IBM T J Watson Research Center Site Surveillance

IBM T. J. Watson Research Center Acknowledgements § Based on work of Vijay Iyengar, Ed Pednault § This material is based upon work supported by the Air Force Research Laboratory(AFRL)/Defense Advanced Research Projects Agency (DARPA) under AFRL Contract No. F 30602 -01 -C-0184. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the AFRL and/or DARPA. Approved for Public Release, Distribution Unlimited (5/3/2004). 2 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center Site-Based Bio-Surveillance The monitoring of a geographically constrained site with a relatively stable population for signs of disease outbreak. § Example of sites could include work sites, university campuses, or military bases § The population need not be present 24 hours a day 3 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center What makes “Site-Based” Bio-Surveillance Different? § Increased data availability – Central authority for permissions – Centralized data collection § “Permissive” – “Sensitive” data more likely to be available § Relatively stable population – May be more homogenous than general population § Geographically constrained – Spatial considerations are greatly reduced or eliminated 4 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center Differential Detection Approach § Define sites (regions) that normally track each other – Determine appropriate model for measured quantities • Quantify normal variation in the tracking § Detect significant deviations in the tracking – Signifies event affecting one of the sites 5 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center Differential Detection Approach Reference Site Target Site Reference

IBM T. J. Watson Research Center Experiments § Monitor phone calling patterns at two IBM sites – Yorktown, Hawthorne (10 miles apart) – Counts of calls/callers to medical facilities – Counts of all calls/callers – Currently being collected on a daily basis – Privacy ensured through • Anonymization of calling number • No reporting of called number 7 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center Method § Assume underlying Poisson process § Define two time windows – History, Test § Model ratio of counts in a window § Use Chi-squared statistic to detect deviations – Empirical variance estimate 8 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation

IBM T. J. Watson Research Center Medically-Related Calls 9 Site Surveillance Using Differential Detection

IBM T. J. Watson Research Center All Calls 10 Site Surveillance Using Differential Detection

IBM T. J. Watson Research Center Issues § Requires – Good tracking – Significant volumes § Can use – Raw counts – Counts adjusted by domain knowledge • If sites respond differently to some phenomenon 11 Site Surveillance Using Differential Detection 2004 -02 -19 © 2004 IBM Corporation