Reducing Bias in a Propensity Score Matched-Pair Sample Using Greedy Matching Techniques

Matching members of a treatment group (cases) to members of a no treatment group (controls) is often used in observational studies to reduce bias and approximate a randomized trial. There is often a trade-off when matching cases to controls and two types of bias can be introduced. While trying to maximize exact matches, cases may be excluded due to incomplete matching. While trying to maximize cases, inexact matching may result. Bias is introduced by both incomplete matching and inexact matching. Propensity scores are being used in observational studies to reduce bias. It has been shown that matching on a propensity score can result in similar matched populations. This paper will describe how to reduce matched-pair bias caused by incomplete matching and inexact matching. Cases will be matched to controls on the propensity score using the presented matching algorithm. SAS/STAT LOGISTIC procedure code will be given to create the propensity score. A user-written SAS macro will be given to create a propensity score matchedpair sample using greedy matching techniques. The results of using the presented code, run on a large observational database of myocardial infarction patients, will be given as an example.