Using and improving Early Aberration Reporting System (EARS) for local public health biosurveillance

Background: The flexibility of the syndrome building process in the Early Aberration and Reporting System (EARS) has proven to be the most useful feature of EARS compared to other biosurveillance tools, but is also the one feature most prone to programming errors. To ameliorate this issue, a collaborative academic/public health partnership was developed to provide an opportunity to study methods to reduce spurious syndrome matches in EARS. Methods: Quantitative and qualitative analysis of EARS' text matching algorithms was conducted to improve the accuracy of daily syndrome counts. This was accomplished by iteratively analyzing and comparing the symptoms as coded by EARS versus the actual chief complaint text and adding in additional text matching logic to eliminate spurious matches. Results: The refined set of algorithms for generating syndrome counts significantly improved the accuracy of the daily syndrome counts. In particular, the algorithms reduced the number of false positive matches between the terms listed in the EARS' symptom_code file and chief complaint free text data. Conclusions: Often, public health does not take advantage of the skills and knowledge available at local research institutions. This collaborative project demonstrated the importance of combining the skills and knowledge of academic scientists with the applied epidemiologic expertise and current, raw public health datasets available to them. It advanced biosurveillance methodologies and contributed to better public health surveillance and preparedness. In addition, this approach could be incorporated into the EARS system to improve the sensitivity and specificity of the system for all users.