Issue 26 - Identifying the Most Useful Model for Syndromic Surveillance

_______________________________________________________________________________

Identifying the Most Useful Model for Syndromic Surveillance
_______________________________________________________________________________

Nationwide, cities and counties across the country are making decisions about how they will conduct syndromic surveillance in their communities. Roles are being identified, personnel assigned, hospitals contacted for data, and software and hardware purchased. Such is the case in the Dallas – Fort Worth Metropolitan Area (DFW).

DFW is a thriving metropolitan area of over 5 million people. The pioneer trading post of 1840 is now the home to many Fortune 500 companies, including American and Southwest Airlines, CompUSA, J.C. Penney, EDS and Texas Instruments. In the 1990’s DFW ranked first in the nation for employment growth. (1) Thus, DFW is an attractive target for terrorists.

The four county health departments (Dallas, Tarrant, Denton, Collin) that encompass the DFW area are not waiting for a potential disaster to unfold. Rather, they are pro-actively making decisions regarding how they will conduct syndromic surveillance within their community. In a meeting recently at the Dallas-Fort Worth Hospital Council, Dr Robert Haley, Professor, Chief of Epidemiology, of the U.T. Southwestern Medical School and Dr Tracy Gustafson, VP, R&D, ICPA, Inc., discussed syndromic surveillance, the likely mortality rate in the DFW area if cases were not identified in the first three or 4 days (the optimal ‘window of opportunity’), and the way data should flow to make syndromic surveillance most useful.

According to Dr Haley, there are two current models for organizing the flow of data in a syndromic surveillance network: Centralized and Hierarchical. In the “Centralized” model, hospitals send data through the internet directly to a private contractor’s server (usually in another state) for compilation and analysis at a national level. The private contractor analyzes the data for all counties and regions, and if a spike is identified the contractor reports to the CDC, as well as local and state health departments. (2)

In the “Hierarchical” model, each hospital has a direct reporting relationship with their local health department. Each hospital automatically transmits data from their emergency department (ED) or billing system directly to the health department every day. The health department analyzes the data from all hospitals in their jurisdiction, investigates flags of problems identified, and transmits an appropriate, non-identified subset of the data to the state health department to contribute to statewide and national syndromic surveillance.

Dr. Haley argued that the Centralized model has serious disadvantages for early identification and intervention in bioterrorism attacks. First, because of HIPAA and related privacy concerns, out-of-state private contractors typically cannot collect identifiable patient information (doing so would require a data use agreement with every hospital, which few hospitals will sign). Therefore, every time a centralized contractor detects a potential outbreak “flag,” they cannot perform an investigation – they can’t even tell the local health authorities which patients are involved in the outbreak. While the local health authorities try to identify the ill patients, they may lose the “window of opportunity” so critical to saving lives.

In contrast, in the Hierarchical model, information is regularly sent directly from the hospitals to the local health department. The health department can receive personally identifiable data because there is a public health exemption in HIPAA. Thus, when a potential outbreak is detected, the health department can immediately identify the involved patients, and arrange for their charts to be reviewed or even contact the patients directly if necessary. This immediate look-back response could not only save lives of those initially stricken, but would also maximize the ‘window of opportunity’ for vaccination and/or prophylactic antibiotic treatment.

Beyond the issue of immediate look-back, there are several other important strengths of the Hierarchical model. Local control, rather than centralized national control, means that data can be “cleaned up” (reviewed for missing data, errors, inconsistencies, etc.) before transmittal. Second, focal outbreaks are easier to detect in a local database, whereas they could easily be overlooked in a huge centralized statewide and/or national database. Third, the hierarchical model allows sensitivity and specificity levels (flag thresholds) to be set locally, rather than at the state or national level. Only local personnel can reasonably determine how many potential outbreaks they are staffed to investigate. Fourth, local control of data flow allows routine automated collection of additional data items completely unrelated to the threat of bioterrorism attacks. A “dual-use” or “multi-use” surveillance system can benefit the hospital, health department, and the entire community. (3)

Finally, the Hierarchical Model promotes a close relationship and cooperation between the local health department and their healthcare facilities. This benefit goes beyond syndromic surveillance. For example, it establishes an automated reporting link for easier reporting of required notifiable diseases.

“Local control is clearly the key to successful syndromic surveillance,” says Dr Haley. “The US is too big to do this on a national level. It has got to be local, because that’s where diseases are prevented and controlled.”

1) http://www.nursingspectrum.com/TravelMoving/TravelNursing/DestinationDallas.htm

2) Dr Robert Haley, taken from a talk presented at the Dallas Fort Worth Hospital Council, August 20, 2004.

3) Dr Tracy Gustafson, taken from a talk presented at the Dallas Fort Worth Hospital Council, August 20, 2004.

Back to the top