Propensity score analysis

  • Statistical methods used to control for treatment selection bias in observational studies
  • Propensity score is the probability of treatment assignment conditional on observed baseline characteristics
  • A propensity score allows one to conduct an observational study in such a way that it might mimic some of the characteristics of a randomised controlled trial
  • This is good - observational studies cannot rely on randomisation to correct for confounding factors
  • If a group of subjects have the same propensity score, they should all have the same baseline characteristics.
  • Thus, propensity score analysis can compare subjects with similar propensity scores, thus attempting to negate the presence of confounding factors


  • Running an observational study with propensity score analysis is cheaper and easier than running an RCT
  • This is a way of studying treatments which cannot be ethically randomised


  • This is a statistical method, so it can only demonstrate an association, not causation
  • Propensity analysis assumes that all possible confounding variables have been measured in the observed population (but of course they have not)
  • Nobody can agree which confounding variables to measure, and when
  • One can include all sorts of meaningless variables, which add complexity without adding any validity