Here we describe the first use of EHR data from veterinary practices to initiate and sustain voluntary antimicrobial prescribing behavioural change in veterinary prescribing. In so doing, we have outlined a data-led and educational support framework by which HPCIA prescription frequency can be significantly reduced, while preserving care quality and clinical autonomy. These insights are now being used to inform a national antimicrobial stewardship scheme, led by RCVS Knowledge (https://knowledge.rcvs.org.uk/home/), firmly demonstrating a profession-wide commitment to responsible usage of antimicrobials.
A primary component of behaviour change is to effectively draw attention to an issue, 37 and departure from ‘social norms’ has been previously demonstrated as a motivator for antimicrobial prescription behaviour change.26 In the absence of evidence, individuals tend to over-estimate perceived negative traits in their peers, thus serving as justification for their own behaviour (e.g. alcohol consumption frequency).38 A similar tendency regarding antimicrobial prescription amongst veterinary surgeons has been previously observed,17 leading us to hypothesise that lack of knowledge of relative frequencies of antimicrobial prescription might in itself serve as a driver for more frequent prescription. Hence, by incorporating an intervention based on social norm messaging of high HPCIA prescription frequency, we ensured that these prescribers were aware of their unusual status.
In the LIG and HIG, use of the SAVSNET antimicrobial benchmarking portal significantly increased post-notification, with increases being most apparent in the two months following notification, suggesting these notifications prompted enhanced portal engagement. However, while post-intervention 75% of HIG practices interacted with the portal, only 40% of LIG practices did so. Further, while engagement waned to at or below pre-intervention levels within two months of notification in the LIG, interest exceeded pre-intervention levels for four months post-notification in the HIG. It is probable that either the additional in-depth benchmarking report provided to practices in the HIG, or post-notification offer of assistance from the hub clinical leads might have enhanced initial interest compared to the LIG. Though the relative contributions of each was not able to be elucidated here, individuals are more likely to re-evaluate existing behaviours if modifying behaviour might bring reward, or not doing so might bring punishment.37 Although the supportive, optional nature of the trial was emphasised throughout, requested hub clinical lead intercession might have nevertheless introduced a perception of potential reward or punishment linked with engagement with the trial. It is also possible that the letter and email sent to LIG practices was not disseminated beyond clinical directors in some cases, whereas practice-wide meetings held in HIG practices ensured engagement of all staff.
There was concern that notification of relatively high HPCIA prescription alone could prompt practice policy changes not reflective of latest clinical evidence, such changes being potentially detrimental to animal welfare or employee wellbeing. Structured reflection and education programmes have been shown to be effective at achieving sustained improvements in anti-infective prescription habits in the medical field,25 and in this trial we compared both a light (LIG) and heavy (HIG) reflection and educational intervention. While significant and sustained reductions in HPCIA prescription frequency were seen in both species in the HIG, no significant decreases were observed in the LIG. Across the veterinary sector impressive reductions have been achieved over the past five years, especially in pigs and poultry,39,40 utilising a variety of statutory41 and voluntary42 improvement measures. Though over this time reductions in antimicrobial prescription frequency in companion animals have been noted,6,10,12 HPCIA use has remained an issue, particularly in cats.6,7,15 Unlike other veterinary sectors, no statutory policies have been introduced to prompt improvements in antimicrobial prescription in companion animals. It is unknown what impact such enforced measures might have on animal welfare, and thus we consider findings presented here to be an encouraging sign that practitioners might be willing to voluntarily engage with improvement efforts, potentially negating need for firmer regulatory approaches.
This study further demonstrated the relative ease by which EHRs can be utilised to both identify participants and monitor key outcomes in near real-time. Such efficiency advantages have been previously outlined in medical research, enabling rapid scaling of interventions to instigate national quality improvement.26 Only comparatively recently have EHRs become available for research and surveillance in the veterinary sector,43 and as such we believe this work serves as an exciting demonstration of what could be achieved using EHR data-led approaches, expanding beyond practitioner-focused interventions to those encompassing owners, or pragmatic efficacy assessment of surgical and pharmaceutical interventions in routine practice, for example.
