|
Activity Number:
|
371
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #309119 |
|
Title:
|
Updating Clinical Prediction Models: Simple Methods Give Promising Results
|
|
Author(s):
|
Kristel J.M. Janssen*+ and Karel G. Moons and Yvonne Vergouwe and Diederick E. Grobbee
|
|
Companies:
|
Julius Center UMC Utrecht and Julius Center UMC Utrecht and Julius Center UMC Utrecht and Julius Center UMC Utrecht
|
|
Address:
|
PO BOX 85500, Utrecht, 3508, , Netherlands
|
|
Keywords:
|
Prediction research ; Updating ; Calibration ; Discrimination ; Prediction model
|
|
Abstract:
|
When prediction models show poor performance in new patients, one often simply redevelop new rule ignoring prior data. We compared five methods to update a model with new data. The model was developed in 1944 pts, validated+updated in 752 pts, tested in 283 pts. In updating method 1: only intercept adjusted; 2: intercept adjusted+regression coefficients multiplied with calibration slope; 3: tested if predictors were different; 4: re-estimated intercept+regression coefficients in validation set; 5: same as 4 in combined derivation+validation set. Calibration original model in validation set substantially improved by all methods. Discrimination was not improved with method 1+2. In test set all updated rules had similar discrimination. Simple recalibration methods improved model calibration similar as extensive revision methods. Discrimination was similar across all updating methods.
|
