Retrospective study shows machine learning significantly improves short-term mortality predictions in CABG patients, suggesting enhanced risk assessment over EuroSCORE II.
Key Points
Machine learning models incorporating preoperative and postoperative data outperform EuroSCORE II in predicting short-term mortality.
The 30-day mortality rate after CABG was 2.5% among 3,483 patients included in the analysis.
Specificity improved significantly for random forest and neural networks with additional preoperative variables.
The most influential postoperative predictors of mortality included kidney failure, pulmonary complications, and myocardial infarction.