Propensity Score Matching: A Step-by-Step Guide to Coding in R and Application in Observational Research Studies.

Although randomized controlled trials are the gold standard approach to identify relationships between an intervention and outcomes, observational studies remain invaluable. They allow for increased study power and efficiency, decreased cost, and demonstrate unique relationships that would be otherwise unfeasible or unethical. However, they are inherently biased by their non-randomized nature. Propensity score matching (PSM) combats this. We outline a step-by-step guide, from PICO question development, database and data processing/analytics software selection, and PSM coding techniques. We demonstrate this through an example evaluating cholecystectomy timing and outcomes in pregnant patients with cholecystitis. We discuss matching methods selected based on data set characteristics. Average Treatment Effect on the Treated (ATT) is applied to evaluate the intervention effect on patients who received the intervention. Balance between the intervention and comparison groups pre- and post-PSM is demonstrated mathematically by calculating standard mean differences and visually with Love Plots. Finally, treatment effect post-PSM is evaluated.