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

Abstract #68724

Carol A. Sniegoski, MS, National Security Technology Department, Johns Hopkins Applied Physics Laboratory, 11100 Johns Hopkins Road, Laurel, MD 20723, 240-228-6333, carol.sniegoski@jhuapl.edu

Syndromic surveillance, in support of either core public health surveillance or bioterrorism defense, aims to achieve early detection of disease outbreaks by monitoring patient data grouped into generalized syndrome categories. Emergency department chief complaint data is highly suited to this method of surveillance. It has the advantages of being routinely generated during normal hospital operations, available the same day the patient is seen, and, in the majority of cases, available in electronic format. A major barrier to using chief complaint data is posed by its unstructured free text format. Hand-coding data into syndrome categories, whether performed onsite in the emergency department or offsite in the public health office, requires significant time and labor. To make chief complaint data more realistically usable for ongoing surveillance, we developed an automated syndromic categorization application. It has been used for the past year to classify electronic chief complaint records collected daily from area hospitals. This presentation will share our experiences with automated chief complaint categorization in the following areas: characterization of the challenging nature of the data (i.e., poor data quality, prevalence of abbreviations and misspellings, context-sensitive vocabulary, inter-hospital variation); usability considerations (i.e., performance, training, intended user base, providing for refining syndrome criteria); an overview comparing the approach taken to other existing approaches and/or implementations; and an analysis of accuracy rates and types of errors seen. This presentation should be of interest to those using or considering using chief complaint data for surveillance and wishing to understand issues in classifying it efficiently and effectively.

Learning Objectives:

• Describe considerations in designing an application to perform automated syndromic categorization of emergency department chief complaint data
• Outline major existing approaches and implementations and understand their advantages and disadvantages
• Describe in detail the success rate of the discussed application and lessons learned.

Keywords: Public Health Informatics, Data/Surveillance

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.