A randomised controlled trial to reduce highest priority critically important antimicrobial prescription in companion animals

The study addresses the challenge of antimicrobial resistance (AMR) linked to frequent antimicrobial use in companion animals and the consequent public health implications due to close human–animal contact. Highest priority critically important antimicrobials (HPCIAs), such as third-generation cephalosporins (e.g., cefovecin) and fluoroquinolones, are still commonly prescribed—especially in cats—often without clear justification or evidence of resistance to first-line drugs. Prior work has revealed large inter-practice variation in prescribing, suggesting cultural and behavioral influences. The authors hypothesized that informing practices about their relatively high HPCIA-prescribing behavior (social norms feedback) combined with educational and benchmarking support would reduce HPCIA prescriptions. The study’s purpose was to generate robust, trial-based evidence on effective antimicrobial stewardship interventions in companion animal practice using electronic health records (EHRs).

Background literature highlights companion animals’ role in AMR development, carriage, and transmission, driven by frequent antimicrobial use. EHR-based studies in the UK show declining overall antimicrobial use but persistently high use of HPCIAs in cats, particularly cefovecin, often without clear clinical justification. Risk factors for antimicrobial prescribing include animal and owner factors, clinical presentations, individual veterinarians, and practice culture, with accredited practices tending to prescribe fewer systemically administered antimicrobials. Prior stewardship efforts in veterinary medicine include guidelines, benchmarking, and audits, but robust evidence for effectiveness is limited. In human medicine, social norms feedback and stewardship programs have effectively reduced inappropriate prescribing. Veterinary qualitative research also indicates misperceptions about peers’ prescribing behaviors that may drive overuse, suggesting benchmarking could leverage social norms to prompt change.

Design: Three-arm randomized controlled trial (RCT) within a single UK corporate veterinary practice group (CVS Group Ltd.). Data source: SAVSNET EHRs from booked consultations, containing anonymized animal/owner data, clinical narratives, and dispensed products. Periods: Pre-intervention baseline August 2018–March 2019; notification March 28–29, 2019; post-intervention follow-up April–November 2019 (8 months). Groups: 60 practices (of 68 eligible from 157 total) randomized equally to control group (CG, n=20), light intervention group (LIG, n=20), and heavy intervention group (HIG, n=20) using block random allocation (randomizr package). Eligibility: Practices in the top 40% for total HPCIA prescribing in both dogs (>0.5% of consultations) and cats (>5.4%) or top 40% in one species and 40–60% in the other. Sample size: 17 practices per group estimated to detect a 10% relative decrease in HPCIA prescribing (SD=10%, power=80%, α=0.05). Ethical approval: University of Liverpool SAVSNET ethical approvals (RETH000964; VREC745). Interventions: - LIG: Notification of above-median HPCIA prescribing; materials included the practice antimicrobial policy aligned with guidance, instructions and interpretive guidance for accessing an anonymized antimicrobial benchmarking portal, and AMR educational videos. - HIG: All LIG materials plus an in-depth benchmarking report with explanatory video, and invitation to participate in a voluntary reflection and education program delivered by hub clinical leads (senior clinicians) through practice reviews and optional follow-up meetings (April–September 2019 window for additional support). Two HIG practices declined the reflection/education program but remained in outcome monitoring. Data definitions: Antimicrobial prescriptions identified via dispensed products using semi-automated rule-based text mining; classified as systemic (oral/injectable) or topical (topical/aural/ocular). HPCIAs defined as fluoroquinolones, macrolides, and third-generation cephalosporins. Each consultation categorized into one of ten main presenting complaints (MPCs). Diagnostics: Orders for cytology and for bacterial culture and susceptibility were counted. Additional practice-level data: veterinary surgeon FTE and locum coverage. Outcomes: Primary outcome was HPCIA prescription frequency as a percentage of total consultations post-intervention versus control, stratified by species. Secondary outcomes included total antimicrobial prescribing (overall, systemic, topical), HPCIA prescribing by MPC, antibiotic class mix, anti-inflammatory prescribing, euthanasia frequency, diagnostic test ordering, and portal engagement (unique practice logins). Analysis: Mixed-effects panel regression models comparing intervention groups (LIG, HIG) to CG across pre- and post-intervention periods, including month-by-month analyses, with evaluation of model fit and temporal terms. Practice-level and consultation-level counts provided context. Portal engagement compared using Fisher’s exact tests. Timeline and engagement: HIG practices undertook initial and follow-up reviews with hub clinical leads (most activity April–July 2019; final reviews August–October 2019). Portal access tracked February–November 2019.

