The importance of integrating multiple data streams in the swine industry
Using production data and off-feed events to detect early signs of PRRSV outbreaks.
by Mafalda Pedro Mil-Homens and Edison Magalhaes
Photo Credit: acceptfoto - stock.adobe.com
Despite the large amount of data collected from sow farms daily, there are significant gaps in utilizing production data for real-time surveillance.
Most sow farms have data record-keeping systems to collect and monitor farm productivity performance.
Photo Credit: dusanpetkovic1- stock.adobe.com
However, there needs to be integration between data collected at the farm and available tools to help the swine production systems identify productivity deviations in real-time that are accurate and sensitive to detect small changes.
A disease surveillance project is being implemented at Iowa State University, where feed and sow production data are collected and analyzed to identify signs related to PRRSV outbreaks.
The preliminary results showed that these signs appeared between one and six weeks before the farm got a positive diagnostic using variables such as off-feed events, number of abortions, and number of dead sows.
Therefore, using multiple data sources routinely collected at the farm enables production systems to implement continuous surveillance programs providing strong support for decision-making in the production systems.
Funding for the disease surveillance project was provided by: AASV - the American Association of Swine Veterinarians and USDA - the National Institute of Food and Agriculture.
The Prosper Platform
The Predictors of Swine Performance (PROSPER) platform project—from Iowa State University—consists of developing an ongoing and automated system-specific platform to:
- Step 1: Consolidate whole-herd big data, from birth to market, matching and merging information from the data stream available in a production system to their respective cohorts of pigs marketed. Algorithms will be built specifically for the system`s data streams, which will allow ongoing data integration and analysis;
- Step 2: Analyze whole-herd big data to identify and measure the association of specific parameters on KPIs (A), here focusing on W2F mortality. Furthermore, predictive models can be applied to forecast the downstream performance of the cohorts based on regression and machine learning algorithms (B).
Once whole herd data has been consolidated into one swine production system, it is possible to measure the effect of specific factors impacting the performance under the conditions of this company, such as the impact of diseases at the sow farm (PRRS, Mycoplasma, PED, and others) on the downstream W2F mortality of the weaned cohorts.
A similar analysis can be conducted to measure the efficacy of interventions, vaccination, and management protocols.
Lastly, it is possible to use information from the sow farm to forecast the downstream nursery mortality of groups, using their retrospective performance in the sow farm as predictors.
Figure 1. Forecasting mortality using PROSPER
Preliminary results from our analyses had an accuracy of 77.14 percent for predicting nursery groups that would have high nursery mortality (>5 percent) (Figure 1).
The development of the PROSPER platform has demonstrated the importance of each production system having the capability to have a near-real-time platform to consolidate and analyze data routinely, supporting data-driven decisions, delivering field-based results, and a whole herd data-driven approach.
Funding: USDA - National Institute of Food and Agriculture, The National Pork Board, and The Foundation for Food and Agriculture.
Mafalda Pedro Mil-Homens
Mafalda graduated in 2020 from the Faculty of Veterinary Medicine, University of Lisbon, Portugal, where she got her DVM and a Master’s in dairy cattle biosecurity. After graduation, she worked in dairy cattle biosecurity and welfare at the Veterinary Associated Services. In 2021 she started a Master’s program at Iowa State University (ISU) under the orientation of Dr. Gustavo Silva, focused on biosecurity in supply entry rooms in swine farms, which she completed in 2022. Currently, Mafalda is pursuing her Ph.D. at ISU, working in disease surveillance.
Edison Magalhaes DVM, MS, PhD(c)
Edison got his DVM in 2015 in Brazil and worked for 4 years in the Brazilian swine industry as a swine practitioner at Brazil Foods Company (BRF). In 2019 he came to the US to do his Master’s degree in Preventive Veterinary Medicine at Iowa State University, which he finished in 2021. Currently, he is pursuing his Ph.D. Degree in Population Sciences in Animal Health at the same University, under the orientation of Dr. Daniel Linhares. He focuses on developing automated models to manage and analyze swine data collected under field conditions, providing insights for production systems that can support data-driven decisions.