In Australia, like elsewhere, there is a socioeconomic gradient in health, with mortality rates highest amongst socioeconomically disadvantaged individuals.[1] Much of this inequality is attributable to causes that are potentially avoidable through effective and timely health care,[2]. the leading cause being ischaemic heart disease (IHD).[3] However, a common view is that the health care system plays little role in generating these health inequalities, particularly in countries like Australia with universal health insurance systems.[4, 5] Indeed, Australia's universal insurance scheme, Medicare, provides free treatment to patients treated in public hospitals and subsidises privately-delivered medical care, aiming 'to give all Australians, regardless of their personal circumstances, access to health care at an affordable cost or at no cost'.[6]

Nevertheless, as the WHO Commission on Social Determinants of Health reported recently, the opportunities for health systems to mitigate the harmful health effects of social stratification are all too often missed, and that in some instances they actually perpetuate inequity.[7] This may be true even in universal systems, particularly those where private insurance and direct private payments play some role in access to services.[8] Indeed, another underlying principle of Australia's universal system is to also allow '...choice for individuals through substantial private sector involvement in delivery and financing.'[6]. This 'choice' principle is somewhat at odds with the equity principle as it is essentially made available to those who can afford private health insurance (PHI), which covers some or all of the out-of-pocket cost of being treated as a private patient, in a public or private hospital. Around half of the population hold private insurance, 76% of households in the highest income quintile, and 23% in the lowest.[9] This, combined with non-system characteristics such as socioeconomic differentials in patient health literacy and expectations,[10, 11] suggests that those who are more socioeconomically advantaged may derive more benefit from the system than those who are less advantaged. This may be particularly true for care where there is more choice involved.

This study examines socioeconomic inequalities in the receipt of high technology health care in patients with IHD in Australia, to examine the possibility that the health care system may actually perpetuate inequalities in IHD. Ischaemic heart disease is the leading cause of avoidable mortality in Australia[12] and the main contributor to the socioeconomic mortality gap.[3] Our main aim is to quantify socioeconomic inequalities in the receipt of coronary procedures, including angiography and revascularisation procedures—angioplasty and coronary artery bypass grafting (CABG)—which are known to improve survival in patients with IHD.[13] A second aim is to explore the hypothesis that inequalities will be greater where there is relatively more discretion around treatment and hence more scope for variation in provision of services.[14] We do this by examining inequalities separately in patients with acute myocardial infarction (AMI), a relatively well-defined clinical condition where the guidelines for intervention are now relatively well established, and in patients with angina, where this is less the case. We also explore inequalities in the angina patient group stratified by emergency and elective admission status as it is likely that emergency patients are largely presenting with unstable angina, where care is perhaps less discretionary compared with elective patients, who are more likely to have chronic stable angina. A third aim is to examine the extent to which private health care and private health insurance explain any inequality in receipt of procedures.

We used administrative hospital and death data from Western Australia (WA). Western Australia comprises a tenth of the total Australian population.[15] In 2004 (the last year of data for the present study), the population was 1.98 million, three-quarters of whom resided in metropolitan Perth.[16] The health system in WA is typical of Australia, however revascularisation rates differ from other states: age-adjusted CABG and total revascularisation rates for WA in 2000 (unadjusted for need) were the lowest of any state or territory, while PTCA rates were the second highest.[17]

Data were extracted from theHospital Morbidity Data System, which contains information on each hospital admission in WA, including sociodemographic characteristics, clinical diagnoses and medical and surgical procedures. The dataset covers all public and private hospitals, including acute care hospitals and day surgeries. In addition, records were extracted from the Mortality Data System, which records all deaths registered in WA. An established population-based record linkage system allows linking of records relating to individual patients both within and between these datasets.

For this study, an AMI case was a patient with an AMI index admission, defined as an admission to hospital between 2001 and 2003 with a principal or co-diagnosis of AMI (ICD-10 code: I2I) and with no previous admissions for AMI recorded for that patient (linked records were available from 1980 onward). For angina cases, an index admission was defined as an admission to hospital between 2001 and 2003 with a principal or co-diagnosis of angina (ICD-10 I20 - 99% of cases; 'other acute IHD' (I24)); or a primary diagnosis of chronic IHD (I25) with a secondary diagnosis of angina (I20), and no previous admissions for AMI. To select the final sample of cases, further exclusions were applied, as shown in Figure 1. The selection criteria and codes used are adapted from established methods.[14] The final sample sizes were 5539 for AMI and 7401 for angina.

