Development and multicenter validation of a machine learning model for predicting in-hospital postoperative cardiac arrest in elderly surgical patients: the critical role of intraoperative hemodynamic disturbances and blood transfusion

BackgroundIn-hospital cardiac arrest (IHCA) is a rare but life-threatening complication in elderly surgical patients. This study aimed to develop and validate a machine learning model for predicting IHCA and to identify key perioperative risk factors.MethodsA multicenter retrospective cohort study was conducted among patients aged ≥65 years who underwent non-cardiac surgery at six tertiary hospitals between January 2020 and January 2025. Clinical, laboratory, and perioperative variables were collected. Five machine learning algorithms, including Extreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM), Multilayer Perceptron (MLP), and K-Nearest Neighbor (KNN), were developed and compared. Model performance was evaluated using receiver operating characteristic (ROC) analysis, calibration, decision curve analysis, cross-validation, and external validation. SHAP analysis was applied to interpret the optimal model.ResultsAmong the five models, XGBoost demonstrated the best predictive performance, with AUCs of 0.967 and 0.960 in the training and validation cohorts, respectively. It also achieved the lowest Brier score and the highest clinical net benefit. External validation confirmed excellent discrimination (AUC=0.971) and calibration. SHAP analysis identified advanced age, intraoperative tachycardia, intraoperative hypotension, intraoperative blood transfusion, intraoperative hypoxemia, elevated postoperative neutrophil-to-lymphocyte ratio, B-type natriuretic peptide, lactate, and D-dimer levels as the most important predictors of IHCA.ConclusionAn interpretable XGBoost-based model demonstrated excellent performance for predicting postoperative cardiac arrest in elderly patients. The identified risk factors reflected perioperative vulnerability, intraoperative hemodynamic instability, and postoperative physiological deterioration.