Primary outcome measures were the proportion of patients whose care met quality of care standards for diagnostic and risk-stratification procedures and management according to the NHF/CSANZ guidelines.1 Compliance with guidelines was assessed by the Care Coordinator according to whether an action, such as medication prescribing, was indicated, and adjusted for stated contraindication. This was determined in conjunction with the principal investigator at the centre, with clarification from the treating physician as required.

We compared treatment between hospitals with interventional facilities and those without. Hospitals were defined as interventional centres if they had facilities for both angiography and percutaneous coronary intervention (PCI). Hospitals without such facilities were defined as non-interventional centres.

The principal investigator at each centre was responsible for obtaining ethics committee approval at that centre. In some hospitals, a formal ethics committee submission was not required for a quality assurance audit of this type.

Our aim was to assess usual practice, and this was reflected in the flexible recruitment criteria. These were not strictly limited to acute coronary syndrome patients; we included all, preferably sequential, patients presenting with chest pain that was reasonably suspected to be cardiogenic. This encompassed all patients admitted to a coronary care unit with chest pain or a step-down unit in a monitored bed.

An independent data analysis company, Statistical Revelations (Melbourne, Vic), conducted the analysis using SAS, version 9 (SAS Institute, Cary, NC, USA). In general, separate results are presented for interventional and non-interventional hospitals, as well as overall results and the difference between the two hospital types.

For proportions, exact 95% confidence limits based on the binomial distribution were used. For continuous variables and for differences between proportions, 95% confidence intervals were based on the t distribution. The median time to events (eg, electrocardiography [ECG], thrombolysis or angiography) and 95% confidence limits were determined using the Kaplan–Meier method. Hazard ratios were determined using a Cox proportional hazards regression analysis.

Logistic regression models were developed for the outcome variables referral for angiography and in-hospital death. Variables considered were age, sex, primary diagnosis, cardiovascular risk factors and comorbidity (renal impairment). Each factor was explored in a univariate logistic regression model. All factors that were significant at the 0.1 level were considered together, and four methods of model selection were pursued: forward selection, backward elimination, stepwise selection, and a best subsets approach (using a score criterion and Akaike’s information criterion to select the best model). The results from these processes were consistent, and a final model was fitted.

Box 5 shows use of medications adjusted for stated contraindication, for all centres during hospital admission, including emergency department, coronary care unit and ward.

The use of clopidogrel, a glycoprotein IIb/IIIa (GPIIb/IIIa) inhibitor, or both across all centres is shown in Box 6.

A logistic regression model for in-hospital mortality included patient age, sex, primary diagnosis, cardiovascular risk factors, renal impairment and interventional versus non-interventional centre. Age was the only predictor of death in hospital. Corrected for age, the odds of dying in hospital were not significantly different (P = 0.78) for patients at an interventional centre (odds ratio, 0.898; 95% CI, 0.423–1.906) compared with a non-interventional centre.

The demographics of patients presenting at interventional and non-interventional centres were significantly different, and may reflect a combination of the community population that is being serviced and patient referral patterns. High-risk acute coronary syndromes, such as myocardial infarction, are more likely to be managed at interventional centres.

We found significant differences in the quality of care between interventional centres and non-interventional centres. Variations were not limited to any single facet of care, and were evident both in procedural treatments and in use of medication, discharge referral and follow-up. These findings were apparent across all centres.

A previous study found little overall difference in quality of care, with regard to use of medical therapies, between hospital types in Queensland.15 However, that study compared tertiary versus non-tertiary hospitals, and did not include some of the largest cardiac centres in Queensland. A further study did find a link between the quality of care and funding initiatives directed towards the implementation of “multiple systematic interventions”.17 In another study, variability in care of patients with acute coronary syndrome depended on whether they experienced STEMI, non-STEMI or unstable angina.14 The CRUSADE initiative in the US demonstrated marked variation in the use of recommended medical therapies between leading (most adherent) and lagging (least adherent) hospitals.5 This variation was most evident with therapies considered recent innovations or more aggressive. If our results are compared with results from these leading and lagging centres, based on acute medication use, it appears that Australian practice varies widely between that of leading and lagging centres in the US, depending on the treatment. For example, overall use of GPIIb/IIIa inhibitors in Australian centres is lower than in the most lagging US hospitals, whereas use of any heparin was similar to the most leading US hospitals.9-11

Not all our indicators favoured interventional centres. For some key indicators, interventional hospitals had lower adherence to guidelines. For example, a higher proportion of patients underwent ECG within the first 10 minutes at non-interventional facilities.

Referral rates for angiography at centres with PCI capability were similar to rates described in GRACE and other registries.18

However, we found a significantly reduced rate of referral for investigation and further evaluation at non-interventional centres than at interventional centres. These findings have been noted in previous audits in Queensland and rural New South Wales. The Queensland study found lower rates of referral for coronary angiography for patients with acute coronary syndromes admitted to non-tertiary centres without interventional facilities (55% v 85%).15 A study in NSW found patients admitted to metropolitan hospitals were more likely to be referred for angiography than patients managed in non-metropolitan hospitals.19

In New Zealand, one study showed a significantly reduced rate of referral for investigation and further evaluation at community hospitals compared with tertiary hospitals with interventional facilities.20 Another New Zealand study showed reduced rates of adherence to medical therapy, referral for angiography and revascularisation in centres without cardiologists.21 The New Zealand Audit Group concluded that patients admitted to hospitals without interventional facilities in general received fewer investigations and less revascularisation than patients admitted to interventional centres.22

The difference in referral rates we observed cannot be attributed to the difference in patient demographics alone. Logistic regression analysis showed that the odds of being referred for angiography are 7.4 times higher at an interventional centre than at a non-interventional centre when adjusted for age, sex, diagnosis and presence of risk factors. We also found men were more likely than women to be referred for angiography. Patients with unstable syndromes were more likely to be referred for angiography than those with simple angina. Patients with hyperlipidaemia were also more likely to be referred for angiography. There was a lower likelihood to refer patients with renal impairment for angiography. With increasing age, the odds of being referred for angiography also decreased. These factors are known to bias physicians in referring patients for angiography.23-25

Rates of referral for invasive assessment may have been influenced by the ascertainment of high-risk acute coronary syndromes at non-interventional centres. The reason for the different rates of referral requires further evaluation, but may include access block, significant delays in transfer of patients, reluctance of patients in rural areas to be transferred, or a lack of adherence to or awareness by local physicians of current guidelines. Regardless of the reasons, this represents an area in which the quality of care could be improved.

Another consistent finding was that significant delays are experienced for patients who require CABG compared with those undergoing PCI. The lengths of stay for patients undergoing CABG were higher, which has direct implications for the cost to the health service, and could increase bed access block, especially to high-dependency beds. There is also significant potential to impair outcomes for patients with high-risk syndromes if revascularisation does not occur early in the course of hospitalisation.

The limitations of our study include the relatively small numbers of patients enrolled and the uneven distribution of patients across states. The total number of patients presenting to the emergency rooms of the individual centres and their outcomes was not recorded. Only patients admitted to monitored beds were studied. This may bias the study sample. Although data were captured in both tertiary and non-tertiary centres, there was no cohesive tracking of patients between the two systems, so transferred patients were not adequately followed. There is potential for significant referral bias between interventional and non-interventional centres. Our study was not powered to adjust for the multiple confounders, especially combined with the low event rate in a cohort that included patients with non-cardiac chest pain. This makes it difficult to draw firm conclusions from the multivariate logistic regression analysis in relation to a comparison of mortality between the two groups. We also acknowledge the limitation of drawing data from case notes and medical records.