Introduction to Propensity Scores

Presented by Dr. Dhruv Sharma, CSTAT

In this workshop we will discuss the merits of using propensity scores in observational intervention based studies. Randomized trials are considered the gold standard approach for estimating the effects of interventions (treatment or control) on outcomes. Random intervention assignment ensures that measured or unmeasured baseline characteristics don’t influence intervention assignment. Whereas in observational studies, intervention assignment is often influenced by subject characteristics which can differ systematically for treated and control subjects leading to biased estimates of treatment effect. The propensity score is the probability of treatment assignment conditional on observed baseline characteristics and is used in observational studies to mimic the process of random assignment of a randomized trial. Conditional on the propensity score, the distribution of observed baseline covariates will be similar between treated and control subjects. Various approaches of propensity scores will be discussed using a real example and implemented in R software language. The R code provided will be easy to implement in the context of the workshop and no knowledge of R in required.