The Duty to Care study was a comprehensive investigation that used population-based data linkage to examine the burden of physical diseases in users of mental health services (MHS) in Western Australia.1-5 The study showed that users of MHS, representing 8% of the general population (mostly with moderate to severe mental illness), were at excessive risk of developing all major physical diseases and had a high case-fatality rate from those illnesses.1-3 This raises the question as to whether lack of physical health care contributes to the poor physical health outcomes associated with mental illness.

In earlier reports, we postulated that part of the problem in Australia may be that individual and structural barriers might preclude the users of MHS gaining adequate access to primary care provided by general practitioners.1-3 Primary care is thought to be one of the foundations for improving the health outcomes of vulnerable populations6 and, for people with mental illness, could be potentially important in facilitating the interface between MHS and specialist medical services for physical conditions.7 This is particularly the case in Australia, where financial barriers to accessing GP services are largely removed by Medicare.8 The transformation of MHS from an institution-based to a community-based care model in Australia has implied an increasing involvement of GPs in the physical health care of people with mental illness.9

However, the extent to which people with mental illness are engaged by the GP sector in Australia is unknown. Previous studies from the United States suggest that people with serious mental illness, particularly schizophrenia and alcohol or drug disorders, receive a lower intensity of primary care.10-11 Homeless people with mental illness are unlikely to receive any medical care at all.12,13

Against this background, our objective was to determine if users of MHS have an appropriate level of access to GP services relative to non-users. Our working hypothesis was that users of MHS, especially those with serious mental illness, visit GPs less often than non-users, which might partly explain their poor physical health outcomes.

We carried out a population-based retrospective cohort study for the period 1 January 1990 to 30 June 2006, using the crossjurisdictional data linkage facility of the Western Australian Data Linkage System (WADLS), which has enabled the integration of datasets held by the Australian federal and Western Australian state governments to support medical and health research.

Study participants comprised two groups (users and non-users of MHS) selected from the Western Australian Mental Health Information System (MHIS) and the electoral roll, respectively. The MHIS collects inpatient data from psychiatric institutions and public and private general hospitals, and outpatient data from public mental health outpatient clinics, community MHS and psychiatric residential units. Ambulatory services provided by private psychiatrists and GPs treating mental disorders were not included.

Users of MHS for this study included all people recorded in the MHIS from 1 July 1966 to 31 December 2004 who were still alive on or after 1 January 1990. These people were categorised as Group A or Group B, based on whether or not they had ever been on the Western Australian electoral roll (Box 1). The electoral roll data represented about 86% of the Western Australian general population aged ≥ 18 years. Non-users of MHS were a random sample of people who were on the electoral roll, but never recorded in the MHIS; these people were categorised as Group C. Group C was matched 2:1 with Group A by 5-year age group and sex.

For users of MHS (Groups A and B), the start of follow-up (T0) was 1 January 1990 for people recorded in the MHIS before 1 January 1990, or the first date of registration with the MHIS for those who were first recorded from 1 January 1990 onwards. For non-users (Group C), T0 was the same as that of their matched person in Group A. People aged less than 18 years, those diagnosed with dementia, overseas visitors, those with a follow-up period of less than 1 day and people recorded in the MHIS with only incomplete mental health outpatient records were excluded from the study (Box 2).

There were 498 803 eligible participants selected: 204 727 users of MHS and 294 076 non-users (Box 1). Characteristics of the study cohorts at T0 are shown in Box 4. Relative to non-users, users of MHS were more socially disadvantaged, more likely to be Indigenous and more likely to live in remote WA. Among users of MHS, Group B was younger, had a higher ratio of males to females (Box 4), and higher proportions of alcohol/drug disorders, schizophrenia and personality disorders compared with Group A (Box 5).

