A BAYESIAN HIERARCHICAL SPATIAL MODEL FOR DENTAL CARIES ASSESSMENT USING NON-GAUSSIAN MARKOV RANDOM FIELDS.

Area of Science:

• Biostatistics
• Dental Public Health
• Epidemiology

Background:

• Dental caries research generates hierarchical data at tooth and surface levels.
• Caries outcomes exhibit spatial dependencies, influencing neighboring teeth and surfaces.
• Mixed discrete outcomes (binary for teeth, trinary for surfaces) complicate unified analysis.

Purpose of the Study:

• To develop a Bayesian two-level hierarchical model for analyzing complex dental caries data.
• To accommodate spatial dependencies and mixed discrete responses within a unified framework.
• To improve the statistical analysis of dental caries, considering tooth and surface-level data.

Main Methods:

• A Bayesian two-level hierarchical model using spatial Markov random field assumptions.
• An autologistic model for tooth-level binary outcomes and a Potts model for surface-level trinary outcomes.
• A double Metropolis-Hastings sampler to address computational challenges in Bayesian estimation.

Main Results:

• The proposed model effectively handles the hierarchical structure and spatial referencing in dental caries data.
• The Bayesian model demonstrates improved performance compared to standard non-spatial models in simulation studies.
• The model was successfully applied to a clinical dataset, illustrating its practical utility.

Conclusions:

• The developed Bayesian hierarchical model offers a robust framework for analyzing complex dental caries data.
• This approach accounts for spatial dependencies and mixed discrete responses, leading to more accurate insights.
• The model provides a valuable tool for researchers and clinicians studying tooth decay.