Application of a Mixed Effects Model for Biosurveilliance of Regional Rail Systems

Robert J. Gallop ； Charles J. Mode ； Kenneth Blank ； Sherri M. Jurgens ； Chad Schaben

《Journal of Data Science》 3卷4期 (2005/10) Pp. 353-370

https://doi.org/10.6339/JDS.2005.03(4).223

並列摘要

Although United States government planners and others outside government had recognized the potential risk of attacks by terrorists, the events of September 11, 2001, vividly revealed the nation's vulnerabilities to terrorism. Similarly, the 2004 terrorist attacks in Madrid illustrated vulnerabilities to terrorism extend beyond the United States. Those attacks were obvious destructive acts with a primary purpose of massive causalities. Let us consider a bioterrorist attack which is conducted subtly through the release of a Chemical/Biological agent. If such an attack occurs through release of a specific biological agent, an awareness of the potential threat of this agent in terms of the number of infections and deaths that could occur in a community is of paramount importance in preparing the public health community to respond to this attack. An increase in biosurveillance and novel approaches to biosurveillance are needed. This paper illustrates the use of mixed effects model for biosurveillance based on commuter size for regional rail lines. With mixed effects model we can estimate for any station on a given rail system the expected daily number of commuters and establish an acceptability criterion around this expected size. If the actual commuter size is significantly smaller than the estimate, then this could be an indicator of a possible attack. We illustrate this method through an example based on the 2001 daily totals for the Port Authority Transportation Company (PATCO) rail system, which serves residents of southern New Jersey and Philadelphia region in the United States. In addition, we discuss ways to put this application in a real time setting for continuous biosurveillance.

並列關鍵字

Best linear unbiased prediction (BLUP) estimates ； biosurveillance,bioterrorism ； effect size ； mixed-effects models