Using probabilistic record linkage and propensity-score matching to identify a community-based comparison population

Area of Science:

• Public Health
• Epidemiology
• Health Services Research

Background:

• Evaluating community interventions is difficult due to challenges in finding high-quality comparison groups.
• Retrospective cohort studies often lack robust control groups for accurate assessment.

Purpose of the Study:

• To present a methodological approach for creating comparison groups in community-based intervention studies.
• To evaluate birth outcomes, specifically interpregnancy interval, for a home visiting program using this method.

Main Methods:

• Utilized probabilistic record linkage to connect home visiting program data with birth certificate records (2005-2015).
• Identified two comparison groups: propensity-score-matched and eligible-but-unenrolled families.
• Analyzed interpregnancy interval (IPI) as a primary outcome to demonstrate the approach.

Main Results:

• Successfully linked 4822 home visiting program families with birth certificate data.
• Created a propensity-score-matched comparison group (14,219 families) with balanced covariates.
• Identified an eligible-but-unenrolled group (1101 families) with poor covariate balance; no significant program effect on IPI was found.

Conclusions:

• Combining propensity-score matching and probabilistic record linkage effectively identifies large, retrospective comparison groups for quasi-experimental research.
• Birth certificate data provides a single source for accessing outcomes across linked and comparison groups.
• Employing multiple comparison groups enhances the reliability of findings by addressing individual method limitations.