- Primary outcome (HPCIA prescribing frequency): • Dogs: Pre-intervention HPCIA prescribing was similar across groups (~0.7% of consultations). Post-intervention, HIG showed a significant 23.5% reduction to 0.49% (95% CI 0.37–0.61; P=0.04) versus CG 0.64% (0.47–0.82). LIG showed no significant change (0.79%; 0.61–0.97; P=0.78). • Cats: Pre-intervention HPCIA prescribing ~7–8% across groups. Post-intervention, HIG decreased by 39.0% to 4.35% (3.41–5.29; P<0.01) versus CG 7.37% (6.02–8.72). LIG decreased by 16.7% to 6.14% (5.29–7.00; P=0.03). - Temporal patterns: In HIG, significant monthly reductions in cats were observed in 8 of 8 post-intervention months, with a marked drop from March to April 2019; in dogs, 3 months showed significant reductions. Linear trends showed significant decreases in HIG for both species (dogs P=0.005; cats P<0.001). LIG showed a significant linear decrease only in cats (P=0.01). - By presenting complaint (MPC): Several feline MPCs (e.g., respiratory, trauma, pruritus) and one canine MPC in HIG showed significant post-intervention reductions in HPCIA prescribing; no consistent MPC reductions in LIG. - Broader antimicrobial prescribing: HIG showed significant decreases in systemic antimicrobial prescribing: dogs −18.9% and cats −17.3% versus CG; total antimicrobial prescribing also decreased significantly in both species. No significant differences between groups in anti-inflammatory prescribing or euthanasia frequencies. - Antimicrobial class mix: In cats (HIG), third-generation cephalosporin proportion tended to decrease while amoxicillin–clavulanate tended to increase, although differences were not statistically significant. - Diagnostics: No significant changes in cytology or bacterial culture and susceptibility test ordering compared to CG; orders remained low (~1–2% of consultations). - Portal engagement: Post-intervention unique practice logins increased significantly: CG n=3, LIG n=8, HIG n=16 (P<0.001 overall; CG vs HIG P<0.001; LIG vs HIG P=0.03). Engagement peaked in April 2019 and waned over time; HIG engagement persisted longer than LIG. - Model performance: Fixed effects explained ~10% of canine and 40% of feline variance for HPCIA models.

The trial demonstrates that EHR-driven social norms feedback combined with education and in-depth benchmarking can reduce HPCIA prescribing in companion animal practice, particularly when augmented with facilitated reflection via hub clinical leads (HIG). Reductions were strongest and more sustained in cats—the species with the highest baseline HPCIA use—indicating that targeted stewardship can address known high-risk prescribing behaviors. The decrease in systemic and overall antimicrobial prescribing in HIG suggests a broader cultural shift beyond HPCIAs, though in cats some switching from HPCIAs to other antimicrobials may have occurred. Increased engagement with the benchmarking portal following notification supports the behavioral mechanism that awareness of deviation from normative prescribing stimulates reflection and change. The absence of changes in anti-inflammatory or euthanasia frequencies suggests no immediate adverse welfare impacts, though more sensitive outcome measures are needed. The lack of increase in diagnostic testing highlights persistent barriers to culture- and evidence-based prescribing. Overall, findings provide robust evidence to guide veterinary antimicrobial stewardship strategies and inform scalable programs, with consideration of feasibility and infrastructure across diverse practice management systems.

An EHR-enabled benchmarking, reflection, and education framework significantly reduced HPCIA, systemic, and overall antimicrobial prescribing in dogs and cats within a large UK veterinary practice group. The heavy intervention, which included in-depth benchmarking and facilitated reflective support, was most effective and sustained, especially in cats. The work offers a robust evidence base for veterinary antimicrobial stewardship and is informing a national scheme with RCVS Knowledge. Future work should: (1) disentangle the relative impact of intervention components; (2) assess long-term sustainability and scalability across diverse practice settings; (3) expand stewardship beyond HPCIAs to optimize overall antimicrobial choice; (4) develop and promote timely diagnostics; and (5) implement more sensitive welfare outcomes to monitor unintended consequences.

- Intervention components were bundled, limiting attribution of effect to specific elements (benchmarking, education, hub clinical lead support). - Conducted within a single large corporate practice group; generalizability to independent or differently structured practices is uncertain. - Follow-up was limited to 8 months; long-term sustainability of behavior change is unknown. - Engagement metrics could not distinguish among individual staff within practices; dissemination of materials in LIG may have been uneven. - Welfare impact assessment used crude proxies (euthanasia frequency); more sensitive clinical outcome measures were not available at scale. - Diagnostic testing remained infrequent and was not significantly influenced; low baseline use and resource constraints may have limited change. - Potential prescribing shifts from HPCIAs to other broad-spectrum agents (e.g., amoxicillin–clavulanate) could sustain or shift resistance pressures. - SAVSNET compatibility across all practice management systems may limit national scalability at present.