Background
Ischaemic heart disease (IHD) remains the leading cause of morbidity and mortality for both Aboriginal and non-Aboriginal populations [
1,
2]. The imperative to prevent the first episode of IHD remains strong given that sudden cardiac death occurs on first presentation for one in five [
3]. Chronic diseases account for 70% of the gap in total disease burden between Aboriginal and non-Aboriginal Australians [
4]. Despite medical advances, cardiovascular disease (CVD) contribute one–fifth of the differential in total disease burden between Aboriginal and non-Aboriginal Australians [
4], with greater burden of chronic diseases, such as diabetes and chronic kidney disease (CKD) in the former subpopulation [
5].
In Australia, private and public health services exist side by side, with private health insurance (covering part costs of private hospital and specialist services) only accessed by those who can afford to pay. Through Australia’s universal health insurance scheme, public hospital inpatient and outpatient services are free to patients. Primary care services are substantially subsidised through Medicare, both in mainstream services and Aboriginal Medical Services [
6], which include additional schemes to better meet Aboriginal health needs [
7,
8], e.g. reduced cost of prescribed medicines through Closing the Gap scheme. In Western Australia, medical specialist services are concentrated in Perth, the capital city, with most of the rural population (including 60% of the Aboriginal population) needing to travel considerable distances for access. Rural services have difficulties in attracting health professionals, with workforce retention a further challenge. Despite the fact that Aboriginal Australians have access to primary care through multiple avenues, a recent cost analysis highlighted that 60% of the health expenditure on Aboriginal Australians was on secondary care in public hospitals [
9]. However, primary health care (Medicare services and medicines) expenditure per person was significantly less than for non-Aboriginal Australians [
9,
10], suggesting underinvestment of funds into prevention, early intervention, secondary prevention and community services for Aboriginal people [
10].
General practitioners (GPs) in primary care serve as the gatekeeper to the health system in Australia. GPs have a critical role in the primary prevention of IHD [
11], and are ideally situated to provide secondary prevention [
11]. The integration between primary, specialist and tertiary care, as a basis of seamless continuity of care is of paramount importance for these patients [
11]. GP shortages are however greater in non-metropolitan regions where the majority of Aboriginal people reside [
12]. Consequently, policies exist to boost GP numbers in rural areas through financial incentives, and international medical graduates (IMGs) are provided restricted registration to work in areas of unmet need [
12]. General practice is recognised as a medical specialty through the General Practice Vocational Register. GPs who have completed their fellowship are vocationally registered (VR GPs) and have access to higher Medicare rebates, while those who are not (Non-VR GPs) utilise lower Medicare rebates except under special programs such as districts of workforce shortage which commonly occurs in disadvantaged and remote areas.
Medicare-rebated (private) specialist services often incur significant co-payment such that patients who experience financial difficulties are under-represented in such practices, relying on hospital out-patient consultations. People who live in rural/remote areas often have limited access to private specialists and are required to travel long distances to access public hospital out-patient services.
Patterns of primary and specialist care in patients leading up to the first admission with IHD potentially impacts on prevention and subsequent outcomes. We aimed to compare the patterns of GP and out-of-hospital specialist consultations, and direct costs of these services in Aboriginal and non-Aboriginal patients in Western Australia (WA) in the 2 years preceding first hospitalisation for IHD. Additionally, we investigated the use of emergency departments (ED) as a substitute for primary health care services.
Methods
Study population
A cohort of WA residents aged 25–74 years was identified from administrative hospital data based on first-ever hospitalisation (15-year clearance period) for IHD as a principal discharge diagnosis (‘index hospitalisation’) during the period 2002–2007. IHD was identified from codes I20-I25 of the International Classification of Diseases 10th edition Australian Modification.
Data sources
Linked state-wide inpatient records from the Hospital Morbidity Data Collection (HMDC) and Emergency Department Data Collection (EDDC) data were obtained from the WA Data Linkage System [
13]. National government-held Medicare Benefits Schedule (MBS) data provided details of professional consultations, procedures and diagnostic tests for citizens/permanent residents that universally claim from Medicare. The MBS data do not cover pharmaceutical claims (separate scheme) or outpatient services provided by public hospitals.
