A Machine Learning Approach to Predictive Modeling of Cardiovascular Events

Acute coronary syndromes (ACS) are a leading cause of morbidity and mortality, emphasizing the need for effective prediction of major adverse cardiovascular events (MACE) to improve patient outcomes, optimize resource allocation, and reduce healthcare costs. This study develops a Random Forest (RF) model to predict MACE within 30-day, 1-year, 2-year, and 3-year using data from 2,721 ACS patients admitted to the Heart Hospital in Qatar between 2018 and 2024. Variables, including demographics, medical history, and clinical parameters, were assessed, with text entries processed using natural language processing. The cumulative MACE prevalence was 22.3%, 34.5%, 45.3%, and 58.1% within 30 days, 1 year, 2 years, and 3 years, respectively. Patients with MACE consistently exhibited higher age, lower left ventricular ejection fraction, and elevated troponin and creatinine levels compared to non-MACE patients (p < 0.05). The RF model achieved Area Under the Receiver Operating Characteristic Curve (AUC) values ranging from 0.817 to 0.865, with sensitivity improving from 75.4% at 30 days to 87.6% at 3 years. This study prevents data leakage by carefully defining the model training steps and avoids excessive variable inclusion, providing a more reliable estimation of MACE risk. The model holds promise for enhancing patient outcomes, optimizing resource allocation, and generating cost savings through informed clinical decisions and timely interventions.