Development of a pig wean-quality score using machine-learning algorithms to characterize and classify groups with high mortality risk under field conditions.

Mortality during the post-weaning phase is a critical indicator of swine production system performance, influenced by a complex interaction of multiple factors of the epidemiological triad. This study leveraged retrospective data from 1723 groups of pigs marketed within a US swine production system to develop a Wean-Quality Score (WQS) using machine learning techniques. The study evaluated three machine learning models, Random Forest, Support Vector Machine, and Gradient Boosting Machine, to classify groups having high or low 60-day mortality, where high mortality groups represented 25 % of the groups among the study population with the highest mortality values (n=431; 60-day mortality=9.98 %), and the remaining 75 % of the groups were of low mortality (n=1292; 60-day mortality=2.75 %). The best-performing model, Random Forest (RF), outperformed the other ML models in terms of accuracy (0.90), sensitivity (0.84), and specificity (0.92) metrics, and was then selected for further analysis, which consisted of creating the WQS and ranking the most important factors for classifying groups as high or low mortality. The most important factors ranked through the RF model to classify groups with high mortality were pre-weaning mortality, weaning age, average parity of litters in sow farms, and PRRS status. Additionally, stocking conditions such as stocking density and time to fill the barn were important predictors of high mortality. The WQS was developed and correlated (r = 0.74) with the actual 60-day mortality of the groups, offering a valuable tool for assessing post-weaning survivability in swine production systems before weaning. This study highlights the potential of machine learning and comprehensive data utilization to improve the assessment and management of weaned pig quality in commercial swine production, which producers can utilize to identify and intervene in groups, according to the WQS.

Risk factors associated with piglet pre-weaning mortality in a Midwestern U.S. swine production system from 2020 to 2022.

Piglet pre-weaning mortality (PWM) is a significant issue in the U.S. swine industry, causing economic losses and raising sustainability and animal welfare concerns. This study conducted a multivariable analysis to identify factors associated with PWM in a Midwestern U.S. swine production system. Weekly data from 47 sow farms (7207 weaning weeks) were captured from January 2020 to December 2022. Initially, 29 variables regarding farm infrastructure, productivity parameters, health status, and interventions were selected for univariate analysis to assess their association with PWM. The initial multivariable analysis included the variables with P < 0.20 in the univariate analyses. A backward stepwise model selection was conducted by excluding variables with P > 0.05, and the final multivariable model consisted of 19 significant risk factors and 6 interaction terms. The overall average PWM for the study population was 14.02 %. Yearly variations in PWM were observed, with the highest recorded in 2020 (16.61 %) and the lowest in 2021 (15.78 %). Cohorts with a pond water source, lower farrowing rate (71.9 %), higher farrowing parity (5.1), shorter gestation length (116.2 days), and using oxytocin during farrowing had increased PWM. The higher productivity parameters such as mummies rate, stillborn rate, and average total born, the higher the PWM was. Additionally, health status and intervention-related factors were associated with PWM, where higher PWM rates were observed in herds facing porcine reproductive and respiratory syndrome virus (PRRSV) outbreaks, porcine epidemic diarrhea virus (PEDV) positive, the weeks before and during feed medication, and weeks without using Rotavirus vaccine or Rotavirus feedback. Altogether, these results corroborate that PWM is a multifactorial problem, and a better understanding of the risk factors is essential in developing strategies to improve survival rates. Therefore, this study identified the major risk factors associated with PWM for groups of pigs raised under field conditions, and the results underscore the significance of data analysis in comprehending the unique challenges and opportunities inherent to each system.

Field Implementation of Forecasting Models for Predicting Nursery Mortality in a Midwestern US Swine Production System.

The performance of five forecasting models was investigated for predicting nursery mortality using the master table built for 3242 groups of pigs (~13 million animals) and 42 variables, which concerned the pre-weaning phase of production and conditions at placement in growing sites. After training and testing each model's performance through cross-validation, the model with the best overall prediction results was the Support Vector Machine model in terms of Root Mean Squared Error (RMSE = 0.406), Mean Absolute Error (MAE = 0.284), and Coefficient of Determination (R2 = 0.731). Subsequently, the forecasting performance of the SVM model was tested on a new dataset containing 72 new groups, simulating ongoing and near real-time forecasting analysis. Despite a decrease in R2 values on the new dataset (R2 = 0.554), the model demonstrated high accuracy (77.78%) for predicting groups with high (>5%) or low (<5%) nursery mortality. This study demonstrated the capability of forecasting models to predict the nursery mortality of commercial groups of pigs using pre-weaning information and stocking condition variables collected post-placement in nursery sites.

