Multimodal Data-Driven Explainable Prognostic Model for Major Adverse Cardiovascular Events Prediction in Patients...

Area of Science:

• Cardiovascular Medicine
• Machine Learning in Healthcare
• Predictive Analytics

Background:

• Heart failure with preserved ejection fraction (HFpEF) and unstable angina (UA) often coexist, creating a high-risk cardiovascular phenotype.
• This dual condition significantly increases the incidence of major adverse cardiovascular events (MACEs).
• Identifying high-risk patients is crucial for improving outcomes and guiding clinical decisions.

Purpose of the Study:

• To develop and externally validate machine learning predictive models for MACEs in patients with coexisting UA and HFpEF.
• To create an online risk calculator for individualized prevention strategies.
• To enhance clinical decision-making for this complex patient population.

Main Methods:

• A multicenter cohort study of 4459 patients with both HFpEF and UA.
• Utilized a hybrid feature selection method (LASSO and Boruta algorithms) to identify key predictors.
• Developed and validated 33 survival models, selecting the best-performing surv.xgboost.cox model for MACE prediction.

Main Results:

• Identified 7 key predictors: diabetes mellitus, platelet count, triglyceride, systemic inflammatory response index, triglyceride-glucose-BMI, NT-proBNP, and atherogenic index of plasma.
• The surv.xgboost.cox model achieved a C-index of 0.788 in the external validation cohort, demonstrating good predictive performance.
• The model showed satisfactory calibration and clinical utility for predicting 40-month MACEs, with AUC values ranging from 0.784 to 0.81.

Conclusions:

• A surv.xgboost.cox-based predictive model for MACEs was successfully developed for patients with concurrent HFpEF and UA.
• The model was implemented as a user-friendly web-based calculator to support clinical practice.
• This tool facilitates individualized risk assessment and prevention strategies for high-risk cardiovascular patients.