A ML Approach for MACEs Prediction in Patients with UA and HFpEF - Web-based risk calculator study

Sep 28, 2026

Machine learning models are developed to estimate major adverse cardiovascular events (MACEs) risk among patients with coexisting unstable angina (UA) and heart failure with preserved ejection fraction (HFpEF). A cohort of 2,638 patients (2015-2021) is split into derivation and geographic validation groups, with clinical, laboratory, and imaging variables extracted from electronic medical records. A hybrid LASSO-Boruta feature-selection identifies seven key predictors, and nine ML algorithms are compared; the XGBoost model performs best and is deployed as a web-based risk calculator.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

Background

Objective

Methods

Results

Conclusion

1044 Innovation in Aging, 2025, Vol. 9, No. S2 Abstract citation ID: igaf122.3518 A ML APPROACH FOR MACES PREDICTION IN PATIENTS WITH UA AND HFPEF Yijun Wang, and Jinhui Wu, West ChinaHospital, Sichuan University, Chengdu, Sichuan, China (People’s Republic) Background Heart failure with preserved ejection fraction (HFpEF) and unstable angina (UA) often coexist in clinical practice, constituting a high-risk cardiovascular phenotype with a markedly increased incidence of major adverse cardiovascular events (MACEs). Objective This study aimed to develop models based on machine learning (ML) algorithms to estimate the risk of MACEs in patients with coexisting UA and HFpEF. Methods This cohort study included 2,638 patients with both HFpEF and UA between January 1, 2015, and December 31, 2021. Patients were divided into the derivation cohort (n=1,367) and validation cohort (n = 1,271) based on geographic regions. Clinical, laboratory, and imaging data were extracted from electronic medical records. Key predictors were identified using a hybrid featureselection method combining LASSO and Boruta algorithms. Nine ML models were developed: XGBoost, LGBM, RF, DT, GBDT, KNN, SVM, MLP, and GNB. Model performance was evaluated for the area under the receiver operating characteristic curve, accuracy, sensitivity, specificity, recall, F1-score, precision-recall curve, Brier scores, calibration curves and decision curve analysis. Innovation in Aging, 2025, Vol. 9, No. S2 1045 The best-performing model was deployed as a web-based risk calculator. Results Using a combination of LASSO regression and the Boruta algorithm, seven key predictors were identified: Diabetes Mellitus, TG/HDL-C ratio, SIRI, TyG-BMI index, Creatinine, NT-proBNP, and Gensini score. The XGBoost-based model exhibited superior performance with an AUC of 0.825 in the derivation cohort and an AUC of 0.781 in the validation cohort based on seven features. Conclusion XGBoost-based predictive model constitutes an innovative approach to risk stratification for the complex dual phenotype of HFpEF and UA.

FAQs

What was the objective of this study?

To develop machine learning models to estimate the risk of MACEs in patients with coexisting unstable angina and HFpEF.

How were patients and cohorts organized?

The cohort included 2,638 patients between 2015-01-01 and 2021-12-31, split into a derivation cohort (n=1,367) and a validation cohort (n=1,271) by geographic regions.

Which model and predictors showed the best performance?

The XGBoost-based model performed best using seven predictors identified by LASSO and Boruta: Diabetes Mellitus, TG/HDL-C ratio, SIRI, TyG-BMI index, Creatinine, NT-proBNP, and Gensini score.

Sep 28, 2026

Upload by Stanford