Bayesian Modeling for Space-Time Surveillance of Disease

• Activity Number: 200
• Type: Roundtables
• Date/Time: Monday, August 2, 2010 : 12:30 PM to 1:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Andrew Lawson*+
• Companies: Medical University of South Carolina
• Keywords: Bayesian ; space-time ; health ; disease ; sequential ; surveillance

Abstract

This discussion will focus on the difference between space-time (ST) retrospective modeling of disease variation and modeling for surveillance purposes. Often retrospective modeling has a different focus from surveillance, in that the ability of models to correctly recover risk is a primary focus in retrospective analysis but the analysis or detection of change is the focus of surveillance. This can lead to quite different approaches to modeling. Retrospective models are often adaptive over time but this adaptation may be not useful for the detection of changes in risk (which by their nature are not usually smooth). Sequential methods of updating are often adopted in the surveillance setting but are largely underdeveloped. Sequential refitting of retrospective models in a surveillance context can lead to biases which increase over time. Some alternatives will be discussed