown, NSW 2560. E-mail: g.isouardATuws.edu.au 1: The FOCUS-PDCA model Find a process to improve Organise a total quality management team select team members team building Clarify laboratory test-requesting process flow charts of test-requesting process identify problems customer requirements Uncover causes of poor quality test use causes of variation collect data Select process improvement determine proposed improvements prioritise Plan process improvement develop improvement strategies education and training development Do the improvement pilot test implement strategies collect and analyse data Check results check lessons learnt determine how effort could be improved Act to hold the gain standardise procedures establish monitoring and feedback processes plan continuous improvement 2: Comparison of characteristics of patients with confirmed acute myocardial infarction, before and during the total quality management intervention Experimental group Control group Pre-TQM intervention periodn=252n=203 Mean age in years (95% CI)60.3 (48.6-74.4)59.2 (47.5-71.8) Men:women (%)71.8:28.277.1:22.9 Median LOS in days (95% CI)8.8 (7.2-12.6)8.0 (6.8-11.4) TQM intervention periodn=253n=211 Mean age in years (95% CI)61.0 (49.1-77.3)59.6 (47.8-72.6) Men:women (%)74.7:25.370.8:29.2 Median LOS in days (95% CI)8.0 (6.3-11.4)7.3 (5.4-10.5) TQM=total quality management. CI=confidence interval. LOS=length of stay in hospital. 3: Changes in the use of clinically indicated tests for the experimental confirmed acute myocardial infarction group Pre-TQM intervention period TQM intervention periodPercentage point increase Number of testsNumber of testsin indicated tests Requested*Indicated†Ratio A‡Requested*Indicated†Ratio B§requested (B-A)P Sodium, potassium, chloride2193339464.6%2631318082.7%18.1<0.01 Urea00--00------ Creatinine17532953.2%28129495.6%42.4<0.01 Glucose31052958.6%37552271.8%13.2<0.01 Calcium, magnesium, phosphate758192.6%242885.7%-6.9<0.01 Cardiac enzyme profile§§1860201692.3%1962202496.9%4.70.18 Liver function tests#14715495.5%778491.7%-3.8<0.01 Cholesterol/triglyceride47950495.0%43850686.6%-8.5<0.01 Free thryoxine, thyrotropin304271.4%202676.9%5.5<0.01 Miscellaneous clinical chemical tests116116100.0%8686100.0%0-- All clinical chemical tests5385716575.2%5894675087.3%12.2<0.01 Automated blood count**1136130487.1%1028114889.6%2.40.20 Prothrombin index252252100.0%253253100.0%0-- Miscellaneous haematology tests2424100.0%4545100.0%0-- All haematology tests1412158089.4%1326144691.7%2.30.15 Urine microexamination1111100.0%1414100.0%0-- Miscellaneous microbiology tests1616100.0%2828100.0%0-- All microbiology tests2727100.0%4242100.0%0-- All tests6800877277.5%7262823888.2%10.6<0.01 TQM=total quality management. *Clinically indicated tests that were requested. †Total number of tests that were clinically indicated but not necessarily requested. ‡Pre-TQM tests requested/pre-TQM tests indicated. §TQM tests requested/TQM tests indicated. §§Creatine kinase and creatine kinase-MB. #Total bilirubin, aspartate aminotransferase, alanine aminotransferase, gamma glutamyltransferase, alkaline phosphatase, protein, albumin. **Haemoglobin, white cell count, red cell count, platelets. 4: Comparison of the changes in the use of clinically indicated tests (A) and non-clinically indicated tests (B) for the experimental and control confirmed AMI groups APre-TQM intervention period TQM intervention periodPercentage point increase Number of testsNumber of testsin indicated tests Requested*Indicated†Ratio A‡Requested*Indicated†Ratio B§requested (B-A)P Experimental group6800877277.5%7262823888.2%10.6<0.01 Control group7340893382.2%7399920080.4%-1.80.19 BPre-TQM intervention period TQM intervention periodReduction in Total number of tests requestedNumber of tests per admissionTotal number of tests requestedNumber of tests per admissionnon-clinically indicated tests per admissionP Experimental group966738.417767.081.7%<0.01 Control group489024.1484823.04.6%0.48 AMI=acute myocardial infarction. TQM=total quality management. *Clinically indicated tests that were requested. †Total number of tests that were clinically indicated but not necessarily requested. ‡Pre-TQM tests requested/pre-TQM tests indicated. §TQM tests requested/TQM tests indicated. 5: Total number of tests requested in the experimental and control confirmed acute myocardial infarction groups Pre-TQM intervention period TQM intervention period TotalIndicated*Not indicated†TotalIndicated*Not indicated† Experimental group1646768009667903872621776 Control group12230734048901224773994848 TQM=total quality management. *Number of clinically indicated tests requested. †Number of non-clinically indicated tests requested.

Godfrey Isouard

Outdoor air pollution and children's respiratory symptoms in the steel cities of New South Wales Peter R Lewis, Michael J Hensley, John Wlodarczyk, Ruth C Toneguzzi, Victoria J Westley-Wise, Trevor Dunn and Dennis Calvert MJA 1998; 169: 459-463 For editorial comment, see Abramson & Beer Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Public and environmental health - ©MJA1998 Abstract Objective: To investigate the relationship between outdoor air pollution and the respiratory health of children aged 8 to 10 years. Design: A cross-sectional survey (between October 1993 and December 1993) of children's health and home environment. Summary measures of particulate pollution (levels of particles with an aerodynamic diameter less than 10 µm [PM10] each 6th day) and SO2 (daily mean and maximum hourly values) were estimated for each area (using air quality monitoring station data from July 1993 to June 1994). Setting and survey participants: Parents of 3023 primary school children (Years 3, 4 and 5) from industrial and non-industrial areas with air quality monitoring stations in the Hunter and Illawarra regions of New South Wales. Main outcome measures: Reported occurrence of four or more chest colds, four or more attacks of wheezing, and night-time cough without a cold for more than two weeks, all within the previous 12 months. Results: 77% response rate, ranging by area from 66% to 88%. The average annual outdoor air pollution for the nine areas was 18.6-43.7 µg/m3 for PM10 and 0.16-0.90 parts per hundred million for SO2. The proportion of children reported to have the main outcome symptoms were: chest colds, 3.0%-9.7%; night cough, 12.3%-30.5%; and wheeze, 3.4%-11.3%. There was no significant association with SO2, but a significant increase in the odds of symptoms per 10 µg/m3 increase in PM10 on chest colds (odds ratio [OR], 1.43; 95% confidence interval [CI], 1.12-1.82) and night-time cough (OR, 1.34; 95% CI, 1.19-1.53), but not wheeze. Passive smoking was significantly associated with chest colds but not with the other symptoms. Maternal allergy was associated with all three respiratory symptoms, most strongly with wheeze. Conclusion: These results provide evidence of health effects at lower than expected levels of outdoor air pollution in the Australian setting. They also suggest differences in contributions of environmental and hereditary factors to cough and chest colds compared with wheeze. Introduction The effect of outdoor air pollution on the respiratory system has been investigated and reviewed often.1,2 Various health outcomes have been studied; for example, the relationship between exposure to elevated sulfur dioxide, total suspended particulates (TSP) and respiratory symptoms;3 the association between increased air pollution and hospital admissions for asthma;4,5 and studies of air pollution and mortality.6,7 These studies consistently report associations between air pollution and health outcomes, and illustrate the coherence of effects8 across the range of outcomes. There have been few studies of the relationship between air quality and health in Australia. Studies of asthma in school children living near power stations in the Hunter region of New South Wales had very low levels of exposure and lacked the power to detect small associations.9,10 Other studies in south-eastern Australia confirmed the high prevalence of asthma in Australia but had no exposure measures.11-13 The lower Hunter and Illawarra regions of New South Wales are on the coast, about 150 km north and 80 km south of Sydney, respectively. The main cities are Newcastle, in the Hunter region, and Wollongong, in the Illawarra region. The estimated resident populations in 1993 for Newcastle and Wollongong were 451 100 and 248 700, respectively.14 Both regions have fully integrated steelworks, combining coke ovens, blast furnaces, steel-making vessels and rolling mills, and the main pollutants are coke oven gases, blast furnace gases and particulates. In addition, lead smelting occurs in the Hunter region, and copper smelting in the Illawarra region. The Hunter Illawarra Study of Airways and Air Pollution (HISAAP) arose out of longstanding community concerns about the health effects of air pollution near heavy industry in Newcastle and Wollongong. The generalisability of overseas studies to Australia was not known, because of differences in climate, geography, population and pollution sources. Public health workers need local data to initiate action on air quality at the community level and to formulate air quality guidelines. This study, the first part of HISAAP, investigated the association between outdoor air quality, as particulates and sulfur dioxide, and respiratory symptoms among primary school children. Methods The study was a cross-sectional survey to measure the prevalence of respiratory and atopic symptoms among primary school children and relate these to measures of ambient air quality. Study areas To make best use of resources, study areas were selected on the basis of existing or proposed air quality monitoring stations. In the lower Hunter region, the study areas close to industrial sources of air pollution were North Lake Macquarie, Mayfield and Stockton, and distant areas were Beresfield and Wallsend (Figure 1a). In the Illawarra region, we estimated that there would be an increasing gradient of exposure to air pollution, from Wollongong, to Albion Park, Kembla Grange and Port Kembla (Figure 1b). Wallsend, Beresfield, Wollongong and Albion Park were expected to have background levels of air pollution. We were able to compare these multiple control areas with those areas close to point sources of air pollution. Sampling strategy All primary schools within a three-kilometre radius of the air quality monitoring sites were approached to be involved in the study. The sampling frame included all children in Years 3 and 4, aged 8 to 10 years, whose parents or guardians approved their child's participation. In areas that had small numbers of children, children in Year 5 were included. We aimed to enrol about 300 children in each of the nine areas. Questionnaire Our questionnaire incorporated features used by previous investigators in the Hunter region and NSW,9,11,12 the International Study of Asthma and Allergies in Childhood,15 the American Thoracic Society questionnaire16 and the Six Cities study.17 It addressed demographic data and parental education; child's health (cough, chest colds, wheeze) and atopic symptoms (hayfever, eczema); child's use of respiratory medications; family's symptoms; and home environmental factors (indoor smoking, heating, cooking, carpets, cats). The questionnaire was piloted in both regions.18The primary health outcomes were night-time cough, chest colds and frequent wheeze. They were the responses to the following questions: In the last 12 months has your child had a dry cough at night, apart from a cough with a cold or chest infection? If yes, has this cough lasted for more than 2 weeks? In the last 12 months how many chest colds did your child have? . . . 4 or more chest colds. How many attacks of wheezing has your child had in the last 12 months? . . . 4 or more attacks. Exposure to indoor tobacco smoke was assessed by asking whether the respondent, other adult or any other person smoked cigarettes inside the home. The survey was conducted from October 1993 to December 1993. Children received consent forms and questionnaires at school. Parents completed the questionnaires, and children returned them to their teachers. If a questionnaire was not returned after two weeks, a copy was posted to the parents for completion and return by reply-paid post. Exposure measurement The lower Hunter and Illawarra Air Pollution Monitoring Networks19 consist of sites operated by the Environment Protection Authority (EPA) and industry self-monitoring sites audited by the EPA. Particulate and sulfur dioxide measures were collected for all study areas, from January 1993 to December 1994. As air pollution data were incomplete for some sites in 1993, the goal was to have representative data for 12 consecutive months close to the reference period of the questionnaire (the 1993 calendar year). The most complete data were found to be from July 1993 to June 1994. Sensitivity analyses using data from adjacent 12-month periods showed very little effect on the results. Particulate pollution as PM10 (particles with an aerodynamic diameter less than 10 µm) was measured for 24 hours every sixth day using high volume air samplers with a size-selective inlet, and these measures were averaged over the 12 months. North Lake Macquarie, Mayfield and Stockton measured only TSP, so local data in which both TSP and PM10 were measured for shorter periods20,21 were used to derive a conversion factor to PM10. In North Lake Macquarie, PM10 was estimated to be 60% of TSP, close to the value used by others.22 In Mayfield and Stockton, PM10 was estimated to be 45% of TSP. Sulfur dioxide measurements were obtained hourly using pulsed fluorescent spectrophotometry. Twelve-month averages of the daily maximum values and of the daily mean values were calculated. There were no major trends in air pollution during this time. Seasonal changes did not affect the relative ranking of the areas in terms of air pollution. Analysis We used SAS software for the analysis.23 Descriptive statistics were produced for participation rates, health measures and environmental measures for each of the nine study areas. Logistic regression analysis using generalised estimating equations (GEEs) was used to model the reported symptom (outcome variable) on the following explanatory variables: PM10 and sulfur dioxide measures, unflued gas heating, adults smoking inside the home, indoor cats, age, sex, and maternal allergy.24 The level of clustering for this analysis was the school (ie, clusters of children within schools), with 35 schools in the study. Although schools were also clustered within study areas, area was not chosen as the clustering level because of the allocation of exposure measure by area. Ethical approval The study received ethical approval from the NSW Education Department, the Regional Catholic Education Office, and the Ethics Committees of the Hunter Area Health Service, the University of Newcastle, the Illawarra Area Health Service, and the University of Wollongong. Results Subjects Questionnaires were given to 3023 primary school children -- 1554 children (from 18 schools) in the Hunter region, and 1469 (from 17 schools) in the Illawarra region. In the Hunter region, 1284 (82%) questionnaires were returned, and in the Illawarra region 1056 (72%) were returned. There were no significant differences in age and sex profiles across the study areas (Table 1). Respiratory symptoms About a quarter of the children were reported to have had at least one of the primary respiratory symptoms in the preceding 12 months (Table 2). While some areas with high proportions of children with symptoms were close to industry (Mayfield and Stockton with children having chest colds), this was not a consistent pattern (Beresfield and children with night cough). The proportion of children who had ever been diagnosed with asthma ranged from 17.4% to 38.4%. Home environment and outdoor air pollution Depending on the study area, a third to half of the children lived in homes where one or more adults smoked cigarettes in the home (Table 3). The proportion of children in homes with gas heating that did not have a flue to the outside varied greatly between the areas. The levels of particulate pollution in Mayfield and Stockton approached the US EPA standard for PM10 of 50 µg/m3 (annual mean) (Table 3) and the Australian standard, at that time set for TSP at 90 µg/m3 (annual mean). The annual mean sulfur dioxide levels (daily means) were below the Australian standard of 2 parts per hundred million (annual mean).19 Within study areas, particulate levels were not necessarily high in high sulfur dioxide areas, and vice-versa. Scatter plots of primary health outcomes by air quality measures suggested some positive associations between PM10 and both night cough and chest colds (Figure 2). Generalised estimating equation analysis For each 10 µg/m3 increase in annual mean PM10, children were 43% more likely to be reported as having four or more chest colds (odds ratio [OR], 1.43; 95% confidence interval [CI], 1.12-1.82). The approximate difference in particulate pollution between the area with the highest PM10 level and the area with the lowest was 25 µg/m3, and children were 144% more likely to be reported to have frequent chest colds per 25 µg/m3 increase in annual mean PM10 (OR, 2.44; 95% CI, 1.33-4.49). There was a consistent significant association between particulate pollution and symptoms of night cough and chest colds (Table 4). No association was found between sulfur dioxide and the three respiratory symptoms. An adult smoking in the home was associated with chest colds, and unflued gas heating was associated with frequent wheeze. Boys were more likely to experience frequent wheeze, and there was a significant association between increasing age and likelihood of frequent wheeze. Maternal allergy was consistently associated with an increased risk for all health outcomes, but strongest for wheeze. To estimate the possible impact of the two areas with the largest particulate pollution (Mayfield and Stockton, in the Hunter region), the GEE analysis was repeated excluding those areas. There was little change in the coefficients for most variables, including maternal allergy, for each outcome. However, the odds ratio for chest colds per 10 µg/m3 increase in annual mean PM10 increased substantially -- with Mayfield and Stockton excluded, the odds ratio was 2.05 (95% CI, 1.46-2.87), compared with 1.43 for the full dataset. Discussion This cross-sectional study of primary school children has shown an important association between relatively low levels of particulate air pollution and respiratory symptoms. The importance is reflected in the finding that the odds ratios for particulates were of the same order as for exposure to indoor tobacco smoke. We found an association with maternal allergy across all three symptom outcomes, but strongest for wheeze. The prevalence estimates of respiratory symptoms are similar to those found in other Australian and New Zealand studies.11-13,25 The association between air pollution and respiratory symptoms in children is also consistent with similar cross-sectional studies from other countries. An Italian study found that children living in polluted areas experienced more cough, rhinitis, pneumonia and early respiratory infections than control subjects,26 but that study did not measure actual air pollution exposure. A British study found that children were more likely to miss school because of respiratory symptoms of cough and wheeze if their school was exposed to a higher coal dust burden (OR, 1.55; 95% CI, 1.17-2.06).27 Our results are directly comparable with those of the Six Cities and 24 Communities studies in North America;17,28,29 effects of particulate pollution on respiratory symptoms and effects of indoor smoking were of the same order in those studies and ours. Two results of particular note are the relationship between frequent wheeze and age, and the effect of removing from the analysis the areas with the highest particulate pollution. The increase in wheeze with age is quite large (OR, 1.38 per year). Stratified analysis revealed similar coefficients for younger and older age groups, suggesting that a sampling bias may explain the strength of the relationship. The substantial increase in the odds ratio of particulate pollution and chest colds after removal of the areas with highest PM10 suggests that the dose-response curve for particulate pollution and chest colds is curvilinear and much steeper at lower pollution levels. Potential limitations of our study include possible reporting bias by residents of industrial areas and the variability in the measurement methods for particulate air pollution. It is not possible to estimate the level of possible reporting bias, but its effect may be reduced by selection bias in that families with children with lung disease may not come to, or may leave, polluted areas. The difference in measurement of particulates was unavoidable. A subsequent diary study obtained comparisons of TSP and PM10 which provided evidence on which to convert the measures for local conditions.20,21 Particularly striking are the results which show differences in the determinants of wheeze compared with coughs and colds. While maternal allergy played a part in all, passive smoking was restricted to colds, and particulates to coughs and colds. This suggests differences in underlying pathology which deserve further exploration. Our data lead us to question the current air quality standards, which are based on the premise that levels at or below recommended ambient levels are not likely to produce a clinically important effect. In particular, recommended levels do not consider increased individual sensitivity caused by a disease such as asthma. While this study cannot infer causation, it adds to other local and overseas work that increasingly indicates the role of air pollution in respiratory health -- alone and synergistically with other aeroallergens.30 This information contributes to the review of Australian air quality guidelines, and to the broader societal debate about lifestyle, transport and energy consumption. Individuals, communities, industry and organisations need to reassess their environmental performance not only for the environment, but also for health. Acknowledgements Our thanks to members of the following organisations which supported the study: Newcastle Environmental Toxicology Research Unit (especially Randall Robertson and Julie Holt), University of Newcastle, Hunter Public Health Unit, Hunter Area Pathology Service, Newcastle City Council, Lake Macquarie Council, Illawarra Public Health Unit, University of Wollongong, NSW Health Department, NSW Environment Protection Authority, Australian Nuclear Science and Technology Organisation, BHP (Newcastle and Port Kembla), Incitec, Pasminco Metals Sulphide, Southern Copper. We are grateful to the schools, parents, and children who participated in this study. HISAAP was partly funded by the NSW Health Department as part of its Health and Air Research Program. Dr P Lewis received an NHMRC Public Health Fellowship for two years. References Committee on the Medical Effects of Air Pollution. Asthma and outdoor air pollution. London: HMSO, 1995. Committee of the Environmental and Occupational Health Assembly of the American Thoracic Society. Health effects of outdoor air pollution. Am J Respir Crit Care Med 1996; 153: 3-50. Lunn FE, Knowelden J, Handyside AJ. Patterns of respiratory illness in Sheffield infant schoolchildren. Br J Prev Soc Med 1967; 21: 7-16. Pope CA III. Respiratory admissions associated with PM10 pollution in Utah, Salt Lake and Cache Valleys. Arch Environ Health 1991; 46: 90-97. Bates DV, Baker-Anderson M, Sizto R. Asthma attack periodicity: a study of hospital emergency visits in Vancouver. Environ Res 1990; 51: 51-70. Anderson HR, Ponce de Leon A, Bland JM, et al. Air pollution and daily mortality in London: 1987-92. BMJ 1996; 312: 665-669. Dockery DW, Pope CA III, Xu X, et al. An association between air pollution and mortality in six US cities. N Engl J Med 1993; 329: 1753-1759. Bates DV. Health indices of the adverse effects of air pollution: the question of coherence. Environ Res 1992; 59: 336-349. Henry RL, Abramson R, Adler JA, et al. Asthma in the vicinity of power stations: I. A prevalence study. Pediatr Pulmonol 1991; 11: 127-133. Henry RL, Bridgman HA, Wlodarczyk J, et al. Asthma in the vicinity of power stations: II. Outdoor air quality and symptoms. Pediatr Pulmonol 1991; 11: 134-140. Bauman A, Mitchell CA, Henry RL, et al. Asthma morbidity in Australia: an epidemiological study. Med J Aust 1992; 156: 827-831. Peat JK, Toelle BG, Gray EJ, et al. Prevalence and severity of childhood asthma and allergic sensitisation in seven climatic regions of New South Wales. Med J Aust 1995; 163: 22-26. Robertson CF, Bishop J, Dalton M, et al. Prevalence of asthma in regional Victorian schoolchildren. Med J Aust 1992; 156: 831-833. Australian Bureau of Statistics. Regional population growth 1991 and 1996. Canberra: ABS, 1997. (Catalogue No. 3218.0.) Asher M, Keil U, Anderson H. International study of asthma and allergies in childhood (ISAAC): rationale and methods. Eur Respir J 1995; 8: 483-491. Ferris BG. Epidemiology standardisation project. Part II. Am Rev Respir Dis 1978; 118: 1-53. Ware JH, Dockery DW, Spiro A III, et al. Passive smoking, gas cooking and respiratory health of children living in six cities. Am Rev Respir Dis 1984; 129: 366-374. Lewis PR, Toneguzzi R, Long K, et al. The Hunter Illawarra study of airways and air pollution: refining the process. NSW Public Health Bulletin 1995; 6: 110-112. State Pollution Control Commission. Quarterly air quality monitoring report No 4, 1990. Sydney: SPCC, 1992. Lewis P, Holt J, Fryer J. Daily particulate monitoring in the Speers Point/Boolaroo area -- Lake Macquarie Research Grant 94-9. Newcastle: University of Newcastle, Newcastle Environmental Toxicology Research Unit, 1997. Environment Section, BHP. Mayfield quality control: measures of particulate pollution. BHP, 1995. Dockery DW, Pope CA III. Acute respiratory effects of particulate air pollution. Annu Rev Public Health 1994; 15: 107-132. SAS for Windows [computer program]. Version 6.03. Cary, North Carolina: SAS Institute, 1988. Karim MR. PC version of GEE1 [computer program]. Baltimore: Department of Biostatistics, The Johns Hopkins University, 1989. Moyes CD, Waldon J, Ramadas D, et al. Respiratory symptoms and environmental factors in schoolchildren in the Bay of Plenty. N Z Med J 1995; 108: 358-361. Corbo GM, Forastiere F, Dell'Orco V, et al. Effects of environment on atopic status and respiratory disorders in children. J Allergy Clin Immunol 1993; 92: 616-623. Brabin B, Smith M, Milligan P, et al. Respiratory morbidity in Merseyside schoolchildren exposed to coal dust and air pollution. Arch Dis Child 1994; 70: 305-312. Cunningham J, O'Connor GT, Dockery DW, Speizer FE. Environmental tobacco smoke, wheezing and asthma in children in 24 communities. Am J Respir Crit Care Med 1996; 153: 218-224. Dockery DW, Speizer FE, Stram DO, et al. Effects of inhalable particles on respiratory health of children. Am Rev Respir Dis 1989; 139: 587-594. Devalia JL, Rusznak C, Herdman MJ, et al. Effect of nitrogen dioxide and sulphur dioxide on airway response of mild asthmatic patients to allergen inhalation. Lancet 1994; 344: 1668-1671. (Received 29 Oct 1997, accepted 21 Jul, 1998) Authors' details Newcastle Environmental Toxicology Research Unit, University of Newcastle, Newcastle, NSW. Peter R Lewis, MPH, FAFPHM, Public Health Research Fellow; Michael J Hensley, MB BS, PhD, Director and Professor of Medicine; John Wlodarczyk, BEc, PhD, Statistician; Ruth C Toneguzzi, RN, DipClinEpi, Clinical Nurse Specialist. Illawarra Public Health Unit, Wollongong, NSW. Victoria Westley-Wise, MPH, FAFPHM, Director; Trevor Dunn, GdDipSc(Nursing), MPH(Occ Health), Project Manager. University of Wollongong, Wollongong, NSW. Dennis Calvert, MD, FRACP, Professor in Medicine and Public Health. Reprints will not be available from the authors. Correspondence: Dr P R Lewis, Newcastle Environmental Toxicology Research Unit, Division of Medicine, John Hunter Hospital, Locked Bag 1, Hunter Regional Mail Centre, NSW 2310. E-mail: plewiATdoh.health.nsw.gov.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Peter R Lewis · Michael J Hensley · John Wlodarczyk · Ruth C Toneguzzi · Victoria J Westley-Wise · Trevor Dunn · Dennis Calvert

