Bayesian Analysis of Surveillance Data and Meteorological Data for Detection of Aerosol Releases of B. Anthracis Spores

• Activity Number: 109
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): William Hogan*+ and Garrick Wallstrom and Gregory F. Cooper and Michael Wagner and Jean-Marc Depinay
• Companies: University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh and University of Pittsburgh
• Keywords: Biosurveillance ; Statistical outbreak detection ; Atmospheric dispersion ; Anthrax ; Emergency department data ; Bayesian data analysis

Abstract

Statistical analysis of data is a key component of biosurveillance. The newer types of data it monitors, such as emergency-department (ED) visits, have variable baseline activity and extracting the signal of an outbreak from this noise is difficult. We present the Bayesian Aerosol Release Detector (BARD), an algorithm to detect large, outdoor aerosol releases of B. anthracis spores as early as possible. BARD estimates the posterior probability of a release--and a posterior distribution over the release location, quantity, and time--from ED visit and meteorological data. It comprises a Bayesian model of inhalational anthrax and background disease and a deterministic atmospheric dispersion equation. In an evaluation, BARD detected simulated releases in a timely, sensitive, and specific manner. Its performance was much less sensitive to varying model parameters than to outbreak size.