Variance estimation when using propensity-score matching with replacement with survival or time-to-event outcomes

Area of Science:

• Epidemiology
• Biostatistics
• Medical Research

Background:

• Propensity-score matching is vital for estimating treatment effects in observational studies.
• Matching without replacement can cause bias due to incomplete matching if controls are limited.
• Existing variance estimators for matching with replacement are not suitable for time-to-event data.

Purpose of the Study:

• To propose a novel variance estimator for the hazard ratio in propensity-score matching with replacement.
• To address the limitations of existing methods for time-to-event outcomes.
• To improve the accuracy of treatment effect estimation in observational studies.

Main Methods:

• Development of a new variance estimator for hazard ratios with matching with replacement.
• Monte Carlo simulations to evaluate the proposed estimator's performance.
• Application to a real-world dataset examining smoking cessation counseling's effect on survival after heart attack.

Main Results:

• The proposed variance estimator performs well in simulations for time-to-event outcomes.
• Matching with replacement, using the new estimator, can effectively reduce bias.
• The method demonstrated utility in analyzing smoking cessation's impact on survival.

Conclusions:

• The developed variance estimator is a valuable tool for propensity-score matching with replacement in time-to-event analyses.
• This approach enhances the reliability of observational studies, particularly in medical research.
• It offers a robust method for estimating treatment effects when control groups are limited.