Sequential Change-Point Detection in the Spread of Infectious Diseases

• Activity Number: 323
• Type: Invited
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Michael Baron*+
• Companies: The University of Texas at Dallas
• Keywords: CUSUM ; delay ; detection ; epidemic trend ; false alarm ; stopping time

Abstract

Several outbreaks of rare infectious diseases and seasonal epidemics of influenza occurred in recent years. A developing area of statistical research is detection of changes in patterns of spread of infectious diseases. Timely detection of such changes can predict the beginning and magnitude of an impendent epidemic and provide important information for environmental health. A desired algorithm must attain sufficient sensitivity to detect changes early and fast, controlling at the same time the probability of false alarms.

Epidemic models deal with stochastic processes that are marked by the presence of nuisance parameters, time-dependence, nonstationarity, and existence of rather complex prior information. Standard change-point detection methods are no longer optimal for such processes.

Under these conditions, we formulate the change-point detection problem and propose stopping rules for its solution. They can be chosen in an optimal way, satisfying constraints on the average detection delay, the rate of false alarms, and probability of non-detection. Sequential tools are applied to the recent data published by the Centers for Disease Control and Prevention.