Propensity score matching is a method used to reduce bias in observational studies by creating two populations that are similar (i.e., balanced) across a number of covariates using a match on only a single scalar, the propensity score. The matched samples are then treated as a quasi-experimental population, allowing for simplified analysis of study outcomes. Whenever a propensity match is performed, the balance between the two samples should be evaluated. Balance checking may be used to compare matches from multiple iterations of the propensity score model or from different matching algorithms and to provide information for any trade-offs between the closeness of the match and final sample size. Additionally, imbalances in the final matched sample should be kept in mind and possibly adjusted for when analyzing study outcomes. The published literature encourages using a variety of methods to evaluate balance. We have developed a SAS macro that will run a series of tests on the pre- and post-matched samples, following published guidelines. This macro evaluates balance independent of the methods used to create the propensity score or perform matching. Macro output includes information about the pre- and post-matched sample sizes, distribution of propensity scores, and the results of tests to compare covariates of interest, including calculating standard differences. The output is in RTF format so that it can easily reviewed by non-programmers. This SAS macro to evaluate balance with easy-toread output allows for thorough testing of balance and contributes to a better final matched sample.
[1]
P. Austin.
Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples
,
2009,
Statistics in medicine.
[2]
Gary King,et al.
Misunderstandings between experimentalists and observationalists about causal inference
,
2008
.
[3]
Peter C Austin,et al.
A critical appraisal of propensity‐score matching in the medical literature between 1996 and 2003
,
2008,
Statistics in medicine.
[4]
D. Rubin,et al.
The central role of the propensity score in observational studies for causal effects
,
1983
.
[5]
R. D'Agostino.
Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group.
,
2005,
Statistics in medicine.
[6]
Gary King,et al.
Matching as Nonparametric Preprocessing for Reducing Model Dependence in Parametric Causal Inference
,
2007,
Political Analysis.
[7]
K Dave,et al.
A CRITICAL APPRAISAL
,
2002
.
[8]
D. Rubin,et al.
Constructing a Control Group Using Multivariate Matched Sampling Methods That Incorporate the Propensity Score
,
1985
.