Modeling longitudinal spatial periodontal data: a spatially adaptive model with tools for specifying priors and...

Area of Science:

• Biostatistics
• Periodontology
• Dental Public Health

Background:

• Attachment loss (AL) is a key indicator of periodontal disease severity.
• Existing spatiotemporal models may not fully capture the complex, non-uniform progression of AL.

Purpose of the Study:

• To develop and validate a nonstationary spatiotemporal conditional autoregressive (CAR) model for monitoring attachment loss progression.
• To improve the statistical modeling of periodontal disease by allowing spatially varying smoothing parameters.

Main Methods:

• Extension of the CAR prior to a nonstationary spatiotemporal CAR model.
• Incorporation of site-specific variance parameters with spatial smoothing.
• Development of a heuristic for selecting priors to ensure parameter identifiability.

Main Results:

• The proposed nonstationary spatiotemporal CAR model demonstrated improved fit compared to the standard dynamic CAR model.
• The model showed a better fit for 90 out of 99 patients in a clinical trial dataset.
• The model effectively handles bursts of large attachment loss values in specific areas.

Conclusions:

• The nonstationary spatiotemporal CAR model offers a more accurate approach to monitoring attachment loss progression in periodontal disease.
• This flexible modeling approach can enhance the understanding and management of periodontal disease.
• The method provides a robust framework for analyzing complex spatiotemporal data in clinical settings.