Evaluation of a chief complaint pre-processor for biosurveillance.

Emergency Department (ED) chief complaint (CC) data are key components of syndromic surveillance systems. However, it is difficult to use CC data because they are not standardized and contain varying semantic and lexical forms for the same concept. The purpose of this project was to revise a previously-developed text processor for pre-processing CC data specifically for syndromic surveillance and then evaluate it for acute respiratory illness surveillance to support decisions by public health epidemiologists. We evaluated the text processor accuracy and used the results to customize it for respiratory surveillance. We sampled 3,699 ED records from a population-based public health surveillance system. We found equal sensitivity, specificity, and positive and negative predictive value of syndrome queries of data processed through the text processor compared to a standard keyword method on raw, unprocessed data.

Duke Scholars

Author Debra Travers School of Nursing