Casemix: moving forward

Despite initial scepticism, particularly by clinicians, casemix has remained integral to healthcare reform in Australia. In the first MJA casemix supplement in 1994 (a year after casemix was introduced in Victoria), clinicians aired their concerns about deficiencies in the acute-care classification, the lack of classifications beyond acute care, and the risks they perceived in casemix-based funding. Following Victoria's lead, the past four years have seen both the introduction and the practical application of casemix in most States and Territories in Australia. At the same time there has been refinement of the acute-care classification, development of classifications for subacute care, ambulatory and emergency care, and the recent implementation of the Australian modification of the International classification of diseases and related health problems, 10th revision (ICD-10-AM). Although only a small core of clinicians were involved in these developmental processes, many have now experienced working in a casemix-funded environment. It is time for these clinicians to comment on the perceived benefits and pitfalls of casemix and their vision for casemix in the future. This second MJA casemix supplement presents these viewpoints. Leading clinicians in the public and private sectors and academics from a range of disciplines discuss the impact and future of casemix in Australia. The intent is not to cover all aspects of casemix, nor to discuss the impact of casemix funding on every discipline, but rather to explore key areas. In the evolutionary phase of casemix since 1994, many clinicians, and particularly clinical managers, have focused their attention on the complexities of casemix and on protecting immediate interests, be these specific to departments, specialties or hospitals. The experience of casemix funding of acute-care services has varied in different States, and refinements are still needed to address inappropriate incentives and ensure fairer funding. It is still not clear what longer term consequences the States' different casemix-funding models will have on the healthcare system as a whole, on hospital funding, teaching and research and, most importantly, on patient care. The Aboriginal and Torres Strait Islander Casemix Study (Fisher et al), and Ruben and Fisher's study of the impost of the current casemix classification on Aboriginal and Torres Strait Islander children, highlight the importance of analysing the application of national casemix classifications in communities with different population profiles. Many factors contribute to the higher resource requirements of Aboriginal and Torres Strait Islander patients, including length of hospital stay, younger age, multiple comorbidities and remote location. These studies clearly demonstrate deficiencies in the current acute-care classification and funding models for health services for indigenous people, and emphasise the need for increased funding for hospitals with a high proportion of these patients. In the broader context, Duckett compares the different States' casemix funding policies and points out the advantages of sharing experiences across State borders in order to develop the best possible casemix-funding model. Of concern is the disparity in costing between the States, suggesting a need for standardisation in defining costs to minimise this potential inequality and facilitate national benchmarking. Phelan et al emphasise the role clinicians can play by increasing their understanding of how patient care is costed. This will allow them to set valid benchmarks and enhance the efficiency of their clinical services while maintaining and improving quality of care. Casemix classifications are appropriate for describing the average patient, but these classifications still perform poorly in distinguishing illness severity in patients in specialist as opposed to general hospitals, or in specific patient groups such as children (Hanson et al). Age and other factors, for example the need for transfer of patients for specialised care (Butt and Shann), are used as proxies for illness severity and higher dependency instead of refining the classification. The value of casemix for addressing the wider issues of healthcare reform has largely been hampered by the lack of suitable classifications. The development of a classification for subacute and non-acute care (AN-SNAP) that reflects the goals of management is a welcome step forward (Lee et al). This scheme will complement the current acute-care classification (AR-DRG-4). The strong clinical acceptance of AN-SNAP augurs well for its implementation. The long-awaited outpatient classification, beyond a simple clinic-based structure, continues to prove elusive (Cleary et al). The attempt at developing a patient-based classification has not been successful. To achieve further progress, the time may have come to adopt one of the State clinic-based classifications, such as the Victorian Ambulatory Classification, as a national standard. Clinicians have expressed concern about the suitability of ICD-9-CM (International classification of diseases, 9th edition, clinical modification) to describe clinical practice and support data quality. The development and implementation of ICD-10-AM through the efforts of the National Centre for Classification in Health, and the Centre's commitment to working with clinicians, will pave the way forward to a more flexible and relevant Australian classification. An important feature of ICD-10-AM is the inclusion of a procedure classification based on the Commonwealth Medical Benefits Schedule, which can be used in both the public and private sectors (Roberts et al). A strong case is made for the development of a distinct classification for nursing by Long and Mann and for allied health by Byron and McCathie. These classifications will facilitate meaningful benchmarking across services and enhance the use of casemix in monitoring quality of care and outcomes. What is still lacking in casemix development is an emphasis on continuum of care. Furthermore, little attention has been paid to the way casemix should be used to improve patient care, or to its limitations in addressing patient outcomes and healthcare quality. The wider use of casemix in determining care paths and in utilisation review is now also emerging in both the public and private sectors, and needs closer clinical scrutiny. This changing focus is reflected in the range of articles in this supplement, and presents a challenge to clinicians from all disciplines (see Maxwell, on implications of the use of AN-DRGs in the private sector; and Hart and Wallace, on casemix and surgery). The foundation of clinical practice is education and research. In its evolution, casemix has focused on core clinical requirements, but Phillips argues that a focus on teaching and research is urgently needed. He suggests that it may be time to promote outcome-based funding of teaching and research in the Australian healthcare system. Surprisingly, so far only a relatively small group of clinicians has taken an active interest in casemix. In view of the impact casemix has had on clinical care, this is difficult to explain. There is a risk that this lack of participation by the broader clinical community not only could leave Australia with patient classification and funding systems with inadequate clinical relevance, but could also affect the financial stability of a range of clinical services, reduce clinical autonomy and potentially compromise quality of patient care. The future lies not just in developing nationally consistent methods of classifying healthcare services and costs, but in determining how these methods can be applied to best suit the evolving model of healthcare and improve the quality of care provided to all groups of patients across the country. Clinicians need to be involved in this process. Ralph M Hanson Education Advisor to the Australian Casemix Clinical Committee The New Children's Hospital, Sydney, NSW

Synopsis The Northern Territory Health Service implemented a casemix system of hospital funding in 1996 using national averages and national cost weights as benchmarks for length of stay and funding. Clinicians and health administrators were concerned about the potential of this model to impair health service delivery, especially to children of Aboriginal or Torres Strait Islander (ATSI) descent, whose current poor health has been well described. Data were collected on children aged under 10 years who were discharged from the Royal Darwin Hospital between 1 July 1991 and 30 June 1996 and assigned one of four DRGs (simple pneumonia, bronchitis and asthma, gastroenteritis, nutritional and metabolic disorders). Data collected included age, sex, ethnicity, duration of hospital stay, location of residence and presence of comorbidities. There were significant differences in the proportion of children with multiple comorbidities between ATSI and non-ATSI children, as well as between rural- and urban-dwelling ATSI children. A higher proportion of ATSI compared with non-ATSI children had prolonged hospital stays (22.6% v. 1.5%), with the variables influencing length of stay in ATSI children including "age < 2 years", "living in a remote area", and "presence of two or more comorbidities". These results confirm clinical impressions about disease patterns and length of hospital stay in ATSI children, and highlight the problems of imposing a casemix classification system for a "typical" Australian population on a region with a high proportion of people of ATSI descent. Introduction In July 1996, the Northern Territory Health Service introduced a casemix-based system of funding to its five acute-care hospitals. Average length of stay was to be benchmarked against national averages and funding based on national cost weights. Clinicians and health administrators were concerned over the potential of this model to impair health service delivery because of the Northern Territory population's high proportion of people of Aboriginal or Torres Strait Islander (ATSI) descent. The poor health of ATSI children has been well described.1 It has been our observation that the disease patterns in hospitalised ATSI children are different and that they have longer lengths of hospital stay. The Top End of the Northern Territory includes the city of Darwin and surrounds with 90 500 people, of whom 8232 (9%) are of ATSI descent, and the Darwin Rural/Remote area with 12 460 people, of whom 9186 (73.7%) are of ATSI descent.2 Of the 17 770 children aged 10 years and under resident in the Top End, 4460 (26%) are of ATSI descent; 2440 (52.7%) of these children live in the rural/remote district. The implementation of casemix funding made it a matter of urgency to consider whether, and to what degree, length of stay and therefore resource consumption was greater in Aboriginal children. Moreover, were there justifiable reasons for longer hospital stays in ATSI children? To answer these questions, data were collected at the Royal Darwin Hospital (RDH), a 270-bed teaching hospital, which is the only public hospital in the Top End of the Northern Territory. Methods Data collected Separation data were obtained for all children aged under 10 years discharged from the Royal Darwin Hospital for the five years 1 July 1991 to 30 June 1996. Our study sample was drawn from cases coded as normally resident in Darwin or the surrounding rural/remote area and assigned one of four Australian national diagnosis-related groups version 2 (AN-DRG-2). These were 180 (Simple pneumonia, age < 10 years), 186 (Bronchitis and asthma, age < 10 years), 335 (Gastroenteritis, age < 10 years) and 533 (Nutritional and miscellaneous metabolic disorders, age < 10 years). These are the four commonest DRGs assigned to children at Royal Darwin Hospital. Data were obtained on age, sex, ethnicity, duration of hospitalisation, location of residence and number and nature of the seven most common comorbidities. Age was grouped into completed years. Ethnicity was coded as reported by the children's parents or guardians as "ATSI" or "other". The duration of hospitalisation was in completed days. Location of residence was categorised according to the Northern Territory Health Service's classification into greater Darwin urban area or Darwin rural/remote area. The comorbidities were defined from the International classification of diseases 9th revision, clinical modification, 4th edition (ICD-9-CM4). The definitions used were malnutrition (260.0-263.9), anaemia (280.0-285.9), pneumonia and influenza, chronic obstructive pulmonary disease and allied conditions (480.0-491.9), fluid and electrolyte disturbance (276.0-276.9), skin infection or infestation (132, 133, 680-686.9), middle-ear disease (382.0-382.9), and intestinal infectious diseases (001-009.3). Statistical analysis As the results were not normally distributed, we calculated the significance of differences in mean length of stay between the two groups using the Kruskal-Wallis H test (equivalent to a chi-squared test). For calculating the significance of difference between proportions of comorbidities we used 95% confidence intervals (CIs), taking as significant any result that did not cross zero. We compared the association of each comorbidity for each DRG between ATSI and non-ATSI children, and between ATSI rural residents and ATSI urban residents. We classified as statistically significant any result in which the 95% CI of the odds ratio did not cross one, unless there was a zero in the cell, in which case we used a chi-squared test with a Yates' correction. To assist comparison between ATSI and non-ATSI children, we defined prolonged length of stay as hospital stay exceeding three times the average for non-ATSI children. We examined the associations of prolonged length of stay in ATSI children, both crude associations and those determined with a multiple logistic regression model using EGRET (Statistics and Epidemiology Research Corporation, Portland, Oregon, USA). We classified as statistically significant any result in which the 95% CI of the odds ratio did not cross one. Results During the study period, there were 10 409 paediatric separations, of which 3587 (34.5%) fulfilled the study admission criteria; 2210 of these (61.6%) occurred in ATSI children. Within this cohort, 421 of the 2210 ATSI separations (19.0%) were from urban areas compared with 1352 of the 1377 non-ATSI separations (98.2%); the 25 non-ATSI rural separations have been included with the non-ATSI cases for analysis. The highest number of separations was seen in DRG 335 (gastroenteritis) for rural ATSI cases and DRG 186 (bronchitis and asthma) for non-ATSI and urban ATSI cases (Box 1). Length of hospital stay Only 21 of 1377 non-ATSI children (1.5%), as compared with 500 ATSI children (22.6%), had prolonged hospital stays; 19 of the 21 non-ATSI children (90.4%) were urban residents. Of the ATSI children, 34 (8.1%) urban residents and 466 (26%) rural/remote residents had prolonged hospital stays. The ratios of mean length of stay for ATSI compared with non-ATSI children varied between 2.0 (for DRG 180 - simple pneumonia) and 3.8 (for DRG 335 - gastroenteritis). For all four DRGs, differences in mean length of stay between urban- and rural-dwelling ATSI children were statistically significant, but not as marked as ATSI/non-ATSI differences (Box 2). 2: Comparison of mean length of stay in days (SD) DRGATSI-ruralATSI-urbanRatio ATS-rural/ATSI-urbanTotal ATSINon-ATSIRatio ATSI/Non-ATSI 1807.4 (4.2)5.5 (4.7)1.4:1 P <0.0016.9 (4.4)3.4 (2.3)2.0:1 P <0.001 1865.8 (3.6)3.5 (3.7)1.7:1 P <0.0015.8 (3.8)2.4 (1.9)2.1:1 P <0.001 3359.4 (5.6)5.1 (4.0)1.8:1 P <0.0018.8 (5.6)2.3 (3.2)3.8:1 P <0.001 53312.5 (6.6)10.3 (7.6)1.2:1 P <0.0512.3 (6.7)5.5 (7.4)2.3:1 P <0.001 Total9.4 (5.9)5.1 (4.9)1.8:1 P <0.0018.6 (6.0)2.6 (2.6)3.3:1 P <0.001 ATSI = Aboriginal and Torres Strait Islander. DRG = Diagnosis-related group (AN-DRG-2). DRG 180 Simple pneumonia, age < 10 years. DRG 186 Bronchitis and asthma, age < 10 years. DRG 335 Gastroenteritis, age < 10 years. DRG 533 Nutritional and miscellaneous metabolic disorders, age < 10 years. Comorbidities There were also significant differences in the proportion of children with multiple comorbidities between ATSI and non-ATSI children, as well as between rural- and urban-dwelling ATSI children, with the most marked differences occurring when two or more comorbidities were present (Box 3). 3: Number of comorbidities (percentage of total for ethnic group) No. of comorbiditiesATSI-ruralATSI-urbanRatio ATSI-rural/ATSI-urbanTotal ATSINon-ATSIRatio ATSI/Non-ATSI 0441 (24.7%)276 (65.6%)0.4:1*717 (32.4%)1224 (88.9%)0.4:1* 1324 (18.1%)69 (16.4%)1.1:1393 (17.8%)119 (8.6%)2.1:1* 2471 (26.3%)47 (11.2%)2.4:1*518 (23.4%)26 (4.8%)4.9:1* 3379 (21.2%)17 (4.0%)5.3:1*396 (17.9%)5 (0.4%)44.8:1* 4150 (8.4%)9 (2.1%)4.0:1*159 (7.2%)3 (0.2%)36.0:1* 524 (1.3%)3 (1.1%)1.2:127 (1.2%)0 ATSI = Aboriginal and Torres Strait Islander *Statistically significant difference in proportions. The small number of non-ATSI children with comorbidities precluded comparing mean length of stay and comorbidities for DRGs 180, 186 and 533. However, for DRG 335 (gastroenteritis) there was a significant difference between the mean length of stay for ATSI versus non-ATSI children (7.5 v. 2.2 days; P < 0.01), and ATSI-rural versus non-ATSI children (8.1 v. 2.2 days; P < 0.01) when a comorbidity was present. Compared with non-ATSI children, ATSI children were significantly more likely to have any of the comorbidities for all the studied DRGs, with the strongest associations being with malnutrition, skin infection/infestation and anaemia (Box 4). Compared with ATSI urban-dwelling children, ATSI rural children were also significantly more likely to have comorbidities, with malnutrition strongly associated (Box 5). Factors influencing length of stay On multiple regression analysis of variables influencing length of stay in ATSI cases, "age less than 2 years", "living in a remote area" and "the presence of two or more comorbidities" were the commonest statistically significant associations on both univariate and multivariate analysis. For both crude and adjusted analyses the comorbidity "malnutrition" was significantly associated with prolonged stay for all applicable DRGs (Box 6). Discussion This study shows that not only are there differences in the casemix of ATSI compared with non-ATSI children, but also that there are differences in the casemix of rural- and urban-dwelling ATSI children. To our knowledge this is the first study quantifying a higher prevalence of comorbidities in a subpopulation and their relationship to average length of hospital stay. DRGs are created at a federal level as a means of bundling inpatient episodes to guide hospital funding. The bundled episodes are homogeneous both clinically and in resource use. Although DRGs have been refined, this has been done using national length-of-stay data. Our study shows the dangers of imposing a classification system for a "typical" Australian population on a region with a high proportion of people of ATSI descent. The Royal Darwin Hospital's 10 409 paediatric separations in the study period included 5180 ATSI children (49.8%), of whom 4221 (81.5%) were from rural or remote areas. These children have quite a different disease course, with a high admission rate, many comorbidities and a clinically justified prolonged length of stay related to the number of comorbidities. The DRG classification's capacity to cater for different populations with the same diagnosis is limited. New classification approaches have been postulated to take into account severity and comorbidities.3 The "clinical complexity level" refinement project currently being undertaken by the Commonwealth Department of Health and Family Services is addressing the impact of complications and comorbidities on DRGs, with the possibility of optimising DRG splits (Development of Australian refined diagnosis-related groups, version 4, volume 2, Commonwealth Department of Health and Family Services, to be published January 1999). Moreover, AR-DRG-4 has an alphanumeric coding system potentially facilitating such splits. For example, in AR-DRG-4 the code for gastroenteritis in children under 10 years is G68 with a split for complicating factors -- G68A/G68B -- with the possibility of adding further letters for further splits. A recent multicentre study has confirmed that ATSI inpatients in remote hospitals in Australia have longer lengths of stay and consume more resources in most of the Major Disease Categories (MDCs)4 (see Fisher et al). Our study supports these data and suggests that improving the health of Aboriginal children by reducing comorbidities would reduce hospital costs. Our study highlights the heterogeneity of Aboriginal patients, with those in remote communities having more comorbidities. For this reason we do not recommend that ethnicity be classified as part of the DRG system, but that the emphasis should remain on clinical criteria. Unfortunately, however, "Aboriginality" remains a marker of "multiple comorbidities" and at least one State funding formula carries a loading for this patient group.5 Our sample was too small to compare ATSI and non-ATSI patients from remote communities. However, other studies have shown lower hospitalisation rates in the non-ATSI remote populations.6 That some patient populations need special consideration and do not fit national averages is not a new message, but to date many studies aiming to make this point have been descriptive.7,8 Clinicians are being urged to review their data and oversee the impact of casemix-based funding,9 but the number of published quantitative studies remains small. Data-based clinical interpretation is imperative if patients are not to be disadvantaged by output measures being linked to and equated with "efficiency". These results confirm the clinical impression that there is potential for inappropriate funding of inpatient Aboriginal children under the current classification system. The Northern Territory Health Service is aware of these issues and is implementing a special cost weight in such DRGs; patients in other DRGs remain vulnerable. References Australian Bureau of Statistics and the Australian Institute of Health and Welfare. The health and welfare of Australia's Aboriginal and Torres Strait Islander peoples. Canberra: ABS/AIHW, 1997. (Catalogue No. 4704.0.) Population estimates. Darwin: Epidemiology and Statistics Branch, Territory Health Services, NT, 1997. Pilla J. Developments of AN-DRGs: meeting the concerns of clinicians. Med J Aust 1994; 161 Suppl Sep 5: S9-S11. Commonwealth Department of Health and Family Services. Report on National Aboriginal and Torres Strait Islander Casemix Study. Adelaide: Brewerton and Associates Pty Ltd, April 1997. South Australian Health Commission. Casemix funding for health -- hospitals. Adelaide: SA Health Commission, 1997-98. Hart FG, Ring I, Runciman C. Public hospital activity, expenditure and staffing levels for indigenous and non-indigenous settlements in remote Queensland. Aust J Public Health 1993; 17: 325-330. Henderson A. Casemix-based funding for Queensland discriminates against hospitals treating very sick patients [letter]. Aust N Z J Med 1996; 26: 421-422. Stoelwinder JU. Casemix payment in the real world of running a hospital. Med J Aust 1994; 161 Suppl Sep 5: S15-S18. Hickie JB. Clinical representation in the development of casemix: measures and applications in Australia. Med J Aust 1994; 161 Suppl Sep 5: S6-S8. Authors' details Royal Darwin Hospital, Darwin, NT. Alan R Ruben, FRACP, FAFPHM, formerly, Community Paediatrician, Territory Health Services; currently, Advisor in Paediatrics, Fiji School of Medicine, Suva, Fiji Islands. Dale A Fisher, FRACP, DTM&H, Physician and Senior Lecturer. Reprints will not be available from the authors. Correspondence: Dr A R Ruben, Fiji School of Medicine, PO Box 11683, Suva, Fiji Islands. E-mail: alan_rATfsm.ac.fj

