Comparative Performance of Machine Learning Models and Logistic Regression for Acute Myocardial Infarction Risk...

Introduction The integration of Artificial Intelligence in healthcare systems offers transformative potential for enhancing healthcare administration and improving insurance risk modeling through more accurate patient stratification. This study evaluates the predictive performance of support vector machines (SVM), gradient boosting (GBM), and random forest (RF) against multivariate logistic regression for Acute Myocardial Infarction (AMI) risk stratification, focusing on its implications for healthcare administration and insurance modeling. Methods A retrospective observational study was conducted on 901 patients at Alexandra Hospital of Athens (January 2021 to December 2023). Inclusion was restricted to first-incident AMI cases, verified through clinical and digital registries; patients with a prior history of AMI were strictly excluded. Predictor variables included standardized clinical data, lifestyle factors, and COVID-19 vaccination status, all verified prior to the AMI event. Models were trained and validated using a 70/30 hold-out split (n=631 training; n=270 testing), a strategy selected to provide a stable testing set for benchmark comparison. Performance was evaluated based on accuracy, sensitivity, and area under the curve (AUC) to assess how predictive precision can enhance organizational risk assessment and resource allocation. Results Among the predictive models, SVM achieved the highest overall accuracy (62.08%), while GBM demonstrated a modest discriminative capacity with an AUC of 0.63 and a sensitivity of 0.76. Crucially, within this specific clinical cohort, COVID-19 vaccination status showed no statistically significant association with AMI risk (OR 1.12, p=0.41). These results represent an exploratory comparison of algorithmic performance for risk prediction rather than a finalized tool for robust clinical stratification. Conclusion While the accuracy gain over logistic regression is incremental, SVM and GBM models provide a more refined classification for high-risk cohorts. These findings suggest that AI-based stratification can optimize risk management frameworks and actuarial modeling, effectively bridging clinical cardiology with insurance science.