HIV disease progression in Australia in the time of combination antiretroviral therapies Patricia K Correll, Matthew G Law, Ann M McDonald, David A Cooper and John M Kaldor MJA 1998; 169: 469-472 For editorial comment, see Stewart Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Infectious diseases and parasitology - ©MJA1998 Abstract Objective: To examine the effect of recent developments in antiretroviral therapy on HIV disease progression and survival. Design: Retrospective cohort study. Participants and setting: Two cohorts of people with HIV were defined retrospectively from the records of a large immunology laboratory. The first cohort were subjects whose CD4+ T cell counts had dropped to 200 x 106/L during 1990, and the second were subjects whose CD4+ T cell counts had dropped to 200 x 106/L in 1994. Main outcome measures: HIV disease progression and survival was determined over a minimum three years of follow-up for each cohort (ie, 1990-1993; 1994-1997). Results: 346 subjects were included in the analysis (193 subjects from 1990 and 153 from 1994). The relative risk of progression to AIDS in the 1994 cohort compared with the 1990 cohort was 0.57 (95% confidence interval, 0.35-0.91; P = 0.018) and the relative risk of death was 0.20 (95% confidence interval, 0.08-0.49; P < 0.001). Conclusions: There were 43% fewer AIDS cases and 80% fewer deaths in the time following the increased availability of combination antiretroviral therapy in Australia. Introduction In recent years there have been major advances in antiretroviral treatments for people with HIV infection. Evidence from clinical trials (DELTA and ACTG 175)1,2 first defined the superiority of combination therapy over monotherapy in delaying progression to AIDS and prolonging survival. Since that time more potent antiretroviral agents have been approved, including lamivudine (widely available in Australia since early 1995) and protease inhibitors (available since late 1995). In clinical trials, these drugs have demonstrated significant ability to reduce disease progression and improve survival when added to existing regimens of antiretroviral treatments.3,4Although there has been anecdotal evidence in Australia suggesting that these treatments have reduced the overall number of AIDS cases and improved survival following AIDS, their effect has not been systematically studied. We therefore decided to compare the rates of progression to AIDS and survival in two cohorts of people between 1990-1993 and 1994-1997, with CD4+ T cell counts of 200 x 106/L (CD4 cell count 200/µL) at entry to the study, as a basis for assessing the impact of new treatments. Methods Subject identification Two cohorts of subjects in this study with CD4 cell counts of 200/µL were defined retrospectively from records of routine CD4 cell counts undertaken at the Centre for Immunology, St Vincent's Hospital, Sydney. St Vincent's Hospital has been responsible for the care of about 25% of people with AIDS in Australia, and its immunology laboratory is used by St Vincent's Hospital inpatient and outpatient services, and by several general practitioner clinics that specialise in managing people with HIV infection. Subjects whose CD4 cell counts fell below 200/µL in 1990 would not have had access to combination therapies, whereas the 1994 cohort was under follow-up when combination therapies with more potent drugs were becoming available in Australia. A CD4 cell count of 200/µL was used as the starting point for follow-up because at this point individuals have not usually developed AIDS-defining illnesses,5 but are considered to be at a point of immune decline where the probability of developing AIDS in the near future has greatly increased.6,7 Subjects included were HIV-positive but without AIDS at entry to the study, with CD4 cell counts declining through 200/µL in the index year (1990 or 1994) or with a single CD4 cell count in the range 180-220/µL. All eligible subjects were included, except subjects from two large clinics which had used the Centre for Immunology laboratory in 1990 but not in 1994. Where available, two or more CD4 cell count results from the period beginning in the middle of the year before the index year and ending at the middle of the following year were recorded so that the rate of CD4 cell change could be estimated. Subjects with multiple CD4 cell counts entered the study on the date of the first test in the index year that was below 200/µL. For subjects with a single CD4 cell count only, entry to the study was on the date of this test, provided the result fell between 180-220/µL. Follow-up Analysis was based on a minimum of three (up to four) years' follow-up of both cohorts through linkage to the National AIDS Registry, which records AIDS diagnoses and deaths following AIDS in Australia. Testing records were linked to Registry data by matching name codes (first two letters of the surname and given name) and dates of birth from the reference laboratory database with the Registry. Because reporting of AIDS diagnoses and deaths to the Registry is subject to delay, subjects from the 1994 cohort who did not appear on the Registry were followed up through their referring doctors. The doctor was also contacted for follow-up information if a name code was not available with the CD4 count records. Information sought included name code, date of birth, whether the subject had progressed to AIDS or died with AIDS, and date of the doctor's most recent contact with the subject. Statistical methods All statistical analyses were performed using SAS.8 Decline of CD4 cells was estimated for subjects with two or more CD4 cell counts by linear regression. Baseline characteristics in the two cohorts were compared by means of Fisher's exact test for categorical variables and the Mann-Whitney rank sum test for continuous variables. We calculated AIDS-free survival and overall survival curves using Kaplan- Meier methods, and assessed the statistical significance of the difference in survival between the two cohorts by the log rank test. Proportional hazards regression was used to assess the effect of age, CD4 cell count at entry, and decline of CD4 counts on AIDS incidence and survival. In survival analysis, subjects without an AIDS diagnosis or death reported to the National AIDS Registry were assumed to be AIDS-free and alive at the end of follow-up. Secondary analyses were also performed based on confirmed AIDS-free and survival times by censoring at the date of last confirmed contact on the National AIDS Registry or with the referring doctor. Results After excluding 109 subjects in 1990 who had been referred from two clinics which did not use the Centre for Immunology laboratory in 1994, the database search identified 214 eligible subjects from 1990 and 169 from 1994, of whom 193 and 153, respectively, could be traced for follow-up and were included in the analysis (Box 1). As shown in Box 1, 60% of subjects who were followed up during 1990-1993 progressed to AIDS and 38% died, compared with 37% who progressed to AIDS and 13% who died during follow-up between 1994-1997. This corresponds to a 43% decrease in AIDS and an 80% reduction in death. As shown in Box 2, the median age at entry of the 1994 cohort was older than in the 1990 cohort, but sex, CD4 cell count at entry and CD4 cell decline were similar. The median age and sex of excluded subjects who came from the two clinics which did not use the Centre for Immunology laboratory in 1994 (age 37.5 years, 96% male) were similar to those included in the 1990 cohort (age 35 years, 98% male). Box 3 shows that progression to AIDS was slower and survival longer in the 1994 cohort than in the 1990 cohort, and that the improvement in both indices appeared to emerge after about one year of follow-up. Adjustment for age, CD4 count and CD4 slope at entry did not materially affect the differences between the two cohorts in progression time to AIDS and survival (Box 4). There were 52 subjects for whom only one CD4 cell count was recorded and rate of CD4 cell decline could not be estimated. A sensitivity analysis excluding these subjects gave very similar results to those for the full cohort (relative risk of AIDS and death in the 1994 group, 0.51 [P = 0.02] and 0.21 [P < 0.01], respectively). When follow-up was censored at the date of last confirmed contact, there remained a 28% decrease in progression to AIDS (P = 0.04) and a 45% reduction in death (P = 0.021). Discussion In this study, progression to AIDS was nearly halved and deaths fell by 80% among people with HIV infection whose CD4 cell count fell below 200/µL in 1994 compared with the corresponding 1990 cohort. The difference between the two groups emerged after about one year of follow-up, which in the 1994 cohort corresponded roughly to the introduction of lamivudine in early 1995, followed by protease inhibitors later in the same year. Although other differences between the cohorts cannot be entirely ruled out, the most likely explanation for the improved outcomes is advances in treatment strategies for people with HIV, including combination antiretroviral therapies. The benefit of combination therapy and new antiretroviral agents has been demonstrated so far largely through randomised controlled trials.9 Only recently has information started to emerge of the effectiveness of new therapies in a community-wide setting. The results of this study concur with recent cohort studies overseas, where substantial decreases in AIDS and deaths have been identified in patients followed up after the introduction of combination therapies compared with those under follow-up in earlier times.10,11 Although information on antiretroviral treatment was unavailable for the individuals in our cohorts, recent evidence suggests a dramatic change in the use of antiretroviral therapies in Australia over the time in which the two cohorts reported in our study were followed up. HIV-infected participants in the Sydney Men and Sexual Health Study (a prospective study of homosexually active men in Sydney) increased their uptake of combination antiretroviral therapy from about 2% in 1993 to more than 70% in 1997.12 Our study relied on the National AIDS Registry, which is subject to both reporting delay and underreporting. It is believed that about 70% of HIV diagnoses are reported to the Registry within six months (the minimum reporting time available in this study) and that nearly 100% are reported within three years.13 It is possible that incomplete reporting of AIDS and AIDS-related deaths to the National AIDS Registry may have resulted in underestimates of AIDS incidence or mortality in the later cohort. Results were similar, however, when analyses were censored at the date of last contact, suggesting that underreporting to the Registry was not in fact an important source of bias. One other potential source of bias is that changes in the natural history of the HIV epidemic may have resulted in a lower proportion of rapid progressors in more recent years, so that patients with CD4 counts of 200/µL in 1990 progressed more rapidly than those in 1994.14 In our analysis, there was little difference between the cohorts in CD4 cell decline (Box 2), and survival analysis adjusted for this gave very similar results. Furthermore, a sensitivity analysis excluding those subjects with only a single CD4 cell count also did not significantly affect the results. This analysis was based on CD4 cell counts collected over two years (mid year before the index year to mid year after), and is robust enough to suggest that both cohorts were subject to similar rates of progression. As the subjects were identified at two different times, it is possible that there were unidentifiable differences between the cohorts that may have influenced the outcomes in this study. However, the possibility of bias was substantially reduced by identifying all subjects from the same laboratory. This study contributes evidence to suggest that the rate of progression of HIV disease has decreased in Australia at the same time as new, more potent combination antiretroviral treatments have become available. It would be useful to study similar cohorts in the future to monitor the evolving pattern of the epidemic at a population level. Acknowledgements The National Centre in HIV Epidemiology and Clinical Research is funded by the Commonwealth Department of Health and Family Services through the Australian National Council on AIDS and Related Diseases. The help of the following medical practitioners in study follow-up is gratefully acknowledged: B Anderson, P Brooke, K Brown, A Carr, B Donovan, N Doong, C Duncombe, W Genn, J Kidd, A Mackie, M McMurchie, A McNulty, R Penny, A Pethebridge, M Robertson. We also thank John Zaunders at the St Vincent's Hospital Centre for Immunology for assistance in identifying the study sample. References Delta Coordinating Committee. Delta: a randomised double-blind controlled trial comparing combinations of zidovudine plus didanosine or zalcitabine with zidovudine alone in HIV-infected individuals. Lancet 1996; 348: 283-291. Hammer SM, Katzenstein DA, Hughes MD, et al. A trial comparing nucleoside monotherapy with combination therapy in HIV-infected adults with CD4+ cell counts from 200 to 500 per cubic millimeter. N Engl J Med 1996; 335: 1081-1090. CAESAR Coordinating Committee. Randomised trial of addition of lamivudine or lamivudine plus loviride-zidovudine-containing regimens for patients with HIV-1 infection: the CAESAR trial. Lancet 1997; 349: 1413-1421. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4+ cell counts of 200 per cubic millimeter or less. N Engl J Med 1997; 337: 725-733. Centers for Disease Control. Conditions included in the 1993 AIDS surveillance case definition. MMWR Morb Mortal Wkly Rep 1992; 41: RR-17. Hoover DR, Rinaldo C, Yanhua H, et al. Long-term survival without clinical AIDS after CD4+ cell counts fall below 200 x 106/L. AIDS 1995; 9: 145-151. Tindall B, Swanson CE, Cooper DA. Development of AIDS in a cohort of HIV-seropositive homosexual men in Australia. Med J Aust 1990; 153: 260-265. SAS statistical software version 6.12. Carey, NC: SAS Institute Incorporated, 1994. HIV/AIDS Clinical Trials and Treatments Advisory Committee of the Australian National Council on AIDS and Related Diseases. Antiretroviral therapy for HIV infection: principles of use. Standard of care guidelines; October 1997. Sydney: HIV/AIDS Clinical Trials and Treatments Advisory Committee, 1997. Egger M, Hirschel B, Francioli P, et al. Impact of new antiretroviral combination therapies in HIV infected patients in Switzerland: prospective multicentre study. BMJ 1997; 315: 1194-1199. Palella F, Delaney K, Moorman A, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. New Engl J Med 1998, 338: 853-860. National Centre in HIV Epidemiology and Clinical Research. HIV/AIDS and related diseases in Australia: Annual Surveillance Report 1998. Sydney: National Centre in HIV Epidemiology and Clinical Research, 1998. National Centre in HIV Epidemiology and Clinical Research. An epidemiological assessment of the HIV epidemic in Australia. Evaluation of the National HIV/AIDS Strategy 1993-1994 to 1995-1996. Canberra: Commonwealth Department of Health and Family Services, 1996. Ioannidis JPA, Cappelleri JC, Schmid CH, Lau J. Impact of epidemic and individual heterogeneity on the population distribution of disease progression rates. Am J Epidemiol 1996; 144: 1074-1085. (Received 13 Jan, accepted 30 Apr, 1998) Authors' details National Centre in HIV Epidemiology and Clinical Research, Sydney, NSW. Patricia K Correll, BN, MPH, Research Assistant; Matthew G Law, MA, MSc, Statistician; Ann M McDonald, BSc, MPH, Senior Research Assistant; David A Cooper, DSc, MD, FRACP, Director; John M Kaldor, PhD, Deputy Director. Reprints will not be available from the authors. Correspondence: Patricia Correll, National Centre in HIV Epidemiology and Clinical Research, Level 2, 376 Victoria Street, Sydney, NSW 2010. Email: pcorrellATnchecr.unsw.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Patricia K Correll · Matthew G Law · Ann M McDonald · David A Cooper · John M Kaldor

Factors involved in presentation of older people with thick melanoma Pauline F Hanrahan, Peter Hersey and Catherine A D'Este MJA 1998; 169: 410-414 For editorial comment, see Kelly Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Oncology - ©MJA1998 Abstract Objective: To examine whether presentation of older people with thick melanoma is a result of the site and histological type or of their reduced ability to identify melanoma. Design and setting: Retrospective analysis of the Newcastle Melanoma Unit patient database. Patients: 2154 patients with melanoma for whom complete data (histological type, thickness and site of melanoma) were available and who presented from February 1981 to April 1997. Main outcome measures: Histological type and site of melanoma in older (≥50 years) versus younger men and women; frequency with which these groups identified melanoma and the first changes of melanoma that were noticed. Results: Patients aged ≥50 years, particularly men, were more likely to present with thicker lesions. Older men and, to a lesser extent, older women were more likely to present with nodular melanoma, which were more frequent on the scalp and face in older, compared with younger, men, and scalp and back in older men compared with older women. Failure to identify melanoma was associated with older age, sites on the scalp and back and histological type of the lesion, but was independent of sex once histological type, age, site and thickness were taken into account. Multivariate analysis indicated that the association of older age of patients with failure to identify melanoma applied irrespective of the site, type and thickness of their melanoma. Conclusion:. The higher frequency of thick melanoma in older people is accounted for by an increased proportion of nodular melanoma and decreased ability to recognise the changes of melanoma. These findings have important implications for examination of older patients by doctors and for design of strategies for early detection of melanoma in this age group. Introduction Survival from melanoma is related to thickness of the tumour. Ninety-six per cent of patients whose lesions are detected early (< 0.75 mm thick) will survive to 10 years,1,2 but this decreases to 50% for patients with lesions 4 mm or greater.3,4 Previous studies have shown that most patients who present with thick lesions are men aged over 50 years.5-7 In New South Wales over 50% of deaths from melanoma involve men over 50, even though this group comprises only 12%-14% of the population.8 Clearly, strategies that facilitate early detection of melanoma in older men may markedly reduce the death rate from this cancer. We have previously investigated the ability of older men to distinguish between benign and malignant pigmented skin lesions in photographs and whether an educational brochure assists this ability;9 the brochure increased knowledge of melanoma, but did not help participants distinguish between benign and malignant pigmented skin lesions. We then investigated whether this age group had inherent difficulties in recognising the changes of melanoma and found that all age groups had difficulty in recognising the appearance of new pigmented lesions and changes in lesions over time.10,11 Older people were no less able than younger people to identify change, but falsely identified changes in unchanged lesions more frequently. In this study we sought to examine factors associated with failure of patients to identify their melanoma and the importance of this in presentation of older patients with thick melanoma. Methods Information for the study was obtained from the database in the Newcastle Melanoma Unit (NMU), which was established in 1981 and is a referral centre for patients from the Hunter region of New South Wales. Over 90% of patients with melanoma in the Hunter region attend the unit. The database contains information obtained by experienced oncology nurses who interview patients, asking questions about changes they had noticed leading to diagnosis of melanoma, for how long they had noticed the changes and whether these were in a pre-existing or new lesion. The age and sex of patients, site of their primary lesion, its histological type and thickness were also recorded. Histological investigation of the lesions was reviewed by at least two pathologists and reported as described previously.12 Patient clinical data were entered into a program developed jointly by the Sydney Melanoma Unit and the Macquarie University using Clinical Reporting Systems software (Brisbane Road, Castle Hill, NSW). We included data for all patients presenting to the NMU between February 1981 and April 1997. The data were categorised for statistical analysis as follows. Patients were divided into two groups by age (< 50 and ≥50). As in a previous study,5 thickness of primary lesions was categorised as < 3 mm, or ≥3 mm. Data on the site of the primary melanoma were divided into seven categories (scalp, face, neck, back, anterior trunk, arm and leg). Melanomas were categorised into seven histological types (superficial spreading melanoma [SSM], nodular melanoma [NM], lentigo maligna [LM], in-situ melanoma, desmoplastic melanoma, unknown, and unclassified) of which we included only the first three relatively common types. Categories for types of changes were: unidentified (including instances where patients were uncertain whether a change had occurred, or said that no change had occurred, but melanoma was detected by their general practitioners or other doctors); size and shape; colour; changes in sensation (eg, itching, tingling); and bleeding. Statistical analysis Data were analysed with SAS13 and STATA14 statistical software. Lesion characteristics (thickness, histological type and site of melanoma) were compared by age and by sex separately using chi-squared analyses. Lesion characteristics and age and sex were also compared for those who had, and had not, identified any changes. Logistic regression analyses were performed to examine factors associated with identification of any changes of melanoma, using any change versus no change, as the outcome variable. All five independent variables were tested for significance by forward stepwise regression analysis. After the final main effects model was obtained, all possible interactions were tested. The likelihood ratio statistic was used to assess statistical significance in the logistic regression analyses. We also tested (by chi-squared analysis) lesion characteristics, age and sex for association with type of changes noticed; 95% confidence intervals (CI) were calculated for each type of change. A significance level of 0.05 was used for all analyses. Subgroups were considered to differ significantly if 95% CIs did not overlap. Results The database included records for 3435 patients who attended the NMU during the study period. Exclusions included: 932 patients for whom there was incomplete data or for whom data could not be appropriately classified; 325 with uncommon, unknown, desmoplastic or unclassified histological types of melanoma; and 22 with melanoma at unusual sites (eg, mucosal regions, buttocks, groin). The final sample of 2154 patients comprised 1144 men (53%) and 1010 women (47%); 1142 patients (53%) were aged 50 years or over and 59% of the men in the study were in this age group, compared with 46% of the women. There were no significant differences in sex or age of patients excluded compared with those included (sex: chi-squared = 0.006, df = 1, P = 0.94; age: chi-squared = 1.7, df = 1, P = 0.18). The site and thickness of melanomas were similar in those excluded and included. Excluded patients had a different distribution of histological types (as this was part of the basis for their exclusion) and a higher proportion of unnoticed changes than those included. More detailed information about excluded patients is available from the authors. Lesion characteristics Thickness: Box 1 shows that patients aged 50 years or over (14.5%; 95% CI, 12%-16%) were significantly more likely to present with thick lesions than younger patients (4.7%; 95% CI, 3%-6%), and men (12%; 95% CI, 10%-14%) were more likely to have thick lesions than women (7%; 95% CI, 6%-9%). Site: Melanoma was found most often on the backs of male patients (47%; 95% CI, 45%-50% v. 24%; 95% CI, 22%-27% for women), whereas the leg was the predominant site in women (38%; 95% CI, 35%-41% v. 16%; 95% CI, 13%-18% for men). The incidence of melanoma on the scalp, face and neck was similar for men and women. Melanoma on the arms was more common in women (19%; 95% CI, 16%-21% v. 12%; 95% CI, 10%-13% for men), while melanoma on the face was more common in older patients and melanoma on the anterior trunk and leg was more common in younger men than older men. Melanomas were more common on the scalp and face in older, compared with younger, men and more common on the scalp and back of older men compared with older women (Box 1). Histological type: SSM was more common in younger patients of both sexes and NM was more common in older male patients. Most thin lesions (80%) were SSM and most thick lesions (72%) were NM. Thicknesses of the three histological types were: SSM, 0.02-10.5 mm (median, 0.63 mm; interquartile range [IQR], 0.4-1.0 mm); NM, 0.2-25 mm (median, 2.46 mm; IQR, 1.4-3.8 mm); and LM, 0.1-8.5 mm (median, 0.4 mm; IQR, 0.25-0.8 mm). Factors associated with failure to identify melanoma Many melanomas were not detected by patients, but were identified by their doctors. Sex and age: Men were more likely to fail to notice changes of melanoma than women (37%; 95% CI, 35%-49% v. 30%; 95% CI, 28%-32%) as were older, compared with younger, people (38%; 95% CI, 35%-41% v. 28%; 95% CI, 25%-31%). As shown in Box 2, a greater proportion of older men failed to identify changes in their lesions than younger men; the same trend was evident in women. Site: Patients most frequently failed to identify changes in lesions on the scalp (42%; 95% CI, 30%-52%) and back (41%; 95% CI, 37%-45%), but even among patients with a lesion on the face, 31% (95% CI, 26%-33%) did not identify it as melanoma. The anterior trunk was the least common site for unidentified change (24%; 95% CI, 20%-28%). Older patients failed to identify changes more often than younger patients for all sites except the face and scalp, where the reverse was the case (Box 2). However, numbers are very small for these groups and results should be interpreted with caution. Histological type: A greater proportion of patients with LM (47%; 95% CI, 40%-55%) failed to identify their melanoma than patients with SSM (33%; 95% CI, 31%-35%) or NM (29%; 95% CI, 25%-33%), although by the time they were detected the NM were much thicker. Older people were less likely than younger people to identify melanoma of all three histological types. Regression analyses: Age, thickness, histological type and site of lesion were significantly associated with identifying changes (Box 3). After these variables were included in the model, sex was no longer statistically significant. After adjustment for other variables in the model, the odds of older people identifying changes was two-thirds the odds of younger people identifying changes, and the odds of those with thicker lesions identifying changes was one-and-a-half times the odds of those with thinner lesions identifying changes. Those with LM had about half the odds of identifying changes as those with NM. As shown in Box 3, those with lesions on the face, arm, anterior trunk and leg had significantly higher odds of identifying changes than those with lesions on the back (reference group). Associations with thick melanoma (≥3 mm) As shown in Box 3, age and nodular melanoma were strongly associated with presentation with thick lesions. Changes of melanoma were more frequently identified in thick lesions, but as shown in Box 2, older people were less able to identify melanoma, irrespective of its thickness. It was also of interest that once the histological type of the lesion was taken into account, the site of the melanoma and sex of the patient were no longer related to thickness (ie, thick melanoma on the back and scalp in men appeared to result from the higher frequency of NM at these sites in men; see Box 3). Types of changes reported by patients Significantly fewer older than younger men reported changes in size and shape (21% v. 30%). The same was true for colour (20% v. 27%), and twice as many older men stated that the first change they were aware of in their skin lesion was bleeding (7% v. 3%). However, these differences were not statistically significant. Similarly, our findings that more younger than older women reported colour changes as the first change noticed (32% v. 25%) and (in contrast to men) older women reported changes in size and shape as the first change as frequently as younger women (30% v. 28%), were also not statistically significant. As shown in Box 4, patients with NM reported changes in colour significantly less frequently than those with SSM, change in sensation significantly more frequently than those with LM, and bleeding significantly more frequently than those with LM and SSM. Changes in size and shape were the most frequent changes reported in thick melanoma (34%), whereas changes in colour were more frequently reported in thin lesions (27%). Bleeding was reported as the first change significantly more frequently in thick than thin lesions (17% v. 4%). Reported changes in sensation were similar for thick or thin lesions. Discussion We excluded approximately one-third of the patients on the NMU database, mainly because of missing information about changes noticed or thickness measurements, or because they had melanoma of unusual histological type or location. However, the excluded patients appeared to have similar age and sex distribution, site and thickness measurements to patients in the study. Nonetheless, our results may not be generalisable to patients with the less common histological types of melanoma. Forty per cent of men and 34% of women aged over 50 failed to identify their melanoma. In contrast, younger patients of both sexes reported changes of melanoma in over 70% of cases. These findings were consistent with those of Koh et al15 who found that 38% of melanomas were not identified by men over the age of 60, compared with 23% in men less than 40 years of age. We also found that failure to detect melanoma was not only related to the age of the patient, but was associated with the site and type of the melanoma. Firstly, older men had lesions at sites which are more difficult to see, such as the back and scalp, with nearly 50% of melanoma in older men (compared with only 20% in older women) occurring on the back. Other series have reported similar differences in site distribution of melanoma between the sexes.16,17 Secondly, compared with younger people, older people, particularly men, had a higher proportion of NM, similar to that reported elsewhere.5,18 Patients also identified NMs when they were much thicker than SSMs (median, 2.46 mm v. 0.63 mm). The higher frequency of bleeding as the first change noted in patients with NM was consistent with this finding.19 The higher proportion of NM in older people and more frequent occurrence in "hard-to-see" sites could solely explain the presentation of older people with thicker melanomas. However, regression analysis indicated that for any given thickness and type of melanoma at a given site, older people were less able to recognise melanoma than younger people. In our examination of the first change that drew the attention of patients to their melanoma, we found that self-reporting of a change in colour was infrequent in patients with thick melanoma, and in patients with NM relative to those with SSM. These results were surprising, as previous studies have shown that changes in colour were one of the most frequent early changes of melanoma.20,21 Most of the thick melanoma were NM, which tend to be relatively homogeneous in colour compared with SSM. Thus, the different patterns of colour between NM and SSM may need to be emphasised in educational material directed at older patients. Bleeding was clearly associated with thick and nodular melanoma, but this is a late rather than an early sign.19 To the best of our knowledge this is the first study that has sought to identify reasons why older people present with thicker melanoma. The finding that older patients had lower ability to detect melanoma than younger people is consistent with our previous findings that older people had difficulty in discriminating early changes of melanoma in photographs of pigmented lesions.10,11 Previous studies have noted that younger people were more likely to rate change in skin lesions as extremely important in diagnosis of melanoma than older respondents.22 Older people may also not inspect their skin as frequently as younger people and this may contribute to their lower rate of detection of melanoma. Our findings raise questions about the likely effectiveness for older people of public health strategies that rely on self-detection of melanoma. It seems that that greater reliance will need to be placed on detection of melanoma in this age group by health professionals, who may need to be made aware of the high frequency of thick melanoma and characteristics of melanoma in this age group. Our previous studies have shown that recognition of change in existing lesions over time or the appearance of new pigmented skin lesions can be very difficult, but is assisted by whole-body photographs.10,11 The feasibility of using whole-body photographs to help general practitioners detect melanoma is currently under evaluation in a project funded by the National Health and Medical Research Council. Acknowledgements We thank the nursing sisters Sue Collins, Debbie Bradley, Donna Owens and Kathy Hall for collection of the data and Vicky Hunter for data entry. Dr Hanrahan was supported in part by the Hunter Melanoma Foundation. References Buttner P, Garbe C, Bertz J, et al. Primary cutaneous melanoma: optimized cutoff points of tumor thickness and importance of Clarks level for prognostic classification. Cancer 1995; 75: 2499-2506. Thorn M, Ponten F, Bergstrom R, et al. Clinical and histopathologic predictors of survival of patients with malignant melanoma: a population based study in Sweden. J Natl Cancer Inst 1994; 86: 761-769. Balch CM, Soong SJ, Shaw HM, et al. An analysis of prognostic factors in 8500 patients with cutaneous melanoma. In: Balch CM, Houghton AN, Milton GW, et al, editors. Cutaneous melanoma. Philadelphia: JB Lippincott Co.; 1992: 165-185. Breslow A. Tumor thickness, level of invasion and node dissection in stage 1 cutaneous melanoma. Ann Surg 1975; 182: 572-578. Hersey P, Sillar R, Howe CG, et al. Factors related to the presentation of patients with thick primary melanomas. Med J Aust 1991; 154: 583-587. McHenry PM, Hole DJ, Mackie RM. Melanoma in people aged over 65 in Scotland, 1979-89. BMJ 1992; 304: 746-749. Cohen HJ, Cox E, Manton K, Woodbury M. Malignant melanoma in the elderly. J Clin Oncol 1987; 5: 100-106. Coates M, Day P, McCredie M, Taylor R. NSW cancer incidence and mortality 1992. In: NSW Central Cancer Registry. Sydney: NSW Cancer Council; June 1995: 87, 97. Hanrahan P, Hersey P, Watson AB, Callaghan TM. The effect of an educational brochure on knowledge and early detection of melanoma. Aust J Public Health 1995; 19: 270-274. Hanrahan P, Hersey P, Menzies SW, et al. Examination of the ability of older people to identify early changes of melanoma in computer-altered pigmented skin lesions. Arch Dermatol 1997; 133: 301-311. Hanrahan P. Early detection of melanoma in older people [PhD thesis]. Sydney: University of NSW, 1998. McGovern VJ, Cochran AJ, Van Der Esch EP, et al. The classification of malignant melanoma, its histological reporting and registration: a revision of the 1972 Sydney classification. Pathology 1986; 18: 12-21. SAS for PC [computer program]. Version 6.11. Cary, NC: SAS Institute Inc., 1995. STATA [computer program]. Version 5.0. Houston, Tx: Stata Corp, 1997. Koh HK, Miller DR, Geler AC, et al. Who discovers melanoma? J Am Acad Dermatol 1992; 26: 914-919. Nguyen HL, Armstrong B, Coates M. Cutaneous melanoma in NSW 1983-1995. In: NSW Central Cancer Registry. Sydney: NSW Cancer Council; June 1997: 40. Green A, MacLennan R, Youl P, Martin N. Site distribution of cutaneous melanoma in Queensland. Int J Cancer 1993; 53: 232-236. Nguyen HL, Armstrong B, Coates M. Cutaneous melanoma in NSW 1983-1995. In: NSW Central Cancer Registry. Sydney: NSW Cancer Council; June 1997: 38. Fitzpatrick TB, Milton GW, Balch CM, et al. Clinical characteristics of primary melanoma. In: Balch CM, Houghton AN, Milton GW, et al, eds. Cutaneous melanoma. 2nd ed. Philadelphia: JB Lippincott Co.; 1992: 225-226. Mackie RM. Clinical recognition of early invasive melanoma. BMJ 1990; 301: 1005-1006. McGovern TW, Litaker MS. Clinical predictors of malignant pigmented lesions. J Dermatol Surg Oncol 1992; 18: 22-26. Baade PD, Balanda KP, Stanton WR, et al. Community perceptions about the important signs of early melanoma. J Am Acad Dermatol 1997; 36: 33-39. (Received 1 Sep 1997, accepted 14 May 1998) Authors' details John Hunter Hospital, Newcastle, NSW. Pauline F Hanrahan, BA, PhD, Research Officer, Newcastle Melanoma Unit and Division of Surgery; Peter Hersey, FRACP, D. Phil, Conjoint Professor of Oncology, Oncology and Immunology Unit, Division of Surgery. Centre for Clinical Epidemiology and Biostatistics, University of Newcastle, Newcastle, NSW. Catherine A D'Este, BMath, PhD, Lecturer in Biostatistics, Faculty of Medicine and Health Sciences. Reprints will not be available from the authors. Correspondence: Dr P Hersey, Room 443, David Maddison Clinical Sciences Building, Cnr King and Watt Streets, Newcastle, NSW 2300. - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Pauline F Hanrahan · Peter Hersey · Catherine A D'Este