The 498 803 participants visited GPs a total of 32 575 719 times during a mean follow-up period of 10.9 years. The crude rate of visits to GPs was higher in Group A (8.22 visits per person-year; Box 6), and lower in Group B (4.31) compared with Group C (5.00). There were no visits to GPs observed in 2.1% of people in Group A, 44.4% in Group B (mainly because 98.3% of people in Group B who never had an address recorded in the WADLS had no visits to GPs recorded at all) and 3.9% in Group C (Box 6).

The large number of visits to GPs produced effect measures with very narrow confidence intervals (and all P values < 0.0001), so rate ratios were reported to three decimal places. Univariate regression analysis showed that users of MHS visited GPs significantly more often than non-users; and Group A participants visited GPs significantly more often, but Group B participants significantly less often, than Group C participants (Box 7). Multivariate regression analysis (Box 7) showed that, compared with non-users of MHS, the overall adjusted rate ratio (ARR) for visits to GPs among users of MHS was 1.622 (95% CI, 1.613–1.631). The ARR was highest in Group A (ARR, 1.663; 95% CI, 1.653–1.672), and higher in Group B (ARR, 1.416; 95% CI, 1.401–1.431), compared with Group C (ARR = 1).

Among users of MHS, ARRs were most elevated in people with alcohol/drug disorders, schizophrenia and affective psychoses (ARRs, 2.404, 1.834 and 1.798, respectively); moderately raised in people with other psychoses, neurotic disorders, personality disorder (ARRs, 1.658, 1.552 and 1.700, respectively); and least elevated, but still higher than for non-users in those with adjustment disorder, simple depression and other mental disorders (ARRs, 1.399, 1.472 and 1.321, respectively). Patients from all categories of mental disorders visited GPs more often than non-users of MHS in metropolitan, rural and remote parts of WA (all ARRs > 1; Box 8).

A notable exception to the general pattern of results was observed in the 4% of users of MHS with no fixed address recorded on any record in the WADLS (mostly those in Group B). This subgroup seldom visited a GP at all (ARR, 0.058; 95% CI, 0.057–0.060).

Contrary to our hypothesis, we found that users of MHS from all categories of mental disorders visited GPs more often than non-users, with the exception of the 4% of users with no fixed address who seldom visited GPs at all. Those with more severe mental illness visited GPs more often than those with less severe mental illness. The results were similar in metropolitan, rural or remote areas.

Our findings differ from those of two large US studies that showed poor access to ambulatory general medical services among people with severe mental illness.10,11 One reason may be the differences between the Australian and US health care systems. The Australian health care system provides universal access to GP services financed by Medicare, while the US system does not. Another reason is that because MBS record data for visits to GPs have information on the length but not the purpose of consultations, we could not distinguish whether a GP visit was for physical or mental health reasons. A proportion of visits to GPs in this study may have been solely for mental health reasons, whereas physical and other preventive and health promotion issues may not have been dealt with during the same consultation. Our finding that the 4% of users of MHS who never had an address recorded in the WADLS (presumably homeless people) were unlikely to visit GPs at all is consistent with other studies.12,13

To our knowledge, this is the first geographically defined, population-based, longitudinal study on the effect of mental disorders on visits to GPs, with a large sample size, long follow-up time, and covering the whole spectrum of mental disorders treated at both hospital or community settings. The high precision of the study is reflected by the narrow 95% confidence intervals of the rate ratios. Three strategies controlled potential confounding: domain restriction to the Western Australian electoral roll, matching between users and non-users of MHS, and multivariate analysis.

One limitation of our study was that the MBS data for visits to GPs did not have information on the purpose of the visit. Another limitation was the lack of data for private psychiatrists and GPs treating mental disorders. This would limit the extrapolation of our findings to all people with mental illness in Australia. Nevertheless, our study included about 40% of people with mental illness (generally with moderate to severe illness) whose physical health disparities were probably greater than the remainder of people with mental illness. Moreover, we used the same definition for our mental health cohort as did the Duty to Care study,2 which ensures the continuity and integrity of our investigations and findings.

