Fig. 3
From: Accounting for treatment use when validating a prognostic model: a simulation study
Calibration curves calculated in a treated validation set, following different approaches to account for the effects of treatment. Scenario 1: P(treatment) increases with risk, fixed treatment effect; scenario 2: randomized treatment, fixed treatment effect; scenario 3: P(treatment) increases with risk, treatment effect increases with risk; scenario 4: 18% baseline risk threshold for treatment, fixed treatment effect. Plots were based on sets of 1 million individuals