Doctors detected self-administering opioids in New South Wales, 1985-1994: characteristics and outcomes Megan Cadman and James Bell MJA 1998; 169: 419-421 For editorial comment, see Breen & Court Abstract - Introduction - Methods - Results - Discussion - Acknowledgement - References - Authors' details - - More articles on Drugs and alcohol - ©MJA1998 Abstract Objective: To describe the characteristics and outcomes of doctors whose drug authorities were withdrawn as a result of self-administering opioids for non-medical purposes. Design: Retrospective review of New South Wales Health Department information relating to all doctors whose authorities to possess, supply, prescribe or administer drugs of addiction had been withdrawn in the period 1985 to 1994 as a result of confirmed self-administration of opioids. Outcome measures: Age, sex, geographical location and practice category at the time of intervention; drugs used; period of opioid use before authority withdrawal; means of detection; and registration status as at August 1995. Results: From 1985 to 1994, 79 doctors had their drug authorities withdrawn (0.4% of the NSW medical profession in 1994). The groups significantly over-represented were general practitioners and those aged 30-39 years. Pethidine was the main drug used (66 doctors; 84%). Drug use for more than two years before detection was reported by 34 (43%) doctors. Community pharmacists were the source of reports leading to detection of 28 (35%) doctors. As at August 1995, 27 (34%) of the study group were not practising; 10 (13%) had died. Conclusion: Outcomes for these doctors were poor. There was substantial attrition from practice and a high mortality rate. Introduction Medical practitioners have been shown to have a higher rate of drug abuse, mental illness and suicide than either matched controls or the general population.1-4 It is claimed that in the United States up to 10% of physicians at some time during their careers become dependent on psychoactive drugs or alcohol to an extent where their practice is impaired; about 1% become opioid-dependent.5 For these doctors, if placed in highly structured and strictly monitored programs, treatment outcomes are substantially better than for other opioid-dependent people.6,7Few data have been published on the prevalence of drug misuse among Australian healthcare professionals, and even less on the problems experienced by drug-dependent doctors and the extent of their professional impairment. Two studies of doctors with serious drug dependency brought to the attention of the Victorian Medical Board, in 1964-1984 and 1984-1990, indicated a prevalence of about 0.5%.8,9 By comparison, in a recent anonymous survey of 1125 doctors in New South Wales, 1% of those surveyed reported drug abuse problems, a figure closer to estimates from other countries.10 Accurate reporting of the prevalence of drug misuse in healthcare professionals is often undermined by the tendency of colleagues to either protect or ignore the impaired individual, and by treating doctors often anxious to avoid "labelling" a colleague with a potentially damaging diagnosis. Our study uses official records to examine the characteristics, and the limited information about outcomes, of those doctors in NSW whose opioid use had triggered official intervention. Methods Data collection The study group comprised all doctors whose authorities to possess, supply, prescribe or administer drugs of addiction (Schedule 8 authorities) were withdrawn by the NSW Health Department as a result of confirmed self-administration of opioids in the period 1985 to 1994, inclusive. The data for this study were obtained from files held by the Pharmaceutical Services Branch of the NSW Health Department. Information relating to sex, geographical location and main type of practice or specialty at the time of intervention, use of other drugs, period of opioid use before authority withdrawal, means of detection, and reports of psychiatric consultations, was collected from these files. Data relating to date of birth and registration status as at August 1995 were obtained from the NSW Medical Board, and data on the age, sex, specialty and geographical location of all medical practitioners in NSW were obtained from the Workforce Planning Unit of the NSW Health Department. Analysis To determine whether any groups were over-represented in the study group, demographic characteristics of the subjects were compared with those of the profession as a whole in NSW. Where available, data for 1990 were used in comparisons, as this represented a mid-point in the study period. For the number of medical practitioners and general practitioners in NSW, 1994 data were used. Confidence intervals and the χ2 test were used in the analysis of the data. Ethical approval The study was approved by the New South Wales Health Department. Strict confidentiality was maintained throughout the study. Results In the period 1985 to 1994, 79 doctors relinquished their Schedule 8 authorities as a result of confirmed self-administration of opioids. This represents 0.4% of the medical profession in NSW in 1994.11 Demographic characteristics The main characteristics of doctors in the study group at the time their Schedule 8 authorities were withdrawn are shown in Table 1. Ages of doctors in the study group at the time of authority withdrawal ranged from 26 to 69 years (mean, 39 years; 95% confidence interval, 36.8-41.2 years). Fifty-six per cent of the group were aged 30 to 39 years, a significant over-representation of this age group (Table 1). Rural practitioners were significantly over-represented in the group according to the chi-squared test (chi-squared, 4.67; P = 0.031) (Table 1). Four practice categories were identified in the study group: general practice (64; 81%), anaesthetics (3; 4%), psychiatry (3; 4%), and hospital salaried staff (9; 11%). Compared with the profession as a whole for 1994, general practitioners were significantly over-represented; anaesthetists and psychiatrists were not. Comparative data for hospital salaried staff in NSW were not available. Five of the nine doctors employed in salaried positions in hospitals were working in emergency departments. Drugs used Pethidine was the main drug used (66 doctors; 84%), with 31 (39%) apparently using nothing else. Commonly, doctors initially obtained pethidine, and to a lesser extent morphine, from their doctor's bag emergency supply, with prescriptions and surgery or hospital stock accessed as drug use escalated. Other drugs used included mixed opioids, barbiturates, ketamine and benzodiazepines. A concurrent alcohol problem was reported by 13 (16%) doctors, and eight (10%) admitted illicit drug use. All of the illicit drug users were men aged 30 to 43 years. Period of drug use Thirty-four (43%) doctors reported using opioids for more than two years before intervention, with four individuals admitting drug use for more than 10 years (Table 2). Two of these four also reported illicit drug use dating from their late teens. Means of detection Community pharmacists were the primary source of reports (Table 3), often alerted by a large number of opioid prescriptions written by a particular doctor. Common scenarios included prescriptions written for patients being either picked up personally from the pharmacy by the doctor, or requested to be delivered to the doctor's surgery, or patients being asked to collect opioids from the pharmacy and take them to the doctor's surgery, where one or two ampoules were administered to the patient and the balance retained by the doctor for personal use. Other reporting sources were routine inspections of community pharmacy records by departmental officers, and reports by colleagues and by the doctors themselves. Self-reports occurred in 14 cases (18%); most of these were due to the prompting of a treating psychiatrist or therapist as part of a rehabilitation program. Thirty-one doctors (39%) denied self-administration at first contact by the department. Of this group, 15 stated that the drugs were for use by others, usually relatives; 11 denied self-administration, with no other explanation offered; and five insisted that the drugs were used to treat a medical condition. At subsequent interviews, all admitted to self-administration for non-medical purposes and relinquished their drug authorities. Morbidity and mortality Twenty-eight (35%) doctors reported having consulted a psychiatrist before initial contact by the department. Twelve of these had consulted a psychiatrist for a drug abuse problem, the other 16 had consulted for non-drug related problems. Eight individuals in the study group reported that they had taken antidepressants. Ten (13%) of the doctors (nine men) are known to have died in the study period. Three of these are known to have committed suicide, one died of an overdose, two died in motor vehicle accidents, one died of a heart attack, and the cause of death of three individuals is not known. There were also five known suicide attempts by subjects during the study period. Registration status Only 54 (68%) of the subjects remained on the medical register as at August 1995 (Table 4). Two of these, although still on the register, were known to have left the profession. Including these two individuals, there was an overall loss from the medical profession in NSW of 27 (34%) individuals. Twenty-two doctors had continuing conditions placed on their registration by the Medical Board. These included, in addition to the continued withdrawal of the drug authority, requirements such as ongoing treatment by a psychiatrist, regular assessment and monitoring including urinalysis, supervised employment, and regular Board review, in line with the Board's Impairment Program protocols. Discussion From 1985 to 1994, 79 doctors had their authorities to possess, supply, prescribe or administer drugs of addiction withdrawn as a result of confirmed self-administration of opioids. Pethidine was the main drug used, and 43% of the doctors had been self-administering drugs for more than two years before detection. There are several limitations to the data in this study. The data were obtained from a retrospective review of file notes, which had been compiled by different interviewers, and did not follow a systematic data collection format. Data on several variables, including the extent of illicit drug use and period of drug use, relied on self-reporting and therefore introduced a "recall bias", either deliberate or inadvertent. A number of confounders may have influenced detection rates during the study period. These include the frequency and areas of prescription monitoring by the department, and changes in awareness and/or diligence of the reporting groups, such as community pharmacists and medical colleagues, who may have been influenced by articles in professional journals or by mailings by the Doctors' Health Advisory Service, as occurred in 1988. These factors may have varied between regions and over time. Another limitation of the study is that the subjects represent a subset of all drug-using doctors -- those who were detected. The data indicate that some doctors can self-administer drugs for prolonged periods without detection, and it is quite likely that sufficiently careful medical practitioners may self-administer opioids regularly and never be detected. Few doctors in the study sample surrendered their Schedule 8 drug authorities on their own instigation. Since detection of a problem was often based on observations by others, particularly community pharmacists, this could introduce systematic bias. Geographical isolation has been identified as a "marker" for substance abuse.12 However, the apparent correlation between rural practice and drug use may simply indicate that a doctor with a drug problem has a greatly increased chance of detection in a rural town than in the relative anonymity of the city. Among doctors detected self-administering opioids there is considerable attrition from medical practice. Of the 79 doctors in the sample, there was a loss of 27 individuals from the profession, with 10 of these known to have died in the study period. In recent years, the NSW Medical Board has adopted policies aimed at rehabilitating doctors who misuse drugs, while maintaining them in practice wherever possible. These policies include placing conditions on registration, such as supervised employment, restrictions on possessing and prescribing drugs of addiction, attendance for treatment and assessment, urine testing, and regular review by the Board. After a period of such supervision, practitioners who comply satisfactorily may return to full registration.11 There are no published data on the outcomes of these interventions. Attrition from the profession is taken as a poor outcome, but this may be an erroneous assumption. One hypothesis concerning medical practitioners who self-administer opioids is that they are ambivalent about medical practice,13 and this was confirmed by four subjects, two of whom had sought psychiatric treatment. It may be, therefore, that leaving practice was an appropriate decision for some individuals. The misuse of opioids described in this study should be seen within the wider context of the overall medical and psychological needs of medical practitioners. Acknowledgement The authors acknowledge the assistance of Pia Salmelainen, Policy Analyst (Research Officer), Pharmaceutical Services Branch, NSW Health. References Ball JRB. Alcohol and drug use and related problems in the medical profession. Aust Drug Alcohol Rev 1986; 5: 29-32. Murray RM. Psychiatric illness in male doctors and controls: an analysis of Scottish hospitals inpatient data. Br J Psych 1977; 131: 1-10. Preven DW. Physician suicide. In: Scheiber SB, Doyle BB, editors. The impaired physician. New York and London: Plenum Press, 1983; 39-47. Schlicht SM, Gordon IR, Ball JRB, Christies DGS. Suicide and related deaths in Victorian doctors. Med J Aust 1990; 153: 518-521. Webster TG. Problems of drug addiction and alcoholism among physicians. In: Scheiber SB, Doyle BB, editors. The impaired physician. New York and London: Plenum Press, 1983; 27-38. Centrelia M. Physician addiction and impairment -- current thinking: a review. J Addictive Diseases 1994; 13: 91-105. Morse RM, Martin MA, Swenson WM, Niven RG. Prognosis of physicians treated for alcoholism and drug dependence. JAMA 1984; 251: 743-746. Serry N, Ball JRB, Bloch S. Substance abuse among medical practitioners. Drug Alcohol Rev 1991; 10: 331-338. Serry N, Bloch S, Ball R, Anderson K. Drug and alcohol abuse by doctors. Med J Aust 1994; 160: 402-407. Pullen D, Cait EL, Lyle DM, et al. Medical care of doctors. Med J Aust 1995: 162: 481-484. New South Wales Medical Board. Annual report for the period ending 31 March 1994. Sydney: NSW Medical Board, 1994. Talbott GD, Gallegos KV, Wilson PO, Porter TL. The Medical Association of Georgia's impaired physician program, review of the first 1000 physicians: analysis of specialty. JAMA 1987; 257: 2927-2930. Modlin HC, Montes A. Narcotics addiction in physicians. Am J Psych 1964; 121: 358-363. (Received 11 Nov 1997, accepted 23 Jun 1998) Authors' details Pharmaceutical Services Branch, NSW Health, Gladesville, NSW. Megan Cadman, BPharm, MPH, Acting Deputy Chief Pharmacist. The Langton Centre, Surry Hills, NSW. James Bell, BA, FRACP, Director. Reprints will not be available from the authors. Correspondence: Ms M Cadman, Pharmaceutical Services Branch, NSW Health, PO Box 103, Gladesville, NSW 1675. E-mail: mcadmATdoh.health.nsw.gov.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Megan Cadman · James Bell

The prevalence of hepatitis C in patients admitted with acute hepatitis to Fairfield Infectious Diseases Hospital, 1971-1975 Jennifer A Thomson, Alison J Rodger, Sandra C Thompson, Damien Jolley, Amanda Byrne, Susan J Best and Nick Crofts MJA 1998; 169: 360-363 Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - More articles on Infectious diseases and parasitology - ©MJA1998 Abstract Objective: To identify and determine trends in the prevalence of hepatitis C virus (HCV) antibody in stored sera from 1971 to 1975 and to determine associations with HCV seropositivity, including markers for other hepatitis infections and possible routes of transmission. Design: A retrospective cross-sectional study. Patients and setting: 1511 adults admitted to Fairfield Infectious Diseases Hospital, Victoria, with a clinical and biochemical diagnosis of hepatitis between 1 January 1971 and 31 December 1975. Main outcome measures: Prevalence over study period of hepatitis A virus antibody (anti-HAV) IgM, hepatitis B core antibody (anti-HBc), hepatitis B surface antigen (HBsAg) and hepatitis C virus antibody (anti-HCV) in stored sera; sociodemographic data and risk factors for blood-borne viruses documented in original medical records. Results: Anti-HCV was detected in 17% of adults admitted with hepatitis from 1971 through 1975. Prevalence increased significantly over this period. Most cases were in young men who had a history of injecting drug use. HCV seropositivity was also significantly associated with markers for hepatitis B infection. Conclusions: Given the 20-30-year period between infection with hepatitis and the development of liver disease, our findings predict significant liver-related morbidity in Australia in the next decade. The increase in prevalence over the five years studied suggests rapid spread of HCV through susceptible populations, principally injecting drug users. Introduction Hepatitis C is an important public health problem in Australia. Approximately 150 000 people in this country are currently infected with the virus, predominantly as a result of injecting drug use.1,2 Acute hepatitis C virus (HCV) infection is generally benign; less than 25% are estimated to be icteric.3 The main significance of HCV infection is that it is strongly associated with the development of chronic liver disease. After 20 to 30 years a significant proportion of those infected will have chronic hepatitis, cirrhosis, liver failure and primary hepatocellular carcinoma.4 In view of this 20-30-year latency period, estimating the prevalence of HCV in the Australian community during the early 1970s may give an indication of the morbidity to be expected from this infection over the next decade. We thus aimed to determine the prevalence of hepatitis C antibody in stored sera from people with acute viral hepatitis admitted to Fairfield Infectious Diseases Hospital (FIDH), Victoria, from 1971 to 1975, to identify possible associations with HCV seropositivity and to detect trends in HCV prevalence over that time. Methods Ethical approval for the study was obtained from the ethics committees at FIDH and the Australian Institute of Health and Welfare. We chose the period 1971 to 1975 because systematic storage of sera began in 1971, the same year diagnostic coding was standardised at FIDH (with the introduction of International classification of diseases, ninth revision coding), and because a previous study had identified HCV antibody in stored sera from that time.5We included all patients admitted to FIDH with biochemical and clinical evidence of hepatitis between 1 January 1971 and 31 December 1975. Those for whom there was no original serum sample for testing and those for whom medical records could not be found were excluded. Patients aged under 16 years at the time of admission were also excluded as they were generally admitted for hepatitis A infection, which was endemic at that time.6 Multiple serum samples had been stored for many patients. For each individual, we tested the last available serum sample associated with their final hepatitis admission and which was adequate for testing. Sera were tested for all hepatitis markers at the Victorian Infectious Diseases Reference Laboratory and the National Serology Reference Laboratory, Australia by the following commercially available immunoassays: hepatitis A virus antibody (anti-HAV) IgM, IMX assay and HAVAB M enzyme immunoassay; hepatitis B core antibody (HBcAb), CORAB radioimmunoassay; hepatitis B surface antigen (HBsAg), AUSRIA II radioimmunoassay; and hepatitis C antibody (HCVAb), second generation enzyme immunoassay (all assays manufactured by Abbott Diagnostics, Abbott Park, Illinois, USA). Results were designated reactive or non-reactive by the laboratories. As previous studies of frozen sera stored for long periods suggested a stringent classification was required to avoid overestimating the prevalence of antibody to HCV,7,8 we used twice the manufacturer's recommended cutoff (as recommended in the literature7) to further increase the specificity of the HCV antibody test. Individuals with weakly positive results (ie, a ratio of 1-2 of sample optical density determined by enzyme-linked immunosorbent assay [ELISA] to cutoff optical density) were recorded as equivocal and excluded to minimise misclassification bias. Medical records were located and information collected on sociodemographics and risk factors for hepatitis. For patients with multiple admissions over the study period, information on risk factors was summarised from all admissions preceding the date of the last available serum sample. We compared our test results for the stored serum samples with the original results recorded for HBsAg at the time the samples were taken using K 9 to measure agreement. In addition, we traced a subset of patients in 1996 and 1997 and obtained information on risk factors before their FIDH admission. Traced patients were also retested for hepatitis B and C markers in 1996 and 1997. The results of tests on their original stored serum samples were compared with these follow-up results for hepatitis C virus antibody (anti-HCV) and hepatitis B core antibody (anti-HBc). Statistical analysis We used the Statistical Package for Social Sciences (SPSS)10 for data analysis, which included descriptive statistics, unpaired t test, chi-squared test, Fisher's exact test, chi-squared test for trend, and measures of risk and agreement. Results Patients aged over 16 and admitted to FIDH with hepatitis during the study period numbered 1798. Medical records were found for 1737 (97%) and 5% were noted to have been admitted with hepatitis more than once over the study period. Serum specimens were found for 1559 of these patients (90%) and testing for all the measured hepatitis markers was completed for 1511 (87%). We were able to trace a subset of 161 patients for follow-up in 1996 and 1997. The distribution of optical density ratios indicating anti-HCV status for the 1511 patients in the study is shown in Box 1. Seven per cent (99) had equivocal anti-HCV results and were excluded. Of the remaining 1412, 17% (238) were anti-HCV positive. Only 15% (37) of those who were anti-HCV positive had no evidence of other hepatitis markers. Those who were anti-HCV positive were significantly more likely to have markers for hepatitis B virus (Box 2), and the sensitivity and specificity of anti-HBc as a surrogate marker for HCV infection were 77% and 57%, respectively. The trends in prevalence of each of the measured hepatitis markers were significant (P < 0.05). The prevalence of anti-HCV increased fourfold, with the biggest increase between 1974 and 1975. The prevalence of anti-HBc and HBsAg also increased, while the prevalence of anti-HAV IgM declined (Box 3). In patients who were HCV seropositive, the prevalence of markers of hepatitis A and B did not change significantly over the study period. Validation of serological tests For the 161 patients we traced, comparing the results of serological testing of their stored sera for anti-HBc and anti-HCV with those of follow-up tests in 1997 showed good overall agreement, with complete agreement for anti-HBc of 91% (K, 0.79; 95% CI, 0.67-0.91) and, for anti-HCV, of 88% (K, 0.74; 95% CI, 0.66-0.82). For all patients, there was moderate agreement between HBsAg results recorded at the time of original admission and our results for the stored sera, with complete agreement for 65% (K, 0.58; 95%CI, 0.43-0.72). Sociodemographic data and risk behaviours Sociodemographic data were available from medical case records for all patients. The 238 individuals who were positive for anti-HCV were significantly younger than those who were anti-HCV negative (mean age at time of original admission, 22 years; SD, 5.3 years v. 29.5 years; SD, 12 years; P < 0.001). Anti-HCV-positive individuals were also significantly more likely to have been born in Australia (80% v. 69%; RR, 1.5; 95% CI, 1.4-1.8; P < 0.001) and to be male (62% v. 52%; RR, 1.4; 95% CI, 1.1-1.9; P < 0.001). Information on risk factors was incompletely recorded in medical records. Only 40% had information recorded for injecting drug use, 69% for contact with someone with hepatitis, 36% for transfusion, 15% for tattooing and 11% for travel. A history of injecting drug use and of contact with someone with hepatitis were the only risk factors significantly associated with HCV seropositivity (Box 2). There was good agreement between original records of injecting drug use in case records on admission and follow-up information obtained from the 161 patients traced in 1996 and 1997 (k, 0.83; 95% CI, 0.74-0.92). Discussion Our findings confirm the presence of anti-HCV among adults admitted with acute hepatitis to FIDH in Victoria in the early 1970s and also that its prevalence increased markedly from 1971 through 1975. Most cases were in young men who had a history of injecting drug use. Most individuals with HCV appear to have been admitted because of subsequent infection with hepatitis A or B. While the actual proportion is unclear, it is likely to be significant as only 16% of HCV-seropositive individuals had no evidence of other acute hepatitis markers. As this implies that individuals with subclinical acute HCV infection were admitted and therefore included in the study, the HCV seropositive group would appear to be reasonably representative of community-acquired HCV infection in Melbourne in the 1970s. The significant association between the presence of HCV antibody and documented contact with a person with clinical hepatitis infection probably results from a spurious association with the hepatitis A or hepatitis B infection that precipitated admission. This also suggests that those infected with HCV were likely to have risk behaviours that exposed them to other hepatitis viruses, in particular to hepatitis B. Significant misclassification of hepatitis status appears unlikely as there was moderate to good agreement between original results for hepatitis A and B with our results. In addition, we used a more stringent cutoff to classify anti-HCV status, to increase specificity and reduce misclassification bias. Another source of potential bias in our study was loss of samples as a result of our exclusion criteria, but given the large sample size this should not have significantly affected our findings. The risk factor information in the original case records was also incomplete and may have been subject to bias. However, the results of follow-up assessment of risk factors in patients we were able to trace equate well with risk factor information documented at the time of original admissions, in particular the presence or absence of a history of injecting drug use. In addition, the association between injecting drug use and HCV seropositivity -- documented in many other studies in the 1990s1,2,11-14 -- was very clear. We did not identify the presence of tattoos and a history of blood transfusion as significant risk factors for HCV seropositivity, and this probably reflects the low prevalence of HCV in the general population in the 1970s. The significant increase in the prevalence of HCV seropositivity over the five-year study period suggests that infection became firmly established within the Australian community, particularly the injecting drug user population, as early as the mid 1970s. Most of the estimated 150 000 people in Australia currently infected with HCV are thought to have become infected in the past two decades, implying that the population of injecting drug users expanded dramatically in the late 1970s and early 1980s.1 Our finding of a twofold increase in the recording of injecting drug use among patients admitted with viral hepatitis from 1971 through 1975 supports this. Given the 20-30-year latency period before the development of liver-related complications, it is likely that the rapid increase in prevalence of HCV infection from the mid 1970s will translate into a rapid increase in HCV-positive individuals presenting with liver disease over the next decade. Health service planning and resource allocation should take this into account. Further research is now urgently required on the natural history of community-acquired, as opposed to transfusion-acquired, HCV infection in the Australian population and on factors that predict the development of hepatic sequelae. Acknowledgements We thank Anna Lanigan (MBCMR), Dr Alan Breschkin, Dr Mike Catton and Associate Professor Steven Locarnini (Victorian Infectious Diseases Reference Laboratory) and Dr Elizabeth M Dax (National Serology Reference Laboratory, Australia). The National Health and Medical Research Council (NHMRC) Public Health and Development Committee provided funding for the study, and Dr J Thomson was an NHMRC PHRDC Research Fellow. Dr Nick Crofts was funded from the Research Fund of the Macfarlane Burnet Centre and Dr Alison Rodger partly funded by the Victorian Health Promotion Foundation. References Crofts N, Jolley D, Kaldor J, et al. Epidemiology of hepatitis C virus infection among injecting drug users in Australia. J Epidemiol Community Health 1997; 51: 692-697. MacDonald M, Crofts N, Kaldor J. Transmission of hepatitis C virus rates, routes, and cofactors. Epidemiol Rev 1996; 18: 137-148. Alter HJ. To C or not to C? These are the questions. Blood 1995; 85: 1681-1695. Seeff L. Natural history of Hepatitis C. Hepatology 1997; 26: 21-28. Moaven L, Crofts N, Locarnini SA. Hepatitis C virus infection in Victorian injecting drug users in 1971. Med J Aust 1993; 158: 574. Lehmann NI, Gust ID. The prevalence of antibody to hepatitis A virus in two populations in Victoria. Med J Aust 1977; 2: 731-732. Lok AS, Ma OC, Chan TM, et al. Overestimation of the prevalence of antibody to hepatitis C virus in retrospective studies on stored sera. Hepatology 1991; 14: 756-762. Moaven L, Cunningham T. Hepatitis C serology. Aust Microbiologist 1994; 12: 321-323. Fleiss J. Statistical methods for rates and proportions. 2nd ed. London: Wiley; 1981. SPSS [computer program]. Version 8. Chicago, Ill.: SPSS Inc, 1998. Van Beek I, Buckley R, Stewart M, et al. Risk factors for hepatitis C virus infection among injecting drug users in Sydney. Genitourin Med 1994; 70: 321-324. Crofts N, Aitken, CK. Incidence of and risk behaviours for blood-borne viruses in a cohort of injecting drug users in Victoria, 1990-1995. Med J Aust 1997; 167: 17-20. Galeazzi B, Tufano A, Barbierato E, et al. Hepatitis C virus infection in Italian intravenous drug users: epidemiological and clinical aspects. Liver 1995; 15: 209-212. Van den Hoek JA, van Haastrecht HJ, Goudsmit J, et al. Prevalence, incidence, and risk factors of hepatitis C virus infection among drug users in Amsterdam. J Infect Dis 1990; 162: 823-826. (Received 3 Feb, accepted 3 Aug, 1998) Authors' details Epidemiology and Social Research Unit, The Macfarlane Burnet Centre for Medical Research, Melbourne, VIC. Jennifer A Thomson, PhD, FAFPHM, Research Fellow; Alison J Rodger, MRCP(UK), MFPHM(UK), Research Fellow; Sandra C Thompson, PhD, FAFPHM, Research Associate; Amanda Byrne, BSc(Hons), MSc, Research Assistant; Nick Crofts, MPH, FAFPHM, Head. Department of Public Health & Community Medicine, The University of Melbourne, Melbourne, VIC. Damien Jolley, MSc(Epidemiol), MSc(Stats), Senior Lecturer in Epidemiology and Biostatistics. National Serology Reference Laboratory, Australia, Melbourne, VIC. Susan J Best, DipMedTech, Senior Scientist. Reprints will not be available from the authors. Correspondence: Dr A J Rodger, The Macfarlane Burnet Centre for Medical Research, PO Box 254, Fairfield, VIC 3078. E-mail: rodgerATburnet.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Jennifer A Thomson · Alison J Rodger · Sandra C Thompson · Amanda Byrne · Susan J Best · Nick Crofts

