The BioSense Initiative to Improve Early Event Detection

RTI International, in partnership with the University of North Carolina at Chapel Hill (UNC-CH), and incollaboration with the North Carolina Division of Public Health (NC-DPH), is submitting this application towork with the Centers for Disease Control and Prevention (CDC) to improve early detection of diseaseoutbreaks of public health significance. Rapid detection of disease outbreaks rests on a foundation ofaccurate classification of patient symptoms early in the course of their illness. The overarching objective ofthis research is to define, evaluate, and standardize a methodology for creating useful case definitionsdesigned for the early detection of intentional and naturally occurring disease outbreaks. The specific aim ofthis research proposal is to develop and test methods for increasing the sensitivity and specificity ofsyndrome definitions using timely emergency department data. Improved case definitions will enhanceCDC's capacity to detect and investigate threats to the health of the population, which CDC undertakes aspart of its mission.Emergency department data may serve as a rich source for early signals of health threats to the population,but case definitions have not been standardized, and new methods are needed to process and use thetextual information found within the emergency record. To address these challenges, we propose aninnovative and iterative research plan that leverages RTI's and UNC-CH's capabilities to best serveCDCand the public health community.We will use emergency department data captured through North Carolina's Bioterrorism and EmergingInfections PreventiveService, the operational syndromic surveillance system used by NC-DPH to monitorthe state. After (1) developing a gold standard data set of ED visits for evaluating syndrometestcharacteristics, we will (2) evaluate natural language processing for preprocessing chief complaints; (3)explore use of semantic networking tools for developing definitions; (4) apply a reverse engineering processusing ICD-9-CMcode groupings; and (5) assess the applicability of early event detection for creatingsituational awareness following detection of an event. These methods will make use of information withinthe emergency record and create syndrome definitions with acceptable sensitivity, specificity, and positisvepredictive value.Valid syndromedefinitions will enable public health officials to operate a national monitoring system that canautomatically detect signals that may represent disease outbreaks or other potential threats to health.Operation of this system will protect the public health and will strengthen the capacity of public healthofficials to investigate and respond to these threats rapidly.