Identification of services and costs
HMDC records were merged with MBS data (2000–2007) to identify the Medicare records for 2 years preceding the index hospitalisation. GP and specialist consultations were identified, and associated costs were compared between Aboriginal and non-Aboriginal patients using the bottom-up itemised MBS costs (scheduled fees). Additional resource costs were estimated by applying costs derived from the MBS for ambulatory community consultations and associated pathology and imaging tests. Person-based costs were aggregated and compared for the two sub-populations. Only direct health care (service provision) costs were examined. The historical consumer price index was used to adjust for changes in costs (2005 as base year).
Similarly, ED presentations in the 2 years preceding the first hospitalisation were identified and relevant data extracted, including demographic and episode fields with presentation time/date and triage scores (≤3 represent immediate/urgent cases; 4 = semi-urgent; 5 = non-urgent) [
14].
Demographic and co-morbidity data
To optimise the estimation of Aboriginal status in routinely-collected data [
15,
16], a patient was defined as being Aboriginal if ≥25% of their historical hospital admissions had been coded as Aboriginal. We have used the same definition in previous publications to acknowledge under ascertainment in hospital administrative data collections while avoiding over-inclusion [
17,
18]. The Charlson comorbidity index was calculated using the modified Deyo algorithm [
19] and individual comorbidities (Table
1) identified using a 5-year look-back period. The Accessibility/Remoteness Index of Australia (ARIA) [
20] classifies five categories of residential remoteness and was included as a covariate in the regression analysis. Separately, the greater Perth metropolitan city definition [
21] dichotomised place of residence into metropolitan and rural residence. Private health medical insurance status recorded in the HMDC was used as a proxy for socio-economic status.
Table 1
Characteristics of the cohort (aged 25–74 years) with a first admission of ischaemic heart disease, as principal discharge diagnosis, in the period of 2002–2007, WA residents
Number of patients (%) | 1269 (4.7) | 25,961 (95.3) | | 832 (65.6) | 6706 (25.8) | < 0.001 | 437 (34.4) | 19,255 (74.2) | < 0.001 |
First-ever, n (%) | 1041 (82.0) | 22,416 (86.3) | < 0.001 | 700 (84.1) | 6074 (90.6) | < 0.001 | 341 (78.0) | 16,342 (84.9) | < 0.001 |
Sub-types of IHD, n(%) |
▪ Unstable angina | 346 (27.3) | 6186 (23.8) | < 0.001 | 224 (26.9) | 1652 (24.6) | < 0.001 | 122 (27.9) | 4534 (23.6) | < 0.001 |
▪ Acute myocardial infarction | 456 (35.9) | 7377 (28.4) | | 312 (37.5) | 2259 (33.7) | | 144 (33.0) | 5118 (26.6) | |
▪ Other IHD | 466 (36.7) | 12,391 (47.7) | | 296 (35.6) | 2793 (41.7) | | 170 (38.9) | 9598 (49.9) | |
▪ Subsequent MI | 1 (0.1) | 7 (0.03) | | 0 | < 5 (0.0) | | < 5 (0.2) | 5 (0.0) | |
Total relevant MBS records (pre-2 years) | 47,047 (3.5) | 1,301,191 (96.5) | < 0.001 | 27,805 (59.1) | 228,849 (17.6) | < 0.001 | 19,242 (40.9) | 1,072,342 (82.4) | < 0.001 |
1st year before index admission, n(%) | 26,365 (56.0) | 733,063 (56.3) | 0.200 | 15,735 (56.6) | 129,341 (56.5) | 0.818 | 10,630 (55.2) | 603,722 (56.3) | 0.012 |
