Classification of atherothrombotic events in myocardial infarctions survivors with supervised machine learning using data from an electronic health record system

The aim was to build a prediction model for subsequent atherothrombotic events for patients who survived a myocardial infarction. The dataset contained 7,582 patients from a national Electronic Health Record. The prediction is a binary outcome (event and no event) in a period of five years after a myocardial infarction. Different classifiers were tested and XGBoost achieved the best F1-score=0.76. Top features are: imd_score, age_at_entry, egfr_ckdepi_base, height, and SBP_base