A Bayesian Hierarchical Spatial Approach for Constructing Disease Risk Maps at a Finer Level Than Is Provided in Publicly Available Data

• Activity Number: 520
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Fu-Chi Hsieh*+ and Theodore R. Holford
• Companies: Yale University and Yale University
• Keywords: Bayesian hierarchical spatial model ; spatial statistics ; misalignment problem

Abstract

A Bayesian hierarchical spatial model is developed to construct disease risk maps using exposure data at a finer areal scale when the outcome variable is at a larger administrative areal level.A Poisson log-linear model with a conditionally autoregressive random effect is employed.The method is illustrated using data on the number of lung cancer cases in Connecticut towns,and the covariates are demographic information from the US Census and a traffic-related pollution exposure estimate for census block groups.The model provides estimates of the SMR at the census block group level,using incident cases reported at town level.Measurement errors associated with exposure assessment are also estimated. The results show traffic-related exposure has a significant positive association with lung cancer,but this may be partially explained by socioeconomic factors associating with cigarette smoking.