2nd year before index, n(%) | 20,682 (44.0) | 568,128 (43.7) | | 12,070 (43.4) | 99,508 (43.5) | | 8612 (44.8) | 468,620 (43.7) | |
Mean age ± SD | 50.2 ± 10.1 | 60.5 ± 9.3 | < 0.001 | 44.4 ± 6.6 | 47.9 ± 5.4 | < 0.001 | 61.4 ± 5.0 | 64.9 ± 5.7 | < 0.001 |
Female sex, n(%) | 574 (45.2) | 7363 (28.4) | < 0.001 | 355 (42.7) | 1621 (24.2) | < 0.001 | 219 (50.1) | 5742 (29.8) | < 0.001 |
Urban/rural, n(%) |
Rural | 817 (64.4) | 5260 (20.3) | < 0.001 | 541 (65.0) | 1485 (22.1) | < 0.001 | 276 (63.2) | 3775 (19.6) | < 0.001 |
Urban | 452 (35.6) | 20,701 (79.7) | | 291 (35.0) | 5221 (77.9) | | 161 (36.8) | 15,480 (80.4) | |
ARIA classification, n(%) |
Highly accessible | 282 (22.2) | 12,641 (48.7) | < 0.001 | 177 (21.4) | 3179 (47.5) | < 0.001 | 103 (23.9) | 9454 (49.2) | < 0.001 |
Accessible | 236 (18.6) | 9362 (36.1) | | 155 (18.7) | 2390 (35.7) | | 81 (18.8) | 6971 (36.3) | |
Moderately accessible | 202 (15.9) | 2723 (10.5) | | 135 (16.3) | 685 (10.2) | | 67 (15.6) | 2034 (10.6) | |
Remote | 42 (3.3) | 416 (1.6) | | 25 (3.0) | 111 (1.7) | | 17 (3.9) | 295 (1.5) | |
Very Remote | 505 (39.8) | 812 (3.1) | | 337 (40.7) | 327 (4.9) | | 163 (37.8) | 469 (2.4) | |
Indeterminate | < 5 (0.2) | 7 (0.03) | | 0 | < 5 (0.0) | | 0 | 5 (0.0) | |
With private medical insurance, n (%) | 34 (2.7) | 11,603 (44.7) | < 0.001 | 21 (2.5) | 2752 (41.0) | < 0.001 | 13 (3.0) | 8851 (46.0) | < 0.001 |
Charlson comorbidity index (≥ 1), n % | 446 (35.2) | 6817 (26.3) | < 0.001 | 285 (34.3) | 1879 (28.0) | < 0.001 | 161(36.8) | 4938 (25.6) | < 0.001 |
Other comorbiditiesa, n (%) |
➢ Hypertension | 680 (53.6) | 11,502 (44.3) | < 0.001 | 402 (48.3) | 2317 (34.6) | < 0.001 | 278 (63.6) | 9185 (47.7) | < 0.001 |
➢ Heart failure | 130 (10.2) | 1390 (5.4) | < 0.001 | 64 (7.7) | 167 (2.5) | < 0.001 | 66 (15.1) | 1223 (6.4) | < 0.001 |
➢ Chronic kidney disease | 167 (13.2) | 789 (3.0) | < 0.001 | 90 (10.8) | 156 (2.3) | < 0.001 | 77 (17.6) | 633 (3.3) | < 0.001 |
➢ Valvular heart disease | 161 (12.7) | 3367 (13.0) | 0.770 | 86 (10.3) | 459 (6.8) | < 0.001 | 75 (17.2) | 2908 (15.1) | 0.235 |
➢ Diabetes | 665 (52.4) | 5217 (20.1) | < 0.001 | 403 (48.4) | 1010 (15.1) | < 0.001 | 262 (60.0) | 4207 (21.9) | < 0.001 |
➢ COPD | 169 (13.3) | 1266 (4.9) | < 0.001 | 83 (10.0) | 173 (2.6) | < 0.001 | 86 (19.7) | 1093 (5.7) | < 0.001 |
➢ Cancer | 59 (4.7) | 3607 (13.9) | < 0.001 | 23 (2.8) | 515 (7.7) | < 0.001 | 36 (8.2) | 3092 (16.1) | < 0.001 |
➢ Cerebravascular disease | 53 (4.2) | 808 (3.1) | 0.034 | 29 (3.5) | 71 (1.1) | < 0.001 | 24 (5.5) | 737 (3.8) | 0.074 |
Mortality, n (%) |
30-day mortality post-index | 24 (1.9) | 332 (1.3) | 0.061 | 12 (1.4) | 45 (0.7) | < 0.001 | 12 (2.8) | 287 (1.5) | 0.034 |
1-year mortality post-index | 70 (5.5) | 788 (3.0) | < 0.001 | 29 (3.5) | 90 (1.3) | < 0.001 | 41 (9.4) | 698 (3.6) | < 0.001 |
Statistical analyses
Baseline characteristics and crude mortality by Aboriginal status were compared using descriptive statistics. Results are provided for broad age groups < 55 and ≥ 55 years, with 55 years as the median age of the patients. Two-tailed t- or Mann-Whitney tests (for continuous variables) and the Pearson chi-squared test (for categorical variables) were used to test for significance. Negative binomial regression log-linear models were used to model count of GP consults or specialist consults as separate