Evidence-appraisal glossary
Propensity score
A propensity score is each person's estimated probability of receiving the treatment given their measured baseline characteristics. In observational studies, researchers use it to compare treated and untreated people who looked equally likely to be treated, mimicking randomization to reduce confounding from those measured factors.
Also called: propensity score matching, propensity score analysis, PSM.
What it is
In a non-randomized study, treated and untreated groups often differ at baseline, so raw comparisons mix the treatment effect with those differences. A propensity score collapses many measured covariates (age, severity, comorbidities) into one number: the modeled probability, usually from logistic regression, that a person would receive the treatment. People with similar scores had similar odds of being treated, so comparing them approximates a randomized comparison on measured factors.
How to use it when reading a study
- Check which method was used: matching, stratification, inverse-probability weighting, or covariate adjustment.
- Look for evidence of balance after adjustment (standardized differences below 0.1) and adequate overlap between groups.
- Remember the key limit: propensity scores only correct for measured confounders. Unmeasured factors (for example, frailty not captured in the data) still bias results, so this is not equivalent to randomization.
- Note which covariates entered the model; omitting a real confounder leaves residual bias.
This is a plain-language methodology definition for reading research. It is general education, not medical advice.