Work

Prediction of Stillborn Piglets From Multiparous Sows

Year: 2025

Type: article

Abstract: Background: Assisting sows during parturition reduces the number of stillborn piglets caused by anoxia. However, in industrial settings with a large number of animals, the capacity for assistance is limited. The development of predictive models based on existing data can enable farms to anticipate stillbirths in sows. Objective: To develop a predictive model to identify factors affecting the presence of stillborn piglets (PSbP), estimate the probability of their occurrence, and establish a classification criterion accordingly. Methodology: Data from 2 415 farrowings in 822 sows (Landrace, Yorkshire, and their crossbreeds) were analyzed. Five variables relating to the current farrowing and five variables related to the preceding one were examined. Our study used cross-validation (groups = 5), modeling the response variable (PSbP, 1: presence, 0: absence). Results: The only factor shown to have a negative effect (p<0.01) on PSbP was litter weight at birth, while litter size at birth and parity (number of farrowings) were seen to have a positive effect (p<0.01). PSbP prevalence during training and testing were 0.297 and 0.296 respectively. The model's estimated probability levels were 0.311 during training and 0.303 during testing, indicating an accurate probability estimation. When categorizing using the optimal cutoff point of 0.395, the predictive efficiency as measured by the area under the Receiver Operating Characteristic (ROC) curve was 0.846 for training and 0.813 for testing. Implications: Implementing this model of information-management software could make it possible to provide swift, efficient technical assistance to sows in need, with a high level of predictive efficiency. Conclusions: The probabilistic model described here based on a Bayesian approach and adjusted based on a categorization criterion showed effective predictive efficiency in the prediction of stillborn piglets.

Source: Tropical and Subtropical Agroecosystems

DOI: 10.56369/tsaes.5658