30-day unplanned readmission or death classification metrics for evaluated models: XGBoost, Sparse logistic regression (LR), Bayesian neural network (BNN), our Piecewise exponential model (PEM).

Quantization refers to the histogram-based bucketization of real-valued features. Area under the receiver operator curve (AUROC) and area under the precision-recall curve (AUPRC) computed on held-out 2012 inpatient episodes. Models trained on 2009–2011 episodes. Interpretability judged according to Fig 1.