Investigating Risk Factors Associated with the Geographical Distribution of Disease Using Aggregate and Individual-Level Data

• Activity Number: 34
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Michelle Ross*+
• Companies: University of Washington
• Keywords: Aggregate data studies ; Ecological bias ; spatial epidemiology ; environmental epidemiology

Abstract

In spatial epidemiology, studies are usually based on count data from geographic areas, known as ecological data. Ecologic data alone is unable to characterize within-area variability in exposures and confounders, resulting in ecologic bias. Further, ecologic bias can cause a mismatch between conclusions drawn from aggregate and individual-level data, known as the ecological fallacy. The only solution to the ecological inference problem is to supplement the aggregate data with individual-level information. We describe a method to alleviate ecological bias when individual-level data is combined with both individual-level and area-level exposures of interest, acknowledging confounding by location through the use of random effects. The methods are illustrated using incident tuberculosis data from Mombasa, Kenya. Case information was obtained from patient data from the tuberculosis programs run by the Division of Leprosy, Tuberculosis and Lung Disease, where all tuberculosis cases diagnosed in 2012 were collected. This data is combined with demographic, socioeconomic and geographic data from the 2009 national census conducted by the Kenya National Bureau of Statistics.