Synopsis With increasing implementation of casemix-based funding for hospitals, quantitative data were needed to confirm the clinical impression that treating Aboriginal (compared with non-Aboriginal) inpatients consumes significantly more resources. Utilisation data, collected over a three-month period in 10 hospitals, were used to determine a cost per inpatient episode, which was grouped according to AN-DRG-3 to give a cost per AN-DRG for Aboriginal and Torres Strait Islander (ATSI) patients and non-ATSI patients. ATSI patients had consistently longer average length of stay and significant variation in relative frequency of admissions, compared with non-ATSI patients, with higher prevalences of infectious diseases. Degenerative and neoplastic conditions were more common in non-ATSI patients. There were significant differences in casemix-adjusted costs per patient episode (ATSI, $1856; non-ATSI, $1558; P < 0.001). Our study has quantified differential resource consumption between two Australian populations, and highlights the need for recognition of some hospitals' atypical populations and special funding requirements. Introduction There is substantial evidence in the medical literature of poor health outcomes for Aboriginal and Torres Strait Islander (ATSI) people despite high hospital utilisation rates.1 Among the reforms designed to improve health outcomes, casemix classification (Australian national diagnosis-related groups, AN-DRGs) for hospital inpatients could, on the contrary, have deleterious effects if its limitations were not appreciated. The principle underpinning casemix systems -- that clinically similar patients consuming similar resources can be grouped into a DRG which will have an equal spread of patients consuming more and less resources -- means that a hospital with an atypical population will be inappropriately funded. Health service providers who treat patients from remote Aboriginal communities believe that treating Aboriginal patients is considerably more expensive for a range of reasons (severity of disease at presentation, comorbidities, and social factors relating to culture, education and remote location), but there are few data quantifying their resource consumption during inpatient care. With increasing implementation of casemix, quantitative data were urgently needed, so that hospitals caring for such populations would receive appropriate funding. The first study attempting to quantify differential resource consumption of Aboriginal and non-Aboriginal patients2 had considerable methodological problems, resulting in the data being of limited use. In 1993 the Australian Casemix Clinical Committee recommended to the (then) Commonwealth Department of Human Services and Health that a multicentre study be conducted to quantify differences in resource consumption patterns between ATSI and non-ATSI inpatients in rural and remote settings. Methods In view of the complexity of the project, a representative steering committee was established to define the scope and provide clinical oversight for the proposed research. After an analysis of Australia-wide hospital morbidity data, including utilisation rates by ATSI patients, a sampling framework was developed. Ten hospitals of more than 30 beds from Western Australia (Kalgoorlie), Northern Territory (Royal Darwin, Katherine and Alice Springs), South Australia (Port Augusta) and Queensland (Cairns, Mount Isa, Cunnamulla, St George and Innisfail) agreed to participate as study sites. External consultants (Brewerton and Associates, Adelaide) were appointed to facilitate data collection and analysis within the guidelines established by the steering committee. Collection and review of data, and consultation Data were collected from each site over a three-month period. Six sites commenced collection on 1 July 1995. The remaining four sites began one month later. For each patient in the study, a range of detailed utilisation data was obtained. Specific proformas were developed to collect details on nursing time, medical time, diagnostic services (pathology and imaging) and therapeutic services (theatre, pharmaceuticals, allied health). Additional information on diagnosis, procedures and morbidity was obtained from the hospitals' information systems. The utilisation data were used to determine a cost per inpatient episode. The costed patient data were grouped according to AN-DRG-3 to produce a cost per AN-DRG for the two populations. Traditional costing studies, which use cost information extracted from the hospital's general ledger and allocated to DRG classes, would not have provided costing information to the required level. Therefore, we used national unit prices to complete the cost allocation process (Box 1). This also overcame the lack of sophistication of many of the hospitals' cost reporting, and avoided the need to make accrual adjustments to hospitals' general ledgers for the three-month period. The national unit prices were based on national and State labour force data, and recently completed national casemix costing and service weight studies and analyses undertaken to generate AN-DRG-3 cost weights.3 This approach also removed idiosyncratic local cost variations and enhanced the reliability of the results. Thus, for the purposes of our study, costs such as those for a unit of nursing time, and individual radiology and pathology tests, were the same for all hospitals. Patients were classified according to AN-DRG-3. To ensure satisfactory coding standards, a random audit of medical records was undertaken in each hospital before the commencement of data collection. Interim results were compiled and presented at a workshop in Alice Springs in April 1996. Attendees included health service providers from the study hospitals and State Health Departments, as well as representatives from consumer groups, such as the National Aboriginal Community Controlled Health Organisation (NAACHO) and the Office of Aboriginal and Torres Strait Islanders (OATSI). As a result of this meeting the data were further refined, allowing for more clinically accurate and culturally appropriate interpretation. A final report was presented to the Commonwealth Department of Health and Family Services in April 1997.4 Ethical approval Participating hospitals were required to consider the ethical implications of the research project, and, in particular, issues of confidentiality. The hospital data and the study report were not to include any information identifying individual patients or communities. At the conclusion of the study, hospitals were provided with their own data in addition to that of the total cohort. No hospital had access to another hospital's data unless by private arrangement. Statistical analysis Collation of data was facilitated by a specially designed application using dBase as the programming tool. SPSS (SPSS Inc, Chicago, Illinois, USA) and standard spreadsheet packages were used for the analyses, which were based on t tests, as comparisons were between two populations with large sample sizes. For both populations, only those AN-DRGs with a sample size exceeding 20 separations were analysed. Results The study collected clinical and demographic data on 31 222 inpatient episodes and utilisation data relating to 128 813 occupied bed-days. These data were trimmed to remove incomplete episodes during the study period. A total of 27 768 separations were analysed in detail (Box 2). It was not possible to standardise the data by sex and age for the total study population, as population data for the hospitals' catchment areas were not available. Standardised data for the Northern Territory (not presented here) revealed higher admission rates for male and female ATSI patients compared with non-ATSI patients in all age groups. While ATSI separations represented 33.9% of the total cohort they represented 66% (115/174) of patients and 67.6% (3157/4673) of separations of those assigned AN-DRG 572, Admit for renal dialysis. Because of the impact this caseload would have had on cost analysis (eg, one population would have a disproportionate number of "day-only" admissions), data relating to dialysis were excluded from relevant sections of the analysis. The average length of stay in the ATSI population was two days shorter when AN-DRG 572 was included in the analysis, but in the non-ATSI cohort it was only 0.3 days shorter. Differences in DRGs The ATSI population was distributed across 426 of a possible 667 (64%) DRGs. In contrast, the non-ATSI population was distributed across 547 DRGs (82%). Box 3 highlights the consistently longer average length of stay of ATSI patients, as well as a significant variation in relative frequency of admissions. For example, DRGs for gastroenteritis and respiratory infections contain more ATSI patients, despite there being twice as many non-ATSI patients in the cohort. In contrast, DRGs for gastroscopy and colonoscopy have a higher proportion of non-ATSI patients. Dental extractions and restorations recorded low separation rates in ATSI patients. The following DRGs were not encountered in ATSI patients during the data collection period: Other major joint and limb reattachment procedures without comorbidities and complications; Major shoulder or elbow procedures, age < 60; and Hip and femur procedures except major joint, age > 54 without comorbidities and complications. Boxes 4 and 5 show the top 20 DRGs by volume for ATSI and non-ATSI patients, respectively. Of note is the prevalence of infectious diseases in the ATSI population compared with the non-ATSI population, whereas the non-ATSI population has a high prevalence of degenerative diseases and DRGs related to neoplastic conditions. Cost differences The unadjusted average cost of an ATSI inpatient episode was $1627 compared with $1545 for non-ATSI inpatient episodes (this difference was not significant). The casemix-adjusted costs, however, showed significant differences (P < 0.001) per episode at $1856 and $1558 for ATSI and non-ATSI patients, respectively (Box 6). Box 7 shows the breakdown of total and average costs and confirms that the cost differential is a result of increased utilisation of most services. Theatre and pathology services are the only areas where costs are higher for non-ATSI patients. Further analysis of the data showed that operating room expenses were higher for ATSI patients. However, the average cost is lower because a significantly smaller number of ATSI patients had operations. The data also confirm that ATSI patients have longer lengths of stay and higher costs in most Major Diagnostic Categories (MDCs) (Box 8). An unexpected observation was the shorter length of stay and cost for this population in MDC 19 (Mental Diseases and Disorders), and MDC 20 (Alcohol/Drug Use and Alcohol/Drug Induced Organic Mental Disorders). Discussion The study confirmed the clinical perception that caring for ATSI inpatients consumed greater resources for the same DRG than caring for non-ATSI inpatients, and demonstrated a 39% overall differential cost. For some DRGs (eg, those including paediatric infectious diseases) the increase in resource consumption was considerable in ATSI patients. In MDCs 19 and 20, non-ATSI patients used slightly more resources. The greater costs in ATSI patients are believed to be related to disease severity on admission as well as comorbidities and complicating factors. Data obtained during the project support this. Other studies have also found that resource utilisation for ATSI patients is lower for mental disorders.5 Easier reintegration of ATSI patients into their community may facilitate shorter lengths of stay. Social networks and supports may also favour outpatient psychiatric care. The same may be true for DRGs associated with alcohol abuse, although given the known prevalence and impact of substance abuse in ATSI patients, we may also be identifying a need for further review of the models of healthcare delivery to ATSI patients. Actual needs were not addressed by our study. We only measured the current state of healthcare provision, which is largely a result of historical funding arrangements. However, many clinicians would argue that current health services for underprivileged groups are inadequate. This is the first study to quantify differential resource consumption between two Australian populations. It highlights the need to recognise potentially confounding factors when a casemix classification funding system is implemented. The Northern Territory, South Australia and New South Wales have recognised the disparity and incorporated funding adjustments for ATSI patients. As with ATSI patients in remote and rural hospitals, other socially disadvantaged groups including Aboriginals in urban settings and immigrant subpopulations may also have a cost and utilisation profile different from the "typical" Australian population. Hospitals caring for a significant proportion of such patients may equally need recognition for their "atypical" population. Appropriate funding of such hospitals can be either through funding adjustments or by an improved classification system. Future versions of AN-DRGs are likely to make greater use of complicating clinical factors (CCFs), which could include indicators of social disadvantage. Notwithstanding these efforts, hospitals caring for atypical populations remain vulnerable because their relatively small number of patients lack statistical importance when national figures are reviewed. One of the great challenges of casemix implementation is to provide the basis by which hospitals can be funded appropriately for appropriate care. If this challenge is not met it is the sickest patients from the most disadvantaged subpopulations who will suffer. The Aboriginal and Torres Strait Islander Casemix Study has demonstrated a genuine risk in this regard. 1: Standard unit of cost Unit cost per minute by nursing level Obtained from:Market Basket Database: CDHS&H Applied to:Patient attributable time by nurse per patient Unit cost per minute by allied health professional level Obtained from:Market Basket Database: CDHS&H Applied to:Patient attributable time by allied health professional by patient Unit cost by banded time range for medical officer Obtained from:Banded ranges and standard cost as specified in the MBS schedule and adopted by the South Australian Health Commission Applied to:Frequency of consultations by time range by medical officer Unit cost per operating minute by procedure Obtained from:National Operating Room Service Weight Study: CDHS&H Applied to:Time spent in theatre and recovery rooms Unit cost per day in intensive care/critical care/neonatal intensive care Obtained from:National Intensive Care Service Weight Study: CDHS&H Applied to:Time spent in intensive care/critical care/neonatal intensive care Unit cost per pathology test by type Obtained from:National Pathology Service Weight Study: CDHS&H Applied to:Each pathology test ordered and undertaken per patient Unit cost per diagnostic imaging service by type Obtained from:National Diagnostic Imaging Service Weight Study: CDHS&H Applied to:Each diagnostic imaging procedure performed per patientUnit cost per pharmaceutical by type Obtained from:Average unit price based upon data provided from the participating sites Applied to:Drug type administered per patient by dosage and frequency Unit cost per prosthesis by type Obtained from:Standard List National Operating Room Service Weight Study: CDHS&H Applied to:Prostheses consumed in theatre Unit overhead rate Obtained from:Development of AN-DRG-3 Cost Weights: CDHS&H Applied to:Each day of stay, covering overhead costs plus each day of stay for a boarder CDHS&H = Commonwealth Department of Human Services and Health (now, Commonwealth Department of Health and Family Services). MBS = Medical Benefits Schedule. 8: Average cost by major diagnostic category (MDC) (including AN-DRG 572) MDC DescriptionATSINon-ATSI SepsALOSAverage cost ($)SepsALOSAverage cost ($) 0 Pre MDC3058.23826.9138710.04229.00 1 Diseases and Disorders of the Nervous System2828.42946.617416.62291.44 2 Diseases and Disorders of the Eye873.21482.152162.21093.66 3 Diseases and Disorders of the Ear, Nose, Mouth and Throat3003.31460.268392.11056.30 4 Diseases and Disorders of the Respiratory System7585.81902.799775.71905.76 5 Diseases and Disorders of the Circulatory System2507.92757.718705.82558.46 6 Diseases and Disorders of the Digestive System4646.42185.7418912.91284.93 7 Diseases and Disorders of the Hepatobiliary System and Pancreas877.62039.80535.32141.82 8 Diseases and Disorders of the Musculoskeletal System and Connective Tissue4137.12964.0616015.42225.52 9 Diseases and Disorders of the Skin, Subcutaneous Tissue and Breast3576.52103.137973.61413.75 10 Endocrine, Nutritional and Metabolic Diseases and Disorders17611.53136.521606.52034.64 11 Diseases and Disorders of the Kidney and Urinary Tract34161.6484.1118821.6514.51 12 Diseases and Disorders of the Male Reproductive System592.71360.241862.41179.23 13 Diseases and Disorders of the Female Reproductive System2004.21753.858192.41250.05 14 Pregnancy, Childbirth and the Puerperium6755.11609.2121153.11104.41 15 Newborns and Other Neonates5296.23310.7110804.52901.82 16 Diseases and Disorders of the Blood and Blood Forming Organs585.72019.621412.8959.71 17 Myeloproliferative Diseases and Poorly Differentiated Neoplasms255.21286.682991.9591.75 18 Infectious and Parasitic Diseases8910.43257.201835.61949.19 19 Mental Diseases and Disorders838.52086.4342210.62542.97 20 Alcohol/Drug Use and Alcohol/Drug Induced Organic Mental Disorders473.7852.04854.51291.02 21 Injury, Poisoning and Toxic Effects of Drugs2254.61714.035863.41487.49 22 Burns7012.95454.901938.93297.14 23 Factors Influencing Health Status and Other Contacts with Health Service4625.81551.5716282.4677.04 Total94174.61627.27183513.91545.65 ATSI = Aboriginal and Torres Strait Islander. Seps = Separations. ALOS = Average length of stay (days). AN-DRG 572 = Admit for renal dialysis. 3: Top 20 DRGs by volume - total study population AN-DRG and descriptionTotalATSINon-ATSI SepsALOSSepsALOSSepsALOS 572 Admit for renal dialysis46731.0131571.0215161.00 943 Other factors influencing health status18402.363955.4914411.49 727 Neonate, admission weight <2499g without significant operating room procedure, without problems9943.702684.257263.50 674 Vaginal delivery without complicating diagnoses8053.451633.846423.35 952 Ungroupable6147.932887.063268.70 683 Abortion with D&C, aspiration curettage or hysterectomy5151.15701.464451.10 332 Other gastroscopy for non-major digestive disease without comorbidities and complications3801.41252.323551.34 686 Other antenatal admission with moderate or no complicating diagnoses3572.44883.572692.07 172 Respiratory infections/inflammation, age <55 without comorbidities and complications3514.782455.241063.72 350 Gastroenteritis age <103385.771888.881501.85 187 Bronchitis and asthma age <50 without comorbidities and complications2662.69563.252102.54 659 Conisation, vagina, cervix and vulva procedures2591.57352.692241.39 780 Chemotherapy2561.23152.602411.14 885 Injuries age <652542.301032.971511.84 491 Cellulitis age <60 without comorbidities and complications2224.14915.371313.29 484 Other skin, subcutaneous tissue and breast procedure2131.98324.501811.53 335 Other colonoscopy without comorbidities and complications2111.6793.442021.59 128 Dental extractions and restorations2081.26321.591761.20 349 Oesophagitis/gastroenteritis/other digestive disease age 10-742082.26482.981602.05 660 Endoscopic procedures, female reproductive system2061.34421.901641.20 ATSI = Aboriginal and Torres Strait Islander. Seps = Separations. ALOS = Average length of stay (days) Acknowledgements This study was funded by the Commonwealth Department of Health and Family Services and sponsored by the Australian Casemix Clinical Committee and received constant support from all members and the then Chair, Professor John Hickie. We would like to acknowledge the cooperation of staff at the study hospitals and State and Territory Health Departments and the assistance of the University of Adelaide Statistics Department. Countless individuals were also major contributors to the study, including Art Huston and Jenni Bowen (Brewerton and Associates) and Alan Browne and Josie Lanza (Commonwealth Department of Health and Family Services). The study also owes its success to the other members of the Steering Committee, Dr Mark Salmon, Mr Peter Woodley, Ms Marian Kickett, Mr Garnett Brady and Dr Chris Wagner. References Plant AJ, Condon JR, Durling G. Northern Territory health outcomes, morbidity and mortality 1979-1991. Darwin: Northern Territory Department of Health and Community Services, 1995. Harkin K. Incremental resource consumption by Aboriginal inpatients: a research project conducted at Alice Springs Hospital from 1 October to 31 May,1992. Report to the Department of Human Services and Health. Darwin: NT Dept of Health and Community Services, 1994. Casemix Development Program. Report on the development of AN-DRG Version 3 Cost weights. Canberra: Commonwealth Department of Human Services and Health, 1995. Commonwealth Department of Health and Family Services. Report on National Aboriginal and Torres Strait Islander Casemix Study. Adelaide: Brewerton and Associates Pty Ltd, April 1997. Jablensky A. The epidemiology of schizophrenia. Curr Opin Psych 1993; 6: 43-52. Authors' details Royal Darwin Hospital, Darwin, NT. Dale A Fisher,* FRACP, DTM&H, Physician and Senior Lecturer. Classification and Payments Branch, Department of Health and Family Services, Canberra, ACT. Jo M Murray,* BSc(Med), Acting Assistant Secretary. Princess Alexandra Hospital, Brisbane, QLD. Michael I Cleary,* FACEM, MHA, Executive Director of Medical Services. Brewerton and Associates, Adelaide, SA. Rita E Brewerton, BSc(MaSc)Hons, Director. Reprints will not be available from the authors. Correspondence: Dr D A Fisher, Royal Darwin Hospital, PO Box 41326, Casuarina, NT 0811. E-mail: dale.fisherATnt.gov.au *Steering Committee members (other members are listed in the Acknowledgements above).

