Twenty Australian teaching and major metropolitan hospitals were recruited for the study. Sixteen hospitals provided data in a timely fashion to permit meaningful analysis. The hospitals were located in all mainland states except South Australia, and comprised those with the highest number of documented separations of patients after MI in each state (data from the Australian Institute of Health and Welfare).5

For each hospital, a 1-month snapshot of prescribing of cardiovascular medications was obtained for consecutive patients with MI admitted to the coronary care unit.

At each site, the coronary care unit coordinator obtained patient data on discharge after hospitalisation for MI. Data were collected on teleform, then faxed and, finally, transferred electronically to a Structured Query Language (SQL) database (Microsoft SQL Server 2000, Microsoft Corp, Redmond, Wash, USA) for statistical analysis. Data quality was verified by double data entry and manual checking of more than 95% of fields.

The data collected on the form included the location of the MI, the type of MI (with or without ST-segment elevation) and any percutaneous coronary interventions performed during the patient’s index hospitalisation.

Heart failure status was recorded before hospitalisation, on admission, during hospitalisation and at discharge. Cardiovascular drug therapies (the main focus of our analysis) were recorded before hospitalisation, during admission and at discharge. Prehospitalisation data were obtained from the patient’s hospital admission and subsequent in-hospital notes. Additional data obtained were presence of risk factors for MI and results of assessment of ventricular function (if available).

• Myocardial infarction was defined according to each hospital’s individual diagnostic criteria, and included MI with and without ST-segment elevation.

• Heart failure was defined according to the heart failure guidelines of the National Heart Foundation and the Cardiac Society of Australia and New Zealand,6 but ultimately left to the discretion of individual investigators at each site.

• Risk factors were defined by the individual hospitals.

Ethics approval for the study as a quality improvement project was obtained from the ethics committees of the participating hospitals.

Differences in prescribing were determined by χ2 analysis, with a two-tailed P value < 0.05 considered to be statistically significant. Age was compared between those with and without heart failure by Student’s unpaired t test. Predictors of heart failure were determined by logistic regression analysis, entering all relevant parameters into a backwards stepwise model. Predictors of drug prescribing in heart failure patients were determined by multivariate logistic regression analysis, entering the main univariate parameters of difference between patients with and without heart failure into the multivariate model.

The age, sex and pre-existing risk factors of the 479 patients in our cohort are summarised in Box 1.

Characteristics of the study patients with and without heart failure during hospitalisation or at discharge after MI are summarised in Box 2. In the group with heart failure (24.2% of all patients with MI), there was a higher proportion of women and patients with prior heart failure. The association between diabetes and heart failure was of borderline statistical significance.

Significantly fewer percutaneous coronary intervention procedures were performed in patients with heart failure; in particular, significantly fewer angiography and angioplasty procedures were performed during the index hospitalisation (Box 2).

The proportion of patients taking the various cardiovascular medications on arrival at hospital (prior to admission) are shown in Box 3, and the proportion of patients with and without heart failure prescribed these medications on discharge are shown in Box 4. Fewer patients in the cohort with heart failure were prescribed an ACE inhibitor or an angiotensin receptor blocker, but these differences did not reach statistical significance. Univariate analysis showed that significantly fewer patients with heart failure, compared with those without heart failure, received β-blockers. Interestingly, fewer patients with heart failure were prescribed aspirin, clopidogrel or statin therapy. However, there was greater use of warfarin, diuretics and spironolactone in patients with heart failure. Spironolactone use, even in patients with established heart failure, was low at 7.8%.

Uncorrected differences in prescribing at discharge according to heart failure status are summarised in Box 5. To assess use of accepted heart failure medication after MI (ACE inhibitor/angiotensin receptor blocker, β-blockers and aldosterone receptor blockade), drug utilisation was adjusted for baseline covariates of age, sex and the presence of diabetes mellitus (these differed significantly [or almost significantly] at baseline between patients with and without heart failure). This showed that differences in prescribing between patients with and without heart failure remained non-significant for ACE inhibitors/angiotensin receptor blockers (P = 0.136), became non-significant for β-blockers (P = 0.106), and remained significant for spironolactone (P = 0.015).

Only prior heart failure was a univariate predictor of heart failure after MI. Hyperlipidaemia, diabetes, hypertension, prior stroke, current smoking and prior acute coronary syndrome/angina did not predict heart failure after MI. Backwards, stepwise multiple logistic regression analysis showed that prior heart failure remained a significant independent predictor (P = 0.001) of heart failure after MI.