Figure 2.
The AUROC/AUPRC/E-statistic performance estimates for five repetitions per design per prediction task. The columns represent the prediction task, with the number representing the number of patients with the outcome during the time at-risk. For example, the first column corresponds to a prediction task where 174 patients had the outcome, whereas the last column corresponds to a prediction task where 20 806 patients had the outcome. The rows correspond to whether CV was used by the design (top row does not use CV) or the number of folds (3, 5 or 10). The internal validation performances of the designs that used a test set are coloured in red, and those not using a test set are blue (dots with vertical lines indicating the 95% confidence interval). The external validation performances for a model are the light grey pointers (MDCD) and black crosses (MDCR) that have the same x-coordinate and fall within the same row/column. AUPRC, area under the precision recall curve; AUROC, area under the receiver operating curve; CV, cross-validation; MDCD, Multi-state Medicaid Database; MDCR, Medicare Supplemental Database.