The PSMATCH Procedure

Overview: PSMATCH Procedure

In a randomized study, such as a randomized controlled trial, the subjects are randomly assigned to a treated (exposure) group or a control (non-exposure) group. Random assignment ensures that the distribution of the covariates is the same in both groups, and the treatment effect can be estimated by directly comparing the outcomes for the subjects in the two groups.

In contrast, the subjects in an observational study, such as a retrospective cohort study or a nonrandomized clinical trial, are not randomly assigned to the treated and control groups. Confounding can occur if some covariates are related to both the treatment assignment and the outcome. Consequently, there can be systematic differences between the treated subjects and the control subjects. In the presence of confounding, statistical approaches are required that remove the effects of confounding when estimating the effect of treatment.

One such approach is regression adjustment, which estimates the treatment effect after adjusting for differences in the baseline covariates. However, this approach has practical limitations, as discussed by Austin (2011a). Propensity score analysis is an alternative approach that circumvents many of these limitations.

The propensity score was defined by Rosenbaum and Rubin (1983, p. 47) as the probability of assignment to treatment conditional on a set of observed baseline covariates. Propensity score analysis minimizes the effects of confounding and provides some of the advantages of a randomized study. The basis for propensity score methods is the causal effect model introduced by Rubin (1974).

The PSMATCH procedure provides a variety of tools for propensity score analysis. The procedure either computes propensity scores or reads previously-computed propensity scores, and it provides the following methods for using the scores to allow for valid estimation of treatment effect in a subsequent outcome analysis:

Inverse probability of treatment weighting and ATT weighting (weighting by odds): The procedure computes weights from the propensity scores. These weights can then be incorporated into a subsequent analysis that estimates the effect of treatment.
Stratification: The procedure creates strata of observations that have similar propensity scores. In a subsequent analysis, the treatment effect can be estimated within each stratum, and the estimates can be combined across strata.
Matching: The procedure matches each treated unit with one or more control units that have a similar value of the propensity score. In a subsequent analysis, the treatment effect can be estimated by comparing outcomes between treated and control subjects in the matched sample. If the outcome values for a study are not available prior to matching, only the matched units are needed for follow-up. Thus, the cost of the trial is reduced (Stuart 2010, p. 2).

The PSMATCH procedure also provides methods for assessing the balance of baseline covariates and other variables in the treated and control groups after matching, weighting, or stratification. The procedure itself does not carry out the outcome analysis, nor does it make use of the outcome variable.

After adequate variable balance has been achieved (as described in the section Process of Propensity Score Analysis), and assuming that no other confounding variables are associated with both the treatment assignment and the outcome, the output data set that is created by the PSMATCH procedure serves as input for an appropriate statistical procedure for the outcome analysis.