The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #63031

Howard S. Burkom, PhD¹, Eugene Elbert, MS², Christina S. Polyak, MPH², and Julie Pavlin, MD, MPH². (1) National Security Technology Department, Johns Hopkins Applied Physics Laboratory, 11100 Johns Hopkins Road, Laurel, MD 20723, 240-228-4361, Howard.Burkom@jhuapl.edu, (2) Walter Reed Army Institute of Research, DoD Global Emerging Infections System (DoD-GEIS), 503 Robert Grant Avenue, Silver Spring, MD 20910

Recently, the public health community has devoted much energy to collecting and analyzing a variety of data sources for biosurveillance. Among the nontraditional data sources being examined, records of health care provider office visits (OV) are a rich source of information that also have the potential for enhanced early alerting capability. In case of an outbreak of a disease whose presentation includes a prodromal period, it is plausible that a substantial number of patients would visit their physicians before their symptoms become acute enough to require emergency care. However, the utility of the OV visits depends upon prompt reporting of daily visit counts so that alerting algorithms can be applied. The U.S. Department of Defense Global Emerging Infections System, the Johns Hopkins University Applied Physics Laboratory, and other partners have developed the ESSENCE II syndromic surveillance system that includes OV data from several sources. The variability of reporting promptness differs with each data source. Given the recent visit counts and the reporting patterns measured in past data, we apply survival analysis techniques and regression strategies to estimate the actual number of patient encounters, including those yet unreported, for each monitored syndrome group. Regressors considered include the appointment day of week, distribution of syndromes, and the type of provider information available from each data source. In some data sources, we restrict attention to counts from reliably early reporters. We evaluate this approach by applying alerting algorithms first to the estimated counts and later to the total counts.

Learning Objectives:

• Discuss methods to overcome late data reporting for timely outbreak alerting.
• List the primary covariates to consider when adjusting for late reporting.

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.

Handout (.ppt format, 1748.5 kb)