Rasouliyan L, Plana E, Aguado J. Considerations in the use of propensity scores in observational studies. Poster presented at the 2016 PhUSE Annual Conference; October 2016. Barcelona, Spain.

The use of propensity scores allows one to reduce the effects of confounding that can occur because of potential differences in the distribution of measured baseline characteristics between treatment groups in observational studies. Many considerations are important in the generation of the propensity score model, such as choosing baseline variables from a list of potential confounders, constructing an appropriate logistic regression model, assessing the balance of baseline variables between treatment groups, and evaluating the distribution of propensity scores between treatment groups. Additionally, once the propensity score model has been specified, several different methods can be used to incorporate the propensity score in the assessment of the treatment effect, including matching, stratification, inverse probability of treatment weighting, and covariate adjustment. These considerations will be discussed, and examples will be provided.