Iowa State develops machine-learning model to predict pig nursery mortality risk before leaving the sow farm

Researchers at Iowa State University are creating a machine-learning approach to identify nursery mortality risks in pigs early, enabling proactive health management and reducing losses in pig production.

Researchers at Iowa State University are working on a machine-learning approach that could flag nursery mortality risk in pigs before they leave the sow farm, giving veterinarians and producers a window for earlier intervention. The SwinalytIQ group says the aim is to estimate which weaning groups are most likely to struggle in the nursery, using information already captured at the breeding farm rather than waiting for clinical problems to emerge.

The model is built from routine herd data, including infectious disease status for PRRS, Mycoplasma pneumoniae and porcine epidemic diarrhoea, alongside reproductive and production measures such as abortion rate, total pigs born, stillbirths, mummified foetuses and wean-to-service interval. In the Iowa State analysis, records from 2,025 commercial weaning groups were used to train and test multiple algorithms, which were then combined to improve stability and reduce error.

Related research points in the same direction. A study published in Animal Bioscience found that forecasting models trained on 1,831 pig groups and 48 variables could predict nursery mortality with useful accuracy, while conference data on a Wean-Quality Score reported strong performance from a random forest model for separating higher-risk from lower-risk groups. Other recent work is also exploring earlier identification of pigs likely to need treatment during the nursery phase.

For practitioners, the value is not prediction alone but timing: targeted monitoring, diagnostics, vaccination review, biosecurity checks and environmental correction can be deployed sooner. The Iowa State team says the model still needs wider validation, but the concept is clear: use existing data to move nursery health management from reactive to anticipatory.

Image Source: Iowa State University