When: Friday, Oct 17, 2003
Time: 12 noon - 1 pm
Where: 1221 CS
Speaker: Greg Cooper, Ph.D., Associate Professor, Dept. of Medicine, University of Pittsburgh

Title: Bayesian Biosurveillance Using Causal Networks

Abstract: 

This talk will describe recent research on a method for detecting
disease outbreaks that is based on Bayesian causal modeling of each
individual in the population. For people who have entered the
healthcare system recently (e.g., visited an emergency department),
symptoms and other evidence may be available (in a de-identified
format). For other individuals, evidence might only include
demographic information, such as age, gender, and home zip code. The
models of individuals are linked by common causes of an outbreak
(e.g., airborne anthrax). The linked models of individuals form a
population model. Detection involves performing inference on the
population model to derive the posterior probabilities of various
types of outbreaks given available evidence. In addition to providing
an overview of the modeling framework being used, this talk will
describe techniques for achieving computational tractability when
performing inference on a population model that contains millions of
variables.