News Articles

 

Structural Models Show Promise in Tracking Bioterror Threats

November 2012

Karen Chang
Karen Cheng, Health Effects and Medical Response Software Director
Spotting potential terrorist attacks in the early stages and acting quickly to prevent them can save lives and ensure the safety of countless innocent civilians. Biosurveillance offers new hope for identifying these attacks in the early stages and may assist medical personnel in tracking an infectious disease in order to halt its spread. In both cases, the use of structural models for anomaly detection can help in analyzing existing biosurveillance data and in making informed predictions about the likely next steps or areas of concern for the ongoing investigation. Because these analytical tools incorporate data in real-time from partially observed epidemics, they can provide highly customized and dynamic results for analysts who need to identify and predict the epidemic for crisis response and medical planning.

A recent article entitled "Structural models used in real-time biosurveillance outbreak detection and outbreak curve isolation from noisy background morbidity levels", recently published in the Journal of the American Medical Informatics Association, illustrates the effective value of structural modeling in identifying both naturally-occurring outbreaks and biological or biochemical terrorist attacks. The article details how researchers constructed a computer modeling scenario in which Miami, Florida, had become the target of an anthrax attack by terrorist forces. The structural model was fit to simulated data of an anthrax attack (in the simulation, victims are infected with different doses, consistent with an anthrax plume dispersal where those closest to the anthrax source receive the highest dose; the incubation period of anthrax is also modeled) superimposed on a real medical dataset from Miami that included patient-level information such as symptoms, and chief complaints.

Unlike previous methods of biosurveillance data evaluation, the structural model was surprisingly effective in identifying that an anthrax outbreak was underway. Perfect detection was even achieved when less than 0.03 percent of the population was infected with anthrax. Structural modeling techniques allowed the team to model in the background noise without filtering the data, and they allow prediction to continue properly even when the biosurveillance data provided was incomplete. The structural model used by the research team allowed the epidemic to be extracted from the background noise for two weeks from the outbreak detection point, providing a clean (background-free) epidemic curve that can be fit to epidemic models for forecasting.