Mediation analysis with multiple mediators under unmeasured mediator-outcome confounding

Area of Science:

• Health and social sciences
• Epidemiology
• Biostatistics

Background:

• Investigating joint mediation effects with multiple mediators is crucial in health and social sciences.
• Traditional methods require no unmeasured confounding, an assumption often violated with numerous mediators.
• Unmeasured mediator-outcome confounding poses a significant challenge in complex mediation analysis.

Purpose of the Study:

• To develop a novel statistical method for assessing joint mediation effects with multiple mediators.
• To address the challenge of potential unmeasured confounding in mediation analysis.
• To provide a practical estimation procedure for complex mediation models.

Main Methods:

• A two-step method of moments estimation procedure is proposed.
• The method leverages heterogeneity in the population exposure effect on mediators for identification.
• Simulations and a real-world application are used for illustration.

Main Results:

• The proposed method successfully assesses joint mediation effects in the presence of potential unmeasured confounding.
• The technique is robust and applicable across various empirical settings.
• The application demonstrated the mediating role of PTSD symptoms between self-efficacy and fatigue.

Conclusions:

• The developed method offers a viable approach to analyze complex mediation pathways.
• It overcomes limitations of traditional methods by accounting for unmeasured confounding.
• This facilitates a deeper understanding of indirect effects in health and social science research.