Application of an automated surveillance-data-analysis system in a laboratory-based early-warning system for detection of an abortion outbreak in mares.

Abstract

OBJECTIVE: To develop an early-warning automated surveillance-data-analysis system for early outbreak detection and reporting and to assess its performance on an abortion outbreak in mares in Kentucky. SAMPLE POPULATION: 426 data sets of abortions in mares in Kentucky during December 2000 to July 2001. PROCEDURES: A custom software system was developed to automatically extract and analyze data from a Laboratory Information Management System database. The software system was tested on data on abortions in mares in Kentucky reported between December 1, 2000, and July 31, 2001. The prospective space-time permutations scan statistic, proposed by Kulldorff, was used to detect and identify abortion outbreak signals. RESULTS: Results indicated that use of the system would have detected the abortion outbreak approximately 1 week earlier than traditional surveillance systems. However, the geographic scale of analysis was critical for highest sensitivity in outbreak detection. Use of the lower geographic scale of analysis (ie, postal [zip code]) enhanced earlier detection of significant clusters, compared with use of the higher geographic scale (ie, county). CONCLUSIONS AND CLINICAL RELEVANCE: The automated surveillance-data-analysis system would be useful in early detection of endemic, emerging, and foreign animal disease outbreaks and might help in detection of a bioterrorist attack. Manual analyses of such a large number of data sets (ie, 426) with a computationally intensive algorithm would be impractical toward the goal of achieving near real-time surveillance. Use of this early-warning system would facilitate early interventions that should result in more positive health outcomes.