Slide 1: Propensity score matching

Slide 2: What is propensity score matching? Propensity score matching (PSM) is a statistical technique used in attempt to reduce bias of observational data.

Randomization is to randomized controlled trial as propensity score matching is to observation study.

Slide 3: Remember, randomization helps to control for bias by ensuring that treatment and control groups are the same. PSM does the same thing retrospectively. Let’s take a closer look…

Slide 4: How does PSM work? First, the researchers decide on a set of covariates to use. Covariates are baseline characteristics such as demographics, medical diagnoses, lab values, vitals, etc.

Next, a propensity score is derived for each subject based on covariates. Here, different scores are represented by different sizes of stars.

Slide 5: Subjects in the exposed group are then matched based on score to subjects in the unexposed group.

Notice that within the matched study group, the exposed group and the unexposed group are the same.

Slide 6: Pearls and Pitfalls. Relative to other adjustments techniques such as regression, PSM reduces bias because selection of matched pairs can be pre-specified before outcome assessment.

Success of the approach relies on adequate sample size. (There must be enough people who are similar enough to be matched in both groups!)

Beware of residual confounding! You can only adjust for the variables that you measure and include in the model.