Modeling Heterogeneity in Health Risk Behaviors With a Dynamic Mixture Model Informed by Textual Occupational Data

Area of Science:

• Public Health
• Biostatistics
• Computational Epidemiology

Background:

• Health risk behaviors, including smoking, poor nutrition, alcohol misuse, and physical inactivity (SNAP), are major drivers of chronic disease and healthcare costs.
• Individual behaviors are influenced by demographic factors and broader socioeconomic and occupational contexts.

Purpose of the Study:

• To develop and apply a novel Bayesian dynamic mixture model for analyzing multiple health risk behaviors.
• To investigate the role of occupational environments in modulating the effects of socio-demographic factors on SNAP behaviors over time.

Main Methods:

• Utilized data from the Italian PASSI surveillance system.
• Employed structural topic modeling to derive occupational groups from free-text descriptions.
• Applied a Bayesian topic-informed dynamic mixture model with a multivariate ordered probit component.
• Incorporated non-local spike-and-slab priors for interpretability and variable selection.
• Developed a sequential Monte Carlo approach for efficient online updating of inferences.

Main Results:

• The model successfully integrates occupational context into the analysis of multiple health risk behaviors.
• Demonstrated that covariate effects on SNAP behaviors vary across occupational clusters and evolve dynamically.
• The framework allows for flexible and interpretable monitoring of behavioral trends.

Conclusions:

• Occupational context significantly modulates the relationship between socio-demographic factors and health risk behaviors.
• The proposed methodology offers a flexible and interpretable framework for longitudinal health surveillance.
• This approach supports the development of targeted public health interventions by elucidating contextual influences on health behaviors.