Skip to main content

Table 3 Number of parameters re-estimated for each update under each update strategy. For the closed testing procedure, the total number of parameters re-estimated equals the sum of the number of parameters re-estimated for each candidate update model; the actual number of parameters that change depend on which particular candidate update model is selected

From: Comparison of dynamic updating strategies for clinical prediction models

Update strategy

Number of parameters re-estimated

Never update

0

Closed testing procedure

7.2*

Always recalibrate intercept

1

Always recalibrate intercept and slope

2

Always refit (revision)

15

  1. *7.2 represents the weighted average of the number of parameters re-estimated under the closed testing pProcedure when the update interval equals 1 quarter. More specifically, the number of parameters re-estimated under recalibrate intercept, recalibrate slope, and refit are 1, 2, and 15, respectively. Multiplying these numbers by the total number of times the closed testing procedure selected each candidate update model (i.e., 3, 4, and 5, respectively; see Tables 4 and 5) and dividing by the total number of updates performed (i.e., 12) yields: \( \frac{\left(1\ast 3\right)+\left(2\ast 4\right)+\left(15\ast 5\right)}{12}=7.2 \)