Using Machine Learning to Predict Swine Movements

A collaborative work between the University of Minnesota, UC- Davis, and Pipestone Veterinary Services was published this past month in the journal Frontiers in Veterinary Science.
Between-farm animal movement, despite being an essential factor of infectious disease spread is not currently recorded in the US. The objective of this project was to create a model to predict animal movement based on between-site distance, ownership, and production type of the sending and receiving farms. The model was able to predict animal movement in the south-central region of the study area with a high aggregation. It also showed an overlap with the distribution of outbreaks of porcine reproductive and respiratory syndrome (PRRS) in this area.


Abstract: Between-farm animal movement is one of the most important factors influencing the spread of infectious diseases in food animals, including in the US swine industry. Understanding the structural network of contacts in a food animal industry is prerequisite to planning for efficient production strategies and for effective disease control measures. Unfortunately, data regarding between-farm animal movements in the US are not systematically collected and thus, such information is often unavailable. In this paper, we develop a procedure to replicate the structure of a network, making use of partial data available, and subsequently use the model developed to predict animal movements among sites in 34 Minnesota counties. First, we summarized two networks of swine producing facilities in Minnesota, then we used a machine learning technique referred to as random forest, an ensemble of independent classification trees, to estimate the probability of pig movements between farms and/or markets sites located in two counties in Minnesota. The model was calibrated and tested by comparing predicted data and observed data in those two counties for which data were available. Finally, the model was used to predict animal movements in sites located across 34 Minnesota counties. Variables that were important in predicting pig movements included between-site distance, ownership, and production type of the sending and receiving farms and/or markets. Using a weighted-kernel approach to describe spatial variation in the centrality measures of the predicted network, we showed that the south-central region of the study area exhibited high aggregation of predicted pig movements. Our results show an overlap with the distribution of outbreaks of porcine reproductive and respiratory syndrome, which is believed to be transmitted, at least in part, though animal movements. While the correspondence of movements and disease is not a causal test, it suggests that the predicted network may approximate actual movements. Accordingly, the predictions provided here might help to design and implement control strategies in the region. Additionally, the methodology here may be used to estimate contact networks for other livestock systems when only incomplete information regarding animal movements is available.

Link to the full article

Reducing the likelihood of a piglet reservoir when dealing with influenza in swine herds

Drs. White, Torremorell and Craft from the University of Minnesota recently published an article in Preventative Veterinary Medicine regarding practices that can decrease the likelihood of creating an endemic piglet reservoir in the case of an infection by swine influenza. Indeed, a stochastic model was developed considering that the pigs were in one of the following categories: Susceptible, Exposed,  Infectious, Recovered, or Vaccinated. Loss of immunity over time and differences between naturally infected and vaccinated animals were taken into account. Several scenarios were evaluated regarding their impact on piglet prevalence: timing of gilt introductions, gilt separation, gilt vaccination upon arrival, early weaning, and sow vaccination strategies.

In this model, homologous mass vaccination and early weaning were the most efficacious interventions. By combining frequent homologous mass vaccination, early weaning, gilt separation, gilt vaccination and longer periods between gilt introductions reduced endemic prevalence overall by 51% relative to the null scenario and the endemic prevalence in piglets by 74%.


Abstract: Recent modelling and empirical work on influenza A virus (IAV) suggests that piglets play an important role as an endemic reservoir. The objective of this study is to test intervention strategies aimed at reducing the incidence of IAV in piglets and ideally, preventing piglets from becoming exposed in the first place. These interventions include biosecurity measures, vaccination, and management options that swine producers may employ individually or jointly to control IAV in their herds. We have developed a stochastic Susceptible-Exposed-Infectious-Recovered-Vaccinated (SEIRV) model that reflects the spatial organization of a standard breeding herd and accounts for the different production classes of pigs therein. Notably, this model allows for loss of immunity for vaccinated and recovered animals, and for vaccinated animals to have different latency and infectious periods from unvaccinated animals as suggested by the literature. The interventions tested include: (1) varied timing of gilt introductions to the breeding herd, (2) gilt separation (no indirect transmission to or from the gilt development unit), (3) gilt vaccination upon arrival to the farm, (4) early weaning, and (5) vaccination strategies of sows with different timing (mass and pre-farrow) and efficacy (homologous vs. heterologous). We conducted a Latin Hypercube Sampling and Partial Rank Correlation Coefficient (LHS-PRCC) analysis combined with random forest analysis to assess the relative importance of each epidemiological parameter in determining epidemic outcomes. In concert, mass vaccination, early weaning of piglets (removal 0–7 days after birth), gilt separation, gilt vaccination, and longer periods between introductions of gilts (6 months) were the most effective at reducing prevalence. Endemic prevalence overall was reduced by 51% relative to the null case; endemic prevalence in piglets was reduced by 74%; and IAV was eliminated completely from the herd in 23% of all simulations. Importantly, elimination of IAV was most likely to occur within the first few days of an epidemic. The latency period, infectious period, duration of immunity, and transmission rate for piglets with maternal immunity had the highest correlation with three separate measures of IAV prevalence; therefore, these are parameters that warrant increased attention for obtaining empirical estimates. Our findings support other studies suggesting that piglets play a key role in maintaining IAV in breeding herds. We recommend biosecurity measures in combination with targeted homologous vaccination or vaccines that provide wider cross-protective immunity to prevent incursions of virus to the farm and subsequent establishment of an infected piglet reservoir.

Link to the full article

Advances in Mycoplasma hyopneumoniae elimination: a podcast series

This past month, the Morrison group invited Dr. Paul Yeske, swine practitioner at the Swine Vet Center (St. Peter, MN), Dr. Amanda Sponheim, PhD candidate at the University of Minnesota and Support Veterinarian at Boerhinger Ingelheim, and Dr. Maria Pieters from the University of Minnesota to discuss the latest progress made in successfully eliminating Mycoplasma hyopeumoniae from swine herds. Dr. Pieters is the head of the MycoLab at the College of Veterinary Medicine and focuses on diagnostics and epidemiology of swine mycoplasms to help veterinarians control associated diseases.

  1. History of Mycoplasma hyopneumoniae herd elimination and practices: podcast
  2. Sampling techniques and protocols to use during the process of elimination: podcast
  3. Starting the elimination: when is day zero? podcast

The podcasts in the press