Modelling risk from a disease in time and space

Area of Science:

• Biostatistics
• Spatial Epidemiology
• Bayesian Statistics

Background:

• Longitudinal and spatial data analysis are crucial in epidemiology.
• Hierarchical Bayesian models offer a flexible framework for complex data structures.
• Investigating environmental factors and disease clusters requires robust statistical methods.

Purpose of the Study:

• To combine longitudinal and spatial data models within a hierarchical Bayesian framework.
• To analyze time- and space-varying covariate effects on disease incidence.
• To re-analyze Ohio lung cancer data (1968-1988) and assess potential environmental influences.

Main Methods:

• Hierarchical Bayesian modeling incorporating time- and space-varying covariates.
• Markov chain Monte Carlo (MCMC) methods for data analysis.
• Two approaches for adjusting unmeasured spatial covariates: random effects and urbanization as a smoking surrogate.

Main Results:

• The proposed methodology effectively integrates longitudinal and spatial data.
• Analysis explored the impact of unobserved heterogeneity and urbanization on lung cancer rates.
• The Ohio dataset was found inadequate for definitively investigating the suspected nuclear facility's impact.

Conclusions:

• Hierarchical Bayesian models provide a robust framework for analyzing complex epidemiological data.
• Adjusting for unmeasured covariates like smoking is essential in spatial disease analysis.
• The Ohio lung cancer data does not support a causal link to the nuclear facility without further investigation.