The previously reported Duty to Care study highlighted the poorer physical health outcomes in users of MHS.1-3 With the opportunity provided by the recent development of the crossjurisdictional data linkage facility of the WADLS, we were able to determine whether the poorer physical health outcomes were partly the result of poorer engagement with the GP sector. The results show that, for the vast majority of users of MHS, this is not the case, at least in terms of the frequency of visits to GPs.

Although it is encouraging that most users of MHS seem able to gain access and use GP services, the poorer physical health outcomes in this group may mean that they have not benefited from their higher level of use of GP services. Promoting even higher levels of GP attendance may be futile. Therefore, policies focusing on how to improve the quality and preventive value of existing contacts between patients with mental illness and their GPs may be a more fruitful approach.

An exception to this is that users of MHS with no fixed address grossly lacked provision of any primary care. Further work is required to assess the need for outreach programs to deliver primary care to homeless people with mental illness.

Furthermore, relying solely on GPs to provide preventive services may be insufficient, given the complex care needs in this group of people. A system-wide, multidisciplinary and coordinated approach, together with building a strong community-based MHS system, which includes appropriate health promotion strategies for people with mental illness, may be part of the eventual solution to health inequities in this vulnerable and sizeable population.

The ability to generalise our findings to the whole of Australia is assisted by the consistent operation of the health system across the Australian states and the fact that the Western Australian population is well representative of Australia as a whole. Our results suggest that the desirable social goal of universal access to GPs in Australia works for most people with mental disorders, with the important exception of those who are homeless. However, while universal access may be a necessary pre-condition, it alone is insufficient to guarantee improved health outcomes.

1 Definition of the study cohorts, comprising 498 803 participants

WA = Western Australia. * Users of mental health services who were not registered on the Western Australian electoral roll. † Users of mental health services who were also registered on the Western Australian electoral roll. ‡ People on the Western Australian electoral roll who were never in the Mental Health Information System, matched 2:1 with Group A by 5-year age group and sex.

2 Details of the selection of study participants

* Patients who had contact with a community mental health service, but for whom clinicians provided no further information on the number of service contacts. These could be referrals, once-only visits or situations in which health services were not compliant in providing service contact data.

7 Unadjusted (univariate analysis) and adjusted (multivariate analysis) rate ratios of visits to general practitioners from negative binomial regression analysis

MHIS = Mental Health Information System. * Users of mental health services who were also registered on the Western Australian electoral roll. † Users of mental health services who were not registered on the Western Australian electoral roll. Multivariate analysis included the outcome variable (GP visits), exposure variable and potential confounders (5-year age group, sex, Indigenous status, social disadvantage, locational disadvantage, Charlson index, calendar year, and visits to specialists). Three models were investigated based on different definitions of the exposure variable (exposed/non-exposed): (i) users of mental health services/non-users, (ii) Group A/Group B/non-users and (iii) categories of mental disorders/non-users. The results for confounding variables shown above were obtained from the second model. The reference groups were: non-users; age < 20 years; males; non-Indigenous people; the least disadvantaged group; metropolitan Western Australia (most advantaged location); index year 1990. Because of the large simple size, all estimates had very narrow 95% confidence intervals that will not show in the graph.

8 Adjusted rate ratios of visits to general practitioners from multivariate analysis, stratified by locational disadvantage

MHIS = Mental Health Information System. * Users of mental health services who were also registered on the Western Australian electoral roll. † Users of mental health services who were not registered on the Western Australian electoral roll. Multivariate analysis included outcome variable (GP visits), exposure variable (Groups A, B or C) and potential confounders (5-year age group, sex, Indigenous status, social disadvantage, Charlson index, calendar year, and visits to specialists) except locational disadvantage. The reference groups were: non-users of mental health services (people on the Western Australian electoral roll who were never in the MHIS, matched 2:1 with Group A by 5-year age group and sex in the index year); age < 20 years; males; non-Indigenous people; the least disadvantaged group; index year 1990. There were 349 813 participants included in the metropolitan model, 103 113 in the rural model and 37 297 in the remote model.