Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Austin Miller and Randy Carter, PhD. Department of Biostatistics, University at Buffalo, School of Public Health and Health Professions, 3435 Main Street, Buffalo, NY 14214, 716-829-3434, am65@buffalo.edu

Public Health managers need an easily implemented early warning system for bioterrorism attacks. Many potential bioterrorism agents manifest themselves with flu-like symptoms. A statistical model that provides daily prediction and a 95% upper tolerance limit for the daily numbers of cases with flu-like symptoms was developed. The tolerance limits are suggested for use as threshold values in a real-time monitoring system. The seasonality of genuine influenza outbreaks, in response to the so-called “winter effect”, indoor crowding under conditions of reduced ventilation, low temperature (low humidity) and higher precipitation, is well known. Also known is that the influenza serial interval ranges between 1 day and 9 days, averaging about 4 days. These facts were considered when building a statistical model to derive the tolerance interval for the number of cases with flu-like symptoms. Data were reported to the Erie County Health Department by Emergency Services facilities in Erie County, New York, between August 1999 and November 2002. A prediction model for the number of cases as a function of previous numbers, temperatures, precipitation and their interactions was derived using Poisson regression methods. It was found that the model errors followed an autoregressive structure. This and lagged dependent variables in the mean function caused the usual iteratively reweighted least squares estimators of model parameters to be asymptotically biased. An instrumental variable approach was used to obtain asymptotically unbiased estimates.

Learning Objectives: At the conclusion of the session, the participant (learner) in this session will be able to

• Recognize the importance of an early warning mechanism for Bioterrorism.
• Articulate the assumptions used in building an early warning system, and discuss the strengths and weaknesses of such a system.
• Describe the statistical process used in developing early warning models.

Keywords: Bioterrorism, Statistics

Presenting author's disclosure statement:

Not Answered