Improving Syndromic Surveillance by Data Integration

点击翻译按钮获取中文摘要

The term "syndromic surveillance" is a generic term applied to a variety of newly developed methods in public health practice. For our purposes, syndromic surveillance refers to the automated collection and analysis in near-real-time of electronic health outcome data. Used in this way, syndromic surveillance sits within a broader category of "biosurveillance", the routine collection and analysis of electronic data falling outside of the classical surveillance paradigm. We propose a research program to improve the performance of aberration detection methods for syndromic surveillance using statistical methods of data integration. Our program focuses on three main areas of potential improvement: temporal modeling, spatio-temporal clustering, and integration of multiple data streams. We also include a research translation component, in order to ensure that the results of research will be of practical use to health departments and other practitioners of syndromic surveillance. Our three specific aims are: 1) To develop and improve temporal modeling for syndromic surveillance, using improved seasonal models and Hidden Markov Models (HMMs); 2) To investigate and evaluate data integration methods, including spatio-temporal clustering and multiple data source integration; 3) To develop PHIN-compliant software for use by local health departments and syndromic surveillance practitioners.