Re-engineering the elective surgical service of a tertiary hospital: a historical controlled trial Gideon A Caplan, Ann Brown, Philip J Crowe, Su-Jen Yap and Shaune Noble MJA 1998; 169: 247-251 For editorial comment, see Hillman Abstract - Introduction - Methods - Patients - Outcome measures - Control group--existing system - Postintervention group--the perioperative system - Operative complications - Statistical analysis - Results - Outcomes - Complications - Patient satisfaction - Discussion - Acknowledgements - References - Authors' details - Figure 1 - Figure 2 - - - ©MJA1998 Abstract Objective: To study the clinical effects of re-engineering the processes associated with elective surgery. Design: A prospective, historical controlled trial. Control patients were enrolled from March 1995 to January 1996, and postintervention patients from February 1996 to October 1996. Setting: A major teaching, tertiary care hospital (Prince of Wales Hospital, Sydney). Patients: 224 patients (123 before and 101 after the intervention) undergoing elective herniorrhaphy or laparoscopic cholecystectomy who lived in the local area. Intervention: Introduction of a re-engineered surgical service consisting of preadmission assessment and education, admission on day of surgery, and postacute care after discharge. There were no changes to the operative methods or infection control procedures. Main outcome measures: Length of stay, operative complications, pain scores and patient satisfaction. Results: The risk of a patient suffering one or more complications was reduced in the postintervention group (postintervention v. control patients: 25.7% v. 38.2%; relative risk [RR], 0.66; 95% confidence interval [CI], 0.44-0.98; P = 0.035) because of a reduced risk of wound infections (5.0% v. 16.3%; RR, 0.30; 95% CI, 0.12-0.78; P = 0.0075). Other complications (perioperative or postoperative) and pain scores were unchanged. Patients treated by the re-engineered service had a significantly shorter length of stay, reported a higher level of satisfaction with the preoperative and postdischarge care, and were more likely to say that they would have the same treatment again (92.9% v 82.6%; P = 0.037). Conclusions: Re-engineering surgical services, with an associated reduction in length of stay, does not lead to a deterioration in care and may decrease postoperative complications and increase patient satisfaction. MJA 1998; 169: 247-251 Introduction Financial pressures from insurance companies1 and governments2 are driving changes in the healthcare systems of many countries. One of the most visible changes is shorter hospital stay, especially for elective surgery.3 Organisational changes include preadmission clinics, enhanced patient education, increased use of day surgery, improved discharge planning and postacute care at home; and clinical changes include less invasive surgical techniques and improved anaesthetic drugs.4 The introduction of a package of these organisational changes amounts to a re-engineering of elective surgery.5Many of these changes have been introduced with little scientific evaluation of their effect on patients, and perceptions of detrimental effects on patient care can make such changes unpopular.1,2 Studies of healthcare changes so far6-10 have focused on changes in surgical procedures and have been unable to identify any clinical effects of initiatives, such as preadmission clinics, patient education, reduction in length of stay, and postacute care. Moreover, in these studies, control patients underwent different procedures,6,7 significant results were not obtained,8,9 or there were worse outcomes in the postintervention group.10 We studied the effects of re-engineering our elective surgical service using a prospective, historical, controlled trial. Our re-engineered service comprised a perioperative unit, preadmission anaesthetic assessment based on self-reported questionnaires, admission on the day of surgery, enhanced patient education, use of clinical pathways, and postacute care (Box 1). 1: A re-engineered surgical service Admission coordinated by Perioperative Unit, which sends out, receives and reviews patients' self-reported health questionnaires, and arranges preadmission assessment, if indicated. Patients admitted to Perioperative Unit on day of surgery. After surgery, day-only patients return to Perioperative Unit, others to the ward. Patient information specific for each operation distributed by the surgeon. Nurses reinforce patients' knowledge. Clinical pathways followed. These provide a diagnosis-specific plan for the episode of care and the role of each member of the multidisciplinary team. Postacute care continues at home after discharge, if required. Methods Patients To minimise variation and ensure adequate patient numbers, we studied prospectively a cohort of patients having one of two general surgical procedures (inguinal herniorrhaphy and laparoscopic cholecystectomy). Patients in the control group were treated before and compared with patients treated after the new elective surgical service was commissioned. Patients who lived outside the local area or were admitted as an emergency were excluded. All patients gave informed consent, and the study was approved by the South Eastern Sydney Area Health Service (Eastern Branch) Ethics Committee. Outcome measures Information collected for comparing the control and postintervention groups included length of hospital stay, level of pain (assessed by a linear analogue pain scale), complications of the operation, and patient satisfaction (assessed by a self-reported questionnaire). Control group -- existing system Patients booked for surgery from the surgeon's rooms or surgical outpatients clinic; Patient information and education provided according to individual surgeon's protocols; Patients admitted to the ward on the day before the procedure, clerked by the surgical intern and assessed by the anaesthetist; Postsurgical care given in the surgical ward; and Discharge plans made by ward staff, subject to individual surgeon's decision. A total of 123 control patients were enrolled prospectively from March 1995. (We had planned to enrol 100, but a delay in commissioning the new system extended the enrolment period.) Eligible patients booked for surgery each day were visited by the study nurse on admission to the ward. After surgery, the nurse obtained patient data, including a pain score on Day 1 after the operation, and arranged to contact the patient by telephone on Day 7. Before discharge, patients were given an anonymous, self-reported patient satisfaction survey, based on the Australian model.11 They were asked to rate their overall treatment, the information provided, preoperative treatment, the operation and care after discharge from hospital on a five-point Likert scale from "very satisfactory" to "very unsatisfactory". They were also asked whether they would have the procedure the same way again, and were given a further copy of the pain scale, and asked to complete both documents on Day 7 after the operation and return them in the prepaid envelope provided. The telephone call on Day 7 reminded patients to return their satisfaction surveys and pain scores. Postintervention group -- the perioperative system A more coordinated system was developed, based on the perioperative system (Box 1).12 The surgeons, nursing staff, operating theatre location and organisation and infection control practices did not change. Although surgical registrars changed surgical units during the study period, they worked with other surgeons whose patients were also studied. The Perioperative Unit included a preadmission anaesthetic clinic where patient health questionnaires were completed, and patients with significant comorbidity were fully assessed. Based on the information provided by the questionnaires, 70% of elective surgery patients were seen by an anaesthetist for the first time on the day of surgery, 15% of patients were seen as outpatients before the day of surgery because of anaesthetist preference, and the remaining 15% required further anaesthetic consultation and one (usually) additional outpatient visit for review. All patients, both day-only and those transferring to the surgical ward after the operation, went to theatre from a day-surgery ward within the Perioperative Unit. Patient care was guided by a clinical pathway developed by the multidisciplinary team to drive and support the system. The role of every member of the multidisciplinary team is specified for each day, with alternative plans available if a patient's condition changes. Improved patient educational material was developed for distribution by the surgeons, and patients' comprehension of the material was reviewed by the study nurse at the preadmission anaesthetic clinic or on the day of surgery. If doubts existed about the safety and suitability of a patient's accommodation for postsurgical recovery, the study nurse made a preadmission home visit. After surgery, study patients were transferred to the surgical ward and an existing postacute care service arranged early discharge with follow-up at home. The new system was introduced at the end of January 1996 and enrolment of the 101 patients in the postintervention group commenced in February 1996. A pain scale was completed on Day 1 after operation. The timing of discharge was entirely at the surgeons' discretion, and the study nurse arranged to visit the patient at home within 24 hours of hospital discharge to assess pain control, wound care needs and general postsurgical progress. This visit generally lasted less than half an hour, and occasional further visits were arranged, as clinically indicated. A contact phone number was provided for use in the event of any complications or concerns. Patients in the postintervention group were also asked to complete the self-reported patient satisfaction survey. Operative complications Data on all complications were collected prospectively and crosschecked with the patients' clinical records. The criteria for wound infections were those of the Centers for Disease Control and Prevention (CDC),13 notably that the infection involves only the skin and subcutaneous tissue of the incision, and at least one of the following: purulent drainage from the superficial incision; organisms isolated from aseptically obtained culture of fluid or tissue from the superficial incision; at least one of the following signs or symptoms of infection -- pain or tenderness; localised swelling; redness or heat; and superficial incision is deliberately opened by the surgeon, unless incision is culture negative; diagnosis of superficial incisional surgical site infection by the surgeon or attending physician. Statistical analysis Statistical calculations were performed using SPSS for Windows.14 Normally distributed continuous variables were compared by t tests, dichotomous variables by χ2 tests. Ordinal variables were compared with the Mann-Whitney U test. Logistic regression was used to obtain a multivariate analysis of factors predictive of wound infection. Results From March 1995 to October 1996, 224 patients were enrolled in the study. Baseline characteristics of the two groups were not statistically significantly different (Box 2). 2: Baseline characteristics of elective surgery patients Postintervention group* (n = 101) Control group* (n = 123) Laparoscopic cholecystectomy 57 (56.4%) 65 (52.8%) Herniorrhaphy 44 (43.6%) 58 (47.2%) Mean time in operating theatre (min, SEM) 100.3 (4.2) 102.1 (3.5) Female : male ratio 46 : 55 54 : 69 Mean age (range) (years) 54.4 (21-85) 54.0 (18-90) Smoker 27 (26.7%) 25 (20.3%) Ischaemic heart disease 8 (7.9%) 9 (7.3%) Requires an interpreter 8 (7.9%) 4 (3.3%) Lives alone 19 (18.8%) 23 (18.7%) Lives upstairs (no lift) 37 (36.6%) 39 (31.7%) *Differences were not statistically significant Outcomes The postintervention group had a significantly shorter length of stay, but there was no difference in pain scores between the two groups (Box 3). 3: Outcomes in the postintervention and control groups Outcome Postintervention group Control group Mean length of stay in days (range) 2.2 (1-8) 3.2 (1-9)* For herniorrhaphy 1.8 (1-4) 3.1 (2-7)þ For laparoscopic cholecystectomy 2.5 (1-8) 3.3 (1-9)§ Postoperative pain score mean (SEM) Day 1 21.7 (1.2) 22.6 (0.9) Day 7 9.4 (1.1) 8.4 (1.0) Number of deaths 0 1 (0.8%) Number readmitted 5 (5.0%) 4 (3.3%) Postintervention v. control group: *P < 0.001; þ P < 0.001; § P = 0.020. Complications The postintervention group had a lower risk of suffering any complication (complications in postintervention v. control patients -- 26 [25.7%] v. 48 [38.2%]; relative risk [RR], 0.66; 95% confidence interval [CI], 0.44-0.98; P = 0.035) because of a lower risk of wound infection (Figure 1). There was no significant difference between postintervention and control patients in intraoperative complications or other postoperative complications (Figure 1). Figure 1: Proportion of patients with each type of complication in the postintervention and control groups, including relative risk (RR) and 95% confidence interval (CI) for comparison between the groups. Intraoperative complications were bile/stone spillage (4 v. 5), conversion to open cholecystectomy (4 v. 1) and, in the control group, one each of cystic duct damaged, cystic artery damaged, unsuccessful exploration of common bile duct, local anaesthetic converted to general anaesthetic, penis injured by towel clips, and aspiration into lungs. Postoperative complications are listed in Box 4. Eight patients suffered two complications and one three complications, so that the total for all types of complications exceeds the number of patients who suffered a complication. 4: Postoperative complications Complication Postintervention group (n = 22) Control group (n = 42) Wound infection 5 20 Cerebrovascular accident, died 0 1 ERCP after operation 2 4 Haematoma/haemorrhage/ooze 4 1 Shoulder tip pain 1 2 Scrotal pain or swelling 2 2 Nausea and vomiting 2 0 Constipation (> 4 days) 2 2 Urinary retention requiring indwelling catheter 1 1 Unstable INR after operation, delayed discharge 0 2 Rash (due to antibiotics) 0 1 Fall Day 1 after operation, low blood pressure 0 1 Delirium 0 1 Difficulty mobilising 0 1 Respiratory infection 3 3 ERCP = endoscopic retrograde cholangiopancreatography. INR = International Normalised Ratio. Multivariate analysis using logistic regression found that the only variables significantly predictive of having a wound infection were being in the control group, length of stay (Figure 2) and the patient having a carer (a proxy for functional dependence). Figure 2: Proportion of patients with wound infection in the postintervention and control groups, by length of stay. Patient satisfaction Completed responses to the satisfaction survey were received from 85 postintervention patients (84.2%) and 89 controls (72.4%). Analysis showed that patients in the postintervention group were more satisfied with both the preoperative (P = 0.0094) and the postdischarge treatment (P = 0.0001), as well as the operation, and were also more likely to want to have the procedure done the same way next time, if required (postintervention v. control: 92.9% v 82.6%; difference, 10.3%; 95% CI, 0.8%-19.8%; P = 0.037). There was no difference between the groups in their ratings of the overall treatment or of the information supplied preoperatively. Although the difference in response rates between the control and postintervention groups was significantly different (P < 0.05), there were no significant differences between respondents and non-respondents in terms of age, sex, type of operation, pain scores, complications, comorbidity, living arrangements, occupational or smoking status. Discussion Patients having elective surgery after a re-engineered system of coordinated care for elective surgery was introduced had shorter hospital stays, a lower risk of wound infection and reported higher levels of satisfaction. Previous studies, rather than focusing on the outcomes of the process, as we did, have examined the impact of elements of our re-engineered service (particularly preadmission clinics) on the efficiency of the elective surgical process.15 Our study found a higher rate of wound infection than is generally reported in surgical series, although the rate was within the range of prevalence of nosocomial infections (3%-21%) in a World Health Organization survey,16 and comparable with rates in an Australian survey which also actively followed up patients for wound infections after discharge.17 Surveys may also identify widely varying wound infection rates if different definitions of infections are used.18 Definitions of wound infections which require the presence of pus19 tend to underestimate the true prevalence, given that most doctors institute antibiotic therapy for earlier signs of infection. With decreasing lengths of hospital stay, most wound infections occur after discharge from hospital, where they are often diagnosed and treated by general practitioners (GP) and not surgeons.17 Thus, it is not surprising for series of laparoscopic cholecystectomies to report infection rates which vary from 1% to 17%.6,17 The 70% reduction in the risk of surgical nosocomial infections which we found is relatively large. We used the CDC definition of infection, and found that most infections were diagnosed and treated by GPs before the wounds developed purulent drainage and without waiting for a positive culture result. The study nurses collected the data in the same manner (ie, prospectively during planned follow-up of patients), but the diagnosis of wound infection was left to the treating GP or surgeon. We tested the observation that there is a dose-effect relationship between iatrogenic complications and length of stay in hospital. A relationship between length of stay and risk of nosocomial infection has been observed in a number of studies.18,19 Cruse and foord retrospectively also found that risk of infection increases with length of preoperative stay,19 but previous studies have not prospectively examined the possibility that decreasing the length of stay will result in lower wound infection rates. Our data demonstrate that a reduction in hospital stay as part of a re-engineered surgical service can decrease the wound infection risk. From the near-parallel slope of the lines in Figure 2, it would appear that the difference between the two groups is due to some factor that occurs early in the admission, most likely through decreasing exposure to hospital microbial flora on the night before surgery, resulting in a shift of the curve to the right. Our study design was a prospective, sequential, before-and-after comparison. Randomised controlled trials are generally considered to be the reference standard for evaluating new treatments. However, when the intervention is a coordinated system of care involving a "culture change" in an organisation, there may be a "Hawthorne effect" -- that is, behaviour changes due to an awareness of being in a study -- and leakage of elements of the "culture change" into the control group, thus reducing the power of a randomised controlled trial.20 Our sequential enrolment of patients more closely mimics the situation in hospital when a surgical service is re-engineered, and makes the study results more relevant. Our study had the active collaboration of the hospital administration, and hospital policy was revised at the start of the treatment phase to require all elective surgery patients to be admitted through the Perioperative Unit. The surgeons and operating theatre location and organisation were unchanged, although junior medical staff were rotated. There was no change in antibiotic prophylaxis or hospital infection control policies, thereby minimising any differences in enrolment patterns and other factors between the two arms of the trial. Our patient satisfaction survey suggested that patients were more satisfied with a shorter length of stay combined with the appropriate support after hospital discharge. A previous study in which surgical patients were not similarly supported after early hospital discharge found patients to be less satisfied with the treatment.21 We have demonstrated that changes in the organisation of elective surgery can produce improved health outcomes and patient satisfaction. This does not mean that any reduction in length of stay results in improved outcomes, but that reduced in-hospital support may be replaced with carefully planned and supervised preadmission assessment, education and postdischarge care, without detriment to the quality of patient care. Acknowledgements We would like to acknowledge the invaluable assistance of the staff of Post Acute Care Services, the Perioperative Unit, and the Departments of Surgery and Anaesthetics, without whose help this study would not have been possible. This study was supported by a grant from the Commonwealth Department of Health and Family Services Hospital Access Program. References Sawyer RB. General surgeons in the world of gatekeepers. Am J Surg 1995; 170: 528-531. Maxwell RJ. Why rationing is on the agenda. Br Med Bull 1995; 51: 761-768. Hoare J. Medicine for managers: day surgery. Health Services Management 1992; June: 12-14. Maddern GJ. The changing pattern of surgery [editorial]. Br J Surg 1996; 83: 145-146. Speer TL. With an eye to the future. Hospitals and Health Networks 1996; 70: 43. Barkun JS, Barkun AN, Sampalis JS, et al. Randomised controlled trial of laparoscopic versus mini cholecystectomy. Lancet 1992; 340: 1116-1119. Stoker ME, Vose J, O'Mara P, Maini BS. Laparoscopic cholecystectomy: a clinical and financial analysis of 280 operations. Arch Surg 1992; 127: 589-595. Adler MW, Waller JJ, Creese A, Thorne SC. Randomised controlled trial of early discharge for inguinal hernia and varicose veins. J Epidemiol Community Health 1978; 32: 136-142. Ruckley CV, Cuthbertson C, Fenwick N, et al. Day care after operations for hernia or varicose veins: a controlled trial. Br J Surg 1978; 65: 456-459. Russell IT, Devlin HB, Fell M, et al. Day case surgery for hernias and haemorrhoids: a clinical, social and economic evaluation. Lancet 1977; 1: 844-847. Hill S, Draper M. The role of patient satisfaction surveys in a national approach to hospital quality management. Canberra: AGPS, 1995. Kerridge R, Lee A, Latchford E, et al. The perioperative system: a new approach to managing elective surgery. Anaesth Intens Care 1995; 23: 591-596. Centers for Disease Control and Prevention, US Department of Health and Human Services. National Nosocomial Infections Surveillance Manual. Atlanta, Ga: USDHHS, May 1994. SPSS for Windows [computer program], version 6.0. Chicago: SPSS Inc, 1993. Kahan E, Carel RS, Hart J. Comparison of two pre-admission testing methods for elective surgery patients. Isr J Med Sci 1991; 27: 141-144. Ayliffe GAJ. Surveys of nosocomial infections. Med J Aust 1988; 149: 571-572. Hardy KJ, Miller H, Fletcher DR, et al. An evaluation of laparoscopic versus open cholecystectomy. Med J Aust 1994; 160: 58-62. Freeman J, McGowan JE. Differential risks of nosocomial infection. Am J Med 1981; 70: 915-918. Cruse PJE, foord R. The epidemiology of wound infection: a 10-year prospective study of 62,939 wounds. Surg Clin North Am 1980; 60: 27-40. Bouchet C, Guillemin F, Briancon S. Nonspecific effects in longitudinal studies: impact on quality of life measures. J Clin Epidemiol 1996; 49: 15-20. Michaels JA, Reece-Smith H, Faber RG. Case control study of patient satisfaction with day-case and inpatient inguinal hernia repair. J R Coll Surg Edinb 1992; 37: 99-100. (Received 5 Jun 1997, accepted 18 May 1998) Authors' details Prince of Wales Hospital, Sydney, NSW. Gideon A Caplan, MB BS, FRACP, Director, Post Acute Care Services. Ann Brown, RN, Nurse Manager, Post Acute Care Services. Philip J Crowe, MB BS, FRACS, Surgeon, Department of Surgery. Su-Jen Yap, MB BS, FANZCA, Director, Perioperative Unit, Prince of Wales Hospital. South Eastern Sydney Area Health Service, Sydney, NSW. Shaune Noble, BHA, MPH, Health Services Development Unit. Reprints: Dr GA Caplan, Director, Post Acute Care Services, Prince of Wales Hospital, Randwick, NSW 2031. E-mail: G. CaplanATunsw.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Gideon A Caplan · Ann Brown · Philip J Crowe · Su-Jen Yap · Shaune Noble

