On variance estimate for covariate adjustment by propensity score analysis

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Propensity score (PS) methods are widely used to control for confounding in observational studies.
• Covariate adjustment by PS is a common but flawed approach due to biased variance estimation.
• Biased variance estimation can lead to invalid statistical inference and erroneous public health conclusions.

Purpose of the Study:

• To propose and validate novel methods for accurate variance estimation in covariate adjustment by PS analyses.
• To address the limitations of conventional variance estimation in propensity score methods.
• To provide robust tools for reliable statistical inference in comparative effectiveness research.

Main Methods:

• A two-stage analytic procedure was developed for valid variance estimation.
• An empirical bootstrap resampling scheme was implemented.
• Both methods were implemented in a publicly available R function.

Main Results:

• Extensive simulations confirmed the bias in conventional variance estimators.
• The proposed variance estimators provided valid and robust estimates, even with complex confounding.
• The methods were successfully illustrated using a post-surgery pain study.

Conclusions:

• The developed methods offer valid variance estimation for covariate adjustment by PS.
• These approaches enhance the reliability of statistical inference in observational research.
• The findings have implications for improving accuracy in public health surveillance and drug safety studies.