Cardiopulmonary bypass parameters improve the prediction of 30-day mortality following cardiac surgery

Abstract

Currently 30-day mortality is commonly used as a quality indicator for cardiac surgery; however, prediction models have not included the role of cardiopulmonary bypass (CPB). We hypothesized that reproducing currently utilised prediction model methods of 30-day mortality using the Australian and New Zealand Collaborative Perfusion Registry (ANZCPR) would identify relevant CPB predictors. Nine centers in Australia and New Zealand collected data using the ANZCPR between 2011–2020. CPB parameter selection was determined by evaluating association with 30-day mortality. Data were divided into model creation ( n = 15,073) and validation sets ( n = 15,072). Bootstrap sampling and automated variable selection methods were used to develop candidate models. The final model was selected using prediction mean square error and Bayesian Information Criteria. The average receiver operating characteristic (ROC), p-value for Hosmer—Lemeshow chi-squared test and MSE were obtained from multifold validation. In total, 30,145 patients were included, of which 735 (2.4%) died within 30 day of surgery. The area under the ROC curve for the model including CPB parameters was significantly greater than preoperative risk factors only (0.829 vs 0.783, p < 0.001). CPB parameters included in the predictive model were CPB time, red blood cell transfusion, mean arterial pressure <50 mmHg, minimum oxygen delivery, cardiac index <1.6 L/min/m2. CPB parameters improve the prediction of 30-day mortality. Randomised trials designed to evaluate modifiable CPB parameters will determine their impact on mortality.