Rotavirus infection and rates of hospitalisation for acute gastroenteritis in young children in Australia, 1993-1996 John B Carlin, Patty Chondros, Paul Masendycz, Helen Bugg, Ruth F Bishop and Graeme L Barnes For editorial comment, see Ferson MJA 1998; 169: 252-256 Abstract - Introduction - Methods - Results - Discussion - Acknowledgments - References - Authors' details - Figure 1 - Figure 2 - - - ©MJA1998 Abstract Objective: To determine rates of hospitalisation of young children for acute gastroenteritis in Australia, and to estimate the proportion of these admissions caused by rotavirus infection. Design: Analysis of hospital admission records, and parallel, prospectively collected data on rotavirus-positive admissions. Setting: Hospitals admitting young children in all Australian States and Territories in 1993-1996. Patients: All children under five years admitted to hospital for acute gastroenteritis (International Classification of Diseases, ninth revision principal diagnosis codes 003.0, 004.0-009.3 and 558.9). Main outcome measures: Rate of hospital admission per 1000 children per year by State, and the proportion of admissions caused by rotavirus infection. Results: There were almost 20 000 hospital admissions annually in Australia for acute gastroenteritis in children under five years, at an average rate of 15/1000. An estimated 50% of these were attributable to rotavirus infection, implying a rate of hospitalisation for rotavirus-related gastroenteritis of 7.5/1000/year. Among children under two years this rate was 11.6/1000. Rotavirus incidence rates generally followed a typical seasonal pattern in temperate regions of the country, with sharp peaks in mid to late winter. Rates of hospitalisation varied markedly, even between States with apparently similar patterns of disease, while the incidence in the Northern Territory was 3-5 times higher than other States. Conclusions: Rotavirus-related gastroenteritis is a major cause of hospital admissions in young children, and large savings to the healthcare system are possible if it can be prevented at reasonable cost. Variation in treatment practices between States may be worth studying in greater detail as another source of potential savings. MJA 1998; 169: 252-256 Introduction Rotavirus is a major cause of severe gastroenteritis in young children. Although discovered only 25 years ago,1 the role of this virus in the huge burden of diarrhoeal disease in developing countries was recognised quickly, and in 1985 it was estimated to be responsible for 870 000 deaths annually.2 In developed countries, mortality from gastroenteritis is low, but many studies have documented the large burden of morbidity caused by rotavirus.3-7Precise estimates of rotavirus-related morbidity are difficult to obtain from routine hospital data because stool samples for microbiological testing are not obtained from all children admitted to hospital with diarrhoea. Even when testing is performed, the findings may not be recorded or coded in the medical record. There was no specific category in the International Classification of Diseases, ninth revision (ICD-9-CM) for rotavirus diarrhoea until 1993, and since then the specified code (008.61) has not been used consistently. However, many studies have shown that at least 30% of hospital admissions for acute gastroenteritis in young children are the result of rotavirus infection and, further, that the incidence of rotavirus infection has a distinctive cyclical pattern with a peak in winter (this is unique to rotavirus among all major pathogens associated with gastroenteritis).3,4,6,8-10 While other Australian studies have examined aspects of rotavirus gastroenteritis,5,11 there are no national population-based estimates of the incidence of hospital admission for acute gastroenteritis in young children. Such estimates are important in assessing preventive measures, and in evaluating the cost-effectiveness of rotavirus vaccines currently under development.12,13 We aimed to provide the first national data on rates of hospitalisation of young children for acute gastroenteritis in Australia, and to estimate the proportion of these admissions that could be ascribed to rotavirus infection. Our results will be used in a later study to estimate the cost-effectiveness of rotavirus vaccination. Methods State and Territory health departments provided computer files with unidentified records for the years 1993 through 1996 of all hospital admissions of children under the age of five with a principal diagnosis in the medical record of acute gastroenteritis. In our analysis, we combined data from the Australian Capital Territory with those from New South Wales. Acute gastroenteritis was identified as ICD-9-CM codes in the range 003-009 (covering infectious gastroenteritis of various known and unknown origins) or code 558.9 ("other and unspecified noninfectious gastroenteritis"). This last code was included for consistency with other studies in the expectation that it might comprise a substantial number of cases of infectious gastroenteritis. Similar files of admission records were obtained from participating paediatric hospitals in each of the States (see Box 1). Records from the Royal Children's Hospital (RCH), Melbourne, and Princess Margaret Hospital (PMH), Perth, also included an identifying hospital record number. 1: Participating hospitals Royal Children's Hospital, Melbourne (VIC) Princess Margaret Hospital, Perth (WA) Women's and Children's Hospital, Adelaide (SA) Alice Springs Hospital (NT) Royal Darwin Hospital (NT) Royal Hobart Hospital (TAS) Sydney Children's Hospital (formerly Prince of Wales Children's Hospital) (NSW) New Children's Hospital, Sydney (Royal Alexandra Hospital for Children) (NSW) Westmead Hospital, Sydney (NSW) Royal Brisbane Hospital (QLD) Microbiology departments at participating hospitals were asked to forward all rotavirus-positive faecal specimens obtained from children under five years who were admitted for acute gastroenteritis to laboratories at RCH, where rotavirus infection was confirmed by enzyme immunoassay.14 All participating hospitals provided specimens over the entire study period, except that Westmead Hospital participated in the study for the first two years only, and was replaced by Sydney Children's Hospital in 1995. Seasonal, temporal, and geographic variation in strains of rotaviruses will be reported elsewhere. We estimated the proportion of gastroenteritis admissions that were caused by rotavirus infection both by direct and indirect means. The direct method involved a linked analysis of admissions and rotavirus data. Admission and laboratory records from RCH (Melbourne) and PMH (Perth) were linked by means of hospital unit record number and date of admission/date of specimen, thus identifying the proportion of admissions that we could directly confirm as involving rotavirus infection. This analysis was supplemented with information from each hospital's pathology records, as it became clear that not every rotavirus-positive specimen was being sent on to our laboratory (usually owing to insufficient size of samples). Further, it was possible to omit patients for whom no faecal test was performed, to give a more appropriate denominator for estimating the fraction admitted because of rotavirus infection (calculated as rotavirus fraction = number of admissions with positive rotavirus test result/number of admissions with faecal specimen tested). The indirect statistical estimation of the rotavirus fraction and the rationale behind this method are described in Box 2. The study was approved by ethics committees in each of the participating hospitals. 2: Statistical estimation of the rotavirus fraction We assumed (i) that rotavirus is the only major cause of childhood gastroenteritis admissions that shows significant seasonal variation; and (ii) that the total number of admissions due to rotavirus in a State is a constant multiple of the number seen in our participating hospital(s). Thus, total admissions per month were expressed as the sum of a constant number (representing non-rotavirus causes) and varying numbers of rotavirus admissions. The resulting linear regression model for the number of admissions in a State in each month was used to estimate the rotavirus fraction (the proportion of hospital admissions for acute gastroenteritis resulting from rotavirus infection). This is shown mathematically below, where: Ni = number of acute gastroenteritis admissions in the State in month i; ni = number of "rotavirus-positive" admissions in the State's participating (index) hospital(s); = monthly number of admissions statewide not caused by rotavirus; and = scaling factor relating the number of "rotavirus-positive" admissions in the State to that in the index hospital(s). The regression model (in which i = random error) is: Ni = + ni + i , which leads to: We followed earlier practice4,5 and used ordinary least squares to estimate , despite the fact that the model does not fit the usual assumptions of linear regression (in particular, the variance of i cannot be expected to be constant). As the assumption that there is perfect correlation between rotavirus incidence recorded at the index hospital and total rotavirus numbers at all hospitals in the State is inevitably untrue, there will be an underestimation of in this model, and consequently some underestimation of the rotavirus fraction. On the other hand, if the assumption that rotavirus is the only agent responsible for seasonal variation is wrong, then the estimation may be biased in either direction. The estimate is also potentially subject to other biases relating to the completeness and regularity of coverage of the contributed rotavirus samples. These biases are likely to outweigh random error, so that conventional confidence intervals would not provide a reliable indicator of uncertainty. Results Total numbers and estimated rates of acute gastroenteritis admissions are shown in Box 3. Of all admissions, 65.7% involved children under the age of two years, with approximately equal numbers in each of the first two years of life, and declining numbers in each of the subsequent three years to age five. The age breakdown was similar in all States except the Northern Territory, where a substantially larger proportion were aged under two (86.1%). Length of stay was also considerably longer in the Northern Territory. There were slightly more boys than girls (53.1% overall), consistent across all States. Of all admissions, 29.4% were coded 558.9, although this proportion declined sharply over the four years of the study as coding practices apparently changed. 3: Acute gastroenteritis in children aged less than five years, 1993-1996. Number and rate of hospital admissions, rate ratio relative to the Victorian rate, and average length of stay (95% confidence intervals in parentheses) State or TerritoryAverage annual admissionsAnnual rate per 1000 children under 5 yearsRate ratio relative to VictoriaAnnual rate per 1000 children under 1 yearAnnual rate per 1000 children aged 1-2 yearsAverage length of stay Victoria29159.2 (9.0-9.3)1.00 14.6 (14.1-15.1)14.4 (13.9-14.8) 1.93 days (1.90-1.96) Tasmania3199.3 (8.8-9.8) 1.01 (0.96-1.07) 15.0 (13.6-16.4) 15.1 (13.7-16.6) 1.79 days (1.71-1.87) Western Australia178614.2 (13.9-14.5) 1.55 (1.51-1.60) 25.3 (24.3-26.2) 24.9 (24.0-25.9) 3.13 days (3.03-3.23) Queensland377316.0 (15.7-16.2) 1.74 (1.70-1.79) 24.0 (23.3-24.7) 26.4 (25.6-27.1) 2.15 days (2.11-2.18) New South Wales*786617.1 (16.9-17.3) 1.87 (1.83-1.91) 25.5 (25.0-26.0) 27.5 (27.0-28.1) 2.11 days (2.09-2.13) South Australia189519.2 (18.8-19.7) 2.10 (2.04-2.16) 32.0 (30.7-33.2) 30.7 (29.5-31.9) 2.09 days (2.03-2.15) Northern Territory85449.8 (48.2-51.4) 5.44 (5.24-5.65) 106.2 (101.0-111.3) 108.1 (102.9-113.3) 9.18 days (8.88-9.47) Total1940815.0 (14.9-15.1) 23.9 (23.6-24.2) 25.0 (24.7-25.3) * Including Australian Capital Territory. Box 4 shows the temporal pattern of all admissions and of rotavirus admissions in index hospitals in the five largest States and the Northern Territory. The seasonal peaks in the eastern States (New South Wales, Victoria, Queensland, and South Australia) were generally coincident and occurred regularly in the colder months from July to September (except for the unusual final year in South Australia). In contrast, the weaker and less regular peaks in Western Australia occurred earlier than in the eastern States, and there was no discernible seasonal pattern in the Northern Territory. Even where the seasonal pattern was less distinct, the peaks in statewide admissions were reflected in peaks in the numbers of rotavirus- positive specimens. 4: Monthly incidence of hospital admission for acute gastroenteritis in children under five years of age in Australian mainland States and the Northern Territory, 1993-96 (solid line) and number of rotavirus-positive specimens received from participating hospitals in that State, rescaled from the hospital to the State level (dashed line), by the "regression method" For the data linkage method, among 1732 children admitted for whom a specimen was tested, we received 767 rotavirus-positive specimens from RCH (Melbourne) and found another 205 admissions with hospital records showing a positive rotavirus test, giving an estimated rotavirus fraction of 56%. The corresponding totals from PMH (Perth) were 1631, 626 and 161, respectively, giving a rotavirus fraction of 49%. Regression estimates of the rotavirus fraction ranged from a low of 22% for the Northern Territory (based on a total of 598 rotavirus-positive specimens) to a high of 41% for Victoria (984 specimens). When data from all States were combined, the correlation between rotavirus numbers and admissions increased, because a more accurate representation of the relationship between total admissions and rotavirus numbers in participating hospitals was obtained by pooling. The resulting estimated rotavirus fraction was 50% (based on 4634 specimens received). Separate estimates obtained for children aged under one year, one year to less than two years, and at least two years were 34%, 60% and 48%, respectively. Discussion This national four-year study has shown that there are almost 20 000 hospital admissions annually in Australia for acute gastroenteritis in children aged under five years -- an average rate of 15 per 1000, 50% of which appear to be attributable to rotavirus infection. The annual rate of hospitalisation for rotavirus-related gastroenteritis in children under five is therefore about 7.5 per 1000. Among children under two years, combining estimates for the two one-year age brackets, the corresponding rate is 11.6 per 1000, or just over one in every 100 children. Of the two methods we used to estimate the proportion of admissions that were caused by rotavirus, the method involving direct linkage of hospital admission records and stool specimen tests is likely to be the more accurate. Estimates of 56% (Melbourne) and 49% (Perth) were obtained for the two centres where it was available. These estimates may be too low, as enzyme immunoassay can fail to detect up to 10% of rotavirus-positive specimens.15 Conversely, they might be too high on a population-wide basis, as they were based on tertiary referral centres where the severity of admitted cases -- and therefore the likelihood of rotavirus involvement -- might be higher (although this was only weakly supported by our data on mean length of stay [LOS], which showed a small difference between RCH [mean LOS 1995-1996, 2.15 days] and the whole of Victoria [1.87 days], and no difference between PMH and WA averages [2.80 v. 2.85]). Our second method, based on a regression model, was subject to statistical fluctuations and assumptions that imply it might produce an underestimate of the rotavirus fraction (see Box name="box2"2). In particular, the coverage of rotavirus specimen collection varied between States and also, at times, within States. For example, only a relatively small number of specimens (457) were ascertained from New South Wales, implying that the rotavirus fraction for that State might be considerably underestimated. In fact, when the regression method was applied to the pooled national data, we found reasonable convergence between the two methods, and believe they both support the conclusion that 50% of all gastroenteritis admissions in this age group are due to rotavirus. A reasonable range of possible alternative estimates around this value would be from 40% to 60%. In other studies, estimates of the proportion of hospital admissions in children under five years attributable to rotavirus have ranged from around 30% to as high as 66%.3,4,6,8,9,16 United States studies from the Centers for Disease Control have estimated this proportion both by a simple "residual" method based on assuming that all diarrhoea admissions in the winter seasonal peaks are rotavirus-related,17 and a more direct method based on an eight-year series of children admitted to a Washington DC hospital.6 Both methods indicated that rotavirus was responsible for about one-third of admissions, and the corresponding annual rate of hospitalisation for rotavirus-related gastroenteritis in children under five years was 2.8 per 1000. In the United Kingdom, estimates have been based on laboratory surveillance and hospital discharge data, with a similar statistically based estimation procedure to ours.4 It was found that 43% of admissions in children under five years could be attributed to rotavirus, corresponding to 5.3 per 1000 admissions per year. A New South Wales study by Ferson also used a regression method, based on a larger sample of rotavirus surveillance data than ours, and estimated an annual rate of hospital admission of 8.7 per 1000.5 Using the 1993-1996 NSW hospital admission rates (Box name="box3"3), this corresponds to a rotavirus fraction of 51%, very consistent with our national estimate. Ferson's study found over twice the rate of rotavirus gastroenteritis in the second year of life compared with the first, a slightly greater difference than suggested by our data. The findings of both studies are consistent with a detailed investigation of children with gastroenteritis over 13 years at RCH Melbourne, which found that no pathogen could be identified in a substantially greater proportion of infants under six months than in older children.18 There were substantial differences in hospital admission rates between the States, with NSW, Queensland and South Australia having almost twice the rate of hospitalisation of Victoria. It seems very unlikely that these differences are the result of different disease patterns, especially as the seasonal patterns in these States were very similar. Nor are they likely to be explained by differences in coding, as our incidence figures are based on an inclusive selection of principal diagnosis ICD-9-CM codes. In particular, the inclusion of code 558.9 ("non-infectious gastroenteritis") identified a substantial number of apparently miscoded cases of viral diarrhoea, consistent with findings in other reports.4,5 The differences may reflect variations in patterns of primary care influencing the numbers of children who present to hospital emergency departments, and/or differences in hospital admission policies. The relative importance of each of these factors warrants further study, as it appears likely that substantial savings could be made if hospitalisation rates nationally could be reduced. There were no major seasonal differences in gastroenteritis incidence between States, except in the Northern Territory and Western Australia, where the relative importance of rotavirus in the gastroenteritis disease burden may be somewhat reduced. There is little evidence in our data of a progressive spread of the rotavirus epidemic across the country each year, in the manner that has been described in North America,19 apart from the fact that the annual peak of rotavirus activity in Western Australia appears generally to precede that in the eastern States. The Northern Territory was the most clearly distinct of all the regions, not only in its substantially higher hospital admission rate and lack of seasonal pattern, but also in the dramatically longer length of stay of patients. Average length of stay in this study (2.2 days, excluding the Northern Territory) was substantially shorter than that reported in the United States (3.9 days20), but very similar to British findings.4 Given the higher admission rates in Australia, shorter length of stay may reflect a less severe spectrum of illness, or possibly better management involving more widespread use of oral rehydration solution. Our results imply that 10 000 hospital admissions occur annually in Australia for rotavirus gastroenteritis. The direct cost of this hospital care is at least $12 million, based on an (Australian national diagnosis-related group-based) estimate of $1244 per admission,21 and may be as high as $15-$18 million (preliminary analysis, RCH [Melbourne] costing data). A successful rotavirus vaccine -- depending on its cost -- could result in substantial savings to the healthcare system. Acknowledgments The study was funded by the Public Health Research and Development Committee of the National Health and Medical Research Council. It would not have been possible without the generous cooperation of numerous individuals in all State and Territory health authorities, and in medical records and hospital information units at the participating hospitals, or without the participation and skilled assistance of the following microbiologists and paediatricians: G Davidson, P Goldwater, T Kok, A Lawrence, L Micken, S Weir (Adelaide); G Clift, J Erlich, J Hagger, R Matters, F Morey (Alice Springs); J Faogali, J Farrah, R Shepherd, M Witt (Brisbane); G Lum, A Lowe, A Ruben, B Way, K Withnall (Darwin); A Carmichael, A Claridge, K Dahlenburg, R Fang, R Tucker (Hobart); B Crawford, G Hogg, B Ross, R Schnagl, P Ward (Melbourne); B Dwyer, R Hill, A May, G O Connor, B Wild (Perth); P Amin, T Borg, A Cunningham, J MacRae, P McIntyre, C McIvor, K McPhie, J Mitchell, J Montgomery, G Sandico (Sydney). References Bishop RF, Davidson GP, Holmes IH, Ruck BJ. Virus particles in epithelial cells of duodenal mucosa from children with acute non-bacterial gastroenteritis. Lancet 1973; 2: 1281-1283. Bern C, Glass RI. Impact of diarrheal diseases worldwide. In: Kapikian AZ, editor. Viral infections of the gastrointestinal tract. New York: Marcel Dekker, 1994: 1-26. Glass RI, Kilgore PE, Holman RC, et al. The epidemiology of rotavirus diarrhea in the United States: surveillance and estimates of disease burden. J Infect Dis 1996; 174 (Suppl 1): S5-S11. Ryan MJ, Ramsay M, Brown D, et al. Hospital admissions attributable to rotavirus infection in England and Wales. J Infect Dis 1996; 174 (Suppl 1): S12-S18. Ferson MJ. Hospitalisations for rotavirus gastroenteritis among children under five years of age in New South Wales. Med J Aust 1996; 164: 273-277. Brandt CD, Kim HW, Rodriguez WJ, et al. Pediatric viral gastroenteritis during eight years of study. J Clin Microbiol 1983; 18: 71-78. Matson DO, Estes MK. Impact of rotavirus infection at a large pediatric hospital. J Infect Dis 1990; 162: 598-604. Konno T, Suzuki H, Imai A, et al. A long-term survey of rotavirus infection in Japanese children with acute gastroenteritis. J Infect Dis 1978; 138: 569-576. Spence L, Singer O, Kibsey P, Fauvel M. Rotavirus infection in children with diarrhoea admitted to a general hospital in metro Toronto. Can J Pub Health 1985; 76: 17-20. Lewis HM, Parry JV, Davies HA, et al. A year's experience of the rotavirus syndrome and its association with respiratory illness. Arch Dis Child 1979; 54: 339-346. Liddle JLM, Burgess MA, Gilbert GL, et al. Rotavirus gastroenteritis: impact on young children, their families and the health care system. Med J Aust 1997; 167: 304-307. Smith JC, Haddix AC, Teutsch SM, Glass RI. Cost-effectiveness analysis of a rotavirus immunization program for the United States. Pediatrics 1995; 96: 609-615. Barnes GL, Lund JS, Adams L, et al. Phase 1 trial of a candidate rotavirus vaccine (RV3) derived from a human neonate. J Paediatr Child Health 1997; 33: 300-304. Coulson BS, Unicomb LE, Pitson GA, Bishop RF. Simple and specific enzyme immunoassay using monoclonal antibodies for serotyping human rotaviruses. J Clin Microbiol 1987; 25: 509-515. Husain M, Seth P, Broor S. Detection of group A rotavirus by reverse transcriptase and polymerase chain reaction in feces from children with acute gastroenteritis. Arch Virol 1995; 140: 1225-1233. Donelli G, Ruggeri FM, Tinari A, et al. A three-year diagnostic and epidemiological study on viral infantile diarrhoea in Rome. Epidemiol Infect 1988; 100: 311-320. Ho MS, Glass RI, Pinsky PF, Anderson LJ. Rotavirus as a cause of diarrheal morbidity and mortality in the United States. J Infect Dis 1988; 158: 1112-1116. Barnes GL, Uren E, Stevens KB, Bishop RF. Etiology of acute gastroenteritis in hospitalized children in Melbourne, Australia, from April 1980 to March 1993. J Clin Microbiol 1998; 36: 133-138. LeBaron CW, Lew J, Glass RI, et al. Annual rotavirus epidemic patterns in North America: Results of a 5-year retrospective survey of 88 centers in Canada, Mexico, and the United States. JAMA 1990; 264: 983-988. Jin S, Kilgore PE, Holman RC, et al. Trends in hospitalizations for diarrhea in United States children from 1979 through 1992: estimates of the morbidity associated with rotavirus. Pediatr Infect Dis J 1996; 15: 397-404. Australian casemix report: on hospital activity, 1995-96. Canberra: Department of Health and Family Services, 1997. (Received 30 Oct 1997, accepted 4 May 1998) Authors' details Clinical Epidemiology and Biostatistics Unit, Royal Children's Hospital, Melbourne, VIC. John B Carlin, BSc(Hons), PhD, Deputy Head, and Associate Professor, Department of Paediatrics, University of Melbourne; Patty Chondros, BSc(Hons), Research Assistant. Department of Gastroenterology and Clinical Nutrition, Royal Children's Hospital, Melbourne, VIC. Paul Masendycz, BAppSc, Research Assistant; Helen Bugg, BSc, Research Assistant; Ruth F Bishop, AO, DSc, PhD, Senior Principal Research Fellow (NHMRC, and Professor, Department of Paediatrics, University of Melbourne; Graeme L Barnes, MD, FRACP, Senior Gastroenterologist, and Associate Professor, Department of Paediatrics, University of Melbourne. Reprints: Associate Professor G L Barnes, Department of Gastroenterology and Clinical Nutrition, Royal Children's Hospital, Parkville, VIC 3052. E-mail: barnesgATcryptic.rch.unimelb.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

John B Carlin · Patty Chondros · Paul Masendycz · Helen Bugg · Ruth F Bishop · Graeme L Barnes

Interval breast cancers in an Australian mammographic screening program Mary T Rickard, Richard J Taylor, Mohamed A Fazli and Nadima El Hassan MJA 1998; 169: 184-187 For editorial comment, see Rodger & Kavanagh Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - ©MJA1998 Abstract Objective: To determine the incidence of interval cancers which occurred in the first 12 months after mammographic screening at a mammographic screening service. Design: Retrospective analysis of data obtained by crossmatching the screening Service and the New South Wales Central Cancer Registry databases. Setting: The Central & Eastern Sydney Service of BreastScreen NSW. Participants: Women aged 40-69 years at first screen, who attended for their first or second screen between 1 March 1988 and 31 December 1992. Main outcome measures: Interval-cancer rates per 10 000 screens and as a proportion of the underlying incidence of breast cancer (as estimated by the underlying rate in the total NSW population). Results: The 12-month interval-cancer incidence per 10 000 screens was 4.17 for the 40-49 years age group (95% confidence interval [CI], 1.35-9.73) and 4.64 for the 50-69 years age group (95% CI, 2.47-7.94). Proportional incidence rates were 30.1% for the 40-49 years age group (95% CI, 9.8-70.3) and 22% for the 50-69 years age group (95% CI, 11.7-37.7). There was no significant difference between the proportional incidence rate for the 50-69 years age group for the Central & Eastern Sydney Service and those of major successful overseas screening trials. Conclusion: Screening quality was acceptable and should result in a significant mortality reduction in the screened population. Given the small number of cancers involved, comparison of interval-cancer statistics of mammographic screening programs with trials requires age-specific or age-adjusted data, and consideration of confidence intervals of both program and trial data. Introduction The principal aim of a mammographic screening service is to reduce mortality from breast cancer. As this outcome cannot be measured effectively for many years after the establishment of screening, and as it applies to spatially defined populations, a range of interim measures are used to evaluate screening services and to determine the likelihood of mortality reduction. Interim measures include screen-detected cancer rates and the rates of prognostic indicators for these cancers, such as size, nodal status and grade.1 Another interim measure of screening performance is the interval-cancer rate.2,3 As survival among women with interval cancers is similar to that among unscreened women, then as the incidence of interval cancers approaches that of the underlying incidence of breast cancer the benefit of screening disappears.3 Interval-cancer rates can be compared with background or underlying rates to better assess the effectiveness of screening. Further, comparing the rate of interval cancers with the underlying rate allows comparison between different screening programs in populations with different incidences of breast cancer. Our aim was to evaluate first-year interval breast cancers after screening at the Central & Eastern Sydney Service of BreastScreen NSW (a mammographic screening and assessment service). We report the incidence and proportional incidence of interval breast cancers during the 12 months after both the first and second rounds of screening, over five years. Methods The Central & Eastern Sydney Service of BreastScreen NSW is located in inner Sydney, and started screening in March 1988 as one of the first pilot mammography screening programs in Australia. Any woman aged 40 years or more was eligible to attend, although during the study period only 50-69-year-old residents in the Central Sydney Area Health Service were actively recruited for two-yearly screening. Since the progressive introduction of systematic, government-funded, population-based screening from 1991, the Service has become a regional component of BreastScreen NSW and BreastScreen Australia.4The study population comprised women who attended the Service for mammographic screening between 1 March 1988 and 31 December 1992. These women were all screened by two-view mammography. The technical quality of the Service's screening mammograms met all requirements for accreditation by BreastScreen Australia.5 All films were read independently by two radiologists and were reported as "normal" for routine rescreen or as "suspicious" and requiring assessment. The design of the screening and assessment Service has been previously described.6 Screen-detected cancers The definition of primary breast cancer used for this study includes invasive cancer and ductal carcinoma-in-situ (DCIS), but excludes lobular carcinoma-in-situ.5 All cases of primary breast cancer diagnosed by the screening and assessment Service in women attending for the first time were classified as prevalent (first-round) screen-detected cancers. Cancers in women attending for their first routine rescreen after two years were classified as second-round screen-detected cancers. Data were analysed according to age at first screen, by 10-year age groups (40-49, 50-59 and 60-69 years). Interval cancers We reviewed the data for interval cancers, defined as primary cancer of the breast diagnosed up to 12 months after a first-round or second-round screening mammogram in women aged 40-69 years at the time of screening. The date of diagnosis of these interval cancers was taken as the date of histological confirmation of cancer. Cases in which clinical and imaging findings were consistent with breast cancer but no surgery had been performed were included in the study and the date of diagnosis was taken as the date of cytological confirmation of cancer. Data were analysed by women's age at their first screen. Some interval cancers were reported to the screening Service and others were identified by linking the Service's screening database with the NSW Central Cancer Registry (NSW CCR), which has operated since 1972 as a population-based registry. Notification of malignant neoplasms has been a statutory requirement for all NSW public and private hospitals, radiotherapy departments and nursing homes since 1972, and for pathology and outpatient departments since 1985. The date of diagnosis used by the NSW CCR was defined as "date of first definitive treatment for cancer" during the period covered by this study, although incidence was designated mostly on the basis of the first notification (usually the biopsy report). We matched records of the screening database with the NSW CCR in late 1994 with the aid of probabilistic linkage7,8 using Automatch.9 Equivocal matches were investigated by individual examination of all details available, and by active follow-up by the screening Service. Positive linkages between the screening database and the cancer registry were investigated by the screening Service to ensure that all data on those interval cancers diagnosed in the first year after a negative screen were available for inclusion in the study. Statistical analysis The age-specific incidence of interval cancers was determined by dividing the number of interval cancers found in women screened between 1988 and 1992 by the age-specific (age at first screen) number of women screened over the same period. Age groups were 40-49, 50-59 and 60-69 years, although the latter two age groups are aggregated for reporting purposes. The underlying incidence of breast cancer from 1988 to 1992 was obtained by dividing NSW reported cases (as reported by the NSW CCR.10,11) by the State female population for the same three 10-year age groups. To express the interval-cancer incidence as a proportion of the underlying breast cancer incidence rate, an indirectly age-standardised incidence ratio12 was calculated using the State age-specific incidences as the standard. Because of small numbers of cases, 95% confidence intervals were derived by the Poisson method.12,13We compared our findings with those of studies from Sweden,3 the Netherlands14 and the United Kingdom15 using the 12-month interval-cancer data from the first two screening rounds for all studies except the UK study, for which only the first-round data were available. Confidence limits for interval-cancer rates from comparison populations were calculated from the published data using the same Poisson method as above. Results During the study period, the Central & Eastern Sydney Service carried out 39 988 first-round and second-round screens on women aged 40-69 years, and 226 screen-detected cancers were diagnosed. In the prevalent (first-round) screens, the screen-detected cancer rates per 1000 screens were 3.0 for women aged 40-49 years, 6.2 for 50-59-year-olds and 9.6 for 60-69-year-olds. Second-round rates were 3.5, 4.2 and 4.6, respectively. Eighteen interval cancers occurred in the first 12 months after screening. All of these interval cancers were invasive. The 12-month interval-cancer incidences are shown in Box 1. The underlying annual breast cancer incidences were 13.8 per 10 000 for the 40-49 years age group and 21.2 per 10 000 for the 50-69 years age group (19.4 for 50-59 years and 23.2 for 60-69 years). The interval-cancer incidences as a proportion of these underlying incidences are shown in Box 2, in which they are compared with those of the Swedish Two-County Study.3 The proportional interval-cancer incidence for women aged 50-69 years in the Central & Eastern Sydney Service (22.0%; 95% CI, 11.7%-37.7%) is compared with that for similar age groups from international studies3,14,15 in Box 3. The proportional interval-cancer incidence point estimate for the Central & Eastern Sydney Service was higher and had wider 95% confidence intervals than that of the Swedish Two-County Study (13.2%; 95% CI, 8.4-19.9), but was lower than that of the UK (31.4%; 95% CI, 24.9-39.1) and Dutch (39.5%; 95% CI, 20.4-69.0) studies. Discussion The screen-detected cancer rates achieved by the Central & Eastern Sydney Service exceed the National Accreditation Requirements and compare favourably with those reported from the Swedish Two-County Study and the other Australian mammographic screening services.5,16,17Our findings indicate that the first 12-month interval-cancer rate (as a proportion of underlying incidence) for women attending the Central & Eastern Sydney Service is higher than that of the Swedish Two-County Study,3 but lower than the rates reported from Nijmegen (the Netherlands)14 and North West Region UK.15 However, there is considerable overlap of 95% confidence intervals, and the conclusion must be that the Central & Eastern Sydney proportional first-year interval-cancer rate is not significantly different from that in these other studies. As a proportion of interval cancers occur because an abnormality has not been detected at the time of screening, the first-year interval-cancer rates reflect the proportion of false negative screens in the screening episode. The Central & Eastern Sydney results therefore indicate acceptable screening sensitivity and, as they are not significantly different from the results of the Two-County Study, they would be expected to indicate a similar future mortality reduction. The interval-cancer data presented in this study are not affected by significant ascertainment bias. Restricting our analysis to interval cancers occurring within the first year after a screen eliminates the difficulties of accurately differentiating interval cancers and screen-detected cancers diagnosed around the 21-27-month rescreen interval. We directed considerable attention to the linkage with the NSW CCR to ensure that no interval cancers were missed. The underlying rate of breast cancer is reliable. This was determined in a population that included the women who were screened. However, during the study period, the Central & Eastern Sydney BreastScreen Service was one of only two pilot mammographic screening and assessment endeavours in New South Wales, and screened a small fraction of the State's female population. Its clientele was not limited to the immediate geographic area. For the busiest years in this period (1991 and 1992) less than 5% of breast cancers in NSW were detected through these two mammographic screening services.18 Completeness of enumeration is difficult to determine precisely for cancer registries, but the standard indicators suggest reasonably good completeness for the NSW CCR,10,19 and its data are accepted for inclusion in Cancer incidence in five continents.20 Comparisons of the Central & Eastern Sydney Service and another pilot mammographic screening service with the NSW CCR for the period 1988-1992 showed 100% enumeration of invasive breast cancer by the NSW CCR (R T, NSW Cancer Council, unpublished data). The design of the Australian mammographic screening program and the comparison studies in this paper differ. The Swedish Two-County Study employed single-view mammography and double reading. The Dutch and UK studies used single-view mammography and single reading. Double-view mammography and double reading, as used in Australia, would be expected to produce better results and fewer interval cancers in the first year than the comparison studies. The confidence intervals of the proportional interval-cancer rates from the Swedish Two-County Study and those reported for North West Region UK do not overlap, and direct statistical testing has shown a significant difference between these rates.15 The authors of the UK article15 and those of the accompanying editorial21 expressed concern that the screening sensitivity in the British National Health Service program may not be sufficient to achieve mortality reduction targets. They noted the increased sensitivity resulting from two-view mammography, good film quality and two independent film readings. These features are part of BreastScreen Australia and of the Central & Eastern Sydney Service. Even when data are collected over reasonable lengths of time, the numbers of interval cancers in many screening services are often small because of the size of the base populations. In the 50-69 years age group, there were 23 interval cancers in the Swedish Two-County Study, compared with 12 in the Nijmegen study and 13 in our study. However, performance can still be evaluated through calculation of interval-cancer rates, provided that interpretation is qualified by consideration of confidence intervals based on exact or Poisson methods. Australian mammography screening services are evaluated by comparison with National Accreditation Requirements standards.5 The point estimate of the Swedish Two-County Study has come to be regarded as a standard, as this study achieved significant mortality reduction. However the Two-County Study did not involve large numbers, and its interval-cancer rates must be interpreted in relation to their statistical confidence intervals. That is, standards cannot be derived from point estimates alone as they emanate from real studies in real populations. Further, comparison of the performance of screening services with such standards also requires that the confidence intervals of the screening service data be taken into account. Evaluations should not be based on point estimates of rates which derive from small numbers with considerable stochastic variation. The standard for Australian National Accreditation Requirements is less than six interval cancers per 10 000 screens occurring in the first 12 months. However, no age range or standardisation is specified, and confidence intervals are not considered. Another consideration in setting performance standards for mammographic screening services is the difference in results from experimental and operational studies. Randomised trials and other specially constructed studies usually attract significant financial resources and interested investigators. It may be more reasonable to derive standards from studies of operational mammographic screening services whose data, when sufficient studies have been reported, could be assessed to determine performance standards which could reasonably be expected. However, until operational mammography screening services have been shown to produce reduced breast cancer mortality in the screened populations, the results of successful trials must set the gold standard. Acknowledgements The Central & Eastern Sydney Service of BreastScreen NSW and BreastScreen Australia is jointly funded by the Commonwealth and the NSW State Governments. The NSW Central Cancer Registry, which is administered by the NSW Cancer Council and funded by the NSW Health Department, provided valuable assistance with data linkage. References Duffy SW, Tabar L, Fagerberg G, et al. Breast screening, prognostic factors and survival -- results from the Swedish two county study. Br J Cancer 1991; 64: 1133-1138. Day NE, Williams DRR, Khaw KT. Breast cancer screening programmes: the development of a monitoring and evaluation system. Br J Cancer 1989; 59: 954-958. Tabar L, Fagerberg G, Day NE, Holmberg L. What is the optimum interval between mammographic screening examinations? An analysis based on the latest results of the Swedish two-county breast cancer screening trial. Br J Cancer 1987; 55: 547-551. Australian Health Ministers' Advisory Council. Breast Cancer Screening Evaluation Committee. Breast cancer screening in Australia: future directions. Australian Institute of Health, Prevention Program Evaluation Series No. 1. Canberra: AGPS, 1990. National Program for the Early Detection of Breast Cancer -- national accreditation requirements: March 1994. Canberra: Commonwealth Department of Human Services and Health, December 1994. Rickard MT, Lee W, Read JW, et al. Breast cancer diagnosis by screening mammography: early results of the Central Sydney Area Health Service Breast X-Ray Program. Med J Aust 1991; 154: 126-131. Fellegi IP, Sunter AB. A theory for record linkage. J Am Statistical Assoc 1969; 64: 1183-1210. Jaro M. Advances in record linkage methodology as applied to matching the 1985 census of Tampa, Florida. J Am Statistical Assoc 1989; 84: 414-420. Jaro M. Automatch. Generalised record linkage system. Silver Spring, Md: Matchware Technologies Inc, USA, 1994. Taylor R, Smith D, Hfyer A, et al. Breast cancer in New South Wales 1972-91. Sydney: NSW Central Cancer Registry and Cancer Epidemiology Research Centre, NSW Cancer Council, September 1994. Coates M, Day P, McCredie M, Taylor R. Cancer in NSW, incidence and mortality 1992. Sydney: NSW Central Cancer Registry and Cancer Epidemiology Research Centre, NSW Cancer Council, 1995. Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford: Scientific Publications, 1994. Lentner C, editor. Geigy scientific tables. Volume 2: Poisson distribution. Basle, Switzerland: Ciba-Geigy, 1982: 152. Peeters PHM, Verbeek ALM, Hendriks JHCL, et al. The occurrence of interval cancers in the Nijmegen screening programme. Br J Cancer 1989; 59: 929-932. Woodman CBJ, Threlfall AG, Boggis CRM, Prior P. Is the three year breast screening interval too long? Occurrence of interval cancers in NHS breast screening programme's north western region. BMJ 1995; 310: 224-226. Rickard MT, Donnellan M. Diagnosis of small sized invasive breast cancer by an Australian mammography screening service: surrogate end points for mortality reduction. Aust N Z J Surg 1998; 68: 426-429. Robinson JI, Crane CEB, King JM, et al. The South Australian Breast X-Ray Service: results from a statewide mammographic screening programme. Br J Cancer 1996; 73: 837-842. Smith D, Oudod V, Supramaniam R, et al. BreastScreen NSW. Statistical Report 1991-1995. Sydney: NSW Cancer Council, 1996. Coates M, McCredie M, Armstrong B. Cancer in NSW, incidence and mortality 1993. Sydney: Cancer Control Information Centre, NSW Cancer Council, 1996. Parkin DM, Muir CS, Whelan SL, et al, eds. Cancer incidence in five continents. Volume VI. World Health Organization (WHO), International Association of Cancer Registries (IACR), International Agency for Research on Cancer (IARC). Lyon: IARC, 1992. (IARC Scientific Publication No. 120.) Field S, Michell M J, Wallis MGW, Wilson ARM. What should be done about interval breast cancers? BMJ 1995; 310: 203-204. (Received 7 Aug 1997, accepted 3 Apr 1998) Authors' details BreastScreen NSW, Central & Eastern Sydney, Sydney, NSW. Mary T Rickard, FRACR, MPH, Director and Radiologist; Mohamed A Fazli, BScEng, ME, Database Manager; Nadima El Hassan, BEc(Hons), Computer Consultant. Department of Public Health and Community Medicine, Faculty of Medicine, University of Sydney, Sydney, NSW. Richard J Taylor, FAFPHM, FRCP, Associate Professor in Public Health. Reprints will not be available from the authors. Correspondence: Dr M T Rickard, BreastScreen NSW, Central & Eastern Sydney, PO Box 1535, Strawberry Hills, NSW 2012. E-mail: MaryATces.bci.org.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Mary T Rickard · Richard J Taylor · Mohamed A Fazli

