The Random Forest machine learning model demonstrated superior predictive performance for 30-day major adverse cardiovascular events after heart valve replacement compared to the traditional EuroSCORE II, achieving a validation AUC of 0.823 versus 0.723 (p=0.028).
Cohort (n=346)
No
Does a machine learning-based prediction model improve the prediction of in-hospital MACEs after heart valve replacement compared to traditional scoring systems?
A Random Forest machine learning model outperformed traditional EuroSCORE II in predicting in-hospital major adverse cardiovascular events after heart valve replacement.
Effect estimate: AUC 0.823 (95% CI 0.715-0.930)
Absolute Event Rate: 0.823% vs 0.723%
p-value: p=0.028
Background:Current risk assessment tools for predicting in-hospital major adverse cardiovascular events (MACEs) after heart valve replacement (HVR) have notable limitations. To address this gap, this study aimed to develop and validate a machine learning (ML) model for predicting such events.Methods:A total of 346 patients who underwent HVR were retrospectively included and divided into a training set (n = 242) and a validation set (n = 104). Patients who experienced in-hospital MACEs were classified as having the complication. In the training set, prognostic indicators were screened using univariate analysis, least absolute shrinkage and selection operator (LASSO) regression, and multivariate logistic regression. Prediction models were constructed using random forest (RF), K-Nearest Neighbors (K Model), and gradient boosting (GB). Model performance was evaluated using the area under the receiver operating characteristic (AUC) curve, calibration curves, and decision curve analysis, and the optimal model was selected. Model interpretability was assessed using SHapley Additive exPlanations (SHAP) values.Results:No statistically significant differences were observed in baseline characteristics between the training and validation sets (p > 0.05). Multivariate logistic regression identified age, European System for Cardiac Operative Risk Evaluation II (EuroSCORE II), cardiopulmonary bypass time, aortic cross-clamp time, left ventricular ejection fraction, and serum albumin as independent predictors of MACEs (all p < 0.05). The RF model demonstrated the highest predictive performance, with AUC values of 0.847 in the training set and 0.823 in the validation set. The RF model achieved a validation AUC of 0.823, which was significantly superior to that of the K model (0.790), the GB model (0.771), and the traditional EuroSCORE II (0.723) (all p < 0.05), establishing the RF model as the optimal predictive approach.Conclusion:This study developed and validated a machine-learning model to predict MACEs after HVR. The RF model showed favorable predictive performance compared with traditional scoring systems. The RF model may serve as a clinical decision-support tool to help identify high-risk patients before surgery, potentially aiding in resource allocation and individualized intervention.
Zhang et al. (2026) conducted a cohort in Heart Valve Replacement (n=346). Random Forest prediction model vs. EuroSCORE II was evaluated on Prediction of 30-day major adverse cardiovascular events (MACE) (AUC 0.823, 95% CI 0.715-0.930, p=0.028). The Random Forest machine learning model demonstrated superior predictive performance for 30-day major adverse cardiovascular events after heart valve replacement compared to the traditional EuroSCORE II, achieving a validation AUC of 0.823 versus 0.723 (p=0.028).