Sensitivity to Unmeasured Confounding - Guidance for Performing Simple 'Rule Out' Analyses
1,330 views · Published 30 November 2015 · 51:40 · Indexed 28 October 2026
Channel: Vanderbilt University · 2015 · Education
Presented by: Robert Alan Greevy, Jr, PhD Associate Professor of Biostatistics Director, Health Services Research Biostatistics Vanderbilt University Medical Center The strength of evidence provided by epidemiological and observational studies is inherently limited by the potential of unmeasured confounding for which researchers cannot control. However, researchers can quantify the aspects of a hypothetical confounder that could nullify their study results. Because sensitivity to unmeasured confounding analyses quantitatively address the primary concern of any observational study, we would expect every study to include one; however, a review of 90 observational studies with statistically significant findings published in 2015 in the Journal of the American Medical Association, the New England Journal of Medicine, and the American Journal of Epidemiology revealed only 4 (4%) included a quantitative sensitivity analysis. This disparity reveals the need for practical guidance and simple tools to help both the medical research community incorporate sensitivity analyses into their papers and the readers of medical research to easily perform such analyses when a paper has not included one. Presenting joint work with Lucy D'Agostino McGowan, MS, this talk introduces the concept of "rule out" sensitivity to unmeasured confounding analyses and offers a new, simple tool for performing them. Follow Vanderbilt on Twitter: https://twitter.com/vanderbiltu, on Instagram: http://instagram.com/vanderbiltu and on Facebook: https://www.facebook.com/vanderbilt. See all Vanderbilt social media at http://social.vanderbilt.edu.