Synopsis Casemix funding was introduced first in Victoria in 1993-94, and since then most States have moved towards either casemix funding or using casemix to inform the budget setting process. The five States implementing casemix have adopted some common funding elements: all use AN-DRG-3; all have introduced capping, most commonly at the hospital level; and all ensure accuracy of diagnosis and procedure coding through coding audits. Two funding models have been developed. The fixed and variable model involves a fixed grant for hospital overhead costs and a payment for each patient treated, covering only variable costs. The integrated model provides an integrated payment to hospitals for each patient treated, covering both the fixed and variable costs. There are different weight setting processes and base prices between the States, which result in marked differences in the price paid for the same type of case treated in similar hospitals. Learning across State boundaries should be encouraged, with knowledge of what is effective and what is ineffective in casemix funding arrangements being used to develop Australian best practice in this area. Introduction In Australia, casemix funding was first introduced in Victoria in 1993-94,1 as part of a program of public sector restructuring to reduce expenditure and improve the efficiency.2,3 South Australia4 followed in 1994-95, with a casemix funding approach modelled substantially on the Victorian scheme5 and also accompanied by significant budget cuts. Since then, Western Australia6 and Tasmania7 have also implemented casemix funding (both in 1996-97), and Queensland8 has commenced a phasing-in process for casemix funding. New South Wales is the only State which has eschewed casemix funding arrangements, instead structuring providers on the basis of an area responsibility for hospitals and other service units (eg, community health centres). Funding is distributed to areas based on their population. Even in New South Wales, policy documents emphasise the importance of casemix in informing budgets for hospitals and in paying for patients across regional boundaries.9 The Northern Territory and the Australian Capital Territory have also incorporated elements of casemix funding, but because of their small populations and small number of distinct providers, funding arrangements are essentially determined individually, even when an elaborated formula is used.10 Initial casemix implementation required an unravelling of hospital activity into the major streams of care: inpatient, outpatient, and teaching and research. Casemix implementation focused first on inpatient services where classification for describing services or "products" was the most sophisticated. Inpatient funding arrangements The essence of casemix funding for inpatient services is quite simple: the budget for a hospital is based on the number and type of patients treated in the hospital. The development of diagnosis-related groups (DRGs) as clinical and resource homogeneous categories for inpatients11 provided a means of grouping types of patients treated, which could be used for payment purposes. The budgets of hospitals could thus be determined primarily on performance or output, rather than negotiation, history or politics. The five States implementing casemix funding have all adopted some common funding elements. Firstly, a common nomenclature is used: all States currently use version 3.1 of Australian national diagnosis-related groups (AN-DRGs). Secondly, as these funding arrangements coincided with budget reductions, all the States have introduced capping, most commonly at the hospital level with hospital-specific targets. In some States, the throughput targets are flexible -- if hospitals exceed these targets, they receive additional funding, albeit at a marginal price. Thirdly, with DRG assignment based on recorded diagnosis and procedure codes, all States have introduced coding audits to ensure accuracy of recording. Other aspects of inpatient casemix funding reveal remarkable variability between the States. The Box compares key elements of inpatient funding arrangements across the States. Funding models Two basic funding models have been developed. The initial Victorian model was based on fixed and variable components, following the recommendations of the 1990 Scotton and Owens review of the prospects of casemix funding in Australia.12 Queensland has also adopted this model. A fixed and variable model involves two elements: a fixed grant to cover hospital overhead costs, and a payment for each patient treated covering only the variable costs of that patient. The theory behind this approach is that efficiency is maximised if the incentives are such that hospitals can treat additional patients up to the point at which marginal treatment cost equals marginal revenue. Marginal revenue is the variable payment made by State health authorities. This fixed and variable model mitigates the incentive for hospitals to maximise admissions. After current capacity limits are reached, additional fixed costs are required (eg, for commissioning new wards), but these are not fully reimbursed by the funding system. Thus, States retain control over growth in system capacity. The alternative model is used in Western Australia and Tasmania, which both provide an integrated payment to hospitals for each patient treated covering both the fixed and variable costs. South Australia also uses an integrated payment system, but if a hospital does not achieve the negotiated volume target, payments are discounted, effectively recognising that the savings to hospitals are at marginal or variable costs. Cost variability It would be expected that the utility of treating a patient in a particular DRG would be constant across all hospitals. Likewise, logically, the payment for that DRG should be the same, regardless of the hospital. However, the different State funding systems recognise that there are differences in costs and four of the State systems (Victoria, Queensland, Western Australia and South Australia) have established several funding subgroups that receive different payments. Interestingly, assumptions about economies of scale vary. The Victorian and Western Australian systems assume economies of scale exist, as the payment for a patient in a particular DRG is less in a larger hospital than in a smaller hospital. On the other hand, in Queensland and South Australian diseconomies of scale are assumed and payments are higher in larger hospitals than in smaller hospitals. Divergence in weights The weight setting process also differs across the country. Three States (Queensland, South Australia and Tasmania) use variants of the weights developed as part of the national cost weight study, which are derived from cost modelling undertaken as part of national costing studies.13 Victoria and Western Australia, on the other hand, use data from clinical costing systems in their own State to set weights. In Victoria, for example, weights are set using a dataset of patient costs from over 0.5 million recorded patient admissions to 15 Victorian public acute hospitals in the financial year two years prior to the payment year.14 Price differences As a result of these different weight setting processes, different relative weights are used across the country for each DRG, as well as different base prices. The different prices and weights mean that there are differences between States in the price paid for the same case treated in a similar hospital. This effect can be seen with AN-DRG 674, Vaginal delivery without complicating diagnoses. In the two States with fixed and variable funding (Victoria and Queensland) the variable price paid for an inlier patient in AN-DRG 674 in a major hospital in 1997-98 varies by more than 20% ($925 in Victoria versus $1121 in Queensland). Similarly, in the integrated funding States there is a difference of more than 40% ($1455 in South Australia versus $2097 in Tasmania, with the Western Australian price lying in between these at $1685). Payment differences of this size need some explanation. Firstly, in some DRGs, these differences might reflect differences in inpatient payment system design (eg, in intensive care payments), but this should not be relevant in this DRG. Secondly, they might be attributable to differences in other aspects of the payment system (eg, in training and development), but this would not be sufficient to account for the magnitude of the payment differences. Finally, the differences might reflect different input costs, a factor taken into account in payment system design in the United States.15 However, most employees in public hospitals are now covered by Federal awards and so input price variation is not a feasible explanation. It is thus difficult to see how cost differences of this order of magnitude can be justified for this reasonably homogeneous DRG. Outliers The DRG classification system has been developed to describe the normal, or typical, case in a DRG, known as an inlier. Outlier cases are those which do not fit the normal pattern and, in terms of distribution, lie outside so called "trim points". The basis for setting trim points, and thus determining outliers in all States, follows work done by McGuire et al on the effect of different trim point methods.16 The common trimming method used in Australia is the L3H3 method: the low trim point is a third of the average length of stay, and the high trim point is three times the average length of stay. Some States use modifications of this approach: Queensland has an extra high trim point based on five times average length of stay; and in South Australia the low trim point is determined parametrically at 3 SD below the mean length of stay (where average length of stay is greater than four days). Two States, Western Australia and South Australia, also have trim points based on cost, which provide for additional funding for cases identified as costing more than $75 000 and $60 000, respectively. The cost-based trim points rely on hospitals having robust clinical costing systems with agreed bases for allocation of costs, as different assumptions about allocation of overhead costs, for example, can significantly affect the recorded cost of a case.14 Intensive care Intensive care funding is a particularly sensitive issue, given that variation in system-wide use of intensive care cannot be fully explained by variation in epidemiological and demographic factors.17 There is considerable variation across States in the funding arrangements for intensive care units, and the payment systems have quite different incentive effects. Within an individual hospital there can also be differential incentives on intensive care unit staff to admit to hospital or retain patients within the unit, depending on the funding structure. Victoria and Tasmania provide no specific additional funding for intensive care units. Western Australia provides block funding for intensive care units on top of the existing funding arrangements. In South Australia intensive care units are effectively funded on a per diem basis, with the DRG cost weight calculation being adjusted to exclude intensive care costs. Private patients Costs of private patients obviously differ from those of public patients. Private patients' medical costs (including pathology and radiology costs) and prostheses costs are met by the patients themselves, and are normally reimbursed by health insurance funds. Hospitals also accrue revenue from these patients. The Victorian and Queensland funding models provide differential payments for public and private patients to take account of these different cost structures. Western Australia provides a block payment to compensate for the differing proportion of public patients. On the other hand, the South Australian and Tasmanian arrangements do not provide differential payments, and although revenue differences are compensated for, there would still be an effective incentive to admit private rather than public patients because of the lower hospital costs for private patients. Conclusion Five States have either implemented or are in the process of implementing casemix funding, but the funding models used have significant design differences. Some of the systems are clearly fairer to hospitals than others, and it is therefore not surprising that recent reviews of both Victorian and South Australian formulas have indicated that providers believe that there are still problems in funding design.18,19 Although the design of a funding system is in part a technical process to ensure that hospitals have appropriate incentives for efficiency, it is also a political process insofar as providers need to be assured that the funding formula is fair. The large variation in prices for the indicator DRG used in this article (DRG 674) also suggests an element of inequality between States in pricing strategies. Design of a casemix funding system involves a number of complex technical choices to maintain appropriate balances between competing policy objectives (eg, minimising waiting lists versus reducing stays in hospital emergency departments). Similarly, maintenance of casemix payment systems needs to take account of changes in health technology and to monitor perverse effects of funding system design. There is thus a strong argument that there should be some form of joint development to facilitate better funding system design. State casemix funding arrangements have evolved in a number of areas, such as in the weight setting processes and in the elaboration of the role of the purchaser (including how volume controls are implemented). Although casemix funding arrangements are characterised by relatively low transaction costs, annual funding policy reviews in each State probably incorporate unnecessary overheads. Differences between the States should not preclude the possibility of learning across State boundaries. As casemix funding enters a more mature phase, knowledge of what is effective and what is ineffective in casemix funding arrangements should be used to develop Australian best practice in this area. National cooperation (and national leadership) produced an agreed national casemix classification. Further national action is warranted to facilitate transfer of the best practice elements of each State's funding systems. This should occur early in casemix funding development to reduce the costs incurred by States "reinventing the casemix funding wheel" each year. Furthermore, there remain several areas in which casemix funding is deficient; for example, in identifying and funding teaching and research activities of hospitals, and in the development of funding policy in ambulatory care. Cooperative national action in these areas is needed. References Duckett SJ. Hospital payment arrangements to encourage efficiency: the case of Victoria, Australia. Health Policy 1995; 34: 113-134. Alford J, O'Neill D, editors. The contract state: public management and the Kennett government. Melbourne: Deakin University Press, 1994. Armstrong A. A comparative analysis: new public management -- the way ahead? Aust J Public Adm 1998; 57: 12-24. Department of Human Services. Victoria -- public hospitals: policy and funding guidelines 1997-98. Melbourne: Department of Human Services, 1997. South Australian Health Commission. Casemix funding for health -- hospitals -- 1997-98. Adelaide: South Australian Health Commission, 1997. Health Department of Western Australia. Western Australian government health system funding 1997/1998: Budget reform. Perth: Health Department of Western Australia, 1997. Department of Community and Health Services. Casemix: managing resources for care (policy paper). Hobart: Artemis, 1997. Queensland Health. Hospital funding model for Queensland public hospitals: policy and technical papers 1997/98. Brisbane: Queensland Health, 1998. NSW Health Department. Implementation of the economic statement for health. Sydney: NSW Health Department, 1996. (PDD No. 96-0081.) Beaver C, Zhao Y, McDermid S, Hindle D. Casemix-based funding of Northern Territory public hospitals: adjusting for 36 severity and socio-economic variations. Health Econom 1998; 7: 53-61. Fetter RB, Shin Y, Freeman JL, Averill RF. Case mix definition by diagnosis related groups. Med Care 1980; 18 (2 Suppl): 1-53. Scotton RB, Owens HJ. Case payment in Australian hospitals: issues and options. Melbourne: Public Sector Management Institute, Monash University, 1990. Commonwealth Department of Human Services and Health. Report on the development of AN-DRG version 3 cost weights. Canberra, The Department, 1995. Jackson T, Wilson R, Watts J, et al. Final Report of the 1997 Victorian Cost Weights Study. Melbourne: Victorian Department of Human Services, 1998. Duckett SJ. Health care in the US: what lessons for Australia? Sydney: The Australian Centre for American Studies, University of Sydney, 1997. McGuire TE, Bender JA, Maskell C. Casemix episodic payment for private health insurance. Canberra: AGPS, 1995. Jackson T, Macarounas-Kirchmann K. Changing patterns of intensive care unit admission and length of stay in five Victorian hospitals. In: Selby-Smith C, editor. Economics and health: 1992. Melbourne: Monash University/NCHPE, 1993: 149-164. Auditor-General of Victoria. Acute health services under casemix: a case of mixed priorities. Melbourne: Victorian Government Printer, 1998. (Special Report No. 56.) Brooker J. An evaluation of casemix funding in South Australia 1994-95. Canberra: Commonwealth Department of Health and Family Services, Casemix Development Program, 1996. Authors' details La Trobe University, Melbourne, VIC Stephen J Duckett, BEc, MHA, PhD, Professor of Health Policy; and Dean, Faculty of Health Sciences. Reprints: Professor S J Duckett, Faculty of Health Sciences, La Trobe University, Bundoora, VIC 3083. E-mail: s.duckettATlatrobe.edu.au

Synopsis The costs of subacute care (palliative care, rehabilitation medicine, psychogeriatrics, and geriatric evaluation and management) and non-acute care (nursing home, convalescent and planned respite care) are not adequately described by existing casemix classifications. The predominant treatment goals in subacute care are enhancement of quality of life and/or improvement in functional status and, in non-acute care, maintenance of current health and functional status. A national classification system for this area has now been developed -- the Australian National Sub-Acute and Non-Acute Patient Classification System (AN-SNAP). The AN-SNAP system, based on analysis of over 30 000 episodes of care, defines four case types of subacute care (palliative care, rehabilitation, psychogeriatric care, and geriatric evaluation and management) and one case type of non-acute care (maintenance care), and classifies both overnight and ambulatory care. The AN-SNAP system reflects the goal of management -- a change in functional status or improvement in quality of life -- rather than the patient's diagnosis. It will complement the existing AN-DRG classification. Introduction The Australian healthcare system is about to implement a new casemix classification system for subacute and non-acute care, the costs of which are not adequately described by traditional diagnostic tools. Subacute care comprises palliative care, rehabilitation medicine, psychogeriatrics, and geriatric evaluation and management. Non-acute care includes nursing home, convalescent and planned respite care. The new casemix classification system, which includes hospital as well as community care, reflects the goal of management -- a change in functional status or improvement in quality of life -- rather than the underlying patient diagnosis. Background Subacute casemix has been evolving for 15 years. In 1983, when the United States Health Care Financing Administration decided that payments for hospital care would be on a prospective payment system, based on acute-care diagnosis-related groups (DRGs), rehabilitation, psychiatric, children's and long-term facilities were specifically excluded. It was recognised that these forms of care, although not acute, were still complex and expensive and required long hospital stays. In 1987, a US Department of Health and Social Services report reiterated that their current DRG system did not adequately take into account the special circumstances of patients requiring long hospital stays.1 Studies in the United States over the following few years not only confirmed that DRGs did not adequately describe costs in one of these areas of care (rehabilitation medicine),2 but that as a consequence quality of care had deteriorated, as measured by changed length of hospital stay, increased readmission rates and a rising number of nursing home admissions.3-5 As casemix development progressed in Australia, Australian studies6-12 also expressed the need for a different approach for costing of rehabilitation,6,8-10,12 geriatric evaluation and management,6,9,12 palliative care7,9,11,12 and psychogeriatrics.6,12 The term subacute care was coined in 199213 to describe "care which is provided for a person who requires health services but whose principal medical diagnosis (modified for factors such as age and procedures) is not adequate in explaining the need for, or the cost of, the services that s/he receives". Goals of subacute and non-acute care In subacute care the predominant goal is enhancement of a patient's quality of life and/or improvement in his or her functional status. In non-acute care the predominant goal is maintenance of a patient's current health and functional status. Because of this difference in goals, it was expected that factors other than diagnosis were more likely to explain the costs of these forms of care. Rehabilitation: Factors contributing to the success of rehabilitation programs have included patient characteristics such as functional status on admission, age, disease site, time from referral to beginning of program, comorbidities such as cognitive function and depression, and availability of resources.14-16 The factor which appears in US and Australian studies to predict cost most accurately in these areas of care is a patient's functional status on admission.12,15-17 Palliative care: Australian clinicians were instrumental in developing a casemix classification system with a primary approach from a clinical perspective. The development involved broad consultation and collaboration. The palliative care classification identified stage of illness or palliative care phase (eg, stable, deteriorating, terminal), symptom severity and acuity level (or nursing dependency) as the major factors explaining costs for this form of care.7 Psychogeriatrics and other aged care: The goals of admission in aged care are improving health status, modifying symptoms and enhancing function, living conditions, behaviour and quality of life.12 Subacute and non-acute care classifications Several classification systems for subacute and non-acute episodes of care have