Objective: There is an important role for accurate risk prediction models in current cardiac surgical practice. Such models enable benchmarking and allow surgeons and institutions to compare outcomes in a meaningful way. They can also be useful in the areas of surgical decision-making, preoperative informed consent, quality assurance and healthcare management. The aim of this study was to assess the performance of the European System for Cardiac Operative Risk Evaluation (EuroSCORE) model on the Australasian Society of Cardiac and Thoracic Surgeons (ASCTS) patient database.
Methods: The additive and logistic EuroSCORE models were applied to all patients undergoing cardiac surgery at six institutions in the state of Victoria between 1st July 2001 and 4th July 2005 within the ASCTS database who have complete data. The entire cohort and a subgroup of patients undergoing coronary artery bypass grafting (CABG) only were analysed. Observed and predicted mortalities were compared. Model discrimination was tested by determining the area under the receiver operating characteristic (ROC) curve. Model calibration was tested by the Hosmer-Lemeshow chi-square test.
Results: Eight thousand three hundred and thirty-one patients with complete data were analysed. There were significant differences in the prevalence of risk factors between the ASCTS and European cardiac surgical populations. Observed mortality was 3.20% overall and 2.00% for the CABG only group. The EuroSCORE models over estimated mortality (entire cohort: additive predicted 5.31%, logistic predicted 8.76%; CABG only: additive predicted 4.25%, logistic predicted 6.19%). Discriminative power of both models was very good. Area under ROC curve was 0.83 overall and 0.82 for the CABG only group. Calibration of both models was poor as mortality was over predicted at nearly all risk deciles. Hosmer-Lemeshow chi-square test returned P-values less than 0.05.
Conclusions: The additive and logistic EuroSCORE does not accurately predict outcomes in this group of cardiac surgery patients from six Australian institutions. Hence, the use of the EuroSCORE models for risk prediction may not be appropriate in Australia. A model, which accurately predicts outcomes in Australian cardiac surgical patients, is required.