AI-Driven Prediction of Multiple Outcomes in Older Adults with Coronary Heart Disease.

BACKGROUND AND OBJECTIVES: Frail older adults with coronary heart disease (CHD) face significantly elevated risks of adverse clinical outcomes, including mortality, prolonged hospitalizations, and frequent readmissions. Conventional risk stratification tools, inadequately account for frailty and multi-morbidity, limiting their effectiveness in geriatric care. To address this gap, we developed and validated the first machine learning (ML) model that integrates frailty into a multi-outcome risk assessment framework, thereby enhancing clinical decision-making in geriatric cardiology. RESEARCH DESIGN AND METHODS: Utilizing electronic health records from hospitalized frail CHD patients, we developed a multinomial prediction model employing advanced ML techniques, including principal component analysis, gradient boosting, and random forest. The model incorporates explainable AI features to enhance interpretability and a clinical applicability, prioritizing key predictors such as biomarkers and comorbidities. RESULTS: The ML model demonstrated superior predictive performance with Receiver Operating Characteristic (ROC) curve analysis (Area Under the Curve [AUC] 0.94, 95% CI 0.88-1.00) for mortality, 0.72 (95% CI: 0.55-0.87) readmission and 0.68 (95% CI: 0.57-0.77) prolonged hospital stay, enabling earlier risk identification and personalized intervention strategies. DISCUSSION AND IMPLICATIONS: This AI-driven approach represents a significant advancement in geriatric cardiology designed for integration into hospital dashboards, providing real-time patient-centred decision support, optimization of clinical workflow and resource allocation. By advancing digital health solutions and AI driven precision medicine this model sets a new standard for digital health innovations in aging care.