Bayesian modelling of inseparable space-time variation in disease risk

Area of Science:

• Biostatistics
• Spatial Epidemiology
• Bayesian Modeling

Background:

• Analyzing spatio-temporal patterns in health data is crucial for understanding disease etiology.
• Existing models often lack flexibility in capturing complex space-time interactions.
• Prior specification in Bayesian frameworks significantly influences the analysis of disease incidence and mortality.

Purpose of the Study:

• To propose a unified Bayesian framework for analyzing spatio-temporal incidence or mortality data.
• To introduce and evaluate four novel prior distributions for space-time interaction parameters.
• To investigate the relationship between urbanization, cancer risk factors, and temporal trends using real-world data.

Main Methods:

• Development of a Bayesian hierarchical model incorporating different priors for space-time interactions.
• Application of Markov chain Monte Carlo (MCMC) simulation for parameter estimation.
• Model comparison using posterior deviance metrics to assess fit and complexity.

Main Results:

• The proposed framework effectively analyzes spatio-temporal health data with various interaction priors.
• Analysis of Ohio lung cancer data (1968-1988) revealed significant spatio-temporal patterns.
• Results support the epidemiological hypothesis regarding the temporal development of urbanization's association with cancer risk factors.

Conclusions:

• The unified Bayesian framework offers a flexible approach for spatio-temporal disease analysis.
• The novel priors provide a means to incorporate varying degrees of prior dependence for interaction effects.
• The study demonstrates the utility of Bayesian methods in confirming epidemiological hypotheses in cancer research.