Abstract
Background Acute coronary syndrome (ACS) is a leading cause of morbidity and mortality worldwide. Despite advances in management, patients with ACS remain at a significant risk of developing ischemic stroke (IS), a serious complication associated with high mortality and long-term disability. The accurate prediction of stroke risk in ACS patients can facilitate timely interventions and improve clinical outcomes. Objective This study aimed to develop and validate machine learning (ML) models to predict ischemic stroke within one year of ACS diagnosis, using electronic medical records (EMRs) from a tertiary care hospital in Indonesia. Methods We conducted a retrospective cohort study using data from 4,789 ACS patients treated at Dr. Sardjito Hospital between 2018 and 2022. Machine learning models, including Logistic Regression, Random Forest, and XGBoost, were trained and validated using patient demographics, comorbidities, and clinical variables. Model performance was assessed using precision, accuracy, sensitivity, specificity, and area under the curve (AUC)-receiver operating characteristic (ROC). Results Among the study cohort, 212 patients (4.4%) developed ischemic stroke within one year. Logistic Regression demonstrated a balanced performance with a sensitivity of 65%, a specificity of 70%, and an AUC-ROC of 0.70. Random Forest and XGBoost models achieved higher sensitivities (94% and 95%, respectively) but had lower specificities (12% each). The most significant predictors of ischemic stroke included ST-segment elevation myocardial infarction (STEMI), age of ≥60 years, atrial fibrillation, hypertension, and chronic kidney disease. Conclusion The Logistic Regression model, with its balanced sensitivity and specificity, offers a reliable tool for predicting ischemic stroke in ACS patients. The implementation of this model in clinical practice could enhance risk stratification and inform personalized treatment strategies. Future studies should focus on prospective validation and the integration of additional clinical variables.