A pragmatic randomized trial of a primary care antimicrobial stewardship intervention in Ontario, Canada

The need to more appropriately use antibiotics in order to limit antibiotic resistance has been widely endorsed [1, 2]. However, progress towards this goal has been slow. In Canada for example, antibiotic utilization has changed little, with 666 community antibiotic prescriptions dispensed per 1,000 persons in 2012 and 658 prescriptions in 2017 [3]. The recognition that 90% of antibiotic use occurs in the community has focused attention on primary care prescribers [4]. A 2019 U.S. study found 59% of outpatient antibiotic prescriptions are likely not needed [5]. Reducing this volume of unnecessary antibiotic use may be critical to limit antibiotic resistance. To achieve this, effective community antimicrobial stewardship (AMS) approaches are needed.

Community-based AMS programs have been tested in Scotland [6], Spain [7], the United States [8], the Netherlands [9], the United Kingdom [10], and Canada [11]. Most involved education and prescribing guidelines combined with audit-and-feedback [8‐11]. Some were limited to either specific organisms [7], short durations [8], respiratory infections [9], effects were small [10], or evaluations lacked control groups, limiting conclusions about effectiveness [11, 12]. Some providers report a lack of confidence in AMS approaches, raising questions about whether community prescribers would implement programs, even where effective [13, 14]. To ensure AMS is relevant to community prescribers, programs may need to ensure they are addressing the prescribing challenges that community providers feel are most important.

One factor may be the clinical uncertainty community providers report in trying to distinguish viral from bacterial infections, owing to their similar clinical presentations [15]. There are also few point-of-care tests to aid in identifying bacterial presentations that might benefit from antibiotics. We previously developed a multi-faceted primary care provider-focused antimicrobial stewardship intervention that addressed clinical uncertainty through use of clinical decision aids, delayed prescriptions and safety netting advice [16]. A pilot evaluation identified competing clinical commitments, perceived pressure to prescribe antibiotics, and a lack of resources for stewardship activities as additional barriers for community clinicians [16]. The objective of this study was to assess the effectiveness of a community-based primary care provider-focused antimicrobial stewardship intervention addressing clinical uncertainty and other barriers on antibiotic prescription practices of primary care providers for common community infections.

There were 1,904 eligible visits from January 1, 2019 until May 31, 2019. After exclusions, 1682 (88.3%) visits were available for analysis; 943(56.1%) in control clinics and 739 (43.9%) from intervention clinics (Fig. 1). The average meeting attendance at intervention clinics was 81% with a 77% average education module completion rate. Overall, 16/34 (47.1%) of these providers completed all intervention elements.

Table 1 compares characteristics of intervention and control clinic visits in the prior winter and during the evaluation period. In both time periods, groups differed in the number of providers, total visits, and visits outside of regular clinic hours (after hours). Patients seen at intervention clinics during the evaluation tended to be older (mean 53.4 years) than at control clinics (47.1 years, p<0.001), more likely to be female (76.9% versus 70.6%, p=0.005), and have an antibiotic allergy (24.0% versus 16.1%, p<0.001). The distribution of infections differed between the groups in both time periods (p<0.001). While there were fewer cases of acute bronchitis at intervention clinics during the evaluation period (6.6%) compared to the baseline period (11.0%), there was no increase in pneumonia diagnoses to suggest diagnostic shifting (8.9% evaluation period vs 10.4% baseline).

Factors associated with receiving an antibiotic prