The swine industry understands the epidemiology of porcine epidemic diarrhea virus (PEDv) and porcine reproductive and respiratory syndrome virus (PRRSv) better than ever. Still, the ability of producers to effectively estimate spatial and temporal variation in disease outbreak risk is lacking. By using data contributed by the Morrison Swine Health Monitoring Project (MSHMP), researchers at the University of Minnesota (UMN) built machine learning algorithms that predict whether a sow farm will break with PEDv two weeks in advance.
This research project intended to generate farm-level forecasts of PEDv risk, considering recent animal movements, present disease distribution, and environmental factors. Producers will benefit from anonymously reported information from others in the region to better understand their farms’ risk. Once a farm is flagged with a high probability, systems and veterinarians can enact prevention and response measures to better navigate an outbreak, minimizing its impact. This work was partially funded by the Swine Health Information Center (SHIC).
The UMN-developed forecasting tool was able to detect approximately 20% of the outbreaks in a region (sensitivity), while 70% of the outbreaks the tool said would occur did indeed happen (positive predictive value).The most important factors related to a higher predicted risk of an outbreak were animal movements both into the sow farm and its neighbors, as well as the PED-status of the origin farm of these movements. Farm density and ambient temperature were also important predictors. Given information from the tool, how could producers react? Informing on-farm staff of predicted risk, review of training protocols, ensuring biosecurity adherence, changing pig flow, stocking adequate supplies for treatment, and preparation for higher mortality were listed by producers as possible responses to elevated risk.
UMN developed this tool using data for a single US swine-dense region and began applying it to a second region of the country, operating with different industry systems and a different environment. This forecasting tool is ready for real-time use; since December 2019, contributing systems have received weekly system-specific predictions of PEDv outbreak risk at farm level. These predictions are made two weeks into the future, which allows systems, veterinarians, and producers to act in case a high probability of an outbreak is forecasted. This provides an important tool for informed decision-making and coordinated actions of producers and practitioners to control or mitigate the impact of PEDv outbreaks.
In the long term, benefits to the pork industry may be more substantial. The forecasting tool was built to be able to handle different diseases, given that enough information on their occurrence is entered. As such, this tool presents a unique opportunity to face challenges the industry may encounter, even in the case of the introduction of foreign animal diseases in the US swine herd. Further development may be needed to ensure important aspects of the transmission of those diseases are adequately being entered in the model (such as the issue with strains/lineages/RFLPs for PRRSv), but the operationalization of this project puts the US industry in better spot to fight these challenges that may emerge.
As the world deals with the COVID-19 pandemic, SHIC continues to focus efforts on prevention, preparedness, and response to novel and emerging swine disease for the benefit of US swine health. As a conduit of information and research, SHIC encourages sharing of its publications and research. Forward, reprint, and quote SHIC material freely. SHIC is funded by America’s pork producers to fulfill its mission to protect and enhance the health of the US swine herd. For more information, visit http://www.swinehealth.org or contact Dr. Sundberg at psundberg@swinehealth.org.