Causal mediation analysis with multiple mediators in the presence of treatment noncompliance

Area of Science:

• Causal inference
• Biostatistics
• Health services research

Background:

• Treatment noncompliance complicates randomized experiments, hindering the identification of mediated intention-to-treat (ITT) effects.
• Existing methods often assume a single mediator, which is insufficient for complex mediating mechanisms.

Purpose of the Study:

• To develop a method for estimating the mediated portion of the ITT effect in the presence of multiple dependent mediators and treatment noncompliance.
• To enable researchers to identify key mediators and strengthen intervention effects despite noncompliance.

Main Methods:

• Proposed a nonparametric estimation procedure for the mediated ITT effect with multiple mediators and noncompliance.
• Conducted a Monte Carlo simulation study to evaluate the method's finite sample performance.
• Performed a sensitivity analysis to assess the robustness of key assumptions.

Main Results:

• The proposed method effectively estimates the mediated ITT effect under complex conditions.
• Simulation results demonstrate the approach's good finite sample performance.
• Empirical analysis of JOBS II data illustrated the method's practical application.

Conclusions:

• The developed method allows for the identification of relevant mediators even with treatment noncompliance and multiple mediating pathways.
• This facilitates informed decision-making to enhance intervention effectiveness in public health and social science research.