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.

Considerations in the use of processing fluids for the detection of PRRSV RNA and antibody.

Surveillance is mandatory for tracking the progress of porcine reproductive and respiratory syndrome virus (PRRSV) control and elimination efforts in breeding herds. Processing fluids, the fluid recovered from tissues collected at castration and/or tail docking, are used for breeding herd surveillance by large segments of the industry, but the basic diagnostic characteristics of processing fluids are largely undescribed. We undertook 3 studies to address this information gap. In study 1, we found no differences among the PRRSV RT-rtPCR results obtained with 4 commercial RNA extraction kits. In study 2, we found that PRRSV RNA was highly stable in processing fluid samples at -20°C or 4°C, but detrimental effects were observed at ≥22°C within 24 h. In study 3, using a modified PRRSV ELISA at a sample:positive cutoff of ≥0.5, we found excellent discrimination in the detection of PRRSV antibody (IgM, IgA, IgG) in processing fluids from herds of known PRRSV status. Judicious handling of processing fluid samples from sow herds, and the use of methods available in veterinary diagnostic laboratories, can provide a foundation for reliable PRRSV surveillance.

Probability of PRRS virus detection in pooled processing fluid samples.

There has been a tremendous increase in recent years of population-based diagnostic monitoring and surveillance strategies in swine populations. One example is the use of processing fluids (PF) to screen breeding herds for porcine reproductive and respiratory syndrome virus (PRRSV) activity. An important question from practitioners using such methods is on how intensively can the sample be pooled. More specifically, processing fluids of how many litters can be pooled into a single sample for diagnostic testing to preserve a high probability of PRRSV RNA detection at low prevalence situations? The objective of this study was to model the effect of pooling PF samples on the probability of PRRSV RNA detection. For this study, a PRRSV-positive PF field sample with a RT-rtPCR quantification cycle (Cq) value of 28 was selected to represent a litter of 11 pigs with a single viremic piglet. PF samples from a PRRSV-naïve herd were used to perform 6 replications of 8 two-fold serial dilutions of the PRRSV-positive sample, thus modeling the pooling effect (dilution). Each two-fold dilution represented an increase in the number of PRRS-negative pigs in the sample by a factor of 2. Samples were tested for PRRSV RNA by RT-rtPCR and the data was analyzed using linear and probit regression models. There was an average increment of 1.37 points in Ct for each two-fold dilution. The estimated probability of testing positive on RT-rtPCR was 43 %, 80 %, and 95 % when there was a single PRRSv-positive piglet among 784, 492, and 323 PRRSv-negative piglets contributing to the sample respectively. Results from this study support the practice of collecting and aggregating PF samples from multiple litters for PRRSV RNA testing.

Practical aspects of PRRSV RNA detection in processing fluids collected in commercial swine farms.

Processing fluid samples are easily collected under field conditions and provide the means to test more piglets more frequently in a practical way, thereby improving PRRSV surveillance. However, a deeper understanding of the diagnostic characteristics of this newly described sample type is still required. Therefore, the objective of this field-based study was to determine the relationship between viremic piglets and the detection of PRRSV RNA in processing fluid samples. In two PRRSV-positive breeding herds, processing fluids (n = 77) and individual piglet serum samples (n = 834) were collected from 77 litters in three sampling events and tested for PRRSV RNA. Among the 77 litters in the study, 55 litters (71.4%) contained no viremic piglets and processing fluids tested negative for PRRSV RNA. Among the 22 (28.6%) litters with ≥1 viremic piglets, 10 litters contained a single viremic piglet and 5 of the 10 processing fluids from this group tested positive for PRRSV RNA. Based on a fitted mixed effects logistic regression model, the probability of detecting PRRSV RNA in processing fluids was highly dependent on the number of viremic piglets contributing to the sample. When the within-litter prevalence was ≥39%, the probability of detecting PRRSV RNA in processing fluids was ≥95%. By extension, the results suggest that pooling processing fluids from several litters increases the probability of PRRSV RNA detection because of the greater likelihood of including multiple litters each with ≥1 viremic piglets. In contemporary breeding herds that use processing fluid samples for PRRSV surveillance, the diagnostic costs associated with testing 100% of the processing-age piglet population can be estimated at €0.077 ($0.086 USD) per pig weaned. In contrast, to achieve an equivalent testing coverage with the use of individual piglet serum samples, the diagnostic costs associated would be €4.48 ($5.00 USD) per pig weaned. Processing fluid represents a practical, reliable and efficient method to surveil breeding herds for PRRSV because it allows for continuous surveillance at a low cost.

Use of processing fluid samples for longitudinal monitoring of PRRS virus in herds undergoing virus elimination.

This was an observational study that prospectively followed 29 breeding herds for 65 weeks in the U.S.A. that became infected with porcine reproductive and respiratory syndrome virus (PRRSv). The herds operated in a four-week batch farrowing system and adopted a load-close-expose strategy using a modified-live virus vaccine to achieve PRRSv stability. The purpose of this study was to describe time to stability (TTS) based on RT-qPCR testing for PRRSv RNA on processing fluid samples in herds undergoing PRRSv elimination, after implementing herd closure and mass exposure to a PRRS modified-live virus (MLV) vaccine. For the purpose of this study, stability was defined as consistently producing PRRSv-negative pigs. Study herds were monitored until two consecutive piglet batches tested PRRSv RT-qPCR negative, then 30 due-to-wean piglet sera from the second batch were tested for PRRSv RNA by RT-qPCR. Once the farm re-opened, sera from incoming naïve gilts were tested for anti-PRRSv antibodies by ELISA at 30- and 60-days post-entry to confirm negative status to PRRSv. Day zero was the day of whole-herd exposure to a commercial PRRS vaccine virus. Twenty-eight of 29 herds (96.55%) achieved TTS within the study period. TTS ranged from 18 to 55 weeks with a median of 27 weeks. Serum from due-to-wean piglets was collected on 28 farms, of which 26 (92.85%) obtained PRRSv RT-qPCR-negative results on the first collection. At the end of the observational period, 16 sow farms successfully re-introduced PRRSv-naïve gilts with no detected serologic response. In conclusion, the median time to achieve TTS in breeding herds being operated in a four-week batch farrowing system undergoing PRRSv elimination using load-close-expose with attenuated virus vaccine was 27 weeks. Also, processing fluid-based monitoring of breeding herds under PRRS elimination was practical and reliable to assess PRRSv stability.