Abstract Background Disease in primary care frequently represents a surveillance blind spot, particularly for diseases affecting farm animals. Methods Electronic health records (EHRs) were collected from four farm animal veterinary practices in Wales (February 2024‒January 2025) as part of a pilot study. Information collected included species treated, date, owner postcode, products sold and clinical free text. Text mining and topic modelling were used to describe treatments and classify syndromes. Results In total, 32,799 records were collected. Antimicrobials were prescribed in 32.6% and 63.8% of cattle and sheep records, respectively. The most frequent antibiotic classes in both species were tetracyclines, macrolides, penicillins and penicillin‒aminoglycoside combinations. There were no recorded category A antimicrobials, and category B antimicrobials were prescribed in only 0.12% and 0.04% of cattle and sheep EHRs, respectively. Text mining and topic modelling seemed efficient methods to identify key syndromes, including mastitis, joint ill, lameness and pneumonia, and how these were treated. Limitations Some EHRs described more than one animal with different diagnoses, obfuscating the attribution of treatment to syndrome. Conclusion The increasing availability of EHRs at scale and in real‐time represents a complementary opportunity to survey disease and treatment on farms. Text mining methods, including artificial intelligence, could efficiently identify important syndromes and provide novel insight into use of antibacterials.
Hopkins et al. (Fri,) studied this question.