Fig. 2

The importance of the 15 most important predictors chosen by each model. Derived from 100% of training samples (n = 31,166). Importance is scaled relative to the most important predictor within each model based on model-specific metrics (LR, z-value; C5.0, tree split usage; RF, Gini importance; SVM, univariate AUROC). Prefixes: Previous = before the first MI; Intake = at hospital/lab arrival; CCU = during the Coronary Care Unit stay; Discharge = at discharge from hospital. Unspecified prefix signifies either a fixed predictor or that the predictor was register at some time-point before hospital discharge. C5.0, Boosted C5.0; LR, Logistic regression; RF, Random Forest; SVM, Support Vector Machine; ACE, Angiotensin-converting-enzyme; ECG, Electrocardiogram; PCI, Percutaneous coronary intervention