Why the study?
Traditional STEMI risk scores developed in Western populations may have suboptimal performance in Asian patients, and predictive models often neglect model explainability and probability calibration.
Do explainable machine learning models improve in-hospital mortality prediction compared to the TIMI risk score in Asian STEMI patients?
Population
49,574 Asian STEMI patients in the Malaysian National Cardiovascular Disease registry
Comparison
Machine learning models vs TIMI risk score
Design
Retrospective cohort study
Key result
A calibrated logistic regression model outperformed the TIMI score for in-hospital mortality prediction in Asian STEMI patients (AUC 0.8884; 95% CI 0.8756-0.9011).
Authors
Loading...
Calibrated ML models outperform TIMI for Asian STEMI in-hospital mortality prediction; extends explainable, calibrated tools for risk stratification in underrepresented populations.
Cohort (n=49,574)
Yes
Do explainable machine learning models improve in-hospital mortality prediction compared to the TIMI risk score in Asian STEMI patients?
Effect estimate: AUC 0.8884 (95% CI 0.8756-0.9011)
A calibrated logistic regression model with SHAP-based explainability significantly outperformed the traditional TIMI score for predicting in-hospital mortality in Asian STEMI patients.
Kasim et al. (2026) conducted a cohort in ST-segment elevation myocardial infarction (STEMI) (n=49,574). Calibrated logistic regression model vs. TIMI risk score was evaluated on In-hospital mortality prediction (AUC 0.8884, 95% CI 0.8756-0.9011). A calibrated logistic regression model outperformed the TIMI score for in-hospital mortality prediction in Asian STEMI patients (AUC 0.8884; 95% CI 0.8756-0.9011).