Machine Learning Model Estimates Three-Year Cardiac Risks Following Bypass Surgery

An Extremely Randomized Trees machine learning model provided externally validated risk estimates for three-year major adverse cardiovascular events in adult patients following coronary artery bypass grafting, according to a multicenter retrospective study conducted across eight cardiac centers in Jiangsu Province, China.

Major adverse cardiovascular events represent a residual risk following coronary artery bypass grafting that adversely affects long-term prognosis, making accurate postoperative risk stratification critical for secondary prevention. To address this risk, researchers analyzed clinical data from 3,930 adult patients who underwent coronary artery bypass grafting between January 2019 and December 2022. The investigators gathered 232 preoperative, intraoperative, and early postoperative clinical variables. A primary cohort of 3,018 patients treated at Nanjing First Hospital was divided into training and internal validation cohorts, while data from 912 patients across seven other medical centers formed an independent external validation set.

Predictive features were isolated using minimum redundancy maximum relevance followed by least absolute shrinkage and selection operator regression, identifying seventeen final predictors from the initial set of 232 variables. The authors trained and compared ten distinct machine learning models using the area under the receiver operating characteristic curve, area under the precision-recall curve, accuracy, sensitivity, specificity, F1-score, Brier score, calibration curves, and decision curve analysis. Based on internal validation performance, the Extremely Randomized Trees algorithm was selected as the final model.

Across the entire study population, the observed three-year incidence of major adverse cardiovascular events was 16.41 percent, accounting for 645 of the 3,930 patients. In validation testing, the Extremely Randomized Trees model achieved an area under the receiver operating characteristic curve of 0.90 with a 95 percent confidence interval of 0.87 to 0.93 in internal validation, and 0.82 with a 95 percent confidence interval of 0.78 to 0.86 in the external validation cohort, alongside an external area under the precision-recall curve of 0.59. At a prespecified classification probability threshold of 0.50, the model yielded an external accuracy of 0.87, a sensitivity of 0.54, a specificity of 0.94, and an F1-score of 0.59. The external validation also demonstrated a Brier score of 0.100, a calibration intercept of minus 0.20, and a calibration slope of 0.80.

Model interpretation using SHapley Additive exPlanations revealed that postoperative left ventricular ejection fraction was the most important predictor of three-year outcomes, followed in importance by preoperative neutrophil percentage and preoperative New York Heart Association functional class. While the model was developed for early postoperative risk estimation once required postoperative variables are available, and a web-based calculator was created for further study, the researchers highlighted that sensitivity remained limited at the prespecified decision threshold. Consequently, the authors concluded that prospective clinical evaluation and rigorous threshold validation are required prior to practical clinical adoption.

Subject of Research: Machine learning prediction of 3-year major adverse cardiovascular events after coronary artery bypass grafting

Article Title: Development and validation of a machine learning-based prediction model for 3-year major adverse cardiovascular events in patients undergoing coronary artery bypass grafting: a multicenter study

Article References: Xue, H., Zhang, R., Fan, R., Wang, R., Huang, F., Chen, W., & Chen, X. (2026). Development and validation of a machine learning-based prediction model for 3-year major adverse cardiovascular events in patients undergoing coronary artery bypass grafting: a multicenter study. BMC Medical Informatics and Decision Making. https://doi.org/10.1186/s12911-026-03908-0

Image Credits: AI Generated

DOI: 10.1186/s12911-026-03908-0

Keywords: coronary artery bypass grafting, machine learning, major adverse cardiovascular events, risk prediction, Extremely Randomized Trees, cardiac surgery, postoperative prognosis, retrospective cohort study, Development, validation, machine, learning-based