Surveillance of Infectious Disease Occurrences in the Community: An Analysis of Symptom Presentation in the Emergency Department

Abstract

Objectives: To determine the effectiveness of a simulated emergency department (ED)‐based surveillance system to detect infectious disease (ID) occurrences in the community. Methods: Medical records of patients presenting to an urban ED between January 1, 1999, and December 31, 2000, were retrospectively reviewed for ICD‐9 codes related to ID symptomatology. ICD‐9 codes, categorized into viral, gastrointestinal, skin, fever, central nervous system (CNS), or pulmonary symptom clusters, were correlated with reportable infectious diseases identified by the local health department (HD). These reportable infectious diseases are designated class A diseases (CADs) by the Ohio Department of Health. Cross‐correlation functions (CCFs) tested the temporal relationship between ED symptom presentation and HD identification of CADs. The 95% confidence interval for lack of trend correlation was 0.0 ± 0.074; thus CCFs > 0.074 were considered significant for trend correlation. Further cross‐correlation analysis was performed after chronic and non‐community‐acquirable infectious diseases were removed from the HD database as a model for bioterrorism surveillance. Results: Fifteen thousand five hundred sixty‐nine ED patients and 6,489 HD patients were identified. Six thousand two hundred eight occurrences of true CADs were identified. Only 87 (1.33%) HD cases were processed on weekends. During the study period, increased ED symptom presentation preceded increased HD identification of respective CADs by 24 hours for all symptom clusters combined (CCF = 0.112), gastrointestinal symptoms (CCF = 0.084), pulmonary symptoms (CCF = 0.110), and CNS symptoms (CCF = 0.125). The bioterrorism surveillance model revealed increased ED symptom presentation continued to precede increased HD identification of the respective CADs by 24 hours for all symptom clusters combined (CCF = 0.080), pulmonary symptoms (CCF = 0.100), and CNS symptoms (CCF = 0.120). Conclusions: Surveillance of ED symptom presentation has the potential to identify clinically important ID occurrences in the community 24 hours prior to HD identification. Lack of weekend HD data collection suggests that the ED is a more appropriate setting for real‐time ID surveillance.