Decision making in CPR: attitudes of hospital patients and healthcare professionals Ian H Kerridge, Sallie-Anne Pearson, Isobel E Rolfe and Michael Lowe MJA 1998; 169: 128-131 For editorial comment, see Ryan Abstract - Introduction - Methods - Results - Discussion - Acknowledgments - References - Authors' details - - ©MJA1998 Abstract Objective: To examine the opinions of patients and healthcare professionals regarding the process of making decisions about cardiopulmonary resuscitation (CPR). Design and participants: A cross-sectional survey of 511 healthcare professionals (doctors, nurses and allied health professionals) (64% response rate) and 152 patients (58% response rate) at the John Hunter Hospital, Newcastle, New South Wales, in June 1994. Main outcome measures: Opinions on who should be involved in CPR decision making; what issues are important when making the decision; and how these decisions should be communicated. Results: 80% (95% confidence interval [CI], 72%-86%) of patients and 99% (95% CI, 98%-100%) of healthcare professionals (P < 0.001) thought patients' views should be taken into account when making CPR decisions. More patients (29%; 95% CI, 22%-38%) than healthcare professionals (14%; 95% CI, 11%-17%) indicated that doctors should be the main decision makers. Two-thirds of respondents regarded the patient's wishes, diagnosis and quality of life as important factors. Most respondents (82%) felt comfortable discussing CPR, but only 29% (95% CI, 22%-37%) of patients and 57% (95% CI, 52%-61%) of healthcare professionals had actually discussed CPR with others (P < 0.001). More than half of all respondents preferred to express their wishes about CPR in writing (47% [95% CI, 39%-55%] of patients, 69% [95% CI, 64%-73%] of healthcare professionals; P < 0.01); the others preferred to tell a family member or close friend. Most patients (60%; 95% CI, 52%-68%) and healthcare professionals (85%; 95% CI, 81%-88%) wanted their views in their medical records (P < 0.001). Conclusion: Most patients want to be involved in CPR decision making and many want some form of advance directive. Although there are some differences in opinions between patients and healthcare professionals, both perceive decision making at the end of life as a shared process, primarily involving the patient and doctor. Introduction Cardiopulmonary resuscitation (CPR) has been used in hospitals for about 30 years.1 Early studies of CPR in highly selected populations demonstrated its effectiveness and CPR soon became routine for any patient who died in hospital. However, as experience accumulated it became apparent that many patients, particularly those with chronic diseases, did not benefit from CPR. Hospitals responded by developing procedures for withholding CPR through the documentation of "do-not-resuscitate" orders. Decisions to initiate or withhold CPR were originally taken by doctors, but changing attitudes towards the doctor-patient relationship and the rise of medical consumerism have challenged this form of unilateral decision making. Indeed, respect for patient autonomy suggests that patients should have the right to contribute to such decisions. No Australian studies reported to date have examined patient preferences regarding CPR decision making. International studies suggest that most patients do not consider discussions about CPR preferences to be cruel or insensitive,2 and most wish to participate in decisions regarding CPR and other life-sustaining therapies.3 However, many patients are unable to participate during the final stages of life-threatening illnesses.4 The use of advance directives has been proposed as a means by which competent patients may ensure that their wishes will be carried out. Advance directives are written or verbal statements in which patients set out their preferences regarding life-sustaining treatment in case they later become incompetent. Advance directives have received considerable attention in the United States,5 where they have widespread legal recognition,6 but have received less publicity in Australia. In the absence of advance directives, CPR decisions are generally made by healthcare professionals in consultation with patients' families.7 The aim of this study was to compare the opinions of patients and healthcare professionals about several aspects of CPR: who should be involved in making the decision, what issues are considered important, and how these decisions should be communicated. Methods Survey We carried out a cross-sectional survey of healthcare professionals (doctors, nurses and allied health professionals) and patients at the John Hunter Hospital, Newcastle, New South Wales, in June 1994. The John Hunter Hospital is a 700-bed teaching hospital in an urban area. The survey formed part of a larger study that aimed to compare patients' and healthcare professionals' attitudes and knowledge about CPR. Survey questions were developed by reviewing published articles, incorporating questions from previously administered surveys,8,9 and consulting experts in the fields of ethics and questionnaire development. After pilot testing, the questionnaire was modified to incorporate the suggestions of a sample of patients and healthcare professionals at a local district hospital. The questionnaire included questions on sociodemographic information (age, sex, marital status, educational, religious and ethnic background) and self-reported health status. Respondents were then asked: whether doctors should ask patients if they would like CPR should they require it; whether patients have the right to choose or refuse resuscitation; who (from a list of people) should be involved in CPR decision making and who should be the main decision maker; which of eight issues listed were important in CPR decision making; whether discussing CPR was cruel or insensitive; whether they had discussed CPR and, if so, with whom; whether they would like patients' views about CPR to be expressed in writing before or soon after admission to hospital; and whether they would like patients' views about CPR to be included in medical records. The survey was administered to all healthcare professionals working over a two-day period at the John Hunter Hospital and to all eligible inpatients over an adjacent two-day period. Self-complete questionnaires were delivered by hand and collected as soon as they had been completed. Patients were initially screened by one of the principal investigators in consultation with a nursing unit manager and excluded if they were regarded as unable to complete the questionnaire or could not understand English. The remaining patients were asked by trained interviewers to participate. After consent was obtained, a Mini Mental State Examination (MMSE) was carried out to assess cognitive status. Patients with an MMSE score less than 24 were excluded, as this may indicate incompetence.10 Statistical analysis Data were analysed with SPSS for Windows.11 Continuity-corrected chi-square analysis was used to compare the sociodemographic variables and the responses of patients and healthcare professionals for each questionnaire item. Ethical approval The Hunter Area Health Service Research Ethics Committee and the University of Newcastle Research Ethics Committee approved the study. Results Respondents Of the 803 questionnaires delivered to hospital staff, 511 were returned -- a response rate of 64%. Of these respondents, 148 were doctors, 312 were nurses and 51 were from other healthcare professions, including physiotherapy, social work, occupational therapy, speech therapy and nutrition and dietetics. There were 443 adult patients in the John Hunter Hospital at the time of the study; 183 were excluded because of dementia, neurological impairment or delirium (110 of those excluded), incompetence (35), visual problems (24), psychiatric illness (9), or because they were undergoing procedures (5). Of the 260 remaining patients, 152 completed the questionnaire -- a response rate of 58%. Healthcare professionals were more likely than patients to be younger, female, have tertiary qualifications and report being healthy. Healthcare professionals and patients did not differ in terms of marital status, religion or ethnic background. Patient involvement in CPR decision making Patients and healthcare professionals differed significantly in their responses to whether doctors should ask patients if they would like CPR should they require it (P < 0.001): 23% (95% confidence interval [CI], 16%-30%) of patients and less than 1% (95% CI, 0-2%) of healthcare professionals thought that doctors should never ask patients, whereas 30% (95% CI, 22%-37%) of patients and 41% (95% CI, 37%-46%) of healthcare professionals thought that they should always ask patients. The remaining respondents were unsure or thought it depended on the patient's circumstances. There was also a significant difference in the way healthcare professionals and patients responded regarding the right to choose or refuse CPR (P < 0.001): 32% (95% CI, 25%-40%) of patients and 1% (95% CI, 0-2%) of healthcare professionals felt that patients should never have this right. Fewer patients (34%; 95% CI, 26%-42%) than healthcare professionals (67%; 63%-71%) believed that patients should always have the right to choose or refuse resuscitation, and the remaining respondents thought it depended on the patient's circumstances. Who should be involved in decision making More than 80% of patients and healthcare professionals indicated that patients and their doctors are important in CPR decision making (Table). Approximately half the respondents regarded the patient's partner and family members as important, but only a third regarded nursing staff as important decision makers. Interestingly, 47% (95% CI, 41%-52%) of nurses believed that they should be involved, whereas only 23% (95% CI, 16%-31%) of doctors and 24% (95% CI, 13%-37%) of allied health professionals believed that nurses should be involved. A significantly lower proportion of patients than healthcare professionals indicated that patients and their partners are important in making the decision. Approximately two-thirds of respondents identified the patient and less than one-third identified the doctor as the person who should be the main decision maker. There was a significant difference in the way healthcare professionals and patients responded. Fewer patients (55%; 95% CI, 47%-63%) than healthcare professionals (85%; 95% CI, 82%-88%) thought the patient should be the main decision maker, whereas 29% (95% CI, 22%-38%) of patients but only 14% (95% CI, 11%-17%) of healthcare professionals thought the doctor should be the main decision maker. The other respondents identified nurses, partners or family members. Important factors in the decision The patient's wishes, diagnosis and quality of life were identified by at least two-thirds of all respondents as important factors in CPR decision making (Figure). Patients were less likely than healthcare professionals to regard patient's wishes (P < 0.001), diagnosis (P < 0.001) and quality of life (P < 0.001) as important, but were more likely to regard age (P < 0.001), family wishes (P < 0.01), mental state (P < 0.03) and social circumstances (P < 0.01) as important. Discussing CPR Most respondents (82% in both groups) did not think discussing CPR was cruel or insensitive. Although 92% (95% CI, 86%-96%) of patients and 96% (95% CI, 94%-97%) of healthcare professionals (P < 0.001) indicated that they had family members who could make decisions for them if they were unable to, only 29% (95% CI, 22%-37%) of patients and 57% (95% CI, 52%-61%) of healthcare professionals (P < 0.001) had discussed with others whether they would like CPR. For all respondents, any discussions that had taken place were with partners (37%; 95% CI, 33%-41%), and/or family members (26%; 95% CI, 23%-30%) and/or doctors (3%; 95% CI, 2%-4%). Advance directives More than half the respondents (47% [95% CI, 39%-55%] of patients, 69% [95% CI, 64%-73%] of healthcare professionals; P < 0.01) preferred to express their wishes about CPR in writing before or soon after being hospitalised. The remaining respondents preferred to tell a family member or close friend about their wishes. About two-thirds of all respondents indicated that they would like their views about CPR to be part of their medical records. Fewer patients (60%; 95% CI, 52%-68%) than healthcare professionals (85%; 95% CI, 81%-88%) wanted their views in their medical records (P < 0.001). The remaining respondents were either uncertain or did not wish to have their views as part of their records. Discussion We found that many patients want to be involved in CPR decision making and many want some form of advance directive. Patients and healthcare professionals generally think that decisions about CPR should be shared by doctors and patients, with patients more likely to feel that doctors should be the main decision maker in such circumstances. This study has several flaws. The study uses a self-report questionnaire, and the low response rate among patients suggests that the group may not have been representative of the population as a whole. Indeed, there is likely to have been a systematic bias towards those who were willing to discuss CPR. Despite this, the findings of this study are of interest, particularly as the study sampled inpatients and had methodological strengths, including a process for the assessment of patient competence and the use of trained interviewers who were not members of the patients' own healthcare team, and thus less likely to influence their participation in the study or their responses to the questions about CPR. International studies have demonstrated that most patients would like to be involved in CPR decision making.12 Layson et al, reviewing international published studies, found that 52%-87% of patients had considered their preferences for life-sustaining therapy (CPR and admission to intensive care) and wished to discuss their preferences with their doctors, but few did so.13 We found that 80% of patients believed they should be involved in CPR decisions, but only 29% had discussed CPR with their doctors or their families. Forty-seven per cent of patients in this study would have liked to express their wishes regarding CPR in writing, either before admission or shortly after. The remainder preferred to tell a family member or a close friend. However, very few patients reported giving advance directives. There are few data to validate routine use of advance directives in hospitals. In the largest study to date that looked at the introduction of advance directives in routine care, their use did not affect patient-physician communication, physician's knowledge of patient preferences, the number of patients receiving ventilation or the use of hospital resources.14 Interestingly, that study was based upon the extensive use of nurses to assess patient preferences and communicate these results to physicians. Hospital staff appear to place more importance upon the role of patients in CPR decision making than patients do. This has been noted in previous studies of patient preferences about decision making,15,16 leading commentators such as Ende et al to suggest that the current emphasis upon patient autonomy arises more from normative ethical reasoning than from patients' actual preferences.16 Other researchers suggest that this is a misconception arising from combining several different aspects of decision making into a single question. Deber and Baumann suggest that shared decision making has two parts:17 problem solving, in which an expert identifies diagnostic and treatment options and expected risks and benefits, and decision making per se, in which the doctor and patient together determine an appropriate course of action.18 According to Deber et al, "given the choice between abiding by the decisions of a paternalistic provider and being handed a stack of medical books and being told to figure out what they wish to do, most patients unsurprisingly opt to hand over control to a trusted provider".18 Both these options are probably inappropriate, and if given a wider choice patients are unlikely to opt for unilateral decision making. Our study suggests that, despite differences between the views of healthcare professionals and patients, the overwhelming majority of both groups support patient and doctor involvement in shared decision making about CPR. Who else should be involved in CPR decision making? Both patients and healthcare professionals believe that partners and some family members should be involved. Surprisingly, only 36% of patients believed that nurses should be involved in these decisions. The reasons for this are likely to be complex. There is no doubt that nurses have a genuine stake in discussions about resuscitation. When a patient has a cardiac arrest, nurses are faced with an ethical decision: to follow their own ethical standards, or to follow the protocols documented in patient notes and hospital procedures, if such documentation exists. Within this context it is interesting to note that in a recent study of American critical care nurses 16% reported that they had participated in euthanasia or assisted suicide, many without the knowledge of physicians, patients or patients' surrogates, and 4% reported that they had hastened a patient's death by only pretending to apply life-sustaining treatment ordered by a physician.19 Nurses are independent moral agents, and there are good reasons for nurses to be involved in CPR decision making. However, there are considerable practical difficulties as nurses care for patients in shifts, and it is not possible to involve every nurse who looks after a patient. It is clearly inadequate to include only one nurse, and then to document "nursing staff in agreement". Perhaps some of this uncertainty filters through to patients. Without continuity of nursing care, patients are unlikely to consider nurses to be major participants in CPR decision making. In addition, the professional image of nurses as moral agents in their own right may not have been communicated adequately to patients or other healthcare professionals. Either way, this finding challenges the notion of "nurse as patient advocate". Differences between staff and patients in preferences about CPR decision making may influence the way decisions are made in hospital and limit the capacity for healthcare professionals to act as patient advocates. This suggests that the proper role for advance directives and the process of CPR decision making requires further study. The effect of determinants such as age, health status, education and religion on CPR decision making also requires further research. Simplistic assertions of autonomy or of the theoretical moral or legal value of advance directives without consideration of the importance of care, compassion or clinical context are clearly inappropriate. This study suggests that both healthcare professionals and patients see decision making at the end of life as a complex, shared process rather than a sterile abstraction of expert information disclosure followed by patient self-determination. Acknowledgements We wish to acknowledge student research assistants from the Faculty of Medicine and Health Sciences for their assistance with data collection and the staff and patients of the John Hunter Hospital for their generous participation in this study. References Robertson GS. Resuscitation and senility: a study of patients' opinions. J Med Ethics 1993; 19: 104-107. Stolman CJ, Gregory OJ, Dunn D, Levine JL. Evaluation of patient, physician, nurse and family attitudes toward do not resuscitate orders. Arch Intern Med 1990; 150: 653-658. Reilly BM, Magnassen R, Ross J, et al. Can we talk? Inpatient discussions about advance directives in a community hospital. Arch Intern Med 1994; 154: 2299-2308. Bedell SE, Pelle D, Maher PL, et al. Do-not-resuscitate orders for critically ill patients in the hospital: how are they used and what is their impact? JAMA 1986; 256: 233-237. Council on Ethical and Judicial Affairs. American Medical Association. Guidelines for the appropriate use of do-not-resuscitate orders. JAMA 1991; 265: 1868-1871. Fisher RH, Meslin EM. Should living wills be legalised? CMAJ 1990; 142: 23-26. Hare J, Pratt C, Nelson C. Agreement between patients and their self-selected surrogates on difficult medical decisions. Arch Intern Med 1992; 152: 1049-1054. Frankl D, Oye RK, Bellamy P. Attitudes of hospitalised patients toward life support: a survey of 200 inpatients. Am J Med 1989; 86: 645-648. Murphy DJ, Burrows D, Santilli S, et al. The influence of the probability of survival on patient's preferences regarding cardiopulmonary resuscitation. N Engl J Med 1994; 330: 545-549. Folstein MF, Folstein SE, McHugh PR. Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: 189-198. SPSSx statistical package for the social sciences [computer program]. Version 6.0. Chicago, Ill: SPSS Inc, 1990. Morgan R, King D, Prajapati C, Rowe J. Views of elderly patients and their relatives on cardiopulmonary resuscitation. BMJ 1994; 308: 1677-1678. Layson RT, Adelman HM, Wallach PM, et al. Discussions about the use of life-sustaining treatments: a literature review of physicians' and patients' attitudes and practices. J Clin Ethics 1994; 5: 195-199. The SUPPORT Principle Investigators. A controlled trial to improve care for seriously ill hospitalised patients. JAMA 1995; 274: 1591-1592. Strull WM, Lo B, Charles G. Do patients want to participate in medical decision making? JAMA 1984; 252: 2990-2994. Ende J, Kazis L, Ash AB. Measuring patients' desire for autonomy: decision making and information seeking preferences among medical patients. J Gen Intern Med 1989; 4: 23-30. Deber RB, Baumann AO. Clinical reasoning in medicine and nursing: decision-making versus problem-solving. Teach Learn Med 1992; 4: 140-146. Deber RB, Kraetschmer N, Irvine J. What role do patients wish to play in treatment decision making? Arch Intern Med 1996; 156: 1414-1420. Asch DA. The role of critical-care nurses in euthanasia and assisted suicide. N Engl J Med 1996; 334: 1374-1379. (Received 14 Jul 1997, accepted 1 Apr 1998) Authors' details Faculty of Medicine and Health Sciences, University of Newcastle, NSW. Ian H Kerridge, BMed, MPhil, Lecturer in Clinical Ethics, Clinical Unit in Ethics and Health Law, and Haematology Registrar, John Hunter Hospital; Sallie-Anne Pearson, BSc(Hons), Professional Officer, Programme Evaluation Committee; Isobel E Rolfe, FRACP, MRCP, Senior Lecturer in Medical Education; Michael Lowe, BMed, Tutor in Clinical Ethics, and Medical Registrar, John Hunter Hospital. Reprints will not be available from the authors. Correspondence: Dr I H Kerridge, Clinical Unit in Ethics and Health Law, John Hunter Hospital, Locked Bag 1, Hunter Regional Mail Centre, Newcastle, NSW 2310. E-mail: ikerridgeATmail.newcastle.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Ian H Kerridge · Sallie-Anne Pearson · Isobel E Rolfe · Michael Lowe

