Combining the regression discontinuity design and propensity score-based weighting to improve causal inference in...

Area of Science:

• Health Services Research
• Econometrics
• Biostatistics

Background:

• The regression discontinuity (RD) design is a powerful quasi-experimental method approximating randomized trials in non-experimental settings.
• RD design leverages a cutoff on a continuous variable for treatment assignment, assuming local exchangeability.

Purpose of the Study:

• To review recent advancements in RD design over the past decade.
• To introduce propensity score weighting as a complementary technique to address baseline imbalances in RD studies.
• To evaluate the effectiveness of a health management intervention using RD and propensity score weighting.

Main Methods:

• Description of key methodological advances in RD design.
• Implementation of RD design using health management intervention data.
• Application of propensity score-based weighting to correct for baseline covariate imbalances in RD analysis.

Main Results:

• The study illustrates RD design implementation with health management intervention data.
• Propensity score weighting demonstrated superior performance over standard regression covariate adjustment in balancing baseline characteristics.
• The weighting technique allows direct inspection of achieved balance, a key advantage over regression adjustment.

Conclusions:

• The RD design, especially when combined with propensity score weighting, is a valuable tool for evaluating health management program effectiveness using observational data.
• Propensity score weighting enhances the validity of RD by mitigating bias from baseline imbalances.
• The combined approach offers a practical and effective alternative to traditional methods for causal inference in health research.