Statisticians have been concerned with surveillance from the early days of the field to today. Timely counting is still today of great importance. This talk will trace the development of surveillance from its earliest to a very important application today, namely disease surveillance, and relate this to biosurveillance. A little history, a little counting, some graphics, some inference, and a little detective work, all in all a day in the life of a biostatistician.
In this talk I summarize the course of anthrax and other possible organisms that might be used for bioterrorism. The course reveals an opportunity for early detection-- early enough to allow prophylaxis for many patients who might otherwise have become symptomatic or required hospitalization. However, the task is not simple. The data that can be used for such surveillance is noisy and is subject to complex, if predictable patterns, as well as real-world complications, such as privacy protection, competing business interests, and bureaucratic territoriality. I discuss a system that accommodates all of these complications and describe the statistical techniques currently employed in the system. The statistical tools employed include a relatively straightforward application of Generalized Linear Mixed Models, a brute-strength computational approach, and a synthesis of the two. Finally, I describe a simulation study undertaken to compare various approaches to the use of such data and discuss what was learned.
organized by Nicholas Horton
Last updated May 31, 2004