This chapter introduced a biosurveillance method that uses causal Bayesian networks to model noncontagious diseases in a population. By making two independence assumptions in the model, both of which appear plausible for noncontagious diseases, and by performing inference using equivalence classes and incremental updating, it is possible to achieve tractable Bayesian biosurveillance in a region with 1.4 million people. We implemented and evaluated an outbreak detection system called PANDA. Overall, the run-time results and the detection performance of this initial evaluation are encouraging, although additional studies are needed and are in process.
Was this article helpful?