Outcomes of cardiac surgery in Indigenous Australians

Data were prospectively collected for 2635 consecutive patients who underwent cardiac surgery at Flinders Medical Centre between January 2000 and December 2005. Flinders Medical Centre is a tertiary referral centre serving the southern region of Adelaide and also providing a cardiac surgical consultation service to the Royal Darwin Hospital. The Flinders cardiac surgical database consists of over 2000 prospectively collected data elements per patient, including baseline patient characteristics, operation details, postoperative complications, and late mortality. Informed consent was obtained for all patients before inclusion in the database.

A structured form was completed for each patient on hospital admission to record age, sex, self-reported Indigenous status (including Torres Strait Islander), cardiovascular risk factors and medical history.

Hypertension was defined by either a documented history of hypertension diagnosed and treated with medication, diet and/or exercise; or a measured systolic blood pressure of > 140 mmHg or diastolic pressure of > 90 mmHg on at least two occasions.

Hyperlipidaemia was defined by a history of hypercholesterolaemia diagnosed by a physician and/or a serum cholesterol level > 5.0 mmol/L.

Diabetes was defined by a previous physician diagnosis or treatment with hypoglycaemic medication.

The European System for Cardiac Operative Risk Evaluation (euroSCORE) is a widely used model for predicting 30-day mortality in patients undergoing cardiac surgery7 (the higher the euroSCORE, the higher the predicted operative mortality). It was developed from outcomes in 128 European centres and has been validated in European, Japanese, and North American populations.8-12 Although the euroSCORE model has previously been shown to overestimate operative mortality in an Australian population, it does provide sufficient discrimination between strata of risk.13,14 Using standard component definitions, we calculated the logistic euroSCORE for all patients preoperatively (Box 1).7

Preoperative renal dysfunction was defined by a serum creatinine level > 200 μmol/L.

Preoperative neurological impairment was defined by a condition affecting ambulation or daily functioning.

Critical preoperative state was defined by preoperative ventricular arrhythmia, acute renal failure, or the need for resuscitation or haemodynamic support.

An assessment of left ventricular systolic function was obtained either by echocardiography or left ventriculography and classified, on the basis of ejection fraction (EF), as moderate systolic impairment (EF, 30%–50%) or severe systolic impairment (EF < 30%).

The type of procedure performed was recorded as the presence of coronary artery bypass grafting, with or without valve replacement or repair. In patients undergoing left anterior descending artery grafting, use of the left internal mammary conduit was recorded.

Operative mortality was defined as mortality that occurred at any time in hospital during the same admission or within 30 days of surgery. Other complications recorded included permanent stroke (defined by a new persistent neurological deficit); postoperative renal failure (defined by acute or worsening renal failure resulting in either an increase in serum creatinine to > 200 μmol/L or twice the baseline creatinine level, or a new requirement for dialysis); and prolonged ventilation (> 24 hours).

Late mortality was defined as all mortality occurring during the period of follow-up, including events within 30 days of surgery. Death after hospital discharge was detected by a search of the National Death Index in December 2006. This national database, maintained by the Australian Institute of Health and Welfare, records data from death certificates. Data linkage was performed between patient records in the Flinders cardiac surgical database and recorded deaths in Australia on the basis of first name, family name, sex, date of birth, postcode of residence and date of most recent contact.

Baseline clinical and procedural variables were stratified by Indigenous status. To control for confounding by preoperative clinical parameters, we used the euroSCORE to model preoperative risk.7 Calibration and discrimination of the euroSCORE in our population were first examined using the c statistic (a measure of concordance) and the Hosmer–Lemeshow goodness-of-fit test. Operative mortality was stratified by Indigenous status in quintiles of the euroSCORE.

Given the marked difference between Indigenous and non-Indigenous people in age of onset of cardiac disease, operative mortality for the entire cohort was first calculated unadjusted, then stratified by age group (< 55 years or ≥ 55 years) and adjusted for euroSCORE as a continuous variable through Cox proportional hazards modelling, after confirming preservation of the proportional hazards assumption. We also tested for interaction between age group, Indigenous status and mortality (operative and late).

For analysis of late mortality, unadjusted Kaplan–Meier survival curves were stratified by age group (< 55 years or ≥ 55 years) and compared by the log-rank test. Adjusted late outcomes in the two groups were compared in patients aged < 55 years using the euroSCORE as a continuous variable as an overall measure of risk in Cox proportional hazard models.

Other in-hospital events were assumed to occur at the time of surgery. These binary events are reported as counts and proportions of the total, and adjusted for euroSCORE and age in logistic regression models. Normally distributed variables are expressed as mean (SD), and non-Gaussian variables as median (interquartile range [IQR]). χ2 tests were used for comparisons of binary outcomes between groups, and t tests for continuous variables. All analyses were performed using Stata software, version 10 (StataCorp, College Station, Tex, USA).

