|
Activity Number:
|
76
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304573 |
|
Title:
|
Bayesian Prescribing Model for Cox-2 Selective NSAIDS
|
|
Author(s):
|
Margaret R. Stedman*+ and M. Alan Brookhart
|
|
Companies:
|
Brigham and Women's Hospital and Brigham and Women's Hospital/Harvard Medical School
|
|
Address:
|
, , ,
|
|
Keywords:
|
Bayesian models ; Hierarchical models ; Pharmacoepidemiology ; Random effects ; Winbugs
|
|
Abstract:
|
We propose a Bayesian approach to study physician prescribing patterns. This involves a model of treatment choice with a random intercept to capture underlying physician preference and random coefficients that allow each physician to respond differently to specific patient characteristics. The model is demonstrated in a study of the factors that influence NSAID drug choice in a US Medicare population. Estimation was accomplished by MCMC methods implemented in WinBUGS. We found that physician prescribing can be explained most strongly by underlying physician preference. Patient characteristics capturing GI risk were also predictive of treatment; however, physicians varied little in how they responded to these variables. Results may be used to target prescription drug quality improvement interventions or develop treatment selection models needed for outcomes research.
|
