Machine learning-based prediction of three-year mortality in elderly inpatients with coronary artery disease combined with heart failure.

OBJECTIVE: Accurate prediction of survival outcome is essential for early intervention and treatment optimization. This study aimed to develop a model utilizing machine learning techniques for predicting three-year mortality in elderly inpatients with coronary artery disease (CAD) combined with heart failure (HF). METHODS: This study enrolled 987 elderly inpatients with CAD. This cohort was randomly divided into the training and validation datasets in a 7:3 ratio. Five machine learning methods, including Logistic Regression, Random Forest, Support Vector Machine, eXtreme Gradient Boosting, and Gradient Boosting Decision Trees, were implemented to construct predictive models. RESULTS: Overall, the median age of this cohort was 85 [81,89] years. Three-year mortality in elderly inpatients with CAD combined with HF was 56.46%. The least absolute shrinkage and selection operator method and five-fold cross-validation identified that ten features were significantly associated with three-year mortality. Logistic Regression showed better performance than other models in the Brier Score, Area Under The Curve, Accuracy, Precision, Recall, and F1 Score of 0.1105, 0.9014, 0.8764, 0.9167, 0.8627, and 0.8889, respectively. The Shapley Additive exPlanations method revealed that age, interventricular septum thickness, gamma gap, serum creatinine, N-terminal pro-B-type natriuretic peptide (NT.proBNP), and neutrophil-to-lymphocyte ratio were identified as risk factors, and mean systolic blood pressure, hemoglobin, albumin, and sodium were protective factors. Age, albumin, and NT.proBNP were three features most associated with three-year mortality. The network application could be available at https://cad-hf-predict.tracebook.org.cn. CONCLUSION: Logistic Regression exhibits excellent predictive performance for predicting three-year mortality in elderly inpatients with CAD combined with HF.