The estimation of direct and indirect causal effects in the presence of misclassified binary mediator

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Mediation analysis quantifies exposure effects via intermediate variables.
• Misclassification of mediators significantly threatens the validity of causal effect estimation.
• Understanding direct and indirect effects is crucial in epidemiological studies.

Purpose of the Study:

• To investigate the impact of non-differential binary mediator misclassification on direct and indirect causal effect estimation.
• To develop and validate methods for correcting mediator misclassification in mediation analysis.
• To apply these correction strategies to a real-world perinatal epidemiological study.

Main Methods:

• Proposed a hybrid likelihood-based and predictive value weighting method for misclassification correction.
• Developed an expectation-maximization algorithm-based approach for correction.
• Incorporated sensitivity analysis to assess the robustness of the findings.
• Required knowledge of sensitivity and specificity parameters for correction.

Main Results:

• The developed methods effectively correct for non-differential binary mediator misclassification.
• Accurate estimation of direct and indirect causal effects is achievable with corrected data.
• The approaches were successfully applied to a study on pre-term birth determinants.

Conclusions:

• Accurate mediation analysis requires addressing mediator misclassification.
• The proposed correction methods enhance the validity of causal effect estimation in the presence of misclassified binary mediators.
• These techniques are valuable tools for epidemiological research, particularly in perinatal studies.