How Can Causal Relationships Be Measured in Observational Studies? Propensity Score Matching: A Tutorial Article

In most observational studies, researchers try to find causal relationships between observations. This correlation is associated with error due to lack of controlling confounding variables. One of the methods used to control the confounding variables is propensity score matching. Therefore, the purpose of this study is to explain the steps of propensity score matching.

Propensity score matching has 5 steps. The first step is to estimate the propensity score, which includes selecting the appropriate model and variables. The second step is to select an appropriate matching method based on the estimated propensity scores. The third step is overlap and common support. At this stage, observations that are outside the range of matching scores are removed. The quality of the match is then evaluated, and finally the sensitivity of the estimated effects must be estimated.

In the propensity score matching, the observed basic variables are balanced between the exposure and non-exposure groups. However, if the statistical model used to calculate the propensity score is not correctly selected, an imbalance between the basic characteristics of the two groups can still exist.

Propensity score matching is useful in cases where the confounding variables of study are high. In observational studies, this method can be used as an alternative to randomization in experimental studies, which makes the estimates more accurate by reducing the selection bias and controlling confounding variables.