been developed, including the Resource Utilisation Groups and the California Long Term Care System.18 The Resident Classification Index19 is an Australian classification system used in nursing homes to classify non-acute episodes of care. In the United States the FIM-FRG system (Functional Independence Measure- Function Related Groups)17 for rehabilitation medicine is the most developed. Studies in Australia have continued to demonstrate that the best predictor of cost for subacute care is the goal of care. The most recent studies are the 1995 Victorian Rehabilitation Casemix Report10 and the 1996 NSW Sub-Acute Casemix Area Network Project.12 AN-SNAP study The Australian National Sub-Acute and Non-Acute Patient Casemix Study20 was conducted in 1996 in 99 hospital and community health sites in all Australian States and Territories and in five sites in New Zealand. Over 30 000 episodes of care were analysed, including overnight, same day, outpatient and community episodes of care. The study established that there are five case types of subacute and non-acute care. Subacute care includes palliative care, rehabilitation, psychogeriatric care, and geriatric evaluation and management; and the final case type -- maintenance care -- is defined as non-acute care. Each of the five case types is defined according to the characteristics of the patient and the goal of care, and not the institution or service in which she or he is treated (eg, a patient may receive geriatric evaluation and management in a hospice, or palliative care in a rehabilitation unit). A critical finding of the study was that across the spectrum of case types and classes there is significant diversity in the cost of subacute and non-acute care for both overnight and ambulatory episodes. For example, there is a 30-fold variation in episode cost and a five-fold variation in per diem cost between the most expensive and the least expensive classes in the overnight classification, thus confirming the necessity for a classification in this area to allow for appropriate output-based funding. AN-SNAP classification system From the study, a national classification for subacute and non-acute care was developed -- the Australian National Sub-Acute and Non-Acute Patient Casemix Classification System, or AN-SNAP classification.20 AN-SNAP version 1 (Box 1)21 classifies both overnight and ambulatory care. It has 134 classes and the classification explains 58% of the variation in all episode costs. Of this 58%, 21% is contributed by episode type and 37% by the classes. The overnight branch has 66 classes and the classification explains 47% of the variance in the cost of overnight care. The ambulatory branch has 68 classes and the classification explains 28% of the variance in the cost of ambulatory care. These results are an improvement on the performance achieved by acute-care DRGs. Analysis of the decision trees for overnight and ambulatory care in Box 1 shows the factors which have been incorporated into the system as predictors of cost. Palliative care -- phase, functional dependence as measured by RUG-ADL (resource utilisation groups - activities of daily living),18 and age; Rehabilitation -- impairment groupings, functional status as measured by FIM (Functional Independence Measure),22 and age; Psychogeriatrics -- psychiatric symptom severity and functional status as measured by the HoNOS (Health of the Nation Outcome Scale);23 Geriatric evaluation and management -- cognitive status in addition to motor capacity and age; and Maintenance care -- functional status. The AN-SNAP study showed that the variables driving costs in the inpatient setting are also important cost drivers in the ambulatory setting. However, community care is inherently more complex than institutional care. Common variables across institutional and community care are necessary, but are insufficient in explaining cost variations. In consequence, the classification makes use of some community variables not required in institution care (eg, provider type and assessment or treatment episode). Implications of AN-SNAP The implementation of this classification has important implications. Firstly, a number of classifications are now available in Australia and policy decisions on the interaction between these classifications are required. Secondly, data on many of the characteristics used in AN-SNAP are currently collected by individual service providers, but most are not routinely collected by existing hospital and community information systems. AN-SNAP, along with its further development, has been endorsed by the Australian Casemix Clinical Committee for adoption as the national classification for sub- and non-acute care. Implementation remains a State and Territory issue which requires a planned, staged approach. Already some States, including Queensland and New South Wales, are implementing AN-SNAP, and others have indicated their intention to do so in the near future. The adoption of the system will complement the existing DRG system, as illustrated in the New South Wales approach (Box 2). References Batavia AI, DeJong G. Prospective payment for medical rehabilitation: the DHSS Report to Congress. Arch Phys Med Rehabil 1988; 69: 377-380. Stineman MG, Escarce JJ, Goin HE, et al. A case-mix classification system for medical rehabilitation. Med Care 1994; 32: 366-379. Evans RL, Hendricks RD, Bishop DS, et al. Prospective payment for rehabilitation: effects on hospital readmission, home care and placement. Arch Phys Med Rehabil 1990; 71: 291-294. Fitzgerald JF, Fagan LF, Tierney WM, Dittus RS. Changing patterns of hip fracture care before and after implementation of the prospective payment system. JAMA 1987; 258: 218-221. Heinemann AW, Billeter J, Betts HB. Prospective payment for acute care: impact on rehabilitation hospitals. Arch Phys Med Rehabil 1988; 69: 614-618. Roberts R, McKinley S, Brooks B, et al. The Australian National Non-Acute Inpatient Project. Aust Health Rev 1993; 16: 300-319. Smith M, Firms P. Palliative Care Casemix Classification -- testing a model in a variety of palliative care settings -- preliminary results. Proceedings of the Sixth Australian Casemix Conference; 1994 Aug 29-31; Hobart. Canberra: Commonwealth Department of Human Services and Health, 1994. Baker W. Casemix in rehabilitation -- is it safe to dip into functionally related groups? Proceedings of the Sixth Australian Casemix Conference; 1994 Aug 29-31; Hobart. Canberra: Commonwealth Department of Human Services and Health, 1994. Lee L, Goor E, Kennedy C, et al. Non-acute casemix in the Illawarra. J Qual Clin Pract 1994; 14: 23-30. Coopers & Lybrand. Rehabilitation Casemix Project. Final Report. Melbourne: Victorian Department of Health and Community Services, 1995. Hindle D. The Victorian Palliative Care casemix project: statistical analysis and funding recommendations. Wollongong: Centre for Health Service Development, University of Wollongong, 1995. Eagar K, Cromwell D, Kennedy C, Lee L. Classifying sub-acute and non-acute patients: results of the NSW Casemix Area Network Study. Aust Health Rev 1997; 20: 56-74. Eagar K, Innes K. Standard definitions and source data for hospitals in Australia, Canberra: Commonwealth Department of Health, Housing and Community Service, 1992. Carey RG, Posavac EJ. Who makes the most progress in inpatient rehabilitation? An analysis of functional gain. Arch Phys Med Rehabil 1988; 69: 337-343. Rondinelli RD, Murphy JR, Wilson DH, et al. Predictors of functional outcome and resource utilisation in inpatient rehabilitation. Arch Phys Med Rehabil 1991; 72: 447-453. Stineman MG, Escarce JJ. Analysis of casemix and the prediction of resource use in medical rehabilitation. Phys Med Rehabil Clin North Am 1993: 4: 451-461. Stineman MG, Escarce JJ, Goin HE, et al. A case-mix classification system for medical rehabilitation. Med Care 1994; 32: 366-379. Fries BE, Cooney LM. Resource Utilisation Groups: a patient classification system for long term care. Med Care 1985; 23: 110-132. Commonwealth Department of Health, Housing and Community Services. Classification of nursing home residents. Handbook for directors of nursing. Canberra: DHHCS, 1992. Eagar K, et al. The Australian National Sub-Acute and Non-Acute Patient Classification (AN-SNAP): report of the National Sub-Acute and Non-Acute Casemix Classification Study. Wollongong: Centre for Health Service Development, University of Wollongong, 1997. Eagar K. The Australian National Sub-Acute and Non-Acute Patient (AN-SNAP) Casemix Classification. Proceedings of the Ninth Australian Casemix Conference; 1997 Sep 7-10; Brisbane. Canberra: Commonwealth Department Health and Family Services, 1997. Center for Functional Assessment Research, Uniform Data Set for Medical Rehabilitation. 1993 Guide to the Uniform Data Set for Medical Rehabilitation (Adult FIM), V4.0. Buffalo: State University of New York, Buffalo, 1993. Wing JK, Beevor AS, Curtis RH, et al. Health of the Nation Outcome Scales (HoNOS). Research and development. Br J Psychiatry 1998; 172: 11-18. Authors' details South Eastern Sydney Area Health Service, Sydney, NSW. Lynette A Lee, FAFRM, FRACMA, Director Clinical Services. Centre for Health Service Development, University of Wollongong, Wollongong, NSW. Kathy M Eagar, MA(Psych), Associate Professor and Director. Neringah Palliative Care Service, Sydney, NSW. Michael C Smith, MB BS, MRACMA, Director. Reprints will not be available from the authors. Correspondence: Dr L A Lee, South Eastern Sydney Area Health Service, PO Box 430, Kogarah, NSW 1485. E-mail: leelATsesahs.nsw.gov.au

Synopsis The Outpatient Costing and Classification Study was commissioned by the Department of Health and Family Services to evaluate the suitability of the Developmental Ambulatory Classification System (DACS). Data on the full range of ambulatory services (outpatient clinics, emergency departments and allied health services) were collected prospectively from a stratified sample of 28 public hospitals. Patient encounters captured in the study represent 1% of the total ambulatory encounters in Australia in one year. Costing per encounter included time spent with the patient, cost of procedures, indirect costs (salaries and consumables), overhead costs and diagnostic costs. The most significant variable explaining cost variation was hospital type, followed by outpatient clinic type. Visit type and presence or absence of a procedure -- major splits for the proposed DACS -- did not produce splits that were consistent across all hospital strata. The study found that DACS is not an appropriate classification for hospital ambulatory services. A clinic-based structure for outpatients and allied health departments is recommended for classifying and funding ambulatory services in Australia. Introduction The Casemix Development Program which commenced in Australia in July 1988 focused on developing and implementing a national inpatient classification system for acute patients (AN-DRGs). Relatively little work was done on classifying and costing ambulatory services. In the early 1990s, however, two projects were conducted -- the National Ambulatory Casemix Project in Sydney1 and the Flinders Medical Centre Ambulatory Encounters Project in Adelaide,2 the latter in conjunction with the Royal Children's Hospital in Melbourne. These were primarily "demonstration" projects, which tested some overseas classifications and identified issues for future ambulatory classification projects. In 1994, the National Ambulatory Care Reform Program focused attention on ambulatory services by funding studies to facilitate health policy development in this area.3 However, none of these projects addressed the need for a nationally consistent ambulatory classification system. Recognising this, the Department of Health and Family Services requested that the Australian Casemix Clinical Committee establish a subcommittee to oversee the development of an ambulatory classification system for use in Australia. This committee reviewed existing classifications for their applicability in Australia,4 concluded that none were appropriate and recommended that a new classification system be developed to complement other patient-based classification systems. This work resulted in the Developmental Ambulatory Classification System (DACS), which was patient-based and structured around Ambulatory Major Diagnostic Categories (AMDC), similar to the Major Diagnostic Categories of the AN-DRG classification. The major splits in the proposed classification were based on whether the patient was making a new or a repeat visit, and whether a significant procedure was performed (Box 1). A specially constructed expert panel identified which outpatient and emergency procedures were significant cost drivers. In contrast to the other major classification systems, this classification was not based on empirical data. DACS needed to be evaluated for its suitability as a national classification. To address this issue an Outpatient Costing and Classification Study was commissioned by the Commonwealth Department of Health and Family Services in 1997. Methods The Outpatient Costing and Classification Study was conducted in two phases. Phase 1: Selecting and defining the data elements to be captured during the study and developing a sampling framework (conducted by Deloitte Touche Tohmatsu).5 Phase 2: Data capture and analysis of the results (conducted by Coopers & Lybrand and the South Australian Health Commission).6 The study aimed to include the full range of ambulatory services provided in public hospitals. For the purpose of the study, the term "ambulatory service" encompassed designated outpatient clinics (irrespective of location), emergency departments and allied health services for non-admitted patients. Because hospitals' recording of patient activity varies, the study also included same-day patients and inpatients treated within the outpatient and emergency departments. Hospitals Data were collected prospectively from a stratified sample of Australian public hospitals. South Australian hospitals were over-represented, because a similar State-based research project was initiated in South Australia before the Commonwealth project. Twenty-eight hospitals participated in the study. They included eight teaching hospitals, two specialist hospitals, two metropolitan hospitals, seven large rural hospitals and nine small rural hospitals (Box 2). Data collection Senior staff from each hospital met with the consultants before study commencement to ensure optimal data collection, and all hospitals employed a project officer to facilitate on-site coordination. To ensure data accuracy, a quality management plan was developed, including tolerance reports and edit checks on the data. Data collection commenced in September 1997 and continued in SA hospitals for three months, and at other sites for one month. Because of the difficulty in collecting detailed patient data in busy emergency departments, the collection period in emergency departments was four weeks in South Australia and two weeks in other States. Detailed utilisation data were obtained for each patient in the study (Box 3). A patient encounter was defined as "an interchange between one or more healthcare providers and one or more patients, for assessment, consultation and/or treatment for intended unbroken period of time". Telehealth consultations (including videoconferencing, telemedicine and telephone contacts) were included if the clinician who had previously seen the patient was present, and when the service was considered to be a substitute for face-to-face contact. Radiology and pathology services and dispensed pharmaceuticals were not considered encounters in their own right, but were subsequently linked to the "primary" encounter (ie, the encounter in which the services were ordered). Reviewing results, dictating letters and making telephone calls, which are generally consistent across all encounters, were included as indirect costs (although some clinicians elected to record the time associated with these activities as direct patient contact time). Telephone calls were recorded if the clinician who had previously seen the patient was present and when the service was considered to be a substitute for face-to-face contact. Indirect encounters related to consultations with key providers and relatives of patients in which the patient was the focus of the encounter. Group encounters were defined as encounters with more than one patient and/or more than one practising clinician present. All hospital-paid staff who were involved in providing patient care were requested to record the amount of time they spent in direct patient contact. This has been previously reported as the most variable aspect of an outpatient encounter.1 Specific proformas were developed to collect details on nursing time, medical time, allied health time, diagnostic services (pathology and imaging) and therapeutic services (pharmaceuticals). Coding Accurate diagnosis and procedure coding are not routinely collected for ambulatory patients in Australia. ICD-9-CM classification to the three-digit level was adopted as the minimum standard for coding during the project. This did not reduce the specificity of the clinical data, with some 4364 different codes being used across the study. Coding to the fourth and fifth digit was permissible if desired by clinicians. Standardising the clinic profile A set of generic outpatient clinics had to be established to standardise the profile of outpatient clinics within Australian hospitals. The use of outpatient clinics as a classification variable had been supported by several ambulatory studies, including the Victorian Ambulatory Classification System,7 the Queensland Health Ambulatory Project,8 and the Flinders Medical Centre Ambulatory Encounters Project.2 From these sources a list of 76 generic clinics was identified. Data collection sites were requested to map their clinics to this list. Some hospitals had difficulty in mapping their very specialised clinics. In these situations additional clinic names were added. With these refinements a final generic clinic list comprising 78 clinics was obtained. Costing data Patient level cost data were used to determine the cost of each encounter in four steps: Direct costs: The cost of direct time spent with a patient and the cost of significant procedures for individual encounters were allocated to the specific encounter. Indirect costs: Salary costs and costs for consumables (derived by deducting direct cost from total expenditure reported in line items in ambulatory cost centres) were dispersed across all ambulatory encounters. Overhead costs: Overhead costs, determined by an approach similar to that employed in COSMOS,9 were dispersed across all ambulatory encounters. This process allows all costs incurred in providing services -- power, cleaning and infrastructure costs as well as direct costs -- to be allocated to an individual encounter. Diagnostic costs: Patient specific utilisation data relating to radiology, pathology and pharmacy were downloaded from hospitals' information systems. Standardised unit prices were adopted for radiology and pathology services. This was set at 85% of the Medical Benefits Schedule (MBS) fee. Pharmacy costs were directly allocated and included Section 100 drugs. These costs were directly allocated to the primary ambulatory encounter. Most hospitals in the study were able to provide detailed costing information, with the exceptions being some of the small hospitals in South Australia. To estimate outpatient cost in these hospitals, a proxy outpatient fraction was derived from information obtained during the National Hospital Cost Data Collection Study. Statistical analyses Statistical analysis of the data measured the significance of the associations between the independent variables and the dependent variable, which in this case was cost. Results The study collected clinical and demographic data on 248 608 patient encounters (Box 4). Additional data were incorporated into the database: two previous emergency department studies (the Flinders Medical Centre Emergency Department Study10, and the Women's and Children's Emergency Department Study11); the Mental Health Classification and Service Costing Project (MH-CASC) relating to ambulatory encounters in the Mental Health Division of the Women's and Children's Hospital;12 and data from Launceston and Burnie Hospitals in Tasmania. The patient encounters captured in the study represent about 1% of the total hospital ambulatory encounters in Australia each year.10 Over 82% of these encounters were referred from three sources: other services within the hospital (33%); community general practitioners (28%) and self-referral (22%). The high proportion of self-referred patients was due to the inclusion of emergency department data. Thirty-four per cent of all encounters were new visits. The average cost of a new visit was $128, and of a repeat visit, $110. There were 10% more female than male patient encounters in the study population, and the number of public patient encounters greatly exceeded other types (86.6% of patients were public, 7.6% private, and 5.7% Department of Veterans' Affairs). Of the patient encounters analysed, 95% were direct encounters, 3.5% were telephone encounters and 1.5% were indirect contacts. The average cost for these encounters was $116 (direct), $129 (indirect), $115 (telephone) and $152 (telemedicine), respectively. There were only 46 telemedicine encounters captured during the study period. This represented 0.02% of total encounters. The costs of providing services to patients in hospital outpatient departments and in the ambulatory service components of hospital allied health departments are given in Box 5. This clinic structure was standardised for all hospitals. Group encounters were partitioned on the basis of hospital type and clinic type in the same manner as one-to-one encounters to facilitate standardised approaches to data collection and reporting. After trimming data to remove outliers, 0.5% of clinic encounters were group encounters. The average per patient cost of a "group encounter" was $82. This was about $20 less than one-to-one encounters. A list of group encounters and costs is given in Box 6. Emergency department analysis incorporating data from two previous studies (as mentioned above), and other studies conducted in Australia, have identified the key resource drivers in an emergency department as being triage, disposition and age.1 Classification analysis The objective was to design, from first principles, an outpatient classification system which could be used to fund ambulatory activity, and in doing so report on the appropriateness of the DACS as a framework for a patient-based classification system. A total of 198 495 episodes were analysed in detail, after removal of incomplete episodes. For selected components, data were trimmed to exclude cost outliers (defined as < 4 or > 5 SD