Dynamic Model for Detection of Animal Disease Outbreak

• Activity Number: 513
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Chellafe Ensoy*+ and Christel Faes and Marc Aerts
• Companies: Hasselt University and Hasselt University and Hasselt University
• Keywords: spatio-temporal model ; random walk ; INLA ; Bayesian

Abstract

A Bayesian spatio-temporal model for outbreak detection for both known and unknown diseases was developed using a standard surveillance system data. Information on cattle mortality and abortion for Belgium from 2005 to 2011 was used as it is believed that sudden increase in number of cases could imply the presence of a deadly disease. The model was developed in a Bayesian framework where different random walk assumptions on the time series structure were explored. The spatial structure of the data including incorporation of cattle movement information was also investigated as through this, detection of outbreak not only in time but also in space should be possible. Model inference is performed using the integrated nested Laplace approximation. The threshold for declaring an outbreak was computed based on the predictive posterior of a new observation with necessary multiple testing correction. The alarms generated by the model were verified with reported cattle outbreaks in Belgium during the prediction period.