Fig. 1

A Data preparation. Data were selected based on number of t2t courses. Variables were selected if the missing rate did not exceed 33%. B Machine learning pipeline: Data was labeled, depending on the outcome of the therapy course. Iterative imputation was applied, on the hold-out-set (test-set) and on the training set. Sampling strategies were applied, and the AUC (area under the curve) was collected for each model configuration. The final, re-trained model was explained via applying SHAP (SHapley Additive exPlanations)