dependent variables in the 2 years preceding IHD admission, with adjustment for covariates at index admission. This method was selected because the distribution of our count data showed that it was over-dispersed (highly skewed) and not suitable for Poisson regression models which assume the variance is equal to the mean. Stratified analyses [by gender, broad age group, metropolitan (vs rural), incident (vs prevalent), diabetes and CKD status] were undertaken to determine the differential in subgroups and to identify where the need for intervention was most critical. STATA version 13.1 was used for analyses.
Discussion
This person-level, population-based study investigated the patterns of Medicare-funded GP and specialist care, and resource utilisation for Aboriginal and non-Aboriginal patients in the 2 years preceding first IHD hospitalisation. Our findings show that, after adjustment, Aboriginal patients were more likely to have had a GP consultation, a long/ prolonged consult and consult a non-VR GP, while substantially fewer received Medicare-rebated specialist consultations. The mean cost differential on Medicare expenditure over 2 years was 43.3% lower in Aboriginal compared to non-Aboriginal people, despite higher prevalence of multimorbidity in Aboriginal patients. This difference persisted in metropolitan subgroups. Notably, the resource differential was greatest in diabetic and CKD patients for whom ongoing specialist expertise for managing complex care needs is arguably more critical. Private specialist consults were the key cost driver in Medicare expenditure differentials between Aboriginal and non-Aboriginal people. For ED presentations, Aboriginal patients, particularly those in rural/remote areas were more likely to use ED than GPs for conditions classified urgent/semi-urgent. These findings suggest a substitution effect of ED with primary care.
Published empirical evidence demonstrates the association between a strong primary care system and better health status, reduced hospitalisation, and all-cause (and CVD) mortality [
22‐
24]. Further, evolving research suggests that a collaborative multidisciplinary approach based on the integration of specialists and GPs could lead to better cardiovascular outcomes in primary care setting, particularly for patients with multimorbidity [
25]. In our study there was an over-representation of Aboriginal patients with no primary care records in the pre-event study period. Additionally, although Aboriginal patients had a marginally higher adjusted-IRR for primary consults, over a quarter of GP consults were rendered by non-VR GPs/IMGs. This might impact on the continuity and cultural appropriateness of care. IMGs are strongly represented in rural/remote Australia (particularly WA) [
12] with many relocating from rural areas once licensing restrictions are satisfied [
12,
26]. In a study of IMGs, Durey et al. [
27] identified the need to better address cross-cultural issues and the importance of effective communication and building community and cross-institutional relationships. A strong physician-patient relationship (particularly developed in a culturally secure context), is a key element enabling continuity of care necessary for optimal management of chronic diseases, and might be limited in the current environment.
