A Novel Secondary-Outcome Approach to Estimating Primary Causal Effects With Unmeasured Confounders

Area of Science:

• Causal Inference
• Biostatistics
• Epidemiology

Background:

• Unmeasured confounding poses a significant challenge in accurately estimating treatment effects.
• Existing causal inference methods struggle to fully address this limitation.
• Valid estimation of average causal effects is crucial in various scientific fields.

Purpose of the Study:

• To develop a novel identification strategy for estimating average causal effects in the presence of unmeasured confounding.
• To construct a proxy confounder using primary and secondary outcomes for inverse probability weighting.
• To provide a robust methodology for causal inference with multiple secondary outcomes.

Main Methods:

• Proposed a new identification strategy leveraging information from primary and secondary outcomes.
• Developed a proxy confounder for inverse probability weighting-type estimation.
• Established formal identification results and asymptotic distribution theory for the proposed estimator.

Main Results:

• Simulation studies demonstrated a significant reduction in confounding bias.
• The proposed method refined causal effect estimation compared to existing approaches.
• Practical application successfully inferred the effects of maternal delivery mode on child's test scores by integrating cognitive secondary outcomes.

Conclusions:

• The novel methodology effectively mitigates unmeasured confounding bias.
• Integrating secondary outcomes enhances the accuracy of causal effect estimation.
• This approach offers a promising solution for causal inference problems with multiple secondary outcomes.