|
Abstract:
|
Propensity-score matching can be used to reduce the effect of treatment-selection bias in the estimation of treatment effects in observational studies. In order to estimate the propensity score, one must model the distribution of the treatment indicator variable given the observed covariates. Once estimated, the propensity score can be used to reduce bias through matching. One-to-many matches allow investigators to have a larger sample size and perhaps more statistical power for the test of the treatment effect. However, we hypothesize that the advantage of having many control individuals matched to each individual in the treatment group diminishes after a certain number of matches. Thus, using a Monte Carlo simulation, we will attempt to develop an optimal number of matched controls for a propensity score matched case-control study. We will also address other pragmatic factors that need to be considered in the implementation of the propensity-score matching process, which include (1) the study sample size vis-à-vis the matching ratio, (2) the handling of missing data on the observed covariates, and (3) the selection of the caliper to use along with the matching algorithm.
|