The stress of metropolitan general practice Peter L Schattner and Greg J Coman MJA 1998; 169: 133-137 For editorial comment, see Douglas & Sibthorpe Abstract - Introduction - Methods - Results - Sex differences - Discussion - Acknowledgements - References - Authors' details - - ©MJA1998 Abstract Objective: To identify the work-related stressors of Australian metropolitan general practitioners (GPs). Design and setting: A descriptive postal survey of metropolitan GPs from all States and Territories selected at random from the Health Insurance Commission database. Participants: 296 of 464 GPs (64%) surveyed in June 1996; 67% were male; 87% worked full-time (more than 6 sessions per week). Main outcome measures: Frequency and severity of work stresssors in general practice; overall feelings of stress at work in the past 12 months; effects of the stressors on work satisfaction; contribution of work stress to overall life stress; responses to the 12-item General Health Questionnaire (GHQ) as potential correlates of occupational stress. Results: "Time pressure to see patients" was the most frequently reported stressor. Threat of litigation was perceived as the most severe stressor. Of the top 10 severe stressors, seven were also in the top 10 for stressor frequency. Work was the major stressor in GPs' lives. The GHQ scores did not correlate significantly with major stress outcome measures, but 12.8% of GPs had scores indicative of severe psychiatric disturbance. Fifty per cent of respondents had considered leaving their current workplace and 53% had considered abandoning general practice because of occupational stress. GPs working 6 or more sessions per week were more likely to be moderately or severely stressed than those working part-time (P < 0.02, Fisher's exact test). Those who had considered leaving their current workplace or careers were also more likely to be moderately or severely stressed (P < 0.0001, Fisher's exact test). Conclusions: The most frequent and relatively severe stressful events in general practice involved time pressures. There are implications for government, which, through remuneration policies, might influence GPs to work at a rate beyond their capacity to cope. Strategies are required to manage or prevent stress in metropolitan GPs. Introduction Stress has been defined as a response to challenging events.1 It is usually thought of as "distress" or an inability to cope with an external factor (the "stressor"). In the workplace it is generally referred to as occupational stress. Australian and international studies suggest that general practice is a stressful occupation.2-7 The extent to which GPs feel stressed by various aspects of their occupation may affect the quality of patient care and may also affect practitioners' health. Health problems experienced by GPs include alcohol and drug abuse,8-10 marital disruption and divorce,3,11,12 anxiety,3,13 burnout,14 depression,15 as well as suicide and attempted suicide.16,17 The specific characteristics that make general practice so stressful are largely unknown, although anecdotal evidence would suggest that time and financial issues are major stress factors for Australian GPs. In our study we aimed to identify the sources, frequency and severity of occupational stress experienced by metropolitan GPs; the importance of occupational stress compared with other life stressors; the effects of the work stressors on GPs' job satisfaction; and the possible effects of these stressors on GPs' psychological health. Rural GPs were not included, as factors such as isolation and difficulty in obtaining locum relief suggest this group should be studied separately.18 Methods General practitioner recruitment and survey distribution A list of 500 potential survey participants was obtained from the Health Insurance Commission database of the Commonwealth Department of Human Services and Health (now the Department of Health and Family Services). Potential participants were randomly selected from the population of GPs in all States and Territories who had capital city practices only and more than 1500 consultations annually (the number generally accepted by the General Practice Evaluation Programme as the definition of "active" GPs). The Department advised that potential participants had not recently been surveyed through this database. Questionnaires were mailed to all 500 GPs in June 1996. Non-respondents to the first mailout were sent a second questionnaire in July 1996, and non-respondents to this prompter were telephoned to request their participation in August 1996. Survey instrument We used a four-part questionnaire to collect data on:(1) GPs' demographic and practice information. (2) The frequency and severity of potentially stressful events in general practice, using a 28-item list developed from an extensive literature review which included validated stress-screening instruments,4,5,6,18 and input from a Melbourne-based GP focus group. A four-point Likert-like scale was used, with 3 representing "frequently, at least weekly"; 2, "occasionally, at least monthly"; 1, "rarely, a few times a year"; and 0, "does not occur". Another four-point Likert-like scale, with 3 representing "severe stress"; 2, "moderate stress"; 1, "mild stress"; and 0, "no stress", was used to assess stress severity for each item. (3) The effects of stressors on GPs' job satisfaction, including asking them to (a) indicate on a four-point scale (from 0 = none to 4 = severe) their overall work stress levels in the previous 12 months; (b) rank six aspects of general practice in order of contribution to work stress levels, from 1 for the greatest contribution to 6 for the least; (c) distinguish between stressors related to job content (ie, clinical duties) and job context (ie, environmental and organisational factors) (see Box 1); (d) rank six potential sources of life stress in order from 1 for the greatest source of stress to 6 for the least stressful. (4) The 12-item General Health Questionnaire (GHQ),19 used to determine the presence of psychiatric disturbance, anxiety or depression as potential correlates of occupational stress. GHQ items have four response options, which were scored 0, 0, 1 or 1. This provides a GHQ score range from 0 to 12. Using this method, scores of less than four represent negligible psychiatric disturbance, scores between four and eight suggest moderate disturbance, and scores greater than eight indicate severe psychiatric disturbance. Data entry and analysis Data were computer-coded and analysed using the Statistical Package for the Social Sciences for Windows.20 Analyses included cross-tabulations of stress outcome measures with the demographic data shown in Box 2. Non-parametric tests were used to test for significant differences at the P = 0.05 level. Potential work stressors were ranked in order of reported frequency and severity. Ethical approval Ethical approval was granted by the Monash University Standing Committee on Ethics in Research on Humans. Results Characteristics of respondents Of the 500 questionnaires mailed, 36 were returned unopened; 296 of the remaining 464 GPs participated (response rate, 64%). Box 2 shows their demographic characteristics, which are similar to the demographic profile of Australian metropolitan general practice.21 General practice stressors Frequency of general practice stressors The most frequent stressors were related to perceived lack of time. "Time pressure to see patients", "phone interruptions during consultations" and "too much work to do in a limited time" were ranked 1, 3 and 4, respectively, in frequency, with "paperwork" the second most frequent stressor (Box 3). Severity of general practice stressors The threat of litigation (actual involvement in litigation proceedings was not an item), although perceived as one of the most stressful events in general practice, was reported relatively infrequently. However, seven of the 10 most stressful events (Box 4) were also in the top 10 for frequency. The GPs appeared to experience a number of stressors which were perceived as mild to moderate, but which occurred frequently. The mean ratings for the three most severe stressors -- "threat of litigation", "too much work to do in a limited time" and "earning enough money in general practice" -- were between "mild" and "moderate". Effects of stress Overall work stress The Figure shows that most GPs (273/296; 92%) reported some feelings of stress, with 241 (81%) labelling it "mild" or "moderate" and 32 (10.8%) "severe". When asked about levels of stress in the past 12 months, 121 (41%) of GPs said they had increased, 118 (40%) said they had stayed the same, and 33 (11%) said they had decreased. Fifty per cent of GPs surveyed had considered leaving their current workplace, and 157 (53%) had considered abandoning general practice because of occupational stress. Factors associated with high overall work stress levels When the GPs' reported experience of work stress was cross-tabulated with the demographic variables reported in Box 2, the only significant association was that GPs working six or more sessions per week were more likely to be moderately or severely stressed than those working part-time (P < 0.02, Fisher's exact test). Those who had considered leaving their current workplace or abandoning their careers were also more likely to be moderately or severely stressed (P < 0.0001, Fisher's exact test). Six major causes of occupational stress Of the six most stressful aspects of general practice high workload was rated worst, followed, in descending order, by economic factors (income, running a business) and "medicopolitical" factors (involvement with professional associations, government pressures). Clinical factors and the effect of work on outside life were rated fourth and fifth, respectively, and the physical working environment was considered the dimension which contributed the least to the stresses of general practice (Box 5). These rankings were based on the means of rank scores; the first two items ranked significantly ahead of the other factors (P < 0.05). Sixty per cent of the GPs said that their experience of stress arose mainly from "job context" rather than from "job content". Comparisons with other life stresses Work was the major source of overall stress in GPs' lives, followed closely by financial concerns (Box 6). General Health Questionnaire data The 12-item GHQ was used to assess respondents' levels of psychiatric morbidity, especially depression and anxiety. The mean GHQ score was 3.01 (95% confidence interval, 2.66-3.35), suggesting that, on average, GPs experience low levels of psychiatric disturbance. However, almost a third (30.7%) of respondents recorded GHQ scores of 4 or more and 38 (12.8%) recorded scores of 8 or more, suggesting that many GPs are clinically depressed, anxious, or experiencing other psychiatric symptoms. The responses to three GHQ questions were of concern. These were Item 5 ("Have you recently felt constantly under strain?"), to which 7.1% responded "much more than usual"; Item 9 ("Have you recently been feeling unhappy and depressed?"), to which 5.1% responded "much more than usual"; and Item 11 ("Have you recently been thinking of yourself as a worthwhile person?"), to which 6.1% responded "not at all". Box 7 shows that, while significant, correlations between GHQ scores and the stress measures were not high. The highest correlation (0.53) was with GPs' overall perception of stress in the past 12 months. Sex differences There were small but statistically insignificant differences in the responses of men and women. Discussion Our study indicates that, while the worst stressors in general practice are perceived to be of only mild to moderate severity, many occur frequently. The study cannot determine whether there is a cumulative effect, in which a series of minor frustrations leads to a pervasive level of significant stress in GPs, but this may be an explanation as to why so many GPs have considered leaving their practices or abandoning general practice altogether. Apart from the threat of litigation, work overload issues, such as time pressure to see patients, phone interruptions and too much work to do in a limited time, were the most common stressors for GPs. Both GPs and their patients have been reported to be dissatisfied with the amount of time spent in consultations.4,7,22-25 Time pressure may be a reason for patients' reporting that GPs are not listening to what they are saying and not explaining things to them properly.26 Effects of time constraints on GPs may include irritability, frustration and, possibly, mistakes in diagnosis and treatment.4 Practice administration issues, which have not been traditionally taught to undergraduates, were rated as causing much more stress than issues of clinical competence. This may add to the pressure of high patient loads. The intrusion of work on family life was a significant stressor. Presumably, GPs would be less stressed if work did not impinge on family time or if they could see more of their family by working fewer hours. Although our list of potential sources of life stress was not exhaustive, we have not found a similar ranking attempt in the other stress studies reviewed. It would be interesting to see how this group of GPs compares with other occupational groups and the community at large in this regard. Concerns about remuneration and other financial concerns, such as the cost of practice overheads, were considered to be among the most frequent and severe stressors in this study (although they were not rated as highly as some might expect). This should be of interest to governments, which continue to emphasise cost-cutting in healthcare expenditure. It is of concern that almost a third of the respondents (91; 30.7%) had GHQ scores of 4 or more, suggesting moderate psychiatric disturbance, and that 38 (12.8%) scored 8 or more, suggesting severe disturbance. To some extent this is an independent finding; work stress cannot be blamed entirely for the presence of psychiatric disturbance in GPs, although it is likely to be a contributing factor. Our 64% response rate and the similarity of our sample's demographic profile to that of Australian metropolitan general practice21 suggest that the survey results are generalisable to urban GPs. It is possible that highly stressed GPs are under-represented because they were too stressed to comply with yet another demand on their time (the completion of a questionnaire). General practice may be no more or less stressful than many other "caring professions".4 However, the specific stressors highlighted in this study do suggest that government, healthcare policymakers, organisers of vocational training and GPs themselves should be aware of the high levels of stress experienced by doctors who choose this vocation, and steps should be taken to institute stress prevention and management. Acknowledgements Our thanks to the administrative staff at the Department of Community Medicine and General Practice at Monash University, as well as Dr Lynne Ham and Dr Chris Peterson for their expert advice on occupational stress, and Dr Con Tsalamandris for statistical analysis. This project was funded by a grant from the General Practice Evaluation Program of the Department of Human Services and Health. References Selye H. The stress of life. New York: McGraw Hill, 1976. Mawardi B. Satisfactions, dissatisfactions and causes of stress in medical practice. JAMA 1979; 241: 1483-1486. Roeske N. Stress and the physician. Psychiatr Ann 1981; 11: 245-258. Porter A, Howie J, Levinson A. Stress and the general practitioner. In: Payne R, Firth-Cozens J, editors. Stress in health professionals. Chichester: Wiley, 1987: 45-70. Makin P, Rout U, Cooper C. Job satisfaction and occupational stress among general practitioners -- a pilot study. J R Coll Gen Pract 1988; 38: 303-306. Sutherland V, Cooper C. Job stress, satisfaction and mental health among general practitioners before and after introduction of new contract. BMJ 1992; 304: 1545-1548. Sutherland V, Cooper C. Identifying distress among general practitioners: predictors of psychological ill health and job dissatisfaction. Soc Sci Med 1993; 37: 575-581. Ball J. Alcohol and drug use and related problems in the medical profession. Aust Drug Alcohol Rev 1986; 5: 29-32. Baldwin D, Hughes P, Conrad S, et al. Substance abuse among senior medical students. JAMA 1991; 265: 2074-2078. Serry N, Bloch S, Ball R, Anderson K. Drug and alcohol abuse by doctors. Med J Aust 1994; 160: 402-407. Bird H. The physician's marriage: joys and sorrows: life transition points. Facets 1979; Summer: 18-20. Miles J, Krell R, Lin T. The doctor's wife: mental illness and marital pattern. Int J Psychiatry Med 1975; 6: 481-487. Cramond W. Anxiety in medical practice: the doctor's own anxiety. Aust N Z J Psychiatry 1969; 3: 324-328. Lemkau J, Purdy R, Rafferty J, Rudisill J. Correlates of burnout among family practice residents. J Med Educ 1988; 63: 682-691. Smith M, Andrasik F, Quinn S. Stressors and psychological symptoms of family practice residents and spouses. J Med Educ 1988; 63: 397-405. Craig A, Pitts F. Suicide by physicians. Dis Nerv Syst 1968; 219: 763-772. Steppacher R, Mausner J. Suicide in male and female physicians. JAMA 1974; 228: 323-328. Dua J. Development of a scale to assess occupational stress in rural general practitioners. Int J Stress Manage 1996; 3: 117-128. Goldberg D, Williams P. Manual of the General Health Questionnaire. Windsor (UK): Nfer-Nelson, 1988. Statistical Package for the Social Sciences [computer program] release 6.0. SPSS, Chicago, Ill, USA, 1996. Commonwealth Department of Health and Human Services. General practice in Australia: 1996. Canberra: AGPS, 1996. Cartwright A, Anderson R. General practice revisited. London: Tavistock, 1981. Bates E. Doctors and their spouses speak: stress in medical practice. Sociol Health Illness 1982; 4: 25-39. Richardson IM, Howie J, Durno D, et al. A study of general practitioner consultations in North-East Scotland. J R Coll Gen Pract 1973; 23: 132-142. Brooks M, Stewart-Weeks M. Integrating consumer views about quality in general practice. Canberra: Consumers' Health Forum of Australia, 1996. Myerson S. Doctors' methods of dealing with "on going" stress in general practice. Med Sci Res 1991; 19: 267-269. (Received 7 Oct 1997, accepted 31 Mar 1998) Authors' details Department of Community Medicine and General Practice, Monash University, Melbourne, VIC Peter L Schattner, MMed BS, FRACGP, Director, Research Unit; Greg J Coman, GradDipAppPsych, MSc, Psychologist. Reprints will not be available from the authors. Correspondence: Dr P L Schattner, Department of Community Medicine and General Practice, Monash University, 867 Centre Road, East Bentleigh, Melbourne, VIC 3165. E-mail: peter.schattnerATmed.monash.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Peter L Schattner · Greg J Coman

Analysing potential harm in Australian general practice: an incident-monitoring study Alice L Bhasale, Graeme C Miller, Sharon E Reid and Helena C Britt MJA 1998; 169: 73-76 For editorial comment, see Kidd & Veale Abstract - Introduction - Method - Results - Discussion - Acknowledgements - References - Authors' details - - ©MJA1998 Abstract Objective: To collect data on incidents of potential or actual harm to general practice patients and to evaluate the possible causes of these incidents. Design: An observational study of incidents of potential harm based on a modified critical incidents technique. A non-random sample of general practitioners (GPs) anonymously submitted incident reports contemporaneously. Setting and participants: Australian general practices between October 1993 and June 1995. During the study period, 324 GPs participated at some time. Main outcome measures: GP-reported free-text descriptions of incidents and structured responses for preventability, potential for harm, immediate consequences, predicted long-term outcomes, type of incident, contributing factors, mitigating factors, and additional resource use. Results: 805 incidents were reported -- 76% were preventable; 27% had potential for severe harm. No long term harm was predicted for 66% of incidents. Incidents could relate to pharmacological management (51 per 100 incidents), non-pharmacological management (42 per 100 incidents), diagnosis (34 per 100 incidents) or equipment (5 per 100 incidents). The most common contributory factors were poor communication between patients and healthcare professionals and actions of others (23 per 100 incidents each) and errors in judgement (22 per 100 incidents). Conclusion: Human errors and preventable system problems were identified. The incident monitoring technique provided useful data which could be applied to incident prevention strategies. Introduction Increased interest in quality improvement in healthcare has led to greater recognition that errors occur. The Harvard Medical Practice Study1,2 and the Quality in Australian Health Care Study3 retrospectively reviewed hospital medical records to identify and quantify adverse events caused by the hospital care provided. These studies have provided important insight into rates of adverse outcomes. In Australia, around 16% of admissions studied were associated with an adverse event.3Following the success of incident monitoring in anaesthetics,4,5 in 1993 the Commonwealth Review of Professional Indemnity Arrangements for Health Care Professionals initiated pilot incident-monitoring studies in six medical specialties. The aim of this study was to test the incident-monitoring method in general practice. An important difference between studies such as the Harvard Medical Practice Study and incident-monitoring studies is that no claims of generalisability are made in the latter. Incident monitoring cannot estimate prevalence. It provides detailed contextual information about specific processes and situations that can lead to adverse patient outcomes. Such information is then used to make specific recommendations. The objectives of this study were to test a mechanism for obtaining incident reports from GPs; to gain an understanding of the type of incidents occurring in general practice and hence areas of potential harm; and to identify areas which might benefit from intervention. Method The method has been described in detail elsewhere.6 Participating general practitioners reported incidents contemporaneously, on purpose-designed reporting forms. An incident was broadly defined as "an unintended event, no matter how seemingly trivial or commonplace, that could have harmed or did harm a patient". This criterion included "near misses" -- where the harm may have been averted, but the potential for harm existed. All GP members of the Australian Sentinel Practice Research Network and the Central, Southern, Western and Wentworth Area Health Sentinel Networks in Sydney were asked by letter to participate. Others volunteered when reporting was offered as a Quality Assurance option through the Royal Australian College of General Practitioners (RACGP). Reports were anonymous and information was protected from subpoena by the Health Insurance Act 1973 (Cwlth).7 The reporting form included both free text and fixed-response questions. Information requested included GP characteristics and the patient's age, sex and health problems associated with the incident. Free text descriptions were elicited of the incident, its consequences, and contributing and mitigating factors; the outcome; and the GP's opinion of the incident's preventability. Fixed-choice responses included the type of incident, contributing factors, mitigating factors, potential for harm, immediate consequences, and predicted long term outcomes. All fixed-choice questions included an "other" category to allow continuing development of categories. Quantitative descriptive analysis was used to describe the main features of the data for all 805 incidents received, using SAS.8 No significance testing was carried out as it would have been invalid because of the nature of the study and sample. Qualitative analysis of textual data was carried out for 500 incidents. This analysis involved examination of groups of events for common causes, situations and contributing factors. Results GP response Of 510 GPs initially approached, 214 accepted. A further 110 GPs volunteered during the study. Between October 1993 and June 1995, 805 reports were received from 324 enrolled participants. As all reports were anonymous, reporting rates could not be calculated. However, a GP-generated personal identification number on most reports allowed estimation of the number of reports received from each identification number. There were 241 GPs who submitted at least one report, and 82 unidentified reports. The proportion of each sex and the proportion from rural practices among the respondents were equivalent to the overall GP population at that time.9 However, only 16% of respondents had less than 10 years' experience, whereas in a national sample in 1990-1991 about 30% of GPs had less than 10 years' experience.10 Patients Of the 805 incidents, 58% involved female patients. Almost one-third of incidents involved patients over 65 years old (median age, 50 years; range, 0-98 years). Younger patients (0-24 years) were least likely to be involved in incidents (Figure). Compared with the overall age-sex distribution of patients attending GP consultations, women older than 75 years and infants were over-represented in the incidents reported.9 Age and sex of patients involved in reported incidents. Incidents Incidents were categorised as pharmacological, non-pharmacological, diagnostic or equipment events (Box 1). Pharmacological events were most frequent. The most common incident type overall was omitted or delayed treatment. A high degree of overlap between non-pharmacological and diagnostic events was noted -- 79 incidents involved both types. GPs thought 76% of the incidents were preventable (Box 2) and 11% were unpreventable. They were undecided about the remaining 13%. More than one-third of the incidents resulted in additional use of resources. While pharmacological incidents were most often reported, they were also largely considered preventable (79%) and the least harmful potentially and in the long term. In contrast, diagnostic events were less preventable and more harmful (Box 2). Of the 38 deaths reported, 30 involved a diagnostic incident. Major immediate consequences were reported in 17% of incidents, and 4% resulted in the patient's death (Box 2). GPs predicted that most incidents would cause no long term harm, but that 7% would result in the future death of the patient. Mitigating and contributing factors Incident outcomes were more frequently mitigated by chance factors than through systematised fail-safe or preventive procedures. The most frequently reported mitigating factors were good fortune, early intervention by the GP, and the patient's good physical condition (Box 3). Ineffective communication and problems in coordination between healthcare services were frequent contributing factors (Box 4). Actions or involvement of others in caring for patients were also frequently cited. Assessment-related mistakes constituted another cluster of contributing factors (eg, errors in judgement, failure to recognise significant signs and symptoms, and inadequate patient assessment). Insufficient use of standard resources and protocols by the GP was also recognised. GPs reported being tired, rushed or late in 10% of incidents, but GP memory lapses and stressful situations were least frequently cited as contributing factors. There were no obvious associations between specific contributing factors and incident type. Qualitative analysis Qualitative analysis of free-text responses identified four broad problem groups: communication, procedural, clinical and external problems. More comprehensive descriptions are available elsewhere.11,12Communication problems: Most communication problems between GPs and patients occurred when patients did not understand some aspect of their condition, the management plan or the treatment. Patients with mental health problems or lacking English language skills were particularly at risk. Some serious pharmacological incidents occurred when patients misunderstood instructions and took newly prescribed drugs in addition to, rather than instead of, a previously prescribed medication. Poor communication also manifested as clinical error in inadequate history taking. There were also problems with communication between healthcare providers. Clinical information about the outcomes of hospital referrals or admissions, the expected role of the GP in post-discharge care, specialists' recommendations for management and results of tests or investigations was often not communicated between GPs and other healthcare providers, or was too late to be of use. Unclear or uninformative medical records caused a number of incidents, and inadequate records resulted in poor communication within practices. Incorrect or unclear labelling of drugs resulted in administration of the wrong drug, and inappropriate drugs were dispensed because of poor handwriting on prescription forms. Such errors were often compounded by similar medication names (eg, Teldane/Feldene, Microlax/ Murelax). Procedural problems: The main procedural error was failure to check medical records. Diagnostically, procedural error manifested in poor application of the differential diagnosis, when the correct diagnosis was inappropriately rejected because of insufficient or incorrect examination, or was not included in the differential diagnoses and was therefore not evaluated. In non-pharmacological incidents, treatment procedures may have been neglected (eg, not checking for anaesthesia before an excision). In equipment incidents, machines were not properly checked before use (eg, leads or tubes were disconnected) or were not well maintained (eg, unfilled emergency oxyviva equipment). Lack of protocols for ensuring action on results of tests and investigations meant that important results were filed unseen or left until the GP who ordered the test was next in the surgery. Recall systems were often inadequate, preventing recall of patients for follow-up tests and investigations and resulting in missed or delayed diagnosis or management. Clinical problems: These consisted of incidents that could only be attributed to human error. In pharmacological incidents, clinical errors occurred where an inappropriate drug was chosen; in non-pharmacological incidents, where the wrong treatment was chosen; in diagnostic incidents, where symptoms and signs were not recognised (eg, chest pain considered to be musculoskeletal); and in equipment incidents, where equipment was used inappropriately. External problems: External problems involved factors outside the GP's control, such as actions of other healthcare providers, equipment faults or unavailability, disease-related factors, and unpredictable drug effects. Patient-related factors included non-compliance, age, level of comprehension, and psychological condition. Discussion The diversity of incidents reflects the heterogeneous nature of general practice and its context within the healthcare system. Nevertheless, it was possible to isolate recurrent themes and errors and to identify incidents that occurred because of system problems or other preventable factors. Pharmacological incidents were most frequent and could be prevented through clearer prescription writing, double-checking of doses and potential interactions, and improved patient understanding. Unexpected drug reactions (eg, to previously well-tolerated drugs) and idiosyncratic complications were largely beyond GPs' control. However, compilation of this information may be useful to reinforce or add to data collected elsewhere. Diagnostic incidents had the most serious outcomes. While management of test results and lack of recall systems contributed to these incidents, many occurred while the patient was under another medical practitioner's care, or were caused by the diagnostic decision-making processes of GPs. Patient misunderstanding was a factor in many incidents. This might be addressed by improved GP awareness to ensure full patient understanding. However, information management problems require "hard-wired" solutions -- structures and systems that ensure better outcomes. In particular, opportunities for errors may be reduced by better recall systems for follow-up and for regular tests and screening,13 better transfer of medical information between healthcare providers, clearer writing of prescriptions, and better maintenance and use of medical records. Information technology could help by streamlining some of the administrative functions, particularly prescription writing, which appears to be adversely affected when GPs are tired, rushed, or running late. Limitations to the validity of our data include the non-random nature of the sample, limited recognition of incidents, selectivity in reporting incidents, and the lack of an alternative perspective. (We did not seek the patients' perspective because of the controversial nature of the data.) However, issues we identified as important have been described by other researchers. Many of the issues pertaining to pharmacological incidents have been described in a study of adverse drug events in United States hospitals.14 Similarly, the problems of multiple prescribing in the elderly have been well summarised, and opportunities for improvement in the management of polypharmacy have been identified.15 Prescription clarity has also been recognised as an area of potential misunderstanding.16 In critical incident studies, incidents are sampled until "saturation" (until no new incident types are reported).17 While these 805 incidents may not reach saturation, they demonstrate at least some of the types of incidents occurring in Australian general practice. Logically, where inadequate systems are widely used, similar events are likely to occur regularly rather than being related to individual characteristics or techniques. Continued collection and classification of events will elucidate more causal factors, particularly for the less common types of incidents. This study also resulted in a number of suggested interventions.18 Incident monitoring studies are intended to be continually used in improving systems. It is important to use the data for formulating interventions. This study has shown that such data can be collected in general practice, but the translation of findings into active change remains to be addressed. Acknowledgements We gratefully acknowledge the funding provided by the Commonwealth Review of Professional Indemnity Arrangements for Health Care Professionals. The project was conducted under the auspices of the RACGP. Recruitment of GPs and administration of the GP participants was carried out jointly by the Family Medicine Research Unit and the RACGP Research and Health Promotion Unit, South Australia. Other members of the research team made considerable contributions to running the project. Thanks must go to the GP participants for so generously sharing their experiences with us. References Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients: results of the Harvard Medical Practice Study I. N Engl J Med 1991; 324: 370-376. Leape LL, Brennan TA, Laird NM, et al. The nature of adverse events in hospitalized patients: results of the Harvard Medical Practice Study II. N Engl J Med 1991; 324: 377-384. Wilson RM, Runciman WB, Gibberd RW, et al. The Quality in Australian Health Care Study. Med J Aust 1995; 163: 458-471. Williamson JA, Mackay P. Incident reporting. Med J Aust 1991; 155: 340-343. Webb R, Currie M, Morgan CA, et al. The Australian incident monitoring study: an analysis of 2000 incident reports. Anaesth Intensive Care 1993; 21: 520-528. Britt H, Miller G, Steven I, et al. Collecting data on potentially harmful events: a method for monitoring incidents in general practice. Fam Pract 1997; 14: 101-106. Commonwealth of Australia. Health Insurance Act 1973. Declaration under section 106L. Commonwealth of Australia Gazette 1993; GN 31: 2269-2272. SAS [computer program]. Version 6.11. Cary, North Carolina: SAS Institute, 1995. General Practice Branch. Commonwealth Department of Health and Family Services. General Practice in Australia: 1996. Canberra: Commonwealth of Australia, 1996. Bridges-Webb C, Britt H, Miles DA, et al. Morbidity and treatment in general practice in Australia 1990-1991. Med J Aust 1992; 157 Suppl: S1-S56. Miller G, Britt H, Steven I, et al. Incident monitoring in general practice: a pilot study. Report to the Professional Indemnity Review, Commonwealth Department of Human Services and Health. Sydney: Royal Australian College of General Practitioners, 1995. Bhasale AL. Collecting data on potentially harmful events: evaluation of an incident monitoring technique in Australian general practice [master's thesis]. Sydney: University of Sydney, 1997. Bhasale AL, Norton KJ, Britt H. Tests and investigations. Indicators for better utilisation. Aust Fam Physician 1996; 25: 680-694. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of adverse drug events. JAMA 1995; 274: 35-43. Teng Liaw S. Drug interactions among the elderly. Aust Fam Physician 1997; 26: 355-357. Liddell MJ, Goldman SP. Attitudes to and use of a modified prescription form by general practitioners and pharmacists. Med J Aust 1998; 168: 322-325. Flanagan JC. The critical incident technique. Psych Bull 1954; 51: 327-359. Bhasale AL, Britt H, Miller G. Developing interventions which could reduce the occurrence of adverse patient incidents in general practice. Report to the Taskforce on Quality in Australian Health Care, Department of Human Services and Health, Canberra. Sydney: Family Medicine Research Unit, University of Sydney, 1995. (Received 4 Aug 1997, accepted 26 Mar 1998) Authors' details Department of General Practice, University of Sydney, Sydney, NSW. Alice L Bhasale, BSc(Psychol), MSc(Med), Senior Research Assistant, Family Medicine Research Unit; Graeme C Miller, MB BS, FRACGP, Medical Director, Family Medicine Research Unit; Sharon E Reid, MB BS, MPH, Lecturer; Helena C Britt, PhD, Director, Family Medicine Research Unit. Reprints will not be available from the authors. Correspondence: Ms A L Bhasale, Acacia House, Westmead Hospital, Westmead, NSW 2145. E-mail: alicebATgenprac.wsahs.nsw.gov.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Alice L Bhasale · Graeme C Miller · Sharon E Reid · Helena C Britt

