Letters Section Editor: Robert M. Golub, MD, Senior Editor.
In Reply: Drs Novikov and Kalter-Leibovici suggest that we include our instrumental variable of regional cardiac catheterization rate in the propensity score. The propensity score is used in analyses to remove confounding. We did not include this variable because it was not independently related to the outcome. Incorporating it into the standard models does not change the results for adjusted relative risk (multivariable model risk adjustment, 0.503; 95% CI, 0.495-0.513; propensity score risk adjustment, 0.517; 95% CI, 0.508-0.526; propensity-based matching, 0.539; 95% CI, 0.515-0.563). This is further evidence that regional catheterization rate is not related to patient health status, and since it is not a confounder, it is not required in the propensity score.
Stukel TA, Fisher ES, Wennberg DE. Analytic Approaches to Observational Studies With Treatment Selection Bias—Reply. JAMA. 2007;297(19):2077-2078. doi:10.1001/jama.297.19.2078-a