Comparisons of timeliness and signal strength for multiple syndromic surveillance data types--San Diego County, July 2003-July 2004

Introduction: San Diego County is the site of the BioNet Project, which seeks to improve the ability of the Navy Region Southwest and San Diego County to respond to a biologic attack on its population and its critical infrastructure by improving, integrating, and enhancing disparate military and civilian detection and characterization capabilities. BioNet is funded by the Department of Homeland Security. One component of this project is the comparison of data sources available in San Diego County to understand their relative strengths and weaknesses for syndromic surveillance purposes. Objectives: This study quantitatively compared the different syndromic data sources (both military and civilian) available in San Diego County both in terms of signal strength and timeliness. Methods: Multiple types of data were compared, including emergency medical services (EMS), school nurse, school absentee, physician outpatient encounters, over-the-counter (OTCs) pharmaceuticals, and prescription pharmaceuticals. Three major historical disease outbreaks are used as points of comparison. The specific outbreaks are respiratory disease caused by a major wildfire event in October 2003, influenza-like illness in December 2003, and a surge of gastrointestinal illness in February 2004. Each data source is separately filtered to bring out the types of symptoms associated with each of the outbreaks. The sources are compared both before and after smoothing with a moving 7-day average, designed to eliminate certain idiosyncratic effects and to reduce noise. Finally, the data sources were compared on the basis of timing and signal-to-noise ratio for their ability to capture these outbreaks. Additional time-series comparisons were also used to determine whether the data sources trend together during nonoutbreak periods. Results: The disease outbreaks are each observable in multiple data sources, but the most useful data source varies with the event. EMS, military ambulatory encounters, OTCs, and school nurse reports were especially useful for different illness events. For example, EMS data indicate the strongest signal-to-noise ratio for disease caused by wildfires; the school nurse data give an early indication of influenza; and the military ambulatory encounter data provide the strongest indication of an outbreak of gastrointestinal illness. Conclusion: These results indicate that a system that integrates multiple syndromic data streams into a single prospective surveillance tool