Utilizing productivity and health breeding-to-market information along with disease diagnostic data to identify pig mortality risk factors in a U.S. swine production system.

Aggregated diagnostic data collected over time from swine production systems is an important data source to investigate swine productivity and health, especially when combined with records concerning the pre-weaning and post-weaning phases of production. The combination of multiple data streams collected over the lifetime of the pigs is the essence of the whole-herd epidemiological investigation. This approach is particularly valuable for investigating the multifaceted and ever-changing factors contributing to wean-to-finish (W2F) swine mortality. The objective of this study was to use a retrospective dataset ("master table") containing information on 1,742 groups of pigs marketed over time to identify the major risk factors associated with W2F mortality. The master table was built by combining historical breed-to-market performance and health data with disease diagnostic records (Dx Codes) from marketed groups of growing pigs. After building the master table, univariate analyses were conducted to screen for risk factors to be included in the initial multivariable model. After a stepwise backward model selection approach, 5 variables and 2 interactions remained in the final model. Notably, the diagnosis variable significantly associated with W2F mortality was porcine reproductive and respiratory syndrome virus (PRRSV). Closeouts with clinical signs suggestive of Salmonella spp. or Escherichia coli infection were also associated with higher W2F mortality. Source sow farm factors that remained significantly associated with W2F mortality were the sow farm PRRS status, average weaning age, and the average pre-weaning mortality. After testing for the possible interactions in the final model, two interactions were significantly associated with wean-to-finish pig mortality: (1) sow farm PRRS status and a laboratory diagnosis of PRRSV and (2) average weaning age and a laboratory diagnosis of PRRS. Closeouts originating from PRRS epidemic or PRRS negative sow farms, when diagnosed with PRRS in the growing phase, had the highest W2F mortality rates. Likewise, PRRS diagnosis in the growing phase was an important factor in mortality, regardless of the average weaning age of the closeouts. Overall, this study demonstrated the utility of a whole-herd approach when analyzing diagnostic information along with breeding-to-market productivity and health information, to measure the major risk factors associated with W2F mortality in specified time frames and pig populations.

A cross-sectional assessment of PRRSV nucleic acid detection by RT-qPCR in serum, ear-vein blood swabs, nasal swabs, and oral swabs from weaning-age pigs under field conditions.

Introduction: The porcine reproductive and respiratory syndrome virus (PRRSV) continues to challenge swine production in the US and most parts of the world. Effective PRRSV surveillance in swine herds can be challenging, especially because the virus can persist and sustain a very low prevalence. Although weaning-age pigs are a strategic subpopulation in the surveillance of PRRSV in breeding herds, very few sample types have been validated and characterized for surveillance of this subpopulation. The objectives of this study, therefore, were to compare PRRSV RNA detection rates in serum, oral swabs (OS), nasal swabs (NS), ear-vein blood swabs (ES), and family oral fluids (FOF) obtained from weaning-age pigs and to assess the effect of litter-level pooling on the reverse transcription-quantitative polymerase chain reaction (RT-qPCR) detection of PRRSV RNA. Methods: Three eligible PRRSV-positive herds in the Midwestern USA were selected for this study. 666 pigs across 55 litters were sampled for serum, NS, ES, OS, and FOF. RT-qPCR tests were done on these samples individually and on the litter-level pools of the swabs. Litter-level pools of each swab sample type were made by combining equal volumes of each swab taken from the pigs within a litter. Results: Ninety-six piglets distributed across 22 litters were positive by PRRSV RT-qPCR on serum, 80 piglets distributed across 15 litters were positive on ES, 80 piglets distributed across 17 litters were positive on OS, and 72 piglets distributed across 14 litters were positive on NS. Cohen's kappa analyses showed near-perfect agreement between all paired ES, OS, NS, and serum comparisons (). The serum RT-qPCR cycle threshold values (Ct) strongly predicted PRRSV detection in swab samples. There was a ≥ 95% probability of PRRSV detection in ES-, OS-, and NS pools when the proportion of positive swab samples was ≥ 23%, ≥ 27%, and ≥ 26%, respectively. Discussion: ES, NS, and OS can be used as surveillance samples for detecting PRRSV RNA by RT-qPCR in weaning-age pigs. The minimum number of piglets to be sampled by serum, ES, OS, and NS to be 95% confident of detecting ≥ 1 infected piglet when PRRSV prevalence is ≥ 10% is 30, 36, 36, and 40, respectively.

Whole-herd risk factors associated with wean-to-finish mortality under the conditions of a Midwestern USA swine production system.