from the mean); 1008 records were excluded on this basis. Analysis of emergency department data was conducted separately. The most significant variables identified were hospital type (teaching, specialist, metropolitan, large rural and small rural), outpatient clinic type, visit type (new or repeat), age and significant procedure. The impact of hospital type was highly significant and became the principal variable producing splits. An analysis of secondary variables producing splits is given in Box 7. Clinic-based classification Clinic type explained 24.05% of the cost variation in untrimmed data, and 31.60% of the cost variation in trimmed data. The variation explained was less significant for teaching hospitals (18.04% for untrimmed data and 23.93% for trimmed data). The variation in teaching hospital costs may have been a consequence of the higher number of junior staff who may have ordered additional diagnostic tests and the variable profile of clinicians attending the same patient. A detailed review was conducted of the variables associated at the next level of the classification tree, testing, in particular, age, visit type and the presence or absence of a significant procedure. This analysis did not produce splits which were consistent across all hospital strata. These factors were not considered to be significant splitting variables. To complete the classification analysis, it was necessary to examine group encounters and telephone contacts. Difficulty in defining telephone calls for funding purposes has resulted in telephone calls being excluded in many casemix-funding models. As the cost differential between face-to-face contacts and telephone contacts is so small, a case could be made for recommending funding these services in the same manner as face-to-face contacts. However, concerns were raised about the gaming potential for this class of encounters. Emergency department system When analysing emergency department episodes on the basis of urgency (as assessed by the National Triage Scale) and disposition, a significant explanation of variance was obtained. This remained at 34.39% for both trimmed and untrimmed data. Box 8 details the proportion and cost of encounters, by triage, disposition and age. The performance of this classification structure in small rural hospitals was extremely poor and produced a 0.93% reduction in variance. The flat average cost across the range of classes within this hospital stratum suggests that these services should be funded at a standard rate. DACS structure The assignment to DACS classes was based on the principal diagnosis coded, using ICD-9-CM codes. Problems occurred in the assignment of patients to DACS classes because there was no unique mapping of ICD-9 CM codes to AMDCs. For example "fracture of facial bones" could be assigned to AMDC 2, 3 or 8 (Eye; Ear, Nose, Mouth and Throat; and Musculoskeletal System and Connective Tissue, respectively). This was not addressed during the design phase of the project, and to resolve this an additional step was incorporated into the grouping process. This step used "clinic type" as a defining variable. This allowed 80% of all encounters to be assigned to a specific DACS class. It is not possible to determine whether the 20% of episodes excluded from the analysis had a significant impact on the result. The DACS explained only 15.32% of cost variation when stratified by hospital type. The performance of this classification system was marginally improved when a secondary split based on professional discipline (allied health, emergency, outpatient) was included (20.12%). Discussion It is imperative to establish a standard classification system for ambulatory patients, as has been done for acute patients. Healthcare funders and providers need to able to describe the ambulatory patient profile. Previous studies attempting to explain the resource variation for ambulatory patients have found that classifications based on the provider, rather than the patient, explain greater variation in patient costs. This is to be expected, as ambulatory care takes place in a relatively constrained environment. Clinicians designate "time slots" for their patients based on criteria relevant to their specialty areas. Patients may also be seen for the same condition by medical specialists and by allied health professionals -- the characteristics of the patient are unchanged, but the treatment regimens and resource use by the provider can vary greatly. Nevertheless, despite the difficulties entailed in development, a patient-based classification is considered the ideal long term classification structure for ambulatory encounters, as it would truly reflect the clinical condition of patients and thus enhance the clinical utility of such a system. The DACS, developed with input from experienced clinicians, was designed with this intent but, before this type of classification can be introduced, hospital outpatient information systems will have to be greatly enhanced. A complex patient-based classification requires the collection of patient activity and clinical data, which would exceed the capacity of existing manual or electronic systems. The study clearly indicates that the proposed DACS, in its current form, is not appropriate for classifying hospital based ambulatory services, and that in future classification development work the AMDC structure should not be considered an appropriate primary classification variable. More importantly the study identifies the generic clinic classification structure, partitioned by hospital type, as the most appropriate classification system for one-to-one encounters in outpatient clinics and allied health departments. Group encounters should also be classified by generic clinic type. Separate cost weights would apply to one-to-one and group encounters. The classification of emergency department presentations has been the subject of extensive research. This project confirms previous reports that triage category and patient disposition should be used to classify one-to-one encounters in emergency departments.9 In the short term, the generic clinic based structure for outpatients and allied health departments and the urgency and disposition based structure for emergency departments are recommended for classifying and funding ambulatory services in Australia. Acknowledgements This study was funded by the Commonwealth Department of Health and Family Services and sponsored by the Australian Casemix Clinical Committee, receiving constant support from all members and its then Chair, Professor John Hickie. We would like to acknowledge the cooperation of staff at the study hospitals. The burden placed on hospital staff in collecting detailed information on individual outpatient encounters cannot be underestimated and the commitment to "see the project through" was a major undertaking. We also acknowledge the assistance of State and Territory health departments, Malcolm Bond from Flinders University, Dr Chris Baggoley from Flinders Medical Centre and Coopers & Lybrand Consultants. This article is based on the Outpatient Costing and Classification Study undertaken by Coopers & Lybrand on behalf of the South Australian Health Commission and the Commonwealth Department of Health and Family Services, April, 1998. References Hindle D, Ligaida R. A casemix classification for hospital-based ambulatory services: a report from the National Ambulatory Casemix Project, New South Wales Department of Health. Sydney: New South Wales Department of Health, 1992. Michael R, Piper K, Heard P. Ambulatory Encounters Project, Flinders Medical Centre, Adelaide. Report for the Commonwealth Department of Health and Family Services, 1991 (available from the Department). Medicare agreement 1993-1998. Canberra: Commonwealth Department of Health and Family Services, 1993. Commonwealth Department of Health and Family Services, Classification and Payments Branch. Ambulatory casemix in Australia: Description of relevant classification systems. Canberra: Commonwealth Department of Health and Family Services, October 1995. Deloitte Touche Tohmatsu. DACS Pilot Study Progress Report. Report for the Commonwealth Department of Health and Family Services, Canberra. Sydney: Deloitte Touche Tohmatsu, 1997. Coopers & Lybrand Consultants. Outpatient Costing and Classification Study incorporating the Developmental Ambulatory Classification System Evaluation. Report for the Commonwealth Department of Health and Family Services, Canberra. Adelaide: Coopers & Lybrand Consultants, 1998. Jackson T, Sevil P, Tate R, Collard K. Development of relative resource weights for non-admitted patients. Melbourne: National Centre for Health Program Evaluation, 1989. Coopers & Lybrand Consultants. Queensland Health Ambulatory Project. Brisbane: Coopers & Lybrand Consultants, 1996. COSMOS [computer program], Version 2.0. Sydney: NSW Health. Erwich-Nijhout MA, Bond MJ, Baggoley C. Costings in the Emergency Department, Flinders Medical Centre, Adelaide. Report for the Commonwealth Department of Health and Family Services, 1996 (available from the Department). Erwich-Nijhout MA, Bond MJ, Raftos J. Costings in the Paediatric Emergency Department, Women's and Children's Hospital, Adelaide. Report for the Commonwealth Department of Health and Family Services, 1996 (available from the Department). Mental Health Classification and Service Costs Project. Developing a casemix classification for mental health services. Final report: Volumes 1 and 2. Canberra: Commonwealth Department of Health and Family Services, Aug 1998. Authors' details Princess Alexandra Hospital, Brisbane, QLD. Michael I Cleary, FACEM, MHA, Executive Director of Medical Services. Department of Health and Family Services, Canberra, ACT. Jo M Murray, BSc(Med), Acting Assistant Secretary, Classification and Payments Branch. Deloitte Touche Tohmatsu Consulting Group, Sydney, NSW. Robin Michael, BSc(Hons), MPH, Partner. South Australian Department of Human Services, Adelaide, SA. Kym Piper, MNIA, Principal Consultant, Health Costing and Evaluation Unit. Reprints will not be available from the authors. Correspondence: Dr M I Cleary, Executive Director of Medical Services, Princess Alexandra Hospital, Woolloongabba, QLD 4102. E-mail: clearymAThealth.qld.gov.au

Synopsis The introduction of casemix funding systems has focused attention on the reliability and validity of coded health data. Defining and classifying medical and health related terms are the core activities of the National Centre for Classification in Health (NCCH), which has recently published the Australian modification of the International statistical classification of diseases and health related problems, 10th revision (ICD-10-AM). An important feature is a classification of procedures (MBS-E) based on the Commonwealth Medical Benefits Schedule. Clinicians have made major contributions to the new classification through a network of 21 Clinical Coding and Classification Groups, which advise the NCCH. Major advantages of ICD-10-AM for clinicians include the ability to update the classification within Australia with continued clinical consultation, the familiarity of the procedure codes based on MBS, and the possibility of having one classification for use in public and private healthcare facilities. Introduction Codes for diseases and procedures are the basic ingredients of the casemix recipe. However, the coding function was not invented for casemix. The need to classify and measure has been around for centuries (Box 1). Allied to this need to impose order by classifying is the need to define the elements of a disease or procedure so that the meaning is clear and classification can take place. O'Rourke highlighted the importance of the meaning of medical terms in the context of doctor-patient communication in cardiology.3 Using codes to describe concepts is a shorthand way of ensuring a common understanding of the definition of that concept. The introduction of casemix funding systems based on the classification of diseases and procedures has meant that disciplined attention has been paid to the reliability and validity of coded health data. The connection between the codes and the health dollar has turned the spotlight on coding previously used only to identify groups of similar patients for research, utilisation studies or quality assurance. National Centre for Classification in Health The twin functions of defining and classifying medical and health related terms are the core activities of the National Centre for Classification in Health (NCCH), located at the University of Sydney, and Queensland University of Technology (Box 2). The Quality Division at La Trobe University, Melbourne, examines issues relating to coding and data quality. The NCCH develops codes and coding standards for use in Australian health services, publishes in hard copy and electronically, and educates clinical coders and clinicians in the application of codes. It also publishes methods of measuring coding quality and assists the Australian Institute of Health and Welfare in its role as WHO Collaborating Centre for Classification of Diseases. The Brisbane site supports the Australian Bureau of Statistics (ABS) in relation to its classification of causes of death. The ABS has been recording cause of death using the classification system International classification of diseases (ICD) and its predecessors since 1907 (Box 1). Australian hospitals and health services have collected ICD data on diagnoses and procedures since 1968. Before 1968, the Standard Nomenclature of Diseases and Operations,4 and sometimes the ICD, were used in hospitals to capture disease and procedure information, mainly for research purposes. The use of codes for casemix classification led to the need for Australian national standards in the application of codes, and eventually to the formation of specific Australian codes and classifications. These Australian Coding Standards have been developed by the NCCH.5 ICD-10-AM The NCCH has recently published the International statistical classification of diseases and related health problems, 10th revision, Australian modification (ICD-10-AM),5 which includes Australian extensions of the WHO codes in ICD-10 and some specific Australian disease codes. An important feature is the addition of a classification of procedures based on the Commonwealth Medicare Benefits Schedule (MBS) of fees for health services. It was a deliberate decision of the Casemix Implementation Project Board in 1995 to create this Australian procedure classification based on the fee schedule so that the classification of procedures in the public and private sectors, as well as in ambulatory situations, would be more consistent. The Australian procedure classification, known as the Medicare Benefits Schedule, Extended (MBS-E), is more specific than MBS, and is organised logically according to body system and site and includes a detailed index. Codes have been added for procedures not currently eligible for benefits, such as cosmetic surgery, obstetrics and allied health procedures. ICD-10-AM was introduced in July 1998 in hospitals and other healthcare agencies in New South Wales, the Australian Capital Territory, Victoria and the Northern Territory. It will be introduced in the remaining States from July 1999. The ABS, because of its commitment to report mortality data to the WHO, will continue to use the WHO version of the ICD classification. It is planned to implement ICD-10 for mortality coding in the year 1999 or 2000; the decision will depend on the availability of the Automated Cause of Death coding software from the United States. Casemix classification and mapping Construction of casemix classifications requires data expressed in the source coding systems. Because the national casemix classification system, AN-DRG, has until now been based on the previous standard classification in Australia, ICD-9-CM, the changeover to ICD-10-AM will require mapping between the classifications so that a version of AN-DRG based on ICD-10-AM can be built. Until data are available in ICD-10-AM from Australian hospitals and health services, AN-DRG allocation must rely on mappings between ICD-9-CM and ICD-10-AM. Mappings for grouping purposes are known as "logical" mappings, while those for longitudinal epidemiological studies are called "historical" mappings. The slight differences between these mappings arise because of differences in specificity of the classifications, especially in situations where one code in the new classification maps to many codes in the previous classification, and the many codes are spread over many DRGs (Box 3). Interaction between clinicians and clinical coders The need for accurate decisions on principal diagnosis and code allocation has led to collaboration between clinical coders and clinicians to interpret the documentation in the clinical record and to come to an agreed decision on appropriate codes for episodes of care. Clinicans have made major contributions to the structure and content of the new Australian disease and procedure classifications through a network of 21 Clinical Coding and Classification Groups (CCCG), which advise both the NCCH and the Australian Casemix Clinical Committee (ACCC) on issues relating to coding and casemix refinement. Many additions to the WHO ICD-10 were made as a result of mapping between ICD-9-CM and ICD-10 to ensure that specificity and new Australian codes introduced to ICD-9-CM were replicated in ICD-10-AM. Examples of notable improvements in ICD-10-AM compared with ICD-9-CM are listed in Box 4. To promote clinician-coder communication, the NCCH has been funded by the ACCC to publish a series of specialty booklets on coding and casemix. A series of 21 booklets on different clinical topics is planned, of which a third is already available. Detailed information of interest to clinicians and epidemiologists about changes in ICD-10-AM is currently being prepared by the NCCH and will be available on the NCCH internet homepage.6 The NCCH also has an education function in keeping clinical coders abreast of annual updates to the coding system and the Australian Coding Standards.5 It will play a major role in educating clinical coders in ICD-10-AM. The NCCH's homepage6 is regularly updated and has links to relevant Australian and overseas organisations. Impact of ICD-10-AM Clinicians: Australian clinicians will benefit greatly from the new Australian classification. Firstly, it makes current the description and classification of diseases, and reinforces the Australian clinical contribution to updating the underlying WHO classification. The mechanism of clinical consultation used in constructing the classification will be continued in the updating process so that the classification remains clinically coherent and relevant. Secondly, by using the Medicare Benefits Schedule as the foundation of the procedure classification, the concepts and labels of the procedure codes will be familiar to clinicians, and the update of MBS-E will proceed in tandem with the update of MBS. Having one Australian procedure classification for use in public and private healthcare facilities, inpatient and ambulatory situations will be more efficient than the existing system (ICD-9-CM in the public sector and MBS in the private sector) and will do away with the need for mapping between MBS and ICD. Clinical coder workforce: Introduction of the new classification will have major implications for the clinical coder workforce. They will not only need to become familiar with ICD-10-AM coding, but will also need an understanding of anatomy and the surgical procedures required by the specificity of the MBS-E. ICD-10-AM coding is expected to take longer initially,7 although no allowances have been made in the deadlines for reporting hospital morbidity data in the States and Territories adopting the new classification in 1998. Health facility managers are becoming more aware of the need for resources for clinical coders to reflect the complexity of casemix through accurate and timely clinical coding. Updating the classification: A major benefit is the ability to regularly update the classification within Australia. We intend to maintain close connections with international disease classification systems so that statistics on causes of death and morbidity are comparable. However, having an Australian centre for health classification develops local expertise and fosters robustness of the Australian classification itself, data quality and efficiency of data collection, as well as a clearer understanding of the meaning of clinical terms and their place in classification hierarchies. Synchronising coding systems: Introduction of ICD-10-AM also provides an opportunity for synchronising coding systems with software designed to support electronic patient records. Without appropriate coding standards, data from these systems cannot be extracted, analysed and stored so that it is retrievable and capable of integration with other related modules and functions (eg, pharmacy and laboratory data). Modifications to health service software systems will be required to accommodate the new composition of the codes. The ICD-10-AM disease codes are alpha-numeric (3-5 characters; eg, Ross River disease [B33.1]) and the procedure codes are numeric (7 characters; eg, endoluminal repair of aneurysm [90228-00]). Analysis of longitudinal data: The change in classifications will affect analysis of longitudinal data by epidemiologists and public health practitioners. They will have the option of mapping forwards from the old to the new classification, or backwards from the new to the old classification. In either case, meaning will be lost when the codes of one classification are more precise or less precise than those of the other. However, there will be benefits in the introduction of appropriate new codes and terminology, particularly for infectious diseases, neoplasms, obstetrics and mental health. Conclusion Considerable effort has always been devoted to coding diseases and procedures in hospitals. Casemix funding systems with their reliance on accurate classification of diseases and procedures provide a major incentive to getting the data right. Systems are in place through the ICD-10-AM codes, the Australian Coding Standards, the clinical coder workforce, and through input from clinicians to ensure that coding is clinically appropriate, efficient, accurate and timely. References Lyons AS, Petrucelli RJ. Medicine: An illustrated history. New York: Harry N Abrams, 1987. History of the development of the ICD. In: World Health Organization. International statistical classification of diseases and related health problems. 10th revision. Vol 2, Ch 6. Geneva: WHO, 1993. O'Rourke MF. What's in a name? Med J Aust 1997; 166: 372-373. Thompson ET, Hayden AC. Standard nomenclature of diseases and operations. 5th ed. New York: McGraw-Hill, American Medical Association; 1961. National Centre for Classification in Health. The international statistical classification of diseases and related health problems, 10th revision, Australian modification (ICD-10-AM). Sydney: National Centre for Classification in Health, Faculty of Health Sciences, University of Sydney, 1998. National Centre for Classification in Health. World Wide Web homepage: http://www.cchs.usyd.edu.au/NCCH/ncch.html Department of Health and Family Services. ICD-10-AM impact assessment project. Final report. Adelaide: Coopers & Lybrand Consultants, November 1997: 30. Authors' details National Centre for Classification in Health, University of Sydney, Sydney, NSW; and Queensland University of Technology, Brisbane, QLD. Rosemary F Roberts, MPH, MBA, Director, National Centre for Classification in Health, Sydney. Kerry C Innes, AssocDip(MRA), Associate Director, National Centre for Classification in Health, Sydney. Susan M Walker, BAppSc(MRA), Associate Director, National Centre for Classification in Health, Brisbane. Reprints will not be available from the authors. Correspondence: Associate Professor R F Roberts, National Centre for Classification in Health, University of Sydney, PO Box 170, Lidcombe, NSW 1825. E-mail: R. RobertsATcchs.usyd.edu.au

