A Binomial Model of Transients
in Daily ED Visits for Detecting Infectious Disease Outbreaks
March 2005
James Dunyak, The MITRE Corporation
Kenneth Mandl, The MITRE Corporation
Mojdeh Mohtashemi, The MITRE Corporation
ABSTRACT
The threat of biological warfare and the emergence of new infectious
agents spreading at a global scale have highlighted the need for major
enhancements to the public health infrastructure. Effective confrontation
of these urgent crises requires rapid and accurate detection of unusual
epidemiologic trends, for which our current surveillance capabilities
are not adequate. Critical for real time surveillance are two components:
real-time data and real-time interpretation of data. Today, most existing
surveillance systems are capable of monitoring and capturing real time
data. However, the state of practice for detecting temporal and spatial
abnormalities in surveillance data remains inadequate. We introduce
a locally stationary binomial model of early detection of epidemiologic
events, applied to real historical data pertaining to the daily number
of visits with respiratory syndromes to the emergency department (ED).
We show that when simulated outbreaks are introduced into the respiratory
data, our uniformly most powerful detection algorithm under a constant
false alarm rate is capable of detecting such irregularities in the
data with high sensitivity, specificity, and in a timely manner.
Additional Search Keywords
N/A
|