Swine wean-to-finish (W2F) mortality is a multifactorial, dynamic process and a key performance indicator of commercial swine production. Although swine producers typically capture the relevant data, analysis of W2F mortality risk factors is often hindered by the fact that, even if data is available, they are typically in different formats, non-uniform, and dispersed among multiple unconnected databases. In this study, an automated framework was created to link multiple data streams to specific cohorts of market animals, including sow farm productivity parameters, sow farm and growing pig health factors, facilities, management factors, and closeout data from a Midwestern USA production system. The final dataset (master-table) contained breeding-to-market data for 1,316 cohorts of pigs marketed between July 2018 and June 2019. Following integration into a master-table, continuous explanatory variables were categorized into quartiles averages, and the W2F mortality was log-transformed, reporting geometric mean mortality of 8.69 % for the study population. Further, univariate analyses were performed to identify individual variables associated with W2F mortality (p < 0.10) for further inclusion in a multivariable model, where model selection was applied. The final multivariable model consisted of 13 risk factors and accounted for 68.2 % (R2) of the variability of the W2F mortality, demonstrating that sow farm health and performance are closely linked to downstream W2F mortality. Higher sow farm productivity was associated with lower subsequent W2F mortality and, conversely, lower sow farm productivity with higher W2F mortality e.g., groups weaned in the highest quartiles for pre-weaning mortality and abortion rate had 13.5 %, and 12.5 %, respectively, which was statistically lower than the lowest quartiles for the same variables (10.5 %, and 10.6 %). Moreover, better sow farm health status was also associated with lower subsequent W2F mortality. A significant difference was detected in W2F mortality between epidemic versus negative groups for porcine reproductive and respiratory syndrome virus (15.4 % vs 8.7 %), and Mycoplasma hyopneumoniae epidemic versus negative groups (13.7 % vs 9.9 %). Overall, this study demonstrated the application of a whole-herd analysis by aggregating information of the pre-weaning phase with the post-weaning phase (breeding-to-market) to identify and measure the major risk factors of W2F mortality.

Porcine reproductive and respiratory syndrome virus RNA detection in tongue tips from dead animals.

The control of porcine reproductive and respiratory syndrome virus (PRRSV) hinges on monitoring and surveillance. The objective of this study was to assess PRRSV RNA detection by RT-PCR in tongue tips from dead suckling piglets compared to serum samples, processing fluids, and family oral fluids. Tongue tips and serum samples were collected from three PRRSV-positive breeding herd farms (farms A, B, and C) of three different age groups: newborns (<24 h), processing (2 to 7 days of age), and weaning (18 to 22 days of age). Additionally, processing fluids and family oral fluids were collected from 2-7 days of age and weaning age, respectively. In farms A and B, PRRSV RNA was detected in tongue tips from all age groups (100 and 95%, respectively). In addition, PRRSV RNA was detected in pooled serum samples (42 and 27%), processing fluids (100 and 50%), and family oral fluids (11 and 22%). Interestingly, the average Ct value from tongue tips was numerically lower than the average Ct value from serum samples in the newborn age. In farm C, PRRSV RNA was only detected in serum samples (60%) and family oral fluids (43%), both from the weaning age. Further, no PRRSV RNA was detected in tongue tips when pooled serum samples from the same age group tested PRRSV RNA-negative. Taken together, these results demonstrate the potential value of tongue tips for PRRSV monitoring and surveillance.

Next Generation of Voluntary PRRS Virus Regional Control Programs.

Porcine reproductive and respiratory syndrome virus (PRRSV) became pandemic in the 1980's and today remains one of the most significant pathogens of the global swine industry. At the herd level, control of PRRSV is complicated by its extreme genetic diversity and its ability to persist in pigs, despite an active immune response. Ultimately, PRRSV control or elimination requires the coordination and active cooperation of producers and veterinarians at the regional level. Early voluntary PRRSV regional control programs focused on routine diagnostic testing and voluntary data-sharing regarding the PRRSV status of participants' herds, but no pre-defined action plans or decision trees were developed to secure project successes (or recover from failures). Given that control of PRRSV is paramount to producer profitability, we propose a coordinated approach for detecting, controlling, and ultimately eliminating wild-type PRRSV from herds participating in regional projects. Fundamental to project success is real-time, multi-platform communication of all data, information, and events that concern the regional project and project participants. New to this approach is the concept of agreed-upon action plans to be implemented by project participants in response to specific events or situations. The simultaneous and coordinated implementation of these strategies allows for early detection of wild-type PRRSV virus introductions and rapid intervention based on agreed-upon response plans. An example is given of a project in progress in the Midwest USA.