SHIC-Funded Study Offers to Predict PEDV Outbreaks
Swine Health Information Center
via FarmJournal's Pork - March 7, 2019
The Swine Health Information Center (SHIC) collaborated with the Morrison Swine Health Information Project to enable a study applying machine-learning to predict porcine epidemic diarrhea virus (PED) outbreaks on sow farms. The researchers were able determine it is possible to predict the probability of an outbreak when considering animal movements and environmental conditions. Another goal was to see if shared producer data could be used to develop critical tools for the prevention of disease spread and implementation of risk mitigation. Further, this work serves as a model for near real-time disease forecasting. The authors hope it will advance disease surveillance and control for endemic swine pathogens in the U.S.
Many mechanisms play an apparent role in the spread of viruses: movement of infectious animals, airborne spread of aerosols, wildlife, contaminated fomites, feed, and personnel. Understanding the complexity of animal movements as a whole – routes, volumes, and frequency – is essential. The broader view used in the study, rather than focusing solely on a specific farm, helps to better understand PED epidemiology and spread by analyzing the cumulative effect of animal movement and environment on infection risk.
Analyzing data from PED outbreaks during 2015 as well as a large, retrospective dataset, the study was able to correctly predict when an outbreak occurred during one-week periods with greater than 80% accuracy. Because they used a neighborhood-based approach, researchers were able to simultaneously capture disease risks associated with long-distance animal movement as well as local neighborhood effects. They defined a neighborhood as the area 10 kilometers around a farm. They evaluated the relative importance of neighborhood effects in determining infection risk. This included animal movements to farms nearby, hog density in the area, environmental factors, and the landscape.
The team used a machine-learning technique to analyze the data and found...
more, including links