Syndromic Surveillance with Multiple Data Streams

• Activity Number: 323
• Type: Invited
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Cmte on Statisticians in Defense and National Security
• Author(s): Martin Kulldorff*+
• Companies: Harvard Medical School and Harvard Pilgrim Healthcare
• Keywords: spatial statistics ; public health ; surveillance ; scan statistics ; permutation tests

Abstract

The diagnosis of disease often requires multiple lab tests that may take anything from a day to a couple of weeks after the first symptoms arise. Whether due to bioterrorism or natural causes, it is thought that syndromic surveillance based on symptoms like vomiting and fever may be more useful for the early detection of disease outbreaks than diagnosed disease events, even though there is more noise in the data. Available syndromic surveillance data sources include hospital emergency visits, ambulance dispatch calls, and over-the-counter pharmacy sales. Most current syndromic surveillance systems evaluate each data stream independently. We will present statistical methods that simultaneously incorporates multiple data streams into a single likelihood function, which is then used with a permutation-based space-time scan statistics for the early detection of disease outbreaks. The method will be illustrated using syndromic surveillance data from New York City.