Using restriction to minimize bias in observational studies.

Overview

Title:

Using restriction to minimize bias in observational studies.

Authors:

Turakhia MP, Heidenreich PA

Journal:

JAMA

Publication date:

2010

Volume:

304

Issue:

21

First page:

2359

Last page:

60

ISSN:

1538-3598

Link to pubmed:

http://www.ncbi.nlm.nih.gov/pubmed/21119082

Publication type:

Journal

Free text

In their Commentary, Drs Psaty and Siscovick ¹ emphasized the importance of restriction to minimize confounding by indication and bias in the conduct of observational comparative effectiveness research. We agree with this strategy for less complicated therapies, such as β-blockers for hypertension, for which the dose-adjusted efficacy would not likely be affected by clinician-level, facility-level, or other nonpatient factors. However, we caution against the broad use of restriction to evaluate therapies or services for which quality of care delivery would impact effectiveness.

For example, there is controversy as to whether aspirin or warfarin is more effective for stroke prevention in low-risk patients with atrial fibrillation because the benefits of warfarin, which requires regular blood testing to monitor anticoagulation, may be offset by the potential harms of major bleeding. ^{2, 3} This evidence gap could be evaluated with electronic health record data in hundreds of thousands of …

Preview

Using restriction to minimize bias in observational studies. Turakhia MP, Heidenreich PA. JAMA 2010; 304(21):2359-60.