STEMI-OP in-hospital mortality prediction algorithms: Frailty-integrated machine learning in older patients...

Area of Science:

• Cardiology
• Artificial Intelligence
• Geriatrics

Background:

• Older patients with ST-elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention (PCI) have high mortality rates.
• Traditional prognostic models may lack accuracy in Southeast Asian populations, particularly for elderly individuals.

Purpose of the Study:

• To develop and validate artificial intelligence (AI)-based risk assessment models, named STEMI-OP algorithms, for elderly STEMI patients (≥60 years) undergoing primary PCI in Vietnam.
• To identify key predictors of in-hospital mortality in this specific demographic using advanced machine learning techniques.

Main Methods:

• Machine learning (ML) models were developed and validated using pre- and post-PCI features, incorporating advanced feature selection.
• SHapley Additive exPlanations and Causal Random Forests were used for model interpretability and identifying causal relationships.
• Key predictors identified included Killip classification, Clinical Frailty Scale, glucose, and creatinine levels.

Main Results:

• The CatBoost model (pre-PCI) and Random Forest model (post-PCI) demonstrated superior performance.
• Achieved AUC values of 92.16% (pre-PCI) and 95.10% (post-PCI), significantly outperforming GRACE 2.0 (83.48%) and CADILLAC (87.01%) scores.
• The models successfully incorporated frailty assessment for more precise risk stratification.

Conclusions:

• The AI-based STEMI-OP algorithms provide more accurate and personalized risk assessments for older STEMI patients undergoing primary PCI.
• These algorithms can enhance clinical decision-making and potentially improve outcomes for this vulnerable patient group.
• The study highlights the importance of AI in addressing limitations of traditional models in diverse populations.