|
Activity Number:
|
143
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statisticians in Defense and National Security
|
| Abstract - #305667 |
|
Title:
|
A Bayesian Approach to Radiographic Surveillance in Children
|
|
Author(s):
|
Namhee Kim*+ and Prem K. Goel and Bema Bonsu and M. Marvin
|
|
Companies:
|
The Ohio State University and The Ohio State University and Columbus Children's Hospital and Children's Hospital
|
|
Address:
|
320 W. 10th Ave., Columbus, OH, 43210,
|
|
Keywords:
|
bivariate dynamic model ; biosurveillance ; time series ; outbreak detection
|
|
Abstract:
|
Recently, there has been a growing demand on timely detection of abnormal pattern in the time-series data with epidemiological interest. The spread of SARS virus in 2003 provides an example to motivate this study. We focus on detection of respiratory diseases with abnormal daily count of children who have to undergo chest radiograph at an emergency department (ED). We use bivariate dynamic time-series model for the daily counts of children coming to ED (census) and those given chest radiographs. To detect abnormal departure from realistic underlying situation, it is crucial to learn the regular pattern from historical data and possible variants that already exist in historical data, but not involving outbreaks. These issues are addressed and a decision rule to suggest the existence of outbreak in the current data is discussed and implemented with artificially attempted outbreaks.
|