The eleven procedure code and procedure date fields were used to identify the outcome event, which was receipt of a coronary procedure—angiography (ICD-10 codes 38215, 38218), PTCA (ICD-10 codes 35304, 35305, 35310) and/or CABG (ICD-10 codes 38497, 38500, 38503, 90201)—within 12 months of the index admission. The main explanatory variable, socioeconomic status (SES), was assigned to each individual using the SEIFA Index of Disadvantage, a census-based measure of area-level SES.[18] The Index was assigned at the level of the Collector District (CD), the smallest census unit available, using the 2001 version of the SEIFA. Where the CD information was missing (17% of cases), the CD of the admission closest in time to the index admission was used (4% of cases), and where this was unavailable postcode was used. A SEIFA score could be assigned to 97% of cases. After assigning a score, patients were classified into population-based quintiles of SES.

Non-SES variables in the hospital dataset that may be associated with both SES and receipt of coronary procedures and hence that may confound the inequality estimates were included in the adjusted analyses. These were: Age Group (ten-year age groups from 35-44 to 75-84 then 85 plus), Sex, Marital Status (single or married/defacto), Aboriginal/Torres Strait Islander (ATSI) status (ATSI or non-ATSI), Country of Birth (Australia/New Zealand or other), Area of Residence (based on Accessibility/Remoteness Index of Australia (ARIA+) categories: major cities, inner regional and more remote), Hospital Area of index admission (metropolitan or rural), and Comorbidity. Comorbidity was measured using the modified Charlson Index, which is a weighted index based on diagnoses for all hospitalisations in the 12 months prior to the index admission.[19, 20] It is included as a potential confounder because lower SES is associated with higher comorbidity, and those with comorbidity are less likely to be offered a procedure.[21]

Two variables—Private Insurance (those holding private hospital insurance or who had hospital cover with the Department of Veteran's Affairs versus none) and Private Hospital (private hospital admission versus public)—were treated as mediating variables, as per the third aim of the study, i.e. to examine the extent to which these variables explain any of the SES variation in receipt of procedures.

Admission-date and procedure-date variables were used to calculate time to procedure. Where procedure dates were missing (11.6%), the admission date for the admission in which the procedure was performed was substituted for the procedure date. The admission-type data element describes whether the admission occurred on an emergency basis (admission should occur within 24 hours) or an elective basis (either from a waiting list or not).[22] In this study, the variable was dichotomised as either emergency or elective and was used to select AMI cases (i.e. emergency admission only, see Figure 1) and to stratify angina cases into emergency and elective admissions.

The probability of receiving a coronary procedure in Western Australia varies by SES, with clear inequalities evident in patients admitted to hospital with angina but not in patients admitted for emergency AMI care. There were two exceptions in the findings, where socioeconomic gradients were reversed: in female patients admitted with AMI, more advantaged women were less likely to receive CABG than less advantaged women, and amongst elective angina patients, the advantaged women were less likely to have an angiogram. This may reflect true differences in need for the procedures, i.e., disadvantaged women may present with more severe disease. However, this cannot be ascertained from the study.

A strength of this study was the use of linked administrative data, which enabled individuals to be followed through time and data to be censored. Nevertheless there are limitations in using these data, which may have biased the results. First, several factors could have lead to either an underestimate or an overestimate of inequality. One is that while administrative data are highly reliable for ascertainment of coronary procedures, and the coding of AMI has been found to be reasonably reliable, this is not the case for angina—specificity and sensitivity is high but positive predictive value is relatively low.[24, 25] Whether this would bias the inequality estimates depends on whether such misclassification is differential with respect to SES, which is unknown. Another possible bias is that while we adjusted for 'need' by limiting the study population to only those patients admitted with AMI or angina, appropriateness of care is complex and it is not possible to capture this complexity. The extent and direction of the potential bias this creates is difficult to predict as the relationship between procedure rates and disease and other characteristics is not straightforward, particularly for angina.[26]

Second, comorbidity may not have been fully accounted for in the models as there is considerable under-reporting of comorbidities in hospital admissions data.[27, 28] This could have lead to an overestimation of inequality because lower SES patients are more likely to have comorbidities, and those with comorbidities are less likely to be offered a procedure.[21]

Third, several influences may have lead to an underestimation of inequality, including the use of area-level SES measures (rather than individual-level measures, which were not available), and the fact that only patients admitted to hospital were included—while this means all coronary procedures are captured, not all people with IHD who could potentially benefit from a coronary procedure are. This selection bias is less of a problem for AMI than for angina as most people who initially survive a heart attack present to hospital.[14] However, it is plausible that amongst those with angina, socioeconomically disadvantaged individuals are less likely to be admitted for investigation than the more advantaged. This is consistent with the study data that showed a relatively higher proportion of higher SES patients in the angina sample compared with the AMI sample.

