Predicting the physiological effects of multiple drugs using electronic health record

Supplementary Data 1. Optimized hyperparameters of the XGBoost classifiers for the 20 selected measurement items.

Supplementary Data 2. Information on 416 active pharmaceutical ingredients (APIs) used to develop the machine learning models for the 20 selected measurement items.

Supplementary Data 3. List of top 20 features with the highest feature importance for each XGBoost classifier according to the SHAP analysis (Fig. 4).

Supplementary Data 4. AUROC of the XGBoost classifiers for the 20 measurement items by using pairwise combinations of feature types as input, and using race-specific and sex-specific MIMIC-IV datasets.