An Introductory Tutorial on Propensity Score Matching in RStudio Using the National Health Interview Survey

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Observational studies often suffer from confounding bias, limiting causal inference.
• Randomization is not always feasible, necessitating alternative methods to control for confounders.
• Propensity score matching (PSM) is a statistical technique to address confounding in such studies.

Purpose of the Study:

• To provide a step-by-step tutorial on creating a propensity-matched cohort using RStudio.
• To demonstrate the application of PSM in analyzing telehealth usage among individuals in non-metropolitan areas.
• To illustrate how PSM can improve covariate balance in observational research.

Main Methods:

• Utilized the 2022 National Health Interview Survey adult dataset.
• Calculated propensity scores using multiple logistic regression, with telehealth usage as the dependent variable.
• Applied PSM to balance covariates including age, sex, race, income, education, and insurance status.

Main Results:

• Propensity score matching significantly improved the balance of measured covariates between treatment and control groups.
• Standardized mean differences (SMDs) for all covariates were reduced below 0.1 after PSM, indicating good baseline balance.
• The analysis highlighted the utility of PSM in mitigating bias from observed confounders in telehealth research.

Conclusions:

• PSM is a valuable tool for reducing bias and improving covariate balance in observational studies.
• This tutorial provides a practical guide for researchers interested in applying PSM using RStudio.
• While effective for pedagogical purposes, the specific results are not intended as a nationally representative analysis.