Spatio-temporal models with errors in covariates: mapping Ohio lung cancer mortality

Area of Science:

• Environmental epidemiology
• Spatial statistics
• Biostatistics

Background:

• Regional morbidity and mortality maps are crucial for spatial disease pattern estimation and environmental equity assessments.
• Hierarchical Bayes methods offer effective smoothing for disease rate maps, particularly in low-population regions.

Purpose of the Study:

• To develop a unified hierarchical model for spatial-temporal disease mapping that accounts for covariate errors.
• To apply this model to county-specific lung cancer rates in Ohio, investigating environmental and socio-economic influences.

Main Methods:

• Integration of spatial-temporal mapping techniques with covariate error handling within a single hierarchical model.
• Utilizing Markov chain Monte Carlo (MCMC) methods for posterior distribution estimation, model evaluation, and selection.
• Incorporation of age-adjusted death rates, smoking prevalence, population density, and socio-economic status.

Main Results:

• The developed hierarchical model effectively smooths disease rates, enhancing accuracy in areas with limited data.
• Application to Ohio lung cancer data (1968-1988) provided insights into factors contributing to elevated rates in specific counties.
• The model facilitated the assessment of potential environmental influences, such as a depleted uranium fuel processing facility.

Conclusions:

• The proposed hierarchical Bayesian framework offers a robust approach for spatial-temporal disease mapping and environmental health investigations.
• Accurate mapping requires integrating diverse data, including demographic, environmental, and socio-economic factors.
• This methodology is vital for understanding disease etiology and informing public health policies, especially concerning environmental exposures.