|
Activity Number:
|
397
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #307747 |
|
Title:
|
Causal Inference of Nonrandomized Treatment Effects
|
|
Author(s):
|
Heejung Bang*+ and Yolanda Barrón and Greg Ridgeway
|
|
Companies:
|
Cornell University and Cornell University and RAND
|
|
Address:
|
411 E. 69th St., New York, NY, 10021,
|
|
Keywords:
|
non-randomized treatment ; causal inference ; marginal structural model ; double-robustness
|
|
Abstract:
|
Randomized clinical trials are the gold standard used to evaluate the "causal" effects of treatments in biomedical research. However, it is not feasible to conduct a clinical trial to answer every clinical question. Nonrandomized treatments can not serve as a valid alternative, but often appear in various contexts within observational studies or even as secondary interventions within randomized clinical trials. We use the causal models (e.g., marginal structural model) in cross-sectional and longitudinal settings for the analysis of a non-randomized intervention and further extend the models for doubly robust estimation. The proposed methods are applied to study the effect of non-randomized antidepressant use by depression suffering patients enrolled in a cardiovascular clinical trial, on myocardial infarction and death.
|
