Machine learning predicts major adverse cardiovascular events following modified Morrow surgery for hypertrophic obstructive cardiomyopathy

Background: Although modified Morrow surgery relieves left ventricular outflow tract obstruction, patients remain at risk of postoperative cardiovascular events. This study aimed to develop and validate an interpretable machine learning (ML) model for predicting 24-month major adverse cardiovascular events [MACEs; comprising cardiovascular death, non-fatal myocardial infarction (MI), non-fatal stroke, and heart failure hospitalization (HFH)] following modified Morrow surgery in patients with hypertrophic obstructive cardiomyopathy (HOCM). Methods: We retrospectively studied 354 adult patients with HOCM who underwent modified Morrow surgery at Beijing Anzhen Hospital (approval No. KS2025079) between July 2015 and July 2024. We developed an ensemble ML model integrating recursive feature elimination (RFE) with a voting classifier [support vector machine (SVM), random forest, decision tree, extreme gradient boosting (XGBoost)]. The model was trained and tuned using five-fold cross-validation, with performance rigorously compared against conventional logistic regression (LR) and least absolute shrinkage and selection operator (LASSO) regression. Model interpretability was achieved through SHapley Additive exPlanations (SHAP) analysis. Results: The ML model demonstrated robust performance in five-fold cross-validation and superior predictive capability on the independent test set compared to conventional LR and LASSO regression. On the test set, the ML model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.88 [95% confidence interval (CI): 0.82–0.94], significantly outperforming LR (AUC: 0.70) and LASSO (AUC: 0.71). It also attained the highest accuracy (0.82 vs. 0.62 and 0.68), highest balanced accuracy (0.74 vs. 0.52 and 0.63), and highest F1-score (0.83 vs. 0.66 and 0.68), indicating excellent discrimination and classification performance. SHAP analysis identified a positive family history of hypertrophic cardiomyopathy (HCM) as the most influential predictor, highlighting the critical role of familial risk phenotype in postoperative prognosis. Conclusions: This study developed and internally validated HOCM-ML Risk, an interpretable ML-based tool for stratifying the risk of MACEs after modified Morrow surgery. By identifying key prognostic predictors such as familial history and modifiable clinical factors, the model provides a data-driven framework to support more personalized postoperative assessment.