Synopsis Paediatric patients (compared with adults) require additional time, effort and skill from hospital staff caring for them. Many suggestions for making successive versions of AN-DRGs more child friendly have not been implemented. Rather than relying on age, the AN-DRG classification should allow a better definition of complexity within DRGs. The two groups of children who place a disproportionate burden on paediatric teaching centres are children under 3 years and those with congenital abnormalities and chronic illness. Cost weights are not specific for paediatric patients. The extra costs of caring for children are reflected in nursing costs, highlighting the importance of including nurse dependency data in any costing study. Introduction In recent years attention has been drawn to the differing healthcare needs of children compared with adults, and the high cost of caring for children in hospital.1-6 Children's less-frequent use of inpatient services reflects in part their general well-being, but also a different approach to their care, with every effort being made to keep them out of hospital or to minimise their length of stay. However, children's shorter stay is counterbalanced by their greater dependence, and the intensity of the care they require increases the cost of their hospital stay. A briefing paper prepared by the National Association of Children's Hospitals and Related Institutions (NACHRI) in the United States clearly outlined the uniqueness of children's healthcare services.1 Their findings -- that children are more likely to require acute care than long term care, but when they do have a chronic illness the costs of care are high -- apply also to other developed countries, including Australia. Classification issues -- making DRGs child friendly Studies in the United States have shown inadequacies in many of the classifications describing paediatric care and also found costs to be higher for paediatric patients, in particular for nursing care.7 However, the APR-DRG (all patient refined DRG) classification, widely adopted in the United States, better reflects paediatric care and illness severity than previous casemix classifications. Similarly, studies in Australia have highlighted the inadequacy of the AN-DRG classification.8 Despite the many changes that have been made to AN-DRGs since they were introduced, they are still not seen as ideal for paediatrics. Age splits and comorbidities AN-DRG versions 1 and 2 had a number of adjacent DRGs with an age split at 10 years. These were eliminated from version 3 because they were not supported by length of stay and nurse dependency data. However, in 1996, the Australian Casemix Classification Committee recommended that, for those DRGs with age splits at 10 years, the complications and comorbidities split should take precedence over the age split, so that hospitals caring for a small number of children with complex illnesses would not be disadvantaged. However, because so few children were involved, this was not implemented and remains a major problem for these hospitals. General anaesthesia in children AN-DRGs do not recognise the need for a general anaesthetic for children having procedures (eg, a dental procedure, an endoscopy or a minor orthopaedic procedure) for which adults do not normally require anaesthesia. Many of these procedures are in a "medical" DRG rather than a "procedural" DRG, resulting in an inadequate cost weight for the care provided to the child. Some allowance has been made for this in AR-DRG-4 (Australian refined diagnosis-related groups), which was released in July this year. A general anaesthetic is recognised as a complication in some DRGs, but this is dependent upon there being a split for complications in the particular adjacent DRG. Children under three years High nursing dependency An Australian study in 1996, using paediatric nursing service weights, showed that children under 3 years require significantly more nursing care than older children.6 In specialist teaching centres, children under 3 years required 37% more nursing time per episode of care than patients aged 3-59 years, and despite their shorter length of stay their use of nursing resources was similar to that of elderly people (Box 1).6 Data from this study also suggest that children under 3 years place higher demands on other hospital services irrespective of their length of stay. For children under 3 years versus those over 3 years, 29 AN-DRGs were identified with a cost variation of greater than 50% and 15 with a cost variation of 25%-49%.6 On the basis of these findings and with the restriction that there were to be 10 more DRGs, recommendations were made for future revisions of AN-DRGs: Additional DRGs should be included to cover the high cost of care of younger patients; The DRG age split at 10 years should be adjusted to an age split at 3 years; and If possible, paediatric patients should be shifted from their current DRGs to the adjacent higher-order DRG, if an age split for older patients already existed. Most of these recommendations were rejected on statistical grounds. A limited number of DRGs with age splits were adjusted to 3 years, but others were removed and replaced with splits based on comorbidities and complications. In the long term, replacement of age splits by splits based on severity of illness is preferred. In AR-DRG-4 the problem of complications and comorbidities is better addressed, but further changes still need to be made (Box 2). Cost weights Currently, the cost weights applied to paediatric patients in many States and Territories in Australia are the same as those applied to all other categories of patients. National DRG cost weights have until now been derived by a cost modelling method which allocates costs to DRGs rather than to individual patients (see Phelan). Consequently, it is not possible to compare the costs of caring for children under 3 years derived from actual resource allocation. However, nursing costs are a good proxy for the increased care these patients require. The "average total nursing time per DRG" is the largest component (about 44%) of the total costs per DRG. It is also the most appropriate and available indicator for comparing costs of caring for children under 3 years with those in other age groups. Nursing costs enable a valid comparison of costs across all Australian States and Territories, irrespective of nursing career structures and award rates of pay.9 On average, nursing costs, regardless of the length of stay, are doubled in young children under 3 years -- they account for almost 40% of the throughput of paediatric hospitals. As nursing salaries represent about 35% of all hospital costs and more than 50% of variable costs, this issue needs to be addressed within the payment system.6,9 Several authors have emphasised the higher costs of teaching hospitals (see Butt and Shann; Hart and Wallace; Phillips). Within paediatrics, however, it is difficult to distinguish between the casemix of specialist and non-specialist teaching centres. Because the AN-DRG classification (both versions 3 and 4) is limited in its capacity to take into account complications and comorbidities, it cannot adequately reflect these differences. In attempting to deal with this inequity, paediatric hospitals in Victoria have lobbied individually to have modifications made to their own hospital's reimbursement to reflect the greater cost of providing care for children. South Australia has adopted a standardised approach to developing paediatric cost weights based on benchmark costs. This method entails replacing the cost components for nursing, medical and allied health in the national cost weights with benchmark paediatric costs for South Australia. Paediatric cost weights are derived from these data. Currently, a second national cost weight study is in progress. Data from patient costing systems are being used to update the paediatric cost weights to ensure they are improved for paediatrics. A study into the neonatal services provided by the two intensive care units in South Australia is also in progress. Congenital abnormalities and chronic illness Advances in technology have significantly improved clinical outcomes for a wide variety of paediatric patients. A relatively small group of children with chronic or congenital illness, including newborns requiring neonatal intensive care, accounts for a significant proportion of the cost of acute inpatient care. Many of these children require ongoing care and rehabilitation, which is both resource intensive and often delivered in an acute care setting. This adds considerably to the number of children requiring lengthy hospital admissions ("long stay outliers") and the overall cost of care. In Australia, there are limited facilities for providing ongoing care for these children outside acute-care institutions. Experience in Victoria has highlighted the complexity of children requiring lengthy hospital admissions. A study by Health Solutions5 pointed out that most of these children are erroneously judged to be nursing home type patients. They do not necessarily cost the same as children with shorter stays ("inliers") during the same phase of care, nor do they necessarily cost less than the average cost per day for shorter stay children when their stay continues past the "high trim point". Other specialist paediatric hospitals in Australia and the United States have reported similar problems.7,10 There is substantial underfunding of children who require lengthy admissions, particularly in the areas of neonatology, oncology, and chronic or congenital diseases. The high cost of paediatric care has resulted in health funds in the United States being reluctant to fund paediatric hospitals with patients likely to require long and complex care.10 This situation could easily arise in Australia. Careful case selection by payers or providers in a competitive market can be used to advantage in better risk management within a health plan. Potentially, this can lead to preference being given to children with less-complex conditions, and barriers to access for children requiring longer and more intensive care. With recognition that a relatively small proportion of high cost paediatric patients accounts for almost two-thirds of the expenditure on acute-care paediatric services, it is time for funders to specifically target these children for separate funding so that specialist paediatric hospitals can be more equitably funded. Future strategies Having recognised the different care requirements of children, the issue now is how to have them accepted by the wider health community. The strategies that need to be pursued are clear: The classification should recognise severity rather than rely on age as a proxy; The true costs of paediatric care must be reflected in the cost weights; and Strategies should be implemented to deal with the select group of high cost patients who pose a unique problem to specialist teaching centres. US experience has shown that dealing with these factors alone can decrease the financial losses of children's hospitals from 30% to about 10%.7 AR-DRG-4 allows for better definition of complexity within DRGs. However, the APR-DRGs advocated by NACHRI provide the extensive benefits of a more universal application of grades of severity. References National Association of Children's Hospitals and Related Institutions. Children's health care needs are different - why one size won't fit all. A NACHRI briefing paper. Alexandria, Va: NACHRI, 1993: 1-16. Vertrees JC, Pollatsek JS. Paying for paediatric inpatient care. Final report of the Universal Access for Children Reimbursement Study Project. Conducted for NACHRI. Alexandria, Va: Solon Consulting Group Ltd, 1993. Berry R. Final report of Children's Hospitals' Casemix Classification Study Project. Conducted for NACHRI. Alexandria, Va: NACHRI, 1986. Miller H. Final report of Paediatric Costing Study. Conducted for NACHRI. Alexandria, Va: Center for Health Policy Studies, 1993. Paediatric Costing Study. Kids casemix - more swings than roundabouts. Melbourne: Health Solutions Pty Ltd, 1994. National Paediatric Nursing Study Phase 2. Adelaide: Paediatric Nursing Study Consortium, 1996: 1-19. Muldoon J. Paediatrics and DRG casemix classification. In: Goldfield N, Boland P, editors. Physician profiling and risk adjustment. Chapter 24. Gaithersburg, Md: Aspen Publishers, 1996: 252-270. Phelan PD. Are casemix developments meeting the needs of paediatrics? Med J Aust 1994; 161 Suppl Sep 5: S26-S29. National Paediatric Nursing Study Phase 1. Adelaide: Paediatric Nursing Study Consortium, 1994: 1-13. Andrews JS, Anderson GF, Han C, Neff JM. Pediatric carve outs. The use of disease-specific conditions as risk adjusters in capitated payment systems. Arch Pediatr Adolesc Med 1997; 151: 236-242. Authors' details New Children's Hospital, Sydney, NSW. Ralph M Hanson, FRACP, FACEM, Chair, Division of Information Services. Women's and Children's Hospital, Adelaide, SA. Meradith A Phythian, RN, RM, Clinical Analyst, Clinical Support Unit. Jenni B Jarvis, RN, Head, Clinical Support Unit. Princess Margaret Hospital, Perth, WA. Cyndy Stewart, RN, BAppSc, Head, Best Practice Unit. Reprints will not be available from the authors. Correspondence: Dr R M Hanson, New Children's Hospital, PO Box 3515, Parramatta, NSW 2124. E-mail: RalphHATnch.edu.au

Synopsis AN-DRGs have some splits which take illness severity and complexity into account. Age is also often used as a proxy for severity of illness. The need to transfer a patient may be a marker of illness severity or complexity and therefore resource utilisation. This is supported by studies of patients transferred to intensive care units. Data on the costs and outcomes of all transferred patients should be collected; depending on the results, refinements of DRGs may be indicated. Introduction Greater accuracy of DRG classification would result in more appropriate healthcare funding. A limited number of complications and comorbidity splits in AN-DRGs take into account illness severity and complexity. In addition, age, in both young and old patients, is often used as a proxy for illness severity. DRG accuracy could be further improved, if other easily applied measures of illness severity could be identified. Patients transferred from one hospital to another because they require specialised treatment may represent a different patient population to those not transferred. A recent study has shown the differential resource utilisation of different populations, with hospital care for Aboriginal and Torres Strait Islander patients estimated to cost 30% more than that for non-Aboriginal and Torres Strait Islander patients with a similar DRG classification1 (see Fisher et al). It is possible that hospital care for transferred patients may also be more costly, because being transferred may be a marker for illness severity or complexity and therefore resource utilisation. Summaries of studies of comparative costs of transferred and non-transferred patients are shown in the Box. The studies indicate that transferred patients generally are sicker, use more resources, have a longer length of hospital stay and an increased risk of death. Costs may be higher, particularly if death occurs after a long illness. Most of the currently available data relate to patients transferred to intensive care units. Definitions of transfer status To apply a patient's transfer status as a measure of severity, definitions need to be standardised. We propose the following categories: Referral: Transfer of a patient to a different hospital for a particular doctor's opinion. Up transfer: Transfer of a patient to another hospital for inpatient specialist treatment not available at the primary hospital. Down transfer: Either return transfer of an inpatient to the primary hospital, or transfer of a patient to another hospital for recovery. Sideways transfer: Transfer of a patient to another hospital, because the required facilities at the referring hospital are fully occupied. Further Australian data will need to be collected prospectively on the costs and outcomes of patients who are transferred from one hospital to another for specific, complex treatment. In the light of these results, further analysis of transfer status will be needed, before it can be applied as a measure of illness severity and resource utilisation. Depending on these results, refinements of DRGs may then be indicated. References Commonwealth Department of Health and Family Services. Report on National Aboriginal and Torres Strait Islander Casemix Study. Adelaide: Brewerton and Associates Pty Ltd, April 1997. Munoz E, Soldano R, Gross H, et al. Diagnosis related groups and the transfer of general surgical patients between hospitals. Arch Surg 1998; 123: 68-72. Jencks SF, Bobula JD. Does receiving referral and transfer patients make hospitals expensive? Med Care 1988; 26: 948-958. Pon S, Notterman DA, Kathryn M. Pediatric critical care and hospital costs under reimbursement by diagnosis-related group: effect of clinical and demographic characteristics. J Pediatr 1993; 123: 355-364. Borlase BC, Baxter JK, Kenny PR, et al. Elective intrahospital admissions versus acute interhospital transfers to a surgical intensive care unit: cost and outcome prediction. J Trauma 1991; 31: 915-918. Authors' details Intensive Care Unit, Royal Children's Hospital, Melbourne, VIC. Warwick W Butt, MD, FRACP, Staff Specialist in Intensive Care. Frank A Shann, MD, FRACP, Director of Intensive Care; and Professor of Critical Care Medicine, University of Melbourne, Melbourne. Reprints will not be available from the authors. Correspondence: Dr W Butt, Intensive Care Unit, Royal Children's Hospital, Flemington Parade, Parkville, VIC 3052. E-mail: buttwATcryptic.rch.unimelb.edu.au

Synopsis An Australia-wide patient classification system for nursing is urgently needed as the health system attempts to develop benchmarks against which to measure and compare services. Standardised measures of demand for nursing care must be developed to allow appropriate reimbursement and to act as proxies for illness severity. Nurses need to identify the outcomes that measure the nursing contribution to episodes of care, and assist in developing outcome goals reflecting the efficacy of treatment and the quality of care. A system of measuring nursing requirements and costs of early discharge and coordinated care programs is required. It must be consistent with nursing classifications and hospital costing systems. Introduction Casemix continues to present challenges to the nursing profession. Initially, the focus was on refining the Australian AN-DRG classification to more accurately reflect clinical practice and, although anomalies still exist, this has largely been accomplished. Currently, four issues pose a particular challenge to nursing and relate to integrating the AN-DRG classification with: An Australian patient classification system for nursing; Measures of illness severity within DRGs; Measures of the nursing contribution to quality and outcomes; and, Measures of nursing requirements and costs of home-based and coordinated care programs. A patient classification system for nursing Because patients within a DRG are not necessarily alike in terms of cost and nursing dependency, there is a need for a patient classification system for nursing.1 This has become more urgent as the healthcare system attempts to develop benchmarks against which to measure and compare services. At present, each State and Territory has a very different system of patient classification for nursing acuity, which means that nursing content, reliability, validity and clinical meaning within AN-DRGs cannot be compared across Australia. There have been problems reaching a consensus on nursing classification issues. If nursing does not understand (and agree on) its own cost structure, or its contribution to the costs of patient care, it cannot hope to negotiate prices for the nursing component of an episode of care.2 A comparison of three nursing classification systems used in New York State found that estimates of nursing costs and nursing intensity (resource requirement and complexity of nursing care) varied because each system measured resource use differently.3 In South Australia, information is now available on nursing hours per patient-day and costs per patient-day. In 1995, a computer-based clinical decision-making program was successfully integrated with a computer-based nurse scheduling program. This system has been used in the major metropolitan hospitals and in a small number of larger country hospitals. Data collected can be aggregated to AN-DRGs and nursing costs compared across South Australian hospitals. Information from this system was used in the initial nurse costing study to develop the first nursing service weights incorporated into AN-DRGs. This system lends itself to continual refinement and innovation. Consensus about Australian nursing costs and what constitutes nursing in Australia cannot be reached, and benchmarks cannot be developed, until the nursing profession develops a patient classification system for nursing which applies to the whole of Australia. Illness severity measures The second issue challenging nursing -- measurement of illness severity in terms of nursing dependency -- is directly related to patient classification. Resource use within DRGs varies, as many AN-DRGs are far from homogeneous. Studies have shown that resource use within DRGs can be influenced by a range of factors including illness severity, disease complexity and comorbidities, as well as the socioeconomic profile of the patient.4-7 The South Australian Department of Human Services deals with the issue of case complexity and severity by giving its