Re-evaluating causal inference: Bias reduction in confounder-effect modifier scenarios

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Propensity Score Matching (PSM) is a common method for estimating causal treatment effects.
• Its effectiveness can be compromised in complex situations involving effect modification.

Purpose of the Study:

• To compare the bias-reduction performance of PSM and Inverse Probability Weighting (IPW).
• To investigate scenarios where a confounder also acts as an effect modifier.

Main Methods:

• Utilized University of California, Berkeley graduate admission data for illustration.
• Conducted simulation studies to assess bias and Mean Squared Error (MSE).
• Analyzed observational data from Mexico's Seguro Popular de Salud program.

Main Results:

• PSM demonstrated biased Average Treatment Effect (ATE) estimates when confounders were effect modifiers.
• IPW consistently yielded less biased estimates with lower MSE compared to PSM.
• IPW proved more accurate and reliable in real-world observational data.

Conclusions:

• PSM is not always suitable for estimating causal effects in the presence of effect modifiers.
• IPW is a superior method for bias reduction in such complex observational studies.
• Findings highlight the importance of choosing appropriate causal inference methods based on study design.