Introduction: In observational studies, a significant difference in the outcomes between treated and untreated patients may be observed in absence of treatment effect and caused by differences in baseline characteristics.
Exegesis: Propensity score analysis is a post hoc adjustment method which consists in deriving the conditional probability of receiving the treatment for a patient given his measured baseline characteristics (i.e., the propensity score). Matching each treated patient to an untreated one who has the nearest propensity score tends to balance baseline characteristics between the two groups and reduce the risk for overt bias. Then, the outcomes can be compared between matched treated and untreated patients.
Conclusion: Propensity score analysis is relevant for clinical conditions and treatments for which randomized controlled trials are unlikely to be conducted. However, propensity analysis cannot adjust for unmeasured characteristics and sensitivity analysis should be performed to assess how sensitive the conclusions are to potential confounding factors.