metropolitan teaching hospitals and regional country hospitals a severity loading. This severity index is calculated on the basis of the number of diagnosis and procedure codes per patient record by AN-DRG, and takes into account length of stay factors and variations in patient acuity and complexity.8 Variables of illness severity have been considered in AN-DRG-3. These include age, specific complications and comorbidities, level of effect of complications and comorbidities, complicating effects of interactions between groups of complications and comorbidities, direct clinical and physiological observations, need for life-support or therapeutic interventions, and admission status.9 AN-DRG-3 addresses some illness severity issues by the inclusion of splits based on three complicating clinical factors (CCFs): complication and comorbidity levels, age, and the presence of malignancy. Studies of illness severity and nursing have concentrated on developing standardised measures of demand for nursing care to allow appropriate reimbursement and to more accurately reflect nursing acuity. A study of the relationship between nursing care hours to DRGs and illness severity found that the demand for nursing resources was associated with illness severity, and that classification of patients by DRG and illness severity produced more homogeneous groups in terms of nursing resources.10 As nursing constitutes such a large proportion of the total expenditure for an episode of care, several tools have been developed to measure the impact of illness severity on nursing care. These include the Patient Intensity for Nursing Index (PINI), Nursing Intensity Weights (NIWs) and Case Mix Index (CMI), which are used across a range of medical and surgical DRGs in the United States, and the Apache (acute physiology and chronic health evaluation) system used in critical care in Australia.2,3,11-13 Because these severity of illness measures do not account for some of the variance in length of hospital stay and costs not explained by DRGs, alternative adjustments have been considered, including functional status indexes (particularly those measuring degree of independence in activities of daily living).14 An accurate method of measuring illness severity within AN-DRGs is yet to be developed. Thus, further work is required to identify whether nursing acuity or therapeutic nursing interventions can be used as surrogate indicators of illness severity. This may be the challenge for casemix and nursing in the future. Quality and outcomes The third challenge for nurses -- quality and the establishment of professionally agreed-upon clinical outcomes -- also presents a challenge to other health professionals. Casemix was originally intended to provide a method of more accurately defining and measuring outcomes within homogeneous groups. DRGs have enhanced the ability to map an episode of care for a group of patients, resulting in a proliferation of clinical pathways or care maps. Nurses have been key players in the development and implementation of these tools.15 Clinical pathways or care maps have been effective in streamlining care and in resource utilisation, and they have improved hospital processes and reduced lengths of stay.16 However, most clinical pathways or care maps fall short in defining and measuring outcomes of episodes of care within an AN-DRG. The same applies to guidelines and clinical protocols. The challenge for nurses is to identify the outcomes that measure the contribution that nursing makes to the episode of care. They need to work with the multidisciplinary team to develop measurable and realistic outcome goals that reflect the efficacy of treatment and the quality of the care given within DRGs. Nurses need to be able to demonstrate the impact of their care on the cost and length of stay of each DRG in order to maintain adequate funding for the nursing component of a DRG. Impact on clinical practice The final issue relates to changes in clinical practice and adapting AN-DRGs, nursing resources and nursing classifications to these changes. Early discharge programs and coordinated care programs for chronically ill people encourage management in the community rather than hospital admission. Casemix funding systems in some States have acted as a disincentive to these hospital-in-the-home initiatives, when DRG funding has been allocated only for hospital stay. A more creative way of determining what constitutes a "hospital bed" needs to be considered, to ensure that these programs are adequately funded. Outcome measures must be developed to determine whether these community-based programs are effective and whether they are accurately costed. In many cases, the nurses who manage hospital-in-the-home patients take on a range of roles, including education of carers, counselling and social work activities. In view of these additional roles, a system of measuring the nursing requirements and costs of home care needs to be considered. This should be compatible with hospital nursing classification and costing systems to enable accurate tracking of costs, outcomes and acuity across a full episode of care within a DRG. References Diers D. Whoa!!. Aust Nurses J 1991; June 20(10): 8-9. Ballard KA, Gray RF, Knauf RA, Uppal P. Measuring variations in nursing care per DRG. Nurs Manage 1993; 24(4): 33-41. Phillips CY, Castorr A, Prescott PA, Soeken K. Nursing intensity: going beyond patient classification. J Nurs Adm 1992; 22: 46-52. The Severity Measurement Protocol. Report to the Clinical Advisory Committee, South Australian Health Commission, 1994. Sharkey PD, Horn SD, Brigham PA. Classifying patients with burns for hospital reimbursement: diagnosis-related groups and modifications for severity. J Burn Care Rehabil 1991; 12: 319-329. Horn SD, Sharkey PD, Buckle JM, Backhofen JE, et al. The relationship between severity of illness and hospital length of stay and mortality. Med Care 1991; 29: 210-220. Rapoport J, Teres D, Lemeshow S, Avrunin JS, et al. Explaining variability of cost using a severity-of-illness measure for ICU patients. Med Care 1990; 28: 338-348. South Australian Health Commission. Severity and cost differentiation. Discussion Paper. Adelaide: SA Health Commission, 1998. Marshall R, Zhang X, Lonergan J. Measures of disease severity in the AN-DRG classification: the case for CC severity level indicators. Proceedings of the Seventh Casemix Conference in Australia; 1995; Jul 31-Aug 2; Adelaide. Canberra: Commonwealth Department of Human Services and Health, 1995. Bostrom J, Mitchell M. Relationship of direct nursing care hours to DRG and severity of illness. Nurs Econom 1991; 9: 105-110. Prescott P. Nursing intensity: needed today for more than staffing. Nurs Econom 1991; 9: 409-414. Adams T. Case Mix Index: nursing's new management tool. Nurs Manage 1996; 27(9): 31. McKinley S. Casemix update: Australian critical care costs and service weights. Part 2. Aust Crit Care 1996; 9(2): 56-59. Kelleher C. Validated indexes: key to nursing acuity standardization. Nurs Econom 1992; 10: 36. Zander K. Physicians, care maps and collaboration. The new definition. Boston, Mass: Centre for Case Management, 1992; 7(1): 2-4. Ferguson L. Casemix issues for nursing. Med J Aust 1994; 161 Suppl Sep 5: S37-S39. Authors' details Royal Adelaide Hospital, Adelaide, SA. Lesley E Long, RGN, BAppSci(Nsg), PhD, MHA Nursing Director, Cancer Centre. Rosemary Mann, RGN, BA, GradDipOrthoNsg, Project Nurse, Nursing Administration. Reprints will not be available from the authors. Correspondence: Dr L E Long, Nursing Administration, Royal Adelaide Hospital, North Terrace, Adelaide, SA 5000. E-mail: llongATcancer.rah.sa.gov.au

Synopsis Casemix has given allied health professionals the opportunity to review their approaches to patient care, contribute to reducing inpatient costs and improve quality of care. The National Allied Health Casemix Committee was formed in 1993 to advance allied health participation in casemix. The Committee has taken the first step in establishing cost weights for allied health through the Australian Allied Health Activity Classification, which defines allied health inputs in terms of clinical care, clinical service management, teaching and training, and research. Work is being done on generic classification of allied health inputs, and studies are examining what allied health activities are accounted for by DRGs and ICD-9-CM. Allied health has taken up the challenge of casemix, but better access to information technology will enhance its continued contribution. Introduction The advent of casemix in Australia has provided allied health practitioners and managers with an opportunity to review their approaches to patient care, contribute to organisational goals of reducing inpatient costs, maximise reimbursement within funding rules, and improve the quality of patient care. Significant achievements have been made at local, State and national levels, despite deficiencies in both the systems and the technology supporting casemix implementation. National Allied Health Casemix Committee (NAHCC) The NAHCC was formed in 1993 to advance allied health participation in casemix. There are 14 professional member organisations of the NAHCC. Allied health casemix groups in all States and Territories are also represented. NAHCC has successfully completed major projects by focusing on areas of commonality, rather than difference. This has also occurred at the State level; for example, in South Australia cooperation between allied health, the South Australian Health Commission and information services staff led to agreement on the requirements for an allied health information management system. The Reference Standards Project1 undertaken by the NAHCC is the first step in developing Australian cost weights for allied health. The lack of appropriate infrastructure, including an agreement on what constitutes inputs and outputs of allied health services, has so far precluded their development. The allied health weights applied in AN-DRGs are a version of the Maryland weights (Maryland [USA] Health Services Cost Review Commission, 1993) crudely modified for Australian use and not reflecting Australian allied health practice. Australian Allied Health Activity Classification2 The Australian Allied Health Activity Classification was an important outcome of the Reference Standards Project. The classification broadly defines inputs in terms of clinical care, clinical services management, teaching and training, and research. More specifically, clinical care is defined as all activities which can be attributed to an individual patient, group or community, thus eliminating the inappropriate notions of "direct" and "indirect" care. This approach was taken up by other classification studies, including the Sub-Acute and Non-Acute Patient Casemix Classification Study. The Reference Standards Project identified that occasions of service alone are not a satisfactory method of measuring outputs of allied health, and that further work in this area needs to be done. Allied health procedure codes In consultation with individual professional bodies and the NAHCC, the National Coding Centre (now the National Centre for Classification in Health [NCCH]) has identified discipline-specific interventions. The previous description of interventions within the ICD-9-CM procedures listing was extremely limited. Several disciplines submitted intervention codes and these were included in the July 1996 revision of ICD-9-CM. These codes have been refined and others added to the procedure listing in the recently developed Australian modification of ICD-10 (ICD-10-AM) (see Roberts et al). Future revision of the codes by the professions and the NCCH will ensure their adequacy in describing allied health inputs and their consistency in cross- discipline application. Coding challenges Projects by various allied health disciplines examining the extent to which their activities are accounted for by DRGs and the ICD-9-CM classification have shown considerable lack of agreement between classifications. A study of the intensity of social-work time in an acute hospital3 found it to be more related to a patient's presenting psychosocial problems, but that these were influenced by diagnosis and complexity. A study for the Dietitians Association of Australia found that most, but not all, of the diagnostic terms used by dietitians in describing patient care matched with terms in ICD-9-CM.4 Items which could not be matched related to risk of malnutrition, risk of side effects of treatment or disease, or lack of diagnostic specificity of a particular disease. This study also established that dietitians, like many allied health practitioners, may see patients for reasons other than the principal reason for admission. Despite limitations of existing casemix classifications to take allied health inputs into account, clear examples exist of allied health's contribution to casemix-funded organisations. In strategically managing occupational therapy services, an approach integrating DRGs with cost-benefit analysis was successful in decreasing average length of stay and improving quality of care.5 Community-based services developed by a rehabilitation team resulted in earlier discharge, improved continuity of care and a high level of patient satisfaction.6 Unfortunately, this service was not continued past the initial funding period, because hospital reimbursement would have been reduced if the alternative outpatient service continued. By contrast, dietitians7 have shown that coding malnutrition as a comorbidity can alter DRG assignment and increase casemix reimbursement. More importantly, diagnosing malnutrition provides an opportunity to give appropriate and timely care, thus reducing the associated costs and length of stay.8 Benchmarking Emphasis is often placed on the use of casemix as a budgeting tool, which overlooks its original use for measuring quality of clinical care.9 Casemix allows clinicians to compare inputs and outputs and measure outcomes in terms of quality, value and resource utilisation. A step towards this was the Best Practice in the Health Sector Program10 to promote international best practice standards of care and workplace organisation throughout the health sector. The program case studies provide an overview of the factors influencing the growth of allied health in Australia, in the context of principles of best practice. In 1997, the Central Sydney Area Health Service established the National Allied Health Benchmarking Consortium. Its task was to identify methods of best practice by comparing allied health resource utilisation. The Consortium currently comprises seven teaching hospitals (three in New South Wales, and one each in Victoria, the Australian Capital Territory, Tasmania and South Australia). The objectives are to establish benchmarks of allied health resources at a national level, and to develop a framework to link benchmarks with inputs, processes and outcomes of allied health services and activities. In Phase I, baseline data were collected and are currently being analysed. Phase II, which will be conducted in close association with the NAHCC, will investigate the highest volume AN-DRGs with allied health inputs and selected outcomes. Future directions Allied health has taken up the challenge of casemix, but better access to information technology will enhance its continued contribution. Many allied health departments still rely on manual data collection to guide decision making. Access to hardware must be extended, and allied health information management systems interfacing with hospital decision support systems must be developed. Moreover, consistent application of classifications, such as the Australian Allied Health Activity Classification, will improve understanding of allied health inputs at the local, State and Territory, and national level. This will provide a platform for costing of allied health, and for developing allied health service weights. In addition, improved information systems will also support the allied health service structures that best manage human resources, meet the needs of smaller referring clinical units, maintain an appropriate skill mix of practitioners, and support training of undergraduates. Interest in casemix among allied health practitioners has been steadily growing. The development of new classifications, such as the Australian National Sub-Acute Non-Acute Patient (AN-SNAP) Casemix Classification (see Lee et al) and the Mental Health Classification and Service Costs (MH-CASC), have further involved allied health professionals working in rehabilitation, mental health and other settings outside acute care. Moreover, community practitioners are becoming casemix-aware with the development of the Community Health Information Management Enterprise (CHIME), which is responsible for the National Codeset Project for community-based health services. References The National Allied Health Casemix Committee. Report to the Commonwealth Department of Health and Family Services on the development of National Reference Standards for allied health disciplines -- clinical terms and minimum data set. Melbourne: The National Allied Health Casemix Committee, 1997. Australian Allied Health Classification System: Version 1. Melbourne: The National Allied Health Casemix Committee, 1997. Badger J, Cleak H, Haywood M. Factors affecting the intensity of social work time in an acute hospital. Allied health and casemix: towards 2000. Melbourne: The National Allied Health Casemix Committee, 1996; 23-27. Barrington V. The development and classification of dietetic diagnoses and interventions. Allied health and casemix: towards 2000. Melbourne: The National Allied Health Casemix Committee, 1996: 20-22. Brandis S. An AN-DRG approach to planning occupational therapy services. Allied health and casemix: towards 2000. Melbourne: The National Allied Health Casemix Committee, 1996: 11-14. Brandis S. The frustration of rehabilitation -- why quality doesn't pay. Proceedings of the Ninth Casemix Conference in Australia; 1997 Sep 7-10; Brisbane. Canberra: Commonwealth Department of Health and Family Services, 1997. Ferguson M, Capra S, Bauer J, Banks M. Coding for malnutrition enhances reimbursement under casemix-based funding. Aust J Nutr Diet 1997; 54: 102-108. Funk KL, Ayton CM. Improving malnutrition documentation enhances reimbursement. J Am Diet Assoc 1995; 95: 468-475. Fetter RB. The history and development of diagnosis-related groups. Proceedings of the Eighth Casemix Conference in Australia; 1996 Sep 16-18; Sydney. Canberra: Commonwealth Department of Health and Family Services, 1996. Commonwealth Department of Health and Family Services. Australian health organizations taking up the best practice challenge: the Best Practice in the Health Sector Program: case studies of the funded projects. Canberra: AGPS, 1996. Authors' details National Allied Health Casemix Committee, Melbourne, VIC. Annette L Byron, BSc, BND, MBA, Chairperson, NAHCC; Chief Clinical Dietitian, Nutrition and food Services, Royal Adelaide Hospital, Adelaide, SA. Helen C F McCathie, PhD, Executive member, NAHCC; Area Director of Psychology, Central Sydney Area Health Service, Concord Repatriation General Hospital, Sydney, NSW 2139. Reprints: Ms A L Byron, Nutrition and food Services, Royal Adelaide Hospital, North Terrace, Adelaide, SA 5000. E-mail: abyronATnadmin.rah.sa.gov.au

Synopsis All private hospitals and clinics must now supply de-identified data, using AN-DRG classification, on all admitted patients to the Private Hospitals Data Bureau. Contracts between health funds and hospitals must also be described on the basis of AN-DRGs, which will enable funds to undertake hospital variance analysis. These data provide the foundation for nationally developed clinical pathways and utilisation reviews which could modify clinical practice, improve standards and reduce health costs. Clinicians must understand and participate in these changes, and adequate safeguards are needed to protect them against loss of their clinical integrity, and against inappropriate discretionary control by private hospitals, healthcare corporations and health insurers. Introduction From February 1998, the Health Legislation (Private Health Insurance Reform) Amendment Act 1995 (Cwlth) requires all private hospitals to supply to the Private Hospitals Data Bureau de-identified data on all admitted patients. The data must include hospital charges and use the AN-DRG classification as part of the Hospital Casemix Protocol.1 Because of the impact this could have on clinicians and the delivery of healthcare in the private sector, it is essential that doctors understand and participate in decisions about use of AN-DRGs by private hospitals and health insurers, particularly in regard to healthcare finance and assessment of variance from predetermined healthcare protocols. Healthcare finance Public hospital costing differs from that in the private sector, mainly because of the medical salaries component, making direct comparison impossible. Current Australian private hospital financing includes a profit margin, but remuneration for medical services comes from the government, health insurance funds and patients. Casemix funding in the private sector would affect payment of doctors if the Commonwealth Medicare Benefits Schedule was altered to describe rebated services as casemix-based episodes of care (including pre- and postadmission care as described in expanded DRGs).2 Payment of doctors would also be affected if the government legislated for health funds to receive Medicare rebates for members' medical services and to incorporate these in AN-DRG-based payments to hospitals. The hospitals would accordingly be allowed to pay clinicians, perhaps under the terms of Hospital-Practitioner Agreements or other employment contracts. These are contracts required by legislation if a private hospital is to receive payments from a health benefit organisation and then pay the practitioner for services rendered to a member of that organisation. Increasing public health sector privatisation is attracting the interest of large healthcare corporations.3 They are looking to build, own and operate institutions, or to provide healthcare contracted out from established public hospitals. Use of the AN-DRG classification would facilitate comparison and allow comparable payments to be made by government for services between the two sectors. Public hospital AN-DRG funding is supplemented by other government budgetary allocations, but private sector institutions are dependent on health funds, which pay only for hospital services rendered to fund members -- they receive no other subsidies. Even the most efficient of these private institutions would be vulnerable if public AN-DRG costings were applied, or if any new costing methods were inaccurate or inadequate. These factors are beyond their control and are potential faults of casemix funding. This vulnerability