Advancing antimicrobial use measurement in U.S. Companion animal practice

The Department of Health and Human Services Food and Drug Administration awarded a Cooperative Agreement valued at $199,999 to the Regents of the University of Minnesota on September 20, 2025, under the Food and Drug Administration Research program (CFDA 93.103). The award funds antimicrobial use measurement in U.S. companion animal practice, leveraging the Companion Animal Veterinary Surveillance Network (CAVSNET), a public-private partnership that compiles non-standardized patient-level data from diverse electronic health record systems into a standardized common data model. The project analyzes antimicrobial use data for dogs, cats, and horses by harnessing CAVSNET data and point-prevalence surveys that conduct manual medical record review across multiple veterinary sites. The work includes application of NLP-PIER, a natural language processing platform, to classify and index unstructured clinical notes for analysis in a secure computing environment, mapping free-text data to structured categories. The goal is to provide actionable targets for antimicrobial stewardship by connecting antimicrobial prescriptions to clinical context including disease processes, diagnostic tests, and patient demographics. Performance occurs in Lauderdale, Minnesota, with a period of performance through August 31, 2030.

Abstract

Antimicrobials are critical for animal and human health, but antimicrobial resistance (AMR) is a threat to their continued effectiveness. Antimicrobial use (AU) is a modifiable risk factor for the emergence and spread of AMR, and tracking AU is needed to improve prescribing. Logistical challenges to AU measurement in the veterinary field include access to prescribing data held within diverse mutually exclusive electronic health record (EHR) systems and lack of standard diagnostic coding. The companion animal veterinary surveillance network (cavsnet), a successful public-private partnership that collects, analyzes, and reports companion- animal veterinary data, overcomes these obstacles through the compilation of non-standardized patient-level data into a standardized common data model (CDM). Cavsnet fulfills a critical need in the fight against AMR, providing data on AU at the point of use (i.e., veterinary clinics) to connect the clinical context, such as disease process, diagnostic tests, and patient demographics, with prescription data. In addition to discrete data fields, unstructured free-text data are available for classification with NLP-pier, a natural language processing (NLP) platform that enables named-entity recognition and indexing in a secure computing environment; these notes can be mapped to structured categories for analysis. Point-prevalence survey (PPS) methodology, which relies on detailed manual review of medical records, is structured to collect standardized data from multiple sites over a specific period of time; PPS are used to provide granular data and augment automated cavsnet data collection to better understand the complexities of relating AU data to clinical condition. The overarching goal of this project is to harness data from cavsnet and complementary PPS to analyze AU for dogs, cats, and horses and provide actionable targets for antimicrobial stewardship. Longitudinal AU tracking provides information needed to target prescribing improvement interventions, can be used to motivate practitioners to change behaviors, and show the profession's progress over time. Cavsnet's public-private partnerships for ongoing and long-term data collection will 1) characterize and address AU data gaps, 2) collect, standardize, and provide annual summaries of AU data, including contextual information and data trends, while maintaining veterinary and client confidentiality, 3) share data with the veterinary community and the public, and 5) use a common data model that lends itself to interoperability. IT will provide a comprehensive picture for U.S. Companion animals by 1) providing yearly national estimates of systemic antibiotic use prevalence in companion animal practices, including contextual demographic information, using cavsnet, 2) generating U.S. Estimates of systemic antibiotic use for specific clinical conditions via cavsnet, and 3) characterizing antibiotic use data gaps in companion animal practice and utilizing methods to address gaps, including NLP- pier and national AU PPS for horses and select surgical procedures in dogs and cats.