Latent Class Dynamic Mediation Model with Application to Smoking Cessation Data

Area of Science:

• Statistics
• Biostatistics
• Health Behavior Research

Background:

• Traditional mediation analysis assumes population homogeneity and constant effects over time.
• These assumptions are often violated in real-world applications, such as smoking cessation studies.
• Existing models may not adequately capture subgroup-specific, time-varying mediation processes.

Purpose of the Study:

• To propose a flexible latent class dynamic mediation model.
• To account for unobserved population heterogeneity and time-varying mediation effects.
• To provide a robust framework for analyzing complex longitudinal data in health research.

Main Methods:

• Utilized a proportional odds model to identify latent subgroups and accommodate subject heterogeneity.
• Employed a Bayesian hierarchical nonparametric time-varying coefficient model for dynamic mediation.
• Allowed for distinct dynamic mediation processes within each identified latent subgroup.

Main Results:

• The proposed latent class dynamic mediation model demonstrated good performance in simulation studies.
• The method effectively estimated time-varying mediation effects across different subgroups.
• Application to smoking cessation data illustrated the model's practical utility.

Conclusions:

• The latent class dynamic mediation model offers a powerful approach for analyzing heterogeneous populations with time-varying mediation.
• This methodology enhances understanding of dynamic processes in health behavior change.
• The model provides a valuable tool for researchers investigating complex mediation in longitudinal studies.