would be compounded if a health insurer arbitrarily reduced either the prospectively determined price or the accepted length of stay of a given DRG in an effort to reduce fund expenditure. Clinical pathways could provide protection against such funding decisions and justify the clinical necessity of the length of stay or the hospital charge, and would thus force funds to provide supporting data. Assessment of variance from predetermined healthcare standards Of increasing interest to health insurers are methods of measuring and ensuring adherence to quality standards. If standards are inadequate, length of stay and readmissions may increase. This would mean rises in fund outlays and premiums, which in turn would have adverse effects on fund viability and membership. To establish comparable standards and clinical protocol guidelines, uniform case classification is required, enabling comparison between clinicians, hospitals and regions. This is now facilitated nationally by reporting of all admissions according to AN-DRGs. The two recognised ways of measuring standards and quality are the use of clinical pathways and utilisation reviews. Clinical pathways Clinical pathways are standardised protocols for given episodes of care, rather than compulsory recipes for clinical management. They improve efficiency of patient management by avoiding delays caused by lack of coordination and communication, and reduce average length of stay and costs without compromising quality.4 With innovations in care, clinical pathways provide a basis for proper comparison with current treatments. Clinical pathways are becoming widespread, not only because of quality issues but because of the profession's increasing awareness of the need for cost effective allocation of finite health resources. Clinicians have acknowledged concerns about clinical pathways: They are concerned that inappropriate protocols may be developed by clinicians with inadequate insight into practice or by health managers who favour cost control over quality; They question standardisation of care because of the inherent variability of patients, their diseases and their responses to treatment; and As responsible clinicians they must often make individual judgements, and they are concerned that lack of adherence to clinical pathways may expose them to litigation if their management results in an adverse outcome. The Health Legislation Act also requires contracts between health funds and hospitals to be described on the basis of AN-DRGs. This will enable fund managers to undertake hospital variance analysis. They can then ask hospital managers to investigate cost variability of particular AN-DRGs. Clinical pathways can provide reasons for variance and empower management to redress the system. Analysis of individual clinician variance could possibly result in suggestions for changes in care, economic punitive action (such as withdrawal of accreditation), or demands that a clinical pathway be adhered to as a condition of future accreditation or agreement between clinician and hospital. With accurate knowledge of hospital costs for AN-DRGs and reasons for interhospital variance, purchasers (health insurers) and providers (hospitals) can undertake properly based negotiations and establish appropriate payments. Hospital use and utilisation review of clinical pathways will be increasingly relevant in the negotiations for Hospital Purchaser-Provider Agreements, which a health benefit organisation may choose to enter into with a hospital for the provision of services to its members. The financial viability of private hospitals is dependent on these agreements. Utilisation review Utilisation review is defined as "A set of information activities which support, monitor and evaluate decisions concerning the allocation of healthcare resources to previous and current patients and potential recipients. The aim is to ensure that resultant allocations are cost-effective and equitable."5 It can be prospective (checking, then approving, modifying or rejecting a proposed management), concurrent (assessing and modifying current care) or retrospective (assessing previous decisions, considering their appropriateness, then acting by providing educational resources, denying payment, threatening or undertaking withdrawal of business). Utilisation review can be used to develop clinical pathways, assess resource allocation, or assess compliance with a stated clinical pathway. It allows comparison of hospitals and practitioners, particularly with regard to resource use (particularly bed-days) in the management of individual AN-DRGs. In managed care in the United States, utilisation review addressed overservicing and cost-cutting. Compulsory management protocols had to be fulfilled to enable prospective fund approval; non-compliant management led to retrospective denial of payments.5 The Australian health insurance industry is concerned about an increase in private hospital utilisation in spite of a decrease in insurance numbers. The industry is also concerned that the growth in day surgery has not decreased overall costs or services in the private sector (Mr R Schneider, Australian Health Insurance Association, paper presented to a Health Summit organised by Australian Investment Conferences, Sydney, 24 March 1997). Health insurers could establish compulsory protocols they regard as being within the scope of accepted clinical practice by AN-DRGs and as a condition of Medical Purchaser-Provider Agreements. These are contracts a health benefit organisation may enter into with medical practitioners to address the provision of services and the legally rebatable fees that can be paid by the organisation to the practitioners. Current and potential funding pathways for applying AN-DRG casemix in the private sector are shown in the Box. Implications for clinicians The two most relevant aspects of casemix for clinicians in the private sector are funding and assessment of adherence to quality criteria. The latter has significant ramifications for resource allocation and control of clinicians' practice and independence. It is unlikely that the medical component of private health costs will be incorporated into AN-DRG funding, but the hospital component could be based on finite bed-day allocation for each DRG (instead of the current open per diem payment) or as a case-based payment. Clinicians should be involved in these decisions. The collection and reporting of admission data based on AN-DRGs by private hospitals (as required by the Health Legislation Act) provide the foundation for nationally developed clinical pathways and utilisation reviews to modify clinical practice, improve standards and reduce health costs. The use of AN-DRG data could result in loss of clinicians' professional independence and make them vulnerable to punitive economic measures by hospitals and funds. These institutions could become de facto arbiters of which doctors are suitable to practise in the private sector. Until now the professional freedom of medical practitioners has been decided by ethics, the law, and standards of clinical care, as judged by the community and the profession itself. Economic management could become the new criterion upon which doctors will be judged by health insurers and private hospitals, regardless of practice standards. This must be fully appreciated and carefully considered by all sectors of the Australian community. The practical application of casemix could have a positive impact on the viability of the private health sector and a beneficial effect on clinical management of patients, but its degree of success will depend on the trust and involvement of clinicians. Health insurance funds are currently in fiscal crisis and one method of addressing this is to reduce fund outlays. The private sector relies on a viable health insurance industry and clinicians need to be aware of the cost implications of their management decisions. However, adequate safeguards are needed to protect clinicians against loss of their clinical integrity and inappropriate discretionary control by hospital owners, healthcare corporations and health insurers. References Commonwealth Department of Health and Family Services. Private Hospitals Bureau -- data flow requirements. 3 December 1997. (Circular HBF No. 513 and PH No. 288.) Health Solutions International. Expanded DRGs. An episode approach. Prepared for the Classification and Payments Branch, Department of Health and Family Services. 1997. (Available from Health Solutions International, East Melbourne, VIC.) Hurst J. Year of health carve up. The Australian Financial Review 1998; 2 Jan: 36. Private Sector Casemix Unit. Clinical pathways: a background for private hospitals and private insurers. 1997. (Available from the Private Sector Casemix Unit, Canberra.) Hindle D. Utilisation review. A discussion paper. Private Sector Casemix Unit. 1997. (Available from Private Sector Casemix Unit, Canberra.) Authors' details Australian Casemix Clinical Committee, Melbourne, VIC. Chris N Maxwell, FRCOG, FRACOG, Senior Obstetrician and Gynaecologist, and Director of Clinical Services Obstetrics, Gynaecology and Paediatrics, The Northern Hospital, Melbourne, VIC. Reprints will not be available from the author. Correspondence: Dr C N Maxwell, Northpark Medical Centre, PO Box 1080, Bundoora, VIC 3083. E-mail: cmaxwellATtnh.vic.gov.au

Synopsis Casemix funding has markedly increased surgeons' awareness of the economies of the activities they undertake. Surgery has become a major focus at all large public hospitals, because of its high earning potential, and this pressure to maximise funding could influence surgical practice. Casemix funding's emphasis on length of hospital stay has encouraged forward planning for earlier discharge after surgical procedures. Patients are now assessed in pre-admission clinics, educated about their condition and their hospital stay, and a plan formulated for their discharge and rehabilitation. Funding for major surgical procedures of long duration in patients with complex conditions should reflect the higher level of resource utilisation. Tertiary referral centres, because of their commitment to training and research and their more severely ill patient population, are less cost-effective and require funding to ensure their viability. The improved information that casemix generates should be used to evaluate outcomes and improve patient care; efficiency must not take precedence over quality of care and compassion. Introduction Casemix has been effective in reducing government spending on health and in improving public hospital efficiency.1 Paying hospitals for current rather than previous practice has proven to be beneficial:2 in 1990, acute hospitals in Australia were costing 31.2 cents of every health dollar;3 that figure has now been reduced to 28 cents in the dollar.4 Most importantly, productivity has been increased by 20% in some hospitals.5 Casemix funding has markedly increased surgeons' awareness of the economies of the activities they undertake, and given greater understanding of where money is being spent and where it is being wasted. It has provided a tool for comparing many widely divergent areas of medical practice within the same institution and between different institutions. We are amassing a vast quantity of valuable information, which will be used to monitor outcomes and improve performance. Casemix funding and surgery Under casemix funding, surgical activity has become a major focus at all large public hospitals because of its high earning potential. Regular casemix meetings are held in many surgical units, with the specific aim of maximising reward for work done and hence maximising funding. There can be drawbacks in such a situation. One criticism has been that hospitals now perform procedures rather than care for the sick.6 Furthermore, the recognition that revenue is likely to be higher if a procedure is performed, could potentially influence surgical practice. For example, if a patient were admitted from the emergency department with suspected appendicitis, it is clearly to the hospital's financial advantage for surgery to be performed. For a patient with suspected appendicitis, a condition with significant morbidity and mortality, such a decision is not bad practice. However, casemix funding is not designed to fund specific DRGs, and neither the surgeons performing this work, nor their units, reap the financial rewards directly. The money is used to subsidise less profitable clinical areas.1 Casemix funding has also stimulated surgical activity in units with forward budget planning where funds are allocated according to a predicted level of specialised surgical activity (eg, complex biliary surgery). However, if a unit's activity exceeds forecast levels and the budget is capped, some operations which cannot be deferred may not be appropriately funded. Capping of budgets destroys incentives, closes beds, increases waiting lists and discourages clinicians and others involved in "coal face" healthcare.5,7 Length of stay An interesting benefit of casemix funding, with its emphasis on length of stay in hospital, has been its encouragement of forward planning. Previously, when patients were admitted to hospital for surgery, little thought was given to length of postoperative stay, and an appropriate discharge plan for the patient was not considered until the time of discharge. Now patients are assessed in pre-admission clinics and any special medical and anaesthetic problems are identified. They are educated about their condition and their hospital stay, perhaps given an exercise program, and a plan is formulated for their discharge and rehabilitation. All these measures have the potential to reduce complication rates, and therefore length of stay. Furthermore, the concept of same-day admissions has been considerably advanced by the advent of pre-admission clinics. To further reduce length of hospital stay, casemix must be extended into areas beyond the acute hospital episode,8 such as "hospital-in-the-home" and ambulatory care. With casemix funding, hospitals are rewarded for patients with clearly defined conditions whose hospital stay is shorter than the average for that disorder and, conversely, penalised for patients whose stay exceeds this average. It has been interesting and illuminating to discover just how early patients may be discharged from hospital after major surgical procedures, but there is the potential for them to be sent home too early. Patients need time to adjust to the physical effects of their procedure and its consequences, and to be educated in management of their condition. At the Sir Charles Gairdner Hospital in Western Australia, the practice of discharging patients with femoral neck fractures to nursing homes three days after surgery had to be discontinued because of an unacceptably high mortality.9 Lack of community resources to support early discharge has been a major problem.1 Complexity of care Casemix funding, while it rewards uncomplicated care, does provide some increased funding for patients with complications after surgery. However, AN-DRGs do not adequately take into account variations in illness severity and comorbidities. Patients who have complications during their postoperative recovery obviously consume more resources, but it is paradoxical that more funding is available for patients who do badly than for patients who do well. Tertiary referral centres often treat the most difficult, taxing and hence resource-intensive patients. These patients are referred to these centres because of the complexity of their problem, or because of postoperative complications after one or more procedures elsewhere. Tertiary referral centres are also involved in research, and undergraduate and postgraduate teaching. Because of this commitment to training and research, and their more severely ill patient population, they are less cost-effective and require funding to ensure their viability. In some areas of surgery, as a consequence of casemix funding, patients who will do well and have few complications are being selected to provide a large turnover of trouble-free patients favourable to fund generation. Careful thought needs to be given to funding formulas for simple, short surgical procedures with very low complication rates, as opposed to major procedures of long duration in patients with complex conditions. One technique may be the introduction into DRG classifications of disease-specific conditions as risk adjusters for disorders with known comorbidities and high complication rates.10 Efficiency versus humanity Casemix and budgetary restrictions have created an impersonal atmosphere, in which efficiency has taken precedence over humanity. There has also been a shift from collegiality to contract arrangements.2 Moreover, there is a belief among general practitioners that some patients (eg, the elderly) are not welcome in the public hospital system.11 The lack of time and money to deal adequately with all the facets of care in public hospitals has led to a sharp rise in complaints from consumers.5 At the Alfred Hospital in Melbourne, attempts are being made to monitor patient complaints, and a complaint officer has been appointed. Other measures to counteract this impersonal atmosphere include direct involvement of general practitioners in hospital activities (eg, in outpatient clinics), and there are future plans for general practitioners to be involved with surgical patients before and after operation. Conclusion Although casemix has made us more aware of the need for efficiency, budgetary constraints, including the capping of activity, are likely to adversely affect important aspects of healthcare, such as quality of care and compassion. We must make use of the information that casemix is generating to fully evaluate outcomes and improve patient care, as well as work towards extending the benefits of casemix to total patient care. We need centres of excellence to set and maintain high standards of patient care. References Phelan PD. Casemix funding in Australia. Time to move on. Med J Aust 1998; 168: 560-561. Braithwaite J, Hindle D. Casemix funding in Australia. Time for a rethink? Med J Aust 1998; 168: 558-560. Commonwealth Department of Health and Family Service/South Australian Health Commission. An evaluation of casemix funding in South Australia, 1994-95. Casemix Development Program. Adelaide: Commonwealth Department of Health and Family Services and the South Australian Health Commission, 1997. Australian Institute of Health and Welfare. Health services expenditure by type of expenditure 1989-90 to 1994-95. Health Expenditure Bull 1997; 13: 3-5. Kennedy JT. Perspectives in casemix based funding in Victoria. Good for governments. Med J Aust 1995; 162: 665-666. Tonti-Fillipini N. Negatives of casemix. Australian College of Midwives Inc. Ninth Biennial Conference Proceedings. Melbourne: Australian College of Midwives, 1995: 456-466. Phillips PA. Perspectives in casemix-based funding in Victoria. Med J Aust 1995; 162: 655. Hanson R. Casemix funding in Australia. Have we come full circle? Med J Aust 1998; 168: 561-562. Sikorski JM, Senior JM. Factors affecting mortality in patients suffering a fracture of the proximal femur. J Bone Joint Surg Br 1998; 79 Suppl IV: 410. Andrews JS, Anderson GF, Han C, Neff JM. Pediatric carve outs. The use of disease-specific conditions as risk adjusters in capitated payment systems. Arch Pediatr Adolesc Med 1997; 151: 236-242. Segal GR. Perspectives in casemix-based funding in Victoria. Some patients are not welcome. Med J Aust 1995; 162: 656. Authors' details Monash University, Alfred Hospital, Melbourne, VIC. John A L Hart, MB BS, FRACS, Clinical Associate Professor of Surgery, and Senior Orthopaedic Surgeon. Royal Children's Hospital and Royal Melbourne Hospital, Melbourne, VIC. David Wallace, MB BS, FRACS, Neurosurgeon. Reprints will not be available from the authors. Correspondence: Professor J A L Hart, Clinical Associate Professor of Surgery, Monash University, Alfred Hospital, Prahran, VIC 3181. E-mail: johnhartATmelb.alexia.net.au

If you have any clinical casemix queries, contact the Australian Casemix Clinical Committee (ACCC), your State or Territory casemix clinical committee or your own professional body or College casemix committee. Australia Australian Clinical Casemix Committee Professor Paddy Phillips Chair, Professor and Head of Medicine Flinders University of South Australia, and Divisions of Medicine, Flinders Medical Centre and Repatriation General Hospital, Adelaide, SA Ph: (08) 8204 4039 Fax: (08) 8204 5268 Ms Naarilla Hirsch Executive Officer PO Box 852, Woden, ACT 2606 Ph: (02) 6289 8499 Fax: (02) 6289 7630 Australian Capital Territory Casemix Steering Committee Mr David Butt Chair Mr Sebastian Rosenberg Secretariat ACT Department of Health and Family Services PO Box 825, Canberra City, ACT 2601 Ph: (02) 6205 0842 Fax: (02) 6205 1373 New South Wales NSW Casemix Clinical Committee Dr Ralph Hanson Chair, Division of Information Services New Children's Hospital Cnr Hawkesbury Road and Hainsworth Street, Westmead, NSW 2145 Ph: (02) 9845 3482 Fax: (02) 9845 3632 Casemix Policy Advisory Committee Mr Jim Pearse Chair, Structural and Funding Policy Branch Level 9, NSW Health Department Locked Bag 961, North Sydney, NSW 2059 Ph: (02) 9391 9613 Fax: (02) 9391 9615 Northern Territory Territory Health Services Casemix Clinical Resource Management Project Steering Committee Ms Carol Beaver Chair, Territory Health Services 4th Floor Health House, Mitchell Street, Darwin, NT 0801 Ph: (08) 8999 2400 Fax: (08) 8922 8995 Queensland Casemix Steering Committee Dr Glen Cuffe Chair PO Box 48, Brisbane, QLD 4001 Ph: (07) 3225 3261 Fax: (07) 3234 0987 South Australia Clinical Advisory Committee Dr Chris Pearson Chair Ms Sheryn Reid CAC Executive Officer 11-13 Hindmarsh Square, Adelaide, SA 5000 PO Box 65, Rundle Mall, SA 5000 Ph: (08) 8226 6289 Ph: (08) 8226 6116(Executive Officer) Fax: (08) 8226 0793 Tasmania Tasmanian Casemix Clinical Committee Dr Maarten Kamp Chair, Acting Director of Medical Services Launceston General Hospital Charles Street, Launceston, TAS 7250 Ph: (03) 6332 7008 Fax: (03) 6332 7825 Casemix Steering Committee Dr Jon Mulligan Chair, Director of Hospital and Ambulance Service GPO Box 125B, Hobart, TAS 7001 Ph: (03) 6233 2106 Victoria Victorian Casemix Advisory Committee Associate Professor John Wilson Chair Ms Penny Sharwood Executive Officer Acute Health Division, GPO Box 4057, Melbourne, VIC 3001 Ph: (03) 9616 7221 Fax: (03) 9616 7764 Clinical Casemix Subcommittee Dr John de Campo Acting Chair Women and Children's Health Care Network, 132 Grattan Street, Carlton, VIC 3053 Western Australia At present there is no formal Casemix committee. Casemix issues are handled by: Ms Elizabeth Rohwedder Operations Division, Health Department of Western Australia PO Box 8172, Perth Business Centre, Perth, WA 6849 Ph: (08) 9222 4195 Fax: (08) 9222 4067