Fig 1.
Cohort selection, patient observation timelines.
Red marker indicates observed outcome event (overdose).
Table 1.
Summary of all model features.
Table 2.
Summary of results of all classifiers.
Table 3.
Confusion matrix for the logistic regression model, threshold = 0.1.
Fig 2.
Risk stratification for XGB model.
Fig 3.
Top features with an increased risk of overdose (Logistic regression).
Fig 4.
Top features with decreased risk of overdose.
Fig 5.
Feature importance for the XGB classifier.
Table 4.
Summary of results.
Fig 6.
Partial dependence plot for the effect of PDP in XGB model.
Fig 7.
Individual conditional expectation (ICE) plot for the effect of PDP in XGB model.