Preoperative Risk Prediction Model for 30-day Mortality Following Cardiac Surgery in Australia

Background: Estimating preoperative risk prior to cardiothoracic surgery is of recognized value for both surgeons and patients. Currently, there is no model for predicting risk of mortality for the Australian cardiac surgery population and this paper presents a new risk prediction model suitable for use in the Australian context. Methods: All patients undergoing cardiac surgery between July 2001 and June 2007 were obtained from the Australian Society of Cardiac and Thoracic Surgeons (ASCTS) database and were included for analysis. Bootstrap sampling and automated variable selection procedures were used to identify risk factors for 30- day mortality. The receiver operating curve (ROC) and Hosmer–Lemeshow p-value were then used to assess the predictive ability of the model. Results: Over the 6-year period, 18,159 patients had cardiac surgery. The 30-day mortality rate was 3.26% (592/18,159). The following 12 variables were selected in the predictive model as independent predictors: urgency of procedures, procedure types, age, gender, previous cardiac surgery, New York Heart Association (NYHA) class, use of inotropes, ejection fraction estimate, hypercholesterolaemia, creatinine level >0.20 mmol/L, peripheral vascular disease and body mass index (BMI). For the validation data, the p-value for the Hosmer–Lemeshow Chi-squared test for goodness-of-fit is 0.2753 and the area under the ROC is 82.40%—this indicates that the model has a high predictive accuracy for the Australian patients. Conclusion: A simple preoperative risk prediction model for 30-day mortality has been developed for theAustralian cardiac surgery population.