The XGBoost machine learning model accurately predicted in-hospital mortality in patients with coronary heart disease and renal insufficiency, achieving an AUC of 0.926.
Observational (n=17,813)
No
Can an XGBoost-based machine learning model accurately predict in-hospital mortality in patients with coronary heart disease and renal insufficiency?
An XGBoost machine learning model using five routine clinical variables accurately predicts in-hospital mortality in patients with CHD and renal insufficiency.
Effect estimate: AUC 0.926 (95% CI 0.9034-0.9464)
Absolute Event Rate: 0.926% vs 0.888%
Background: Coronary Heart Disease (CHD) with renal insufficiency is a significant global health issue. This study aimed to develop and validate a predictive model for in-hospital mortality to enable early risk identification in these patients. Methods: We analyzed data from 11,830 CHD patients with renal insufficiency treated at Tongji Hospital, Huazhong University of Science and Technology, Wuhan, Hubei Province, China (1994-2023). Among 113 clinical variables, five key features-age, high-sensitivity C-reactive protein (hs-CRP), estimated glomerular filtration rate (eGFR), creatine kinase (CK), and blood urea-were selected using Recursive Feature Elimination. Six machine learning models (Random Forest, XGBoost, Decision Tree, Neural Network, Logistic Regression, and Support Vector Machine) were developed and assessed for discrimination, calibration, and clinical utility. Temporal validation was performed using data from May 16, 2023 to October 31, 2024. SHapley Additive exPlanations (SHAP) were used for model interpretation. Results: Of the 11,830 patients, 694 (5.9 %) died during hospitalization. Among the six models, XGBoost showed the best overall performance in the test set, achieving the highest AUC (0.926), lowest Brier score (0.034), highest accuracy (0.957), and balanced sensitivity (0.381) and F1 score (0.512). Decision curve analysis confirmed its superior clinical utility. In a temporally independent validation cohort of 5983 patients, XGBoost maintained strong predictive performance (AUC = 0.901), demonstrating excellent robustness and generalizability. Conclusions: The XGBoost-based model accurately predicts in-hospital mortality in CHD patients with renal insufficiency, supporting early risk stratification and clinical decision-making.
Li et al. (2025) conducted an observational in Coronary heart disease and renal insufficiency (n=17,813). XGBoost predictive model vs. Logistic Regression (and other machine learning models) was evaluated on Area under the curve (AUC) for predicting in-hospital mortality (AUC 0.926, 95% CI 0.9034-0.9464). The XGBoost machine learning model accurately predicted in-hospital mortality in patients with coronary heart disease and renal insufficiency, achieving an AUC of 0.926.