Dynamic Latent Trait Models with Mixed Hidden Markov Structure for Mixed Longitudinal Outcomes

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Health Research Methodology

Background:

• Longitudinal health studies often involve mixed outcomes (continuous and categorical).
• Categorical health data can suffer from differential misclassification, biasing results.
• Accurate modeling is crucial for understanding disease progression and risk factors.

Purpose of the Study:

• To develop a general Bayesian joint modeling framework for mixed longitudinal outcomes.
• To explicitly address and correct for differential misclassification in categorical health data.
• To jointly analyze asthma status and lung function measurements in children.

Main Methods:

• Utilized generalized linear mixed-effects models for error-free outcomes.
• Employed mixed hidden Markov models (MHMM) for misclassified categorical outcomes.
• Incorporated a transition modeling structure to account for time-dependent covariates and previous states.
• Applied a Bayesian approach for parameter estimation, including prevalence, transition, and misclassification probabilities.

Main Results:

• The proposed joint modeling approach demonstrated superior performance compared to traditional methods in simulation studies.
• MHMMs effectively captured cluster-level heterogeneity in health trajectories.
• The application to the Children Health Study (CHS) provided enhanced insights into the asthma-lung function relationship.

Conclusions:

• The Bayesian joint modeling framework offers a robust solution for analyzing complex longitudinal health data with misclassification.
• This approach enhances the understanding of the interplay between disease states and physiological measurements.
• The methodology has significant implications for epidemiological research and clinical studies.