Should we be worried? Investigation of signals generated by an electronic syndromic surveillance system--Westchester County, New York.

INTRODUCTION: In January 2003, the Westchester County Department of Health (WCDH) began conducting electronic syndromic surveillance of hospital emergency department (ED) chief complaints. Although methods for data collection and analysis used in syndromic surveillance have been described previously, minimal information exists regarding the responses to and investigations of signals detected by such systems. This paper describes WCDH's experience in responding to syndromic surveillance signals during the first 9 months after the system was implemented. OBJECTIVES: The objectives of this analysis were to examine WCDH's responses to signals detected by the county's syndromic surveillance system. Specific goals were to 1) review the actual complaints reported by hospital EDs to determine whether complaint data were accurately identified and classified into syndrome categories, and provide feedback from this review to data collection and analysis staff to refine text terms or filters used to identify and classify chief complaints; 2) develop procedures and response algorithms for investigating signals; 3) determine whether signals correlated with reportable communicable diseases or other incidents of public health significance requiring investigation and intervention; and 4) quantify the staffing resources and time required to investigate signals. METHODS: During January 27-October 31, 2003, electronic files containing chief-complaint data from seven of the county's 13 EDs were collected daily. Complaints were classified into syndrome categories and analyzed for statistically significant increases. A line listing of each complaint comprising each signal detected was reviewed for exact complaint, number, location, patient demographics, and requirement for hospital admission. RESULTS: A total of 59 signals were detected in eight syndrome categories: fever/influenza (11), respiratory (6), vomiting (11), gastrointestinal illness/diarrhea (8), sepsis (7), rash (7), hemorrhagic events (3), and neurologic (6). Line-listing review indicated that complaints routinely were incorrectly identified and included in syndrome categories and that as few as three complaints could produce a signal. On the basis of hospital, geographic, age, or sex clustering of complaints, whether the complaint indicated a reportable condition (e.g., meningitis) or potentially represented an unusual medical event, and whether rates of hospital admission were consistent with medical conditions, 34 of 59 signals were determined to require further investigation (i.e., obtaining additional information from ED staff or medical providers). Investigation did not identify any reportable communicable disease or other incidents of public health significance that would have been missed by existing traditional surveillance systems. Nine staff members spent 3 hours/week collectively investigating signals detected by syndromic surveillance. CONCLUSIONS: Standardized sets of text terms used to identify and classify hospital ED chief complaints into syndrome categories might require modification on the basis of hospital idiosyncrasies in recording chief complaints. Signal investigations could be reasonably conducted by using local health department resources. Although no communicable disease events were identified, the system provided baseline and timely objective data for hospital visits and improved communication among county health department and hospital ED staff.