|
Activity Number:
|
173
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #307836 |
|
Title:
|
Propensity Score Calibration with Validation Data
|
|
Author(s):
|
Til Stürmer*+ and Sebastian Schneeweiss and Kenneth J. Rothman and Jerry Avorn and Robert Glynn
|
|
Companies:
|
Brigham and Women's Hospital and Brigham and Women's Hospital and Brigham and Women's Hospital and Brigham and Women's Hospital and Brigham and Women's Hospital
|
|
Address:
|
Harvard Medical School, Boston, MA, 02120,
|
|
Keywords:
|
bias ; confounding ; research design ; propensity score ; regression calibration ; validation study
|
|
Abstract:
|
Propensity scores (PS) can adjust for selection bias caused by observed risk factors for the outcome. PS calibration (PSC) can adjust for selection bias by additional risk factors based on the joint distribution of these risk factors in a validation study. We assessed the performance of PSC simulating large cohorts in which information on a risk factor is only available for a small proportion of observations. When the PS estimated based on main study risk factors (error-prone) is independent of the outcome given the PS including additional risk factors from the validation study (gold standard), PSC leads to a bias reduction between 73% and 110%. Otherwise, PSC can lead to an increase of bias. PSC is valid when the direction of confounding due to selection bias by the additional risk factors in the validation study is the same as the one based on risk factors available in the main study.
|