Though the intervention package provided in this trial represented a comprehensive approach to encouraging evidence-based behaviour change, it did limit our ability to determine which individual components might have been of greatest impact. Interestingly, though both LIG and HIG practices received a high feline HPCIA prescription frequency notification at the beginning of the trial, no significant reductions were subsequently noted in LIG practices. However, all HIG practices, including both HIG practices that refused engagement with the reflection and education programme, reported decreases post-notification. Though variation in scale of reduction was evident, circumstantially, these findings might suggest that hub clinical lead involvement was a motivational factor in prompting behavioural change. In either case, the aforementioned refusals do indicate a limiting factor in intervention scalability to a wider audience. Further, this trial benefited from utilising existing quality improvement management structures within a single large practice group for the HIG. Thus, there remains a question as to whether this intervention would be feasible amongst other practices, including those that are relatively infrequent HPCIA prescribers, over a longer period of time than the six months observed here.
More broadly, several and two MPCs in cats and dogs respectively in the HIG were associated with significant HPCIA prescription frequency reductions, a number of which (e.g. respiratory, trauma) were previously associated with frequent HPCIA prescription,6,21 despite often lacking clear clinical justification for their prescription.21 These findings suggest a generalised culture change not necessarily being restricted to reflection on individual disease presentations. This result is supported by significant reductions in systemic and overall antimicrobial prescription frequency in both species. In HIG dogs, these wider reductions were greater than that contributed by HPCIA reductions alone, suggesting that the trial had a wider impact on discouraging antimicrobial prescription more generally. However, in HIG cats the opposite was seen; HPCIA reductions were greater than overall decreases, suggesting a tendency of some practitioners to move from prescribing a HPCIA to prescribing another non-HPCIA antimicrobial, instead of avoiding prescription altogether.
Whilst no prescription choice comparisons were significantly different, 3rd generation cephalosporin feline prescription does appear to have decreased to a degree in the HIG, while clavulanic acid potentiated amoxicillin prescription increased, suggesting a preferred alternative to 3rd generation cephalosporins. Clavulanic acid potentiated amoxicillin is an authorised, widely used antimicrobial in veterinary practice.6 However, like 3rd generation cephalosporins, use of clavulanic acid potentiated amoxicillin has also been associated with resistance development,44 and is only infrequently prescribed in medical practice.45 Thus, future stewardship efforts will need to expand scope beyond HPCIAs to also consider how to promote responsible use of all antimicrobials, and indeed other medicines too.
For instance, we have previously reported a tendency for antimicrobials and anti-inflammatories to be prescribed at the same time, despite perhaps limited clinical evidence to suggest necessity for both pharmaceutical agents.13 However, we have also noted a recent reversing trend for respiratory disease whereby antimicrobial prescription frequency has decreased whilst anti-inflammatory prescription frequency has increased.10 These findings perhaps reflect increasing recognition of frequent non-infectious mediators for respiratory disease,46 or increased attention to prescribing guidance.16 It was interesting to note that no significant variation in anti-inflammatory prescription was observed here, perhaps demonstrating more generalised ‘de-coupling’ of anti-inflammatory and antimicrobial prescription. Though measuring frequency of use represents a relatively simple method for demonstrating change, reduced use is not necessarily representative of more responsible use. Hence, it is probable that more nuanced methods for ascertaining whether a pharmaceutical agent has been prescribed appropriately will be needed. As part of these developments, we would advocate increased attention on use of other pharmaceutical agents that might form effective alternatives to antimicrobial prescription, whilst also satisfying the recognised need of a practitioner to provide a clear demonstration of action via provision of a therapeutic product, to the client.17
Though prescribers retained full autonomy to prescribe what they considered best for the animal under their care, we incorporated euthanasia frequency as a relatively crude measure of any increase in adverse health effects associated with change in prescription decision-making prompted by this trial. While no significant increases were observed in either intervention group, compared to the CG, we recognise that this method lacks sensitivity, not taking into account a range of potential sub-optimal outcomes that might compromise animal welfare. Though effectively and efficiently quantifying such adverse effects from EHRs at the scale required for this trial presents a significant challenge,47 we recommend further development of text-mining and statistical methodologies to explore such nuances, both for this and subsequent trials.
Use of bacterial infection-associated diagnostic tests was not significantly affected in this trial. Low frequency of use of such tests has been identified as a barrier to effective stewardship,48 likely reflecting low confidence in their ability to provide timely, useful clinical insights.49 Indeed, across all groups, test orders were low, indicating a preference for empirical antimicrobial prescription throughout the trial. Used correctly, these tests do play an important role in correctly managing a patient;50 however, there is clearly more work needed to convince practitioners – and owners – of the benefits of regularly pursuing these diagnostic routes. That said, during this trial a lack of equipment and training for cytological examinations within practices was identified, which resulted in wide-scale equipment and training provision (unpublished observations). Thus, there is hope that significant impact will be generated beyond the confines of this trial.