Suicide among young Australians, 1964-1993: an interstate comparison of metropolitan and rural trends Michael J Dudley, Norman J Kelk, Tony M Florio, John P Howard and Brent G H Waters MJA 1998; 169: 77-80 See also Rosenman Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - ©MJA1998 Abstract Objectives: (i) To compare suicide rates in 15-24 year old men and women; and (ii) for 15-24 year old men, to investigate differences in suicide rates between metropolitan and rural areas, and changes in method-specific suicide rates and, in particular, firearm and hanging suicide rates in rural and metropolitan areas. Design: Retrospective analysis of Australian Bureau of Statistics (ABS) suicide data (1964-1993). Setting: All Australian States. Subjects: Young women and men aged 15-24 years who died by suicide. Results: Male youth suicide rates rose substantially over the 30 years in all Australian States, whereas female rates did not increase. Increases in suicide rates in young men in small rural towns consistently exceeded those in metropolitan areas in all Australian States. Metropolitan rates in 1964 were higher than those in small rural towns, but by 1993 the position was reversed. Medium-sized cities were the only areas where there was no consistent interstate trend. Differences were noted in suicide base rates in different States. High car exhaust suicide rates were noted in Western Australia, and high firearm suicide rates in Tasmania and Queensland. The ratio of firearm suicide rates in small rural areas to those in metropolitan areas rose in all mainland States, but the same ratio for hanging suicide rates changed little. Conclusions: All Australian States reflect national suicide trends in relation to sex and residential area. In some States, particular suicide methods predominate. A decreasing trend in overall firearm suicide rates in young men in all States from 1984 to 1993 conceals substantial increases in firearm suicide rates in small rural areas in all mainland States over the 30-year period. This, together with the marked rate ratio difference in firearm suicides between metropolitan and small rural areas, suggests that particular risk factors for suicide are operating in small rural areas. The fact that hanging rate ratios changed little suggests that more general factors in male youth suicide are also operating in all areas. A better understanding of similarities and differences in health risks faced by metropolitan and rural youth is required. Introduction Suicide rates for 15-24 year old Australian men have trebled since the early 1960s. However, these rate increases have not been uniform. In metropolitan areas they have doubled, but they have increased as much as 12-fold in towns with fewer than 4000 people. The rates for suicide with a firearm in 15-24 year old men have declined overall, and in metropolitan areas since the late 1980s, but they have continued to rise in rural areas. Rates of suicide by hanging have risen in all locations.1We aimed to answer questions about whether these national suicide trends are replicated at the State level. We wanted to know whether the trends for the larger States dominate the national picture, or whether the same fundamental trends are occurring in all States. Interstate analysis is also important, because health, education and welfare services are often organised and funded at a State level. Differences among the States have long been part of the Australian culture, and firearms legislation, in particular, is at present a State prerogative. In this article, we report suicide rates in young men and women (aged 15-24 years) in all Australian States and Territories, as well as suicide rates in young men by method of suicide, and by residential area, in all States, for 1964 to 1993. Methods Our methods are described in more detail elsewhere.1 Data on suicides and corresponding population data for subjects aged 15-24 years were sought from the Australian Bureau of Statistics (ABS) for the years 1964-1993. These data were de-identified, and an undertaking was given not to attempt to identify the subjects or to contact their families. Data concerning usual area of residence were available for all but 219 of 8537 people who had died by suicide. Subjects' usual place of residence was classified as: metropolitan areas of capital cities of each State; cities with populations >100 000; cities with populations >25 000; towns with populations >4000; or towns with populations <400. These cut-offs, chosen after consulting social geographical sources,2,3 approximate Australian urban groupings in order of size. Definitions of "metropolitan" and "rural" also follow guidelines given in these sources. Rates throughout are expressed as numbers of suicides per 100 000 of the at-risk population. Data were aggregated in 5-year epochs to enable longer-term trends to be discerned. Results Sex-specific suicide rates Suicide rates in 15-24 year old men rose substantially over the past 30 years in all Australian States and Territories, but rates in women of the same age did not (Figure 1). We will consider suicide rates in 15-24 year old men only in the remainder of the article. Method-specific suicide rates Considering the whole 30-year period, for 15-24 year old men, a clear trend exists, both nationally and for individual States, of a reduction in firearm suicide rates in the past 10-15 years, and of an increase in hanging suicide rates over the whole period at a substantially greater rate than the increase by all other methods. Car exhaust suicide rates also increased at a greater rate than the overall increase, especially in Western Australia, but to a lesser extent than suicide by hanging. Suicides by poisoning did not increase proportionally in any States or nationally. Hanging is now the most common method of youth suicide in most States, exceeding firearms as a means of suicide in all States except Queensland and Tasmania. In the most recent epoch, firearm suicide rates were highest in Tasmania and Queensland (Figure 2). Residence-specific suicide rates In all States, suicide rates in 15-24 year old men have increased substantially in metropolitan areas, but the greatest increases have been recorded for towns with a population under 4000. In 1964 in four States (New South Wales, Victoria, Queensland, Western Australia), metropolitan suicide rates exceeded those of small rural areas, but by 1993 the position was reversed (Figure 3). New South Wales Suicide rates in 15-24 year old men rose consistently in all residential categories except cities with over 25 000 people. Sydney's rates rose least (1.6-fold), smaller towns (population, <4000) rose the most (9.9-fold), and Newcastle- Wollongong (2.7-fold) and towns with populations over 4000 (3.8-fold) were intermediate. Victoria Suicide rates in 15-24 year old men rose in all residential areas. Melbourne's rates began from a low baseline, and in 1964 rate differences between Melbourne and smaller towns (population, <4000) were less marked than in NSW. Melbourne's rates rose least (4.2-fold), smaller towns with fewer than 4000 people the most (34.5-fold), and Geelong (5.7-fold) and larger towns (population, >4000) (5.5-fold) were intermediate. Queensland Suicide rates in 15-24 year old men in Brisbane trebled, and rates for the Gold Coast and Sunshine Coast, areas with a high population growth, which both now exceed 100 000 people, peaked in the mid 1970s, but then fell. Rates in cities (population, >25 000 and larger towns (population >4000) increased 1.7- and 1.9-fold, respectively, but in smaller towns (population, <4000) the increase was 31.6-fold. South Australia Differences in suicide rates in 15-24 year old men between metropolitan and small rural areas were less marked in States with lower populations. Adelaide's suicide rate for young men rose 2.7-fold, and in towns with populations under 4000 it rose 5.5-fold. Rates in larger towns (population, >4000) increased 1.8-fold. High rates in cities with populations >25 000 were based on small numbers. Western Australia Suicide rates in 15-24 year old men in Perth rose 2.5-fold and rates in smaller towns (population, <4000) rose 7.0-fold. Rates in larger towns (population, >4000) fluctuated, rising 3.2-fold overall. Tasmania Tasmania's suicide rates in 15-24 year old men have been consistently high. While Hobart's rates have more than doubled, rates in smaller towns (population, <4000) have risen 3.6-fold. Trends elsewhere were inconsistent, although high rates have been recorded. Firearm and hanging suicide rates for metropolitan and small rural areas Suicide with firearms or by hanging, the most common methods, account for between a half and two-thirds of suicides in 15-24 year old men in each State, so these have been analysed by place of residence. In all States the ratio of the firearms suicide rate in small rural areas to the firearms suicide rate in metropolitan areas rose substantially over the 30-year period. In contrast, this ratio for suicide rates by hanging remained much the same or rose minimally over the same period. Thus, the national trend for a reduction in firearm suicide rates conceals a continuing rise in firearm suicide rates in rural areas in all States. These increases were especially pronounced in Victoria and Queensland (41.5- and 36.7-fold respectively), while Tasmania's rates remained high throughout the period (Figure 4). Discussion National youth suicide trends according to sex of the subject are reflected in the rates of all the individual States: suicides in 15-24 year old men rose substantially, whereas those in 15-24 year old women did not change in the past 30 years. The disproportionate increase in suicide rates in young men in small rural areas nationally is also occurring consistently in all Australian States. Certain data limitations should be noted. Coroners often must exercise subjective judgements in determining verdicts, which may lead to possible systematic differences in rates between different States and over time; Errors may also arise from unstandardised methods of recording suicides, especially earlier in the study period;1 Population data may fluctuate in minor ways with changing residential area boundaries, notably with some towns being absorbed into larger cities; and The data do not allow inferences about particular local geographical areas (see for comparison National Injury Surveillance Unit data4). However, it is unlikely that changing coronial verdicts account for the magnitude or generality of the changes we noted. Cantor and Coory5 found that Queensland's provincial cities and rural towns of modest size had negligible rate increases compared with metropolitan and major urban areas from 1986 to 1991. They questioned whether there was a rural suicide problem in Queensland. Our study confirms their observations for Queensland cities with populations over 25 000, but notes a massive rate increase in towns with fewer than 4000 people, with substantial numbers supporting this finding. Further, the Queensland picture does not appear to be atypical. There are differences in suicide base rates in the various States. Whether Melbourne's low base rate in 1964 is coincidental, or a result of Victorian coroners showing less willingness to reach a verdict of suicide, requires further investigation. Tasmania's relative isolation, demographically stable population, and easy firearm access may explain its consistently high suicide rates. Suicide methods differ somewhat between States. The high rates for car exhaust suicides in Western Australia may reflect its relative isolation and this method's cultural acceptability. Carbon monoxide sensors, exhaust modifications and changes to catalytic converters may be effective prevention measures.6 High firearm suicide rates among 15-24 year old men in Tasmania and Queensland,7 rising metropolitan firearm suicide rates in those States, and rising rates in small towns (population, <4000) in all States, but especially Victoria and Queensland, probably relate to falling populations in small rural areas with stable or rising suicide numbers, the easy availability of firearms, and the more liberal firearm legislation in Tasmania and Queensland. Applying methods to prevent car exhaust suicide and reducing the availability of guns may reduce suicide rates by these two methods, and possibly total suicide rates.8 While rates for suicide by various methods may differ in different States, residential area trends for firearm and hanging suicides were pronounced in all States. A trend for overall firearm suicide rates to decrease in all States from 1984-1993 conceals substantial firearm rate increases in small rural areas in all mainland States over the 30-year period. Differences in rates between metropolitan and small rural areas, and the fact that the ratio of firearms suicide rates in small rural areas to those in metropolitan areas rose substantially in all mainland States, suggest that particular risk factors (directly and indirectly related to firearms) are operating in small rural areas. Restrictive legislation may differentially affect firearm suicides in metropolitan areas, where availability is lower.9 The rural economic downturn, the strain on small rural communities, and a major population exodus in all States from towns with fewer than 4000 people have meant added health burdens for those communities, with possibly specific risk factors.1 The fact that the equivalent ratio for hanging rates changed little suggests that more general factors are also operating across all areas in relation to suicide in young men. Particular States have local issues which require specific attention, but the fact that all Australian States reflect the national trends in youth suicide in relation to sex and residential area supports a national approach to youth suicide prevention. Effective national approaches will depend in part on acquiring a better understanding of the similarities and differences in health risks faced by metropolitan and rural youth.1 Acknowledgements Thanks are due to Caroline Haski and Margaret Alcock for their painstaking work as research assistants, and to the NSW Institute of Psychiatry, which partly funded this study. References Dudley M, Kelk N, Waters B, et al. Suicide among young rural Australians 1964-1993: a comparison with metropolitan trends. Soc Psychiatry Psychiatr Epidemiol 1997; 32: 251-260. Department of Primary Industries and Energy and Department of Human Services and Health. Rural, remote and metropolitan areas classification, 1991 Census edition. Canberra: AGPS, 1994. Holmes J. In: Jeans D, editor. Space and society. Chapters 2 and 3. Sydney: Sydney University Press, 1988: 24-74. Moller J. An atlas of injury death in Australia 1990-1992. Adelaide: National Injury Surveillance Unit, 1995. Also available online: <www.nisu.flinders.edu.au/data/atlas/atlas.html> Cantor C, Coory M. Is there a rural suicide problem? Aust J Public Health 1993; 17: 382-384. Australian Institute for Suicide Research and Prevention. Access to means of suicide by young Australians; a background report to the Commonwealth Department of Health and Family Services Youth Suicide Prevention Advisory Group. Brisbane: Australian Institute for Suicide Research and Prevention, 1996. Cantor C, Lewin T. Firearms and suicide in Australia. Aust N Z J Psychiatry 1990; 24: 500-509. Cantor C, Baume P. Access to methods of suicide: what impact? Aust N Z J Psychiatry 1998; 32: 8-14. Cantor C, Slater P. The impact of firearm control legislation on suicide in Queensland: preliminary findings. Med J Aust 1995; 162: 583-585. (Received 12 May 1997, accepted 17 Apr 1998) Authors' details University of New South Wales, Sydney, NSW. Michael J Dudley, MB BS, FRANZCP, Lecturer, School of Psychiatry. Norman J Kelk, PhD, Social Worker, School of Community Medicine. Prince of Wales Hospital, Sydney, NSW. Tony M Florio, MPsych, Clinical Psychologist. Macquarie University, Sydney, NSW. John P Howard, PhD, Clinical Psychologist. PO Box 474, Edgecliff, NSW. Brent G H Waters, MD, FRANZCP, Psychiatrist. Reprints will not be available from the authors. Correspondence: Dr M J Dudley, Department of Child and Adolescent Psychiatry, Sydney Children's Hospital, Randwick, NSW 2031. E-mail: m.dudleyATunsw.edu.au - Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

Michael J Dudley · Norman J Kelk · Tony M Florio · John P Howard

Prostate-specific antigen testing in Australia and association with prostate cancer incidence in New South Wales David P Smith and Bruce K Armstrong MJA 1998; 169: 17-20 For editorial comment, see McCredie & Cox Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - ©MJA1998 Abstract Objective: To describe patterns and trends in prostate-specific antigen (PSA) testing in Australia and assess its role in the increasing incidence of prostate cancer. Design: Descriptive analysis of (i) Medicare records of PSA testing in Australia, and (ii) prostate cancer recorded incidence in New South Wales. Data: (i) Medicare data for all males who received a Medicare-reimbursed PSA test between August 1989 and December 1996. (ii) NSW Central Cancer Registry data for all males in NSW with prostate cancer diagnosed between 1988 and 1995. Main outcome measures: (i) Number of PSA tests, age-standardised rates of PSA tests by State and Territory, and proportions of males who had a PSA test. (ii) Recorded incidence of prostate cancer in NSW. Results: (i) More than 2.2 million PSA tests were done on more than 1.1 million Australians between 1989 and 1996. The annual number of males tested increased fivefold in this period and peaked in 1995. Twenty-seven per cent of Australian men aged 50 years or over had at least one PSA test in 1995 or 1996; 33% of men aged 60-69 years had a test in this period. (ii) In NSW the number of PSA tests per quarter was highly correlated with the number of new cases of prostate cancer (R2 = 0.92). Conclusions: Although no organised program for prostate cancer screening exists, and despite repeated advice against it, opportunistic screening has been occurring at high rates. There was a high correlation between PSA testing and prostate cancer incidence between 1990 and 1995 in NSW. Introduction Prostate cancer is the most common cancer in Australians after non-melanocytic skin cancers, and accounts for about a third of all newly diagnosed internal cancers in males.1,2 It is the second most common cause of cancer death among Australian males.3 During the late 1980s and early 1990s, recorded prostate cancer incidences increased substantially in Australia, while mortality from prostate cancer remained stable.1,2,4-6 Much of the increase in incidence has been attributed to detection of latent cancers by increased use of the prostate-specific antigen (PSA) test, transurethral resection of the prostate, prostatic biopsy and radical prostatectomy.7There are considerable arguments for8,9 and against10,11 screening for prostate cancer. In 1996 the Australian Health Technology Advisory Committee reviewed the evidence and recommended against screening. However, the committee recognised that de facto screening occurs in the community and stated a need to monitor and review the evidence when important developments occur.12 Data collected by the Health Insurance Commission provide the best available means to monitor trends and patterns in PSA testing. We analysed these data to determine the association between PSA testing and reported incidence of prostate cancer in Australia and in New South Wales. Methods Data PSA tests: The Commonwealth Department of Health and Family Services Medicare Estimates and Statistics Unit provided de-identified data, extracted from the national dataset of all services rendered on a fee-for-service basis for which a Medicare benefit has been paid. The data included all tests itemised under the Medicare Benefits Schedule codes that included PSA tests. PSA tests first appeared in the Schedule in August 1989, but were included with 20 "other" biochemical tests until November 1993. Since then they have been itemised together with prostatic acid phosphatase (PAP) tests. PSA tests could not be differentiated from PAP tests with the available information. To estimate the proportion of non-PSA tests in the dataset, we investigated the patterns of use of the other tests, using data provided by the Commonwealth Department of Health and Family Services. We calculated retrospective projections using exponential regression on the number of tests by age group and annual quarter to give an adjusted number of PSA and PAP tests for the period between August 1989 and November 1993. This study included data for all males who had at least one PSA or PAP test between 1989 and 1996 reimbursed by Medicare. The data included a unique identification number for each person, age, date of service, postcode, and fee charged for the service. The identification number allowed individuals to be linked over time to quantify those who had multiple tests, but was not linkable to any other identifying data. Prostate cancer: Data on prostate cancer incidence in NSW were obtained from the NSW Central Cancer Registry, a population-based register which began collecting data in 1972. Notification of malignant neoplasms has been a statutory requirement for all NSW public and private hospitals, radiation oncology departments and nursing homes since 1972, and for all pathology and outpatient departments since 1985.1 Population: The Australian Bureau of Statistics provided the estimated resident populations of Australia by five-year age group and State or Territory for the years 1989 to 1996.13 Analysis We calculated the rate of testing and the number of males tested by age group, State or Territory, and year from August 1989 to December 1996. We calculated the number of males having multiple tests from January 1995 to December 1996 and the proportion of males tested in each age group in this period. We used a two-year period for these calculations to reduce errors introduced by men moving up age groups. We compared the number of tests carried out in NSW by quarter with the number of prostate cancers diagnosed between 1990 and 1995 (the most recent year for which prostate cancer data were available). All rates, standardised to the total male and female Australian 1991 population, are expressed per 100 000 males. We used SAS software for statistical analysis.14 Results PSA testing in Australia From August 1989 to December 1996 more than 2.2 million PSA or PAP tests were reimbursed by Medicare in Australia. About 1.1 million males were tested during this period. Eighty-eight per cent of the tests were for men aged 50 years or over, with the largest proportion (34%) for men aged 60-69 years (Table 1). Age-standardised rates for males having one or more PSA/PAP tests per year increased fivefold between 1990 and 1996, and the greatest increases occurred between 1993 and 1995. There was substantial variation in the rates of testing by State and Territory (Figure 1). Except in the Australian Capital Territory, the rates peaked in 1995, when the highest rates were in Western Australia (8668 tests per 100 000) and the Australian Capital Territory (8284 tests per 100 000) and the lowest in the Northern Territory (3270 per 100 000). From January 1995 to December 1996, 709 523 Australian males had at least one PSA/PAP test reimbursed by Medicare. Most of those tested (73%) had one test, 17% had two tests, 5% had three tests and 5% had four or more tests. Older men were more likely to have had multiple tests (Table 2). In this period, 27% of Australian men aged 50 years and over had one or more PSA/PAP tests reimbursed by Medicare. This figure peaked at 33% in men aged 60-69 years (Table 3). In 1996 the Medicare schedule fee for a single PSA/PAP test was $19.90 or, where two or more tests were undertaken at the same consultation, $36.65. The overall amount reimbursed by Medicare for PSA/PAP tests in 1996 was $10 675 880 (mean per test, $20.73; mode, $19.90; range, $11.00 to $36.65), not including the fee that may have been charged for the accompanying consultation. PSA/PAP tests and prostate cancer in NSW Between 1990 and 1995, more than 625 000 PSA/PAP tests were carried out in NSW, and 20 120 prostate cancers were diagnosed. The number of tests was highly correlated with the number of prostate cancers diagnosed (R2 = 0.92) (Figure 2). The ratio of number of tests performed to number of new cases of prostate cancer diagnosed increased from about 19 in 1990 to 45 in 1995. In 1995, in NSW, 150 479 males had one or more PSA/PAP tests. Figure 3 shows the age-specific rates of testing and reported incidence of prostate cancer. The number of PSA/PAP tests per prostate cancer detected in 1995 varied between age groups from a high of 579 in men aged 40-49 years to fewer than 24 in men aged 70 years or over. Discussion More than 2.2 million PSA tests were carried out in Australia from 1989 to 1996. More than 1.1 million males were tested in this period, and the annual number of males tested peaked in 1995. Data from this study support the hypothesis that the rising incidence of prostate cancer is associated with increased PSA testing. In NSW, the number of PSA tests was highly correlated with the number of new cases of prostate cancer. The PSA test is a blood test used in diagnosis and monitoring of prostate disease. First used in Australia in the late 1980s to monitor clinically identified disease, it has since been used in the diagnosis of relevant symptoms and as a screening test for asymptomatic men. It was not possible to identify from Medicare data the reasons why the tests were ordered. However, recent research found that 67 of 118 PSA tests (57%) were ordered for screening.15 Although we adjusted for other tests included under the same Medicare Benefits Schedule item from 1989 to 1993, we could not adjust for PAP tests, which are used to monitor the clinical progress of prostate cancer. However, in a continuous six-month period the ratio of PAP tests to PSA tests processed by a large, representative private pathology laboratory in NSW (covering city, suburban and regional centres) was less than 2% (Dr G Caldwell, Pathologist, Douglass Hanly Moir Pathology, personal communication). Data from a large public pathology laboratory in South Australia indicate that the proportion of PAP tests to the total PSA and PAP tests fell from 50% in 1991 to 6% in 1996 (Dr H A Morris, Manager, Endocrine Unit, Institute of Medical and Veterinary Science, personal communication). At their peak in 1995, the rates of PSA/PAP testing in Australian males ranged from 3270 per 100 000 in the Northern Territory to 8668 per 100 000 in Western Australia. These are probably underestimates because Medicare data do not include services provided free to public patients in public hospitals, to Veterans' Affairs patients and to men offered screening under the research activities of centres such as the Perth-based Urological Research Centre. In the one-year period April 1993 to March 1994, 39 626 PSA tests were done on 30 739 veterans.16 Data from the Department of Veterans' Affairs show about 50 000 tests were done each year in Australia between 1994 and 1996, which would have contributed a further 10% to the number of Medicare-reimbursed PSA tests. In South Australia 72 000 PSA tests that would not appear in Medicare statistics were undertaken between 1990 and 1996 by a public laboratory (Dr H A Morris, personal communication). These and the tests done on veterans would have accounted for an approximate under-enumeration of 33% annually in South Australia. Nationally, considering all these extra sources of PSA tests, we estimate that Medicare data underenumerate PSA tests by 14%. In a recent study of self-reported rates of prostate cancer screening in the Central Sydney Area Health Service, about one in five men aged 50 years or over reported being screened in the previous 12 months.17 This agrees quite closely with our results, which show that during the two years 1995 to 1996 about one in four Australian men aged 50 years or over had a PSA test, and in 1995 one in six (17%) had a test. More prostate cancers would result in more PSA tests used for monitoring. However, the overall effect of this on PSA test-ordering is thought to be small. More than 70% of males tested in 1995 and 1996 had only one test, suggesting that most tests were for screening rather than monitoring disease activity. A further possible indication that most tests were undertaken for screening rather than for monitoring or diagnosis is the increase in the ratio of tests to newly diagnosed prostate cancers in New South Wales. This ratio continued to increase in 1995, when the reported incidence of prostate cancer had begun to fall. Increasing recorded incidences of prostate cancer have been reported from the United States,18-22 France23 and elsewhere in Australia.1,2,4-6 Incidence figures began rising earlier in the United States than in Australia, and appeared to peak in 1992 and 1993.20 South Australian and Western Australian age-standardised recorded incidences peaked in 1994 and fell by 22% and 13%, respectively, between 1995 and 1996.2,4,5 The rate of PSA testing peaked in 1995 in both States and fell 10% and 16%, respectively, in 1996. These data and the high correlation between PSA tests and newly diagnosed prostate cancers in NSW support the hypothesis that the rising incidence figures for prostate cancer in the early 1990s were a direct result of screening.7 Based on these trends and correlations, it is likely that the recorded incidence for prostate cancer in most Australian States and Territories will continue to fall after 1995. Given that screening for prostate cancer has never been recommended in Australia, the rates of de facto screening in men aged over 50 years, and especially those aged between 60 and 69 years, are quite remarkable. These findings have important implications for public health policy and for patient and practitioner education aimed at reducing prostate cancer screening. Acknowledgements We would like to thank Mr Ross Saunders, Director of the Medicare Statistics Section of the Department of Health and Family Services, for supplying the data. References Coates M, Armstrong B. Cancer in New South Wales. Incidence and mortality 1994. Sydney: NSW Cancer Council, 1997. South Australian Cancer Registry. Epidemiology of cancer in South Australia. Incidence, mortality and survival 1977 to 1996, incidence and mortality 1996 analysed by type and geographical location. Twenty years of data. Adelaide: South Australian Health Commission, 1997. Australian Bureau of Statistics. Causes of death, Australia, 1995. Canberra: AGPS, 1996. (Catalogue No. 3303.0.) Threlfall T, Whitford M, Thompson J. Cancer incidence and mortality in Western Australia 1992 to 1994. A report of the Western Australian Cancer Registry. Perth: Health Department of Western Australia, 1996. Threlfall T, Thompson J. Cancer incidence and mortality in Western Australia, 1995. A report of the Western Australian Cancer Registry. Perth: Health Department of Western Australia, 1997. Shugg D, Dwyer T, Blizzard L. Cancer in Tasmania. Incidence and mortality 1994. Hobart: Menzies Centre for Population Health Research, 1997. McCredie M, Coates M, Churches T, Rogers J. Rising incidence of prostate cancer in Australia: a result of 'screening'? J Epidemiol Biostatistics 1996; 1: 99-105. Lange PH. Is screening for prostate cancer the current gold standard? -- "Yes". Eur J Cancer 1997; 33: 354-356. Kaye KW. Prostate cancer: enthusiasm for screening. Med J Aust 1995; 162: 540-541. Kramer BS, Gohagan JK, Prorok PC. Is screening for prostate cancer the current gold standard? -- "No". Eur J Cancer 1997; 33: 348-353. Hirst GHL, Ward JE, Del Mar CB. Screening for prostate cancer: the case against. Med J Aust 1996; 164: 285-288. Australian Health Technology Advisory Committee. Prostate cancer screening. Canberra: AGPS, 1996. Australian Bureau of Statistics. Estimated resident population by sex and age: States and Territories of Australia 1996. Canberra: Australian Bureau of Statistics, 1997. (Catalogue no. 3201.0.) SAS [computer program]. Version 6.12. Cary, North Carolina: SAS Institute, 1996. Ward JE, Gupta L, Taylor NJ. Do general practitioners use prostate-specific antigen as a screening test for early prostate cancer? Med J Aust 1998: 169; 29-31. Parkes AJ, Killer GT. Prostate-specific antigen -- is it already being used as a screening test? [letter]. Med J Aust 1994; 161: 722-723. Ward JE, Hughes A-M, Hirst GHL, Winchester L. Men's estimates of prostate cancer risk and self-reported rates of screening. Med J Aust 1997; 167: 250-253. Stephenson RA, Smart CR, Mineau GP, et al. The fall in incidence of prostate carcinoma. On the down side of a prostate specific antigen induced peak in incidence -- data from the Utah Cancer Registry. Cancer 1996; 77: 1342-1348. Jacobsen SJ, Katusic SK, Bergstralh EJ, et al. Incidence of prostate cancer diagnosis in the eras before and after serum prostate-specific antigen testing. JAMA 1995; 274: 1445-1449. Merrill RM, Potosky AL, Feuer EJ. Changing trends in U. S. prostate cancer incidence rates. J Natl Cancer Inst 1996; 88: 1683-1685. Gann PH. Interpreting recent trends in prostate cancer incidence and mortality. Epidemiology 1997; 8: 117-120. Lu-Yao GL, Greenberg R. Changes in prostate cancer incidence and treatment in USA. Lancet 1994; 343: 251-254. Menegoz F, Colonna M, Exbrayat C, et al. A recent increase in the incidence of prostatic carcinoma in a French population: role of ultrasonography and prostatic specific antigen. Eur J Cancer 1995; 31A: 55-58. (Received 28 Nov 1997, accepted 24 Apr 1998) Authors' details Cancer Control Information Centre, New South Wales Cancer Council, Sydney, NSW. David P Smith, BA, MPH, Research Coordinator, Cancer Epidemiology Research Unit; Bruce K Armstrong, DPhil, FRACP, Director. Reprints will not be available from the authors. Correspondence: Mr D P Smith, Cancer Epidemiology Research Unit, NSW Cancer Council, PO Box 572, Woolloomooloo, NSW 2011. E-mail: dsmithATnswcc.org.au Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/>

David P Smith · Bruce K Armstrong