There are few other studies with which the SES inequalities estimated for angina patients in this study can be directly compared. No previous Australian studies, and few international ones, have examined procedure rates in angina patients. Those that have, [29–31] like the current study, found evidence of inequalities. However, unlike the current study, the one study that compared inequalities in procedure rates across AMI and angina patients (Finland, 1995-98), found they were similar across the two groups.[30]

With regard to AMI patients, earlier Australian studies found overall coronary procedure rates to be higher in private than public hospitals (Victoria, 1995-1997)[32] and that socioeconomically advantaged patients were more likely to undergo angioplasty, but not CABG, than disadvantaged patients (Queensland, 1998).[33] Inequalities in procedure rates were also found in a study of patients with IHD followed up in a clinical trial of lipid-lowering medication (1990-1997).[34] Findings have been similar in international studies, though some studies have found no inequalities in procedure rates.[35–37] Most report inequalities in catheterised procedures (angiography and PTCA) and total revascularisation procedures, but not necessarily CABG. [29, 30, 38–42] Notably, at the time these other studies were carried out, percutaneous procedure rates were not used widely in patients with AMI. For example, in the earlier Australian studies the probability of angioplasty was less than 10%, compared with nearly 50% in this study. Limitations in directly comparing the earlier and the current studies notwithstanding, the difference in findings are not inconsistent with the inverse equity hypothesis, which predicts that inequalities will appear when there is still a relatively low rate of use in the population (as in the earlier studies), but will decrease as the intervention becomes more commonly used (as in the current study).[43]

That there was no clear evidence of socioeconomic inequality in coronary procedure rates in patients seeking emergency care following AMI should not be surprising. In Australia there are now relatively clear guidelines for the use of these procedures in this patient population, utilisation is relatively high, and there is free access to public hospital care—an environment that should present few financial barriers to receiving care. In the same context, that inequality exists in the receipt of coronary procedures in patients presenting with angina is perhaps not unexpected. The use of procedures in this population is more discretionary, a large proportion of patients are admitted electively, and a relatively large proportion of procedures are performed in private hospitals.

One of the possible mechanisms underlying socioeconomic inequalities amongst the angina patients—PHI—was explored in this study. As expected, higher SES patients were more likely to hold PHI and this increased the likelihood of receiving a procedure, although PHI did not fully account for the inequality in procedure rates. Notably, inequality was also evident in waiting times. Among the elective angina patients, lower SES patients were more likely to be admitted from a waiting list than higher SES patients (the percentages of patients admitted from a waiting list for Q1 (low SES) to Q5 (high SES), respectively, were: 64%, 51%, 44%, 36%, and 23%). This in turn was related to patients' PHI status, with nearly all (94%) patients without this insurance having to wait for the procedure, while the opposite was true for patients with private insurance (9% having to wait).

Patient and doctor characteristics, not examined in this study, may also explain the inequalities in procedure rates. First, there may be contraindications for receiving a procedure that are more prevalent in lower SES patients, but that were unmeasured in this study, including smoking,[44] obesity [44–46] and late presentation to hospital.[47, 48] Second, disadvantaged patients may be less likely to see a specialist,[49] and specialists may be more likely than non-specialist doctors to recommend a coronary procedure.[50] Third, patients' preferences to seek care and undergo procedures may vary by SES—disadvantaged patients may have lower expectations[51] and be less willing to undergo a procedure[52] than more advantaged patients; and doctor's decisions may vary, either intentionally or unintentionally, depending on the social class of the patient, with higher SES patients at an advantage in this regard.[44, 51, 53]

That inequalities appeared for more discretionary care raises the question whether or not the higher procedure rates in advantaged individuals represent overuse, or whether they represent underuse in disadvantaged individuals. While these two possibilities have different implications for health inequalities, either state can be considered inequitable. In the case of underuse, disadvantaged individuals are not receiving health care from which they could benefit. In the case of over-use, this poses an overall problem for equity in a system with limited resources: where increases in health spending are increasingly going to more discretionary care, this leaves those with a greater capacity to benefit without, or having to wait longer for, much needed care, while the relatively 'well off' perhaps make more marginal gains.

The findings suggest that universal health care systems such as Australia's, with mixed public/private funding and delivery allowing for 'choice' in heath care, may actually perpetuate health inequity. While such systems might ensure equity for patients with AMI, where guidelines for treatment are relatively well established, this is not the case for care of patients with angina, where high technology health care may be less urgent and more discretionary.