Skip to main content
. 2019 Aug 9;26(12):1448–1457. doi: 10.1093/jamia/ocz127

Table 3.

General patterns of our testing procedure’s updating recommendations

Situation Most common recommendation
Changes in outcome prevalence, similar volumes of training and updating data Intercept correction
Changes in case mix, similar volumes of training and updating data No updating or recalibration
Changes in predictor-outcome associations, regardless of sample sizes Refit the model
Updating set substantially larger than training set Refit the model