High performing general practices have lower IHD admission and mortality rates, with the association strongest for practices serving populations of high levels of economic deprivation [
28]. This may lead to a reduction in the health inequalities noted in this analysis. Despite the initiatives to strengthen the GP sector in very remote areas, poor access to GP and both Medicare-funded and public hospital specialist services persists in very remote areas where the majority of Aboriginal people live. This highlights a critical need to create more innovative models of care.
Aboriginal patients, despite their younger average age, have significantly greater co-morbidity, adding considerably to complexities in managing the primary disease. This is often accompanied by complex social circumstances resulting from the unfortunate historical and political legacy of colonialism, with adversities including living in under-resourced remote locations (where access to medical care may be limited), poverty, unhealthy environments and higher rates of mental illness, imprisonment, poor educational attainment and family troubles [
29]. This social, clinical and logistical complexity may explain the higher proportion of long/prolonged visits. Higher ED use among Aboriginal patients may reflect poor access to and/or out-of-pocket costs of primary care (where ED is used as a proxy for primary care) as well as high clinical need. Comorbidities were all independent predictors of increased rates of GP consultations (Table
4). Consequently, Aboriginal (vs non-Aboriginal) patients have poorer 30-day and 1-year mortality outcomes, consistent with other reports [
16,
17].
Rural disparities in Medicare expenditure for the financial year 2006–2007 were reflected in a total Medicare deficit of $811 million [
30], translating into 12.6 million fewer services during that year for the people of regional and remote areas, attributed largely to poorer access to health professionals. We have shown similar under-expenditure across the broad MBS service categories among Aboriginal patients, particularly in poorer access to specialist care. As Medicare-rebated specialist care does not cover public hospital out-patient services, the exact quantification of differential specialist utilisation could not be captured. Nevertheless, access to specialist care remains a significant issue for Aboriginal cardiac patients, often reflecting specialist shortages in rural areas, under-developed care pathways and weaknesses in communication/information flow between primary and acute and specialist care professionals – all of which impact on quality of care. Since 2010, Medicare has introduced a wide range for MBS items and rebates, including those for telehealth consultations for medical specialists in other locations, as a means to improve access. It is possible that greater uptake of such initiatives could address some of this gap.
More innovative models of care, incorporating appropriate telehealth, are needed to overcome the problems of costs and accessibility related to geographical location in WA. Additionally, this analysis shows structural reform of PHC services is needed to ensure better integration of care and management for Aboriginal people, over half of whom reside in rural Australia. Analysis of more contemporary data will be valuable to show how service utilisation has changed with recent health reforms.
Strengths and limitations
This study uses MBS data linked to a diagnostic cohort identified through hospital data, allowing individualised person-based utilisation of specific services and associated costs to be examined. Medicare data represent the most accurate and reliable source of health care attendance in Australia, although visits to an Aboriginal Medical Service could not be differentiated in provided data from mainstream consults. MBS data do not include outpatient consultations at public hospitals and the use of MBS data alone limits the true representation of specialist services provided. As Aboriginal patients are over-represented in the (mainly metropolitan) public outpatient clinics in WA and some specialist visits occur through alternative funding mechanisms, the difference in receipt of specialist ambulatory consults may be over-estimated. Improvement of out-patient hospital data to linkage standard is urgently required. Further, the MBS data is relatively old due to the long lead time needed for the requisition of Commonwealth data; regardless, the current analysis provides a base from which to gauge further improvements with ‘Closing the Gap’ initiatives undertaken by the government.
Acknowledgements
The study was supported by a NHMRC project grant (#1031057). Judith Katzenellenbogen is funded by an NHMRC Early Career Fellowship (#037429). The Western Australian Centre for Rural Health receives funding from the Department of Health and Ageing. The authors thank the Western Australian Department of Health and the Australian Department of Health for providing the cross-jurisdictional linked data used in this study. We also thank the WA Data Linkage Branch and the staff of the Inpatient Data Collections for the provision of the data and the Registry of Births, Deaths and Marriages for providing the death data, and to the data custodian for the Emergency Department Data Collection. The inputs on Aboriginal MBS item codes from Dr. Marianne Woods and Ms. April Rutkay are much appreciated.