Biosecurity practices useful for porcine reproductive and respiratory syndrome control and eradication on commercial swine farms using machine learning models.

It is well known that porcine reproductive and respiratory syndrome (PRRS) decreases herd productivity and leads to economic loss, and it is believed that biosecurity practices are effective for the control and eradication of PRRS. This study aimed to prioritize biosecurity practices useful for PRRS control and eradication by using a machine learning algorithm. Biosecurity practices on 258 Japanese breeding farms were assessed using a biosecurity assessment tool. Random forest analysis was used to assess relevant features based on feature importance. In study 1, each farm was divided into two groups (PRRS controlled or not controlled), and the most relevant biosecurity practices for PRRS control were investigated. Seven features were selected as associated with PRRS control, and results identified the prioritization of semen management and the maintenance of a controlled barn environment to be critical elements for PRRS control. In study 2, each farm was divided into two groups (PRRS-free or not free of PRRS), and the most relevant biosecurity practices for PRRS eradication were investigated. Fourteen features were identified as critically associated with PRRS eradication, and results showed that the stringent management of replacement gilts, the exclusive use of semen from PRRS-free sources, and a comprehensive awareness of farm location were essential elements for PRRS eradication. In conclusion, this research underscores the significant divergence in biosecurity strategies required for effective PRRS control compared with those for eradication.