This Article  • PDF  • Send to a friend  • Save in My Folder  • Save to citation manager  • Permissions  Citing Articles  • Citing articles on Web of Science (7)  • Contact me when this article is cited  Related Content  • Similar articles in JAMA  Topic Collections  • Viral Infections  • Public Health, Other  • Violence and Human Rights  • Bioterrorism  • Infectious Diseases  • Alert me on articles by topic  Social Bookmarking   What's this?

To the Editor: Syndromic surveillance, based on analysis of clinical or administrative data to detect patterns consistent with emerging diseases, could allow for early recognition of attacks with biological or chemical weapons. Since the 2001 terror attacks, many institutions, encouraged by public health authorities and accreditation agencies, have sought to develop their own surveillance systems. However, the sensitivity and specificity of such single-institution systems have not been validated.

Methods
The emergence of West Nile Virus (WNV) as a widespread cause of neurological disease during the summer of 2002 offered a unique opportunity to test the performance of single-institution syndromic surveillance. Between August and October of 2002, more than 700 WNV cases were reported in the state of Illinois, of which more than 500 were in the Chicago area.¹

We reviewed monthly International Classification of Diseases, Ninth Revision (ICD-9) coding data for July 1998 through October 2002 to determine whether the 2002 WNV epidemic would have been detected had a simple syndromic surveillance system been operating prospectively at our hospital. For each month studied, the number of patients seen in the emergency department or inpatient units assigned ICD-9 codes for encephalitis (323.9), aseptic meningitis (047.9), unspecified meningitis (322.9), bacterial meningitis (320.9), or Guillain-Barré syndrome (357.0) were recorded. Any patient assigned at least 1 of the codes was considered a potential case of WNV. Results were expressed as the number of potential cases detected per 1000 patient encounters. The ² test was used to compare data between years using the results for July through October (the period in which all 2002 WNV cases occurred).

Results
During the year 2002, there were 1.63 potential cases detected per 1000 patient encounters (95% confidence interval [CI], 1.13-2.13 per 1000). This is slightly lower than the mean value for the same period during the preceding 4 years (1.83 potential cases detected per 1000 encounters; 95% CI, 1.54-2.12 per 1000; P = .50) (Figure 1). In contrast, during the year 2001, 3.1 potential cases were detected per 1000 patient encounters (95% CI, 2.39-3.82 per 1000), a statistically significant increase from that seen during the 3 previous years (1.36 potential cases detected per 1000 encounters; 95% CI, 1.07-1.65 per 1000; P<.001). Even when the high number of potential cases detected during 2001 is excluded, the number of potential WNV cases detected during 2002 is not statistically increased from the 1998-2000 baseline (P = .34).

View larger version (31K): [in this window] [in a new window] Figure. Number of Potential West Nile Virus Cases per 1000 Patient Encounters as Detected by Single-Institution Surveillance Horizontal bars indicate the mean number of potential cases per 1000 patient encounters during the period July-October of each year.

Comment
A syndromic surveillance system based exclusively on ICD-9 coding data at our hospital would not have detected the emergence of WNV in Chicago during 2002. Moreover, the same system would have sounded an alert for an outbreak during the summer of 2001, during which time no WNV cases or outbreaks of other neurological disease were known to have occurred.

Although a number of methodological factors are likely to have contributed to the poor performance of the surveillance system described in this report, we believe that our experience illustrates fundamental shortcomings in the reliability of single-institution systems in general. Although an excessive number of WNV cases were seen in Chicago in 2002, only a small fraction of these patients presented to our hospital. The limited sensitivity of the surveillance system failed to detect the emerging epidemic. Moreover, low local incidence limited the positive predictive value of the system, which generated a false-positive result in the year 2001.

Given that a future bioterrorism attack might initially be characterized by a relatively small number of individuals seeking care in a nonuniform distribution across the affected region, the number of cases presenting to a single institution would be small. Therefore, a significant attack could go undetected even by a well-designed single-institution surveillance system. Furthermore, the limited positive predictive value of such a system could cause public health resources to be needlessly exhausted in the investigation of false-positive results.

With this in mind, rather than investing in potentially flawed single-institution systems, resources might be better directed to the validation of city-wide and regional surveillance or the training of individual clinicians in the recognition of potential bioterrorism-related syndromes.

Stephen G. Weber, MD, MS; David Pitrak, MD
Department of Medicine
University of Chicago Hospitals
Chicago, Ill

1. Illinois Department of Public Health. West Nile Virus: 2002 Human Case Data. Available at: http://www.idph.state.il.us/envhealth/wnvsurveillance_humancases_02.htm. Accessed January 14, 2003.

Letters Section Editor: Stephen J. Lurie, MD, PhD, Senior Editor.

JAMA. 2003;290:596-598.

CiteULike    Connotea    Del.icio.us    Digg    Reddit    Technorati    Twitter     What's this?