The study was approved by Flinders Clinical Research Ethics Committee and the Aboriginal Health Research and Ethics Committee of the Aboriginal Health Council of South Australia.

Of the 2635 patients who underwent cardiac surgery, 283 (10.7%) were Indigenous. The mean age of the cohort was 63 years (SD, 13 years), with Indigenous patients being younger than non-Indigenous patients (mean, 47 [SD, 14] years v 65 [SD, 12] years; P = 0.001) and more likely to be women (40.3% v 28.0%; P < 0.001). The difference between mean ages at operation remained significant for patients undergoing coronary artery bypass grafting alone (Indigenous patients, 52 [SD, 9] years v non-Indigenous patients, 65 [SD, 10] years; P < 0.001).

The age difference in the operative cohort was clearly demonstrated by an analysis of the proportion of patients in each age group: 202/283 Indigenous patients (71.4%) and 458/2352 non-Indigenous patients (19.5%) were aged < 55 years.

Non-Indigenous patients were more likely to have the cardiovascular risk factors of hypertension and hypercholesterolaemia, whereas Indigenous patients were more likely to have diabetes and renal dysfunction and to be current smokers (Box 2). Indigenous patients were also more likely to present with moderate or severe left ventricular dysfunction, active endocarditis and pulmonary hypertension. The median euroSCORE was lower for Indigenous patients than non-Indigenous patients (2.08 [IQR, 1.49–3.92] v 2.49 [IQR, 1.51–5.13]; P < 0.001), owing to the inclusion of age in this score.

Indigenous patients were more likely to undergo either single or double valve surgery. In patients undergoing mitral valve replacement, Indigenous patients were less likely to undergo implantation of a mechanical prosthesis (Box 2).

Forty-two patients (1.6%) died either within 30 days of cardiac surgery or during the same admission (Box 3). In the overall cohort, euroSCORE was able to discriminate risk of operative mortality (c statistic, 0.75), but overestimated events in all strata (Hosmer–Lemeshow goodness-of-fit, P < 0.001). The increasing frequency of operative mortality associated with increasing euroSCORE is evident from the graph in Box 4. There was a non-significant trend towards higher operative mortality in Indigenous patients than non-Indigenous patients (2.5% v 1.5%), whether unadjusted or adjusted for euroSCORE (Box 3).

There was no significant excess in new postoperative renal failure (hazard ratio [HR], 1.3 [95% CI, 0.8–2.0]); ventilation time > 24 hours (HR, 1.2 [95% CI, 0.9–1.7]), or postoperative stroke (HR, 0.3 [95% CI, 0.0–2.1]) in the Indigenous group.

At a median follow-up of 45 months (IQR, 25–60 months), there were 293 deaths (11.1%) (Box 3). Indigenous status was associated with a borderline-significant excess in late mortality (Indigenous patients, 12.7% v non-Indigenous patients, 10.9%; HR, 1.4 [95% CI, 0.99–2.0]) that became significant when adjusted for euroSCORE (HR, 1.5 [95% CI, 1.03–2.1]).

Survival at 1 and 5 years was 94.0% and 80.6%, respectively, for Indigenous patients compared with 96.7% and 87.7%, respectively, for non-Indigenous patients (Box 5).

The mean age of Indigenous patients was substantially lower than the mean age of non-Indigenous patients. When the analysis was restricted to patients aged < 55 years, there was a significant excess in operative mortality among Indigenous patients (Box 6). The difference persisted when adjusted for euroSCORE (Box 3). The operative mortality of Indigenous patients aged < 55 years (3.5%) was in fact higher than that for non-Indigenous patients aged ≥ 55 years (1.7%).

In patients aged < 55 years, at a median follow-up of 45 months, death occurred in 27/202 Indigenous patients (13.4%) and 26/458 non-Indigenous patients (5.7%) (Box 6). Unadjusted survival curves of patients in the two age groups (< 55 years and ≥ 55 years) (Box 7) showed excess late mortality among Indigenous patients compared with non-Indigenous patients in the younger cohort (HR, 3.0 [95% CI, 1.8–5.3]), which persisted despite adjustment for euroSCORE (HR, 6.9 [95% CI, 1.4–33.5]). By contrast, differences in survival between Indigenous and non-Indigenous patients were not significant in the older cohort, whether unadjusted (HR, 1.2 [95% CI, 0.6–2.3]) or adjusted (HR, 0.9 [95% CI, 0.5–1.8) for euroSCORE. This was primarily due to a low number of Indigenous patients (and hence events) in the older age group (interaction P value for age < 55 years, Indigenous status and mortality, 0.005).