Evaluation of the degree to which the logistic regression model and random forest classifier model can predict inhibitor development. Model discrimination (ie, the degree to which a model assigns a higher risk to an inhibitor-positive patient vs an inhibitor-negative patient) was assessed by plotting ROC curves and by calculating the area under the curve (AUC). The AUC varies between 0.5 (no discrimination) to 1 (perfect discrimination). (A-B) The ROC curves of the LASSO logistic regression model and the random forest model are shown in (A) and (B), respectively. Model calibration (ie, the degree to which the predicted cumulative incidence of inhibitor development matched the observed cumulative incidence) was assessed using a calibration plot. For each quintile of predicted cumulative incidence, we plotted the mean predicted cumulative incidence of inhibitor development in a group against the observed cumulative incidence of inhibitor development in that group. In addition, we plotted a LOESS (locally estimated scatterplot smoothing) line in the same figure to assess model calibration across the full risk range. Ideally, all points should lie exactly on the diagonal line (which represents perfect agreement between predicted and observed values). (C-D) The calibrations plots of the LASSO logistic regression model and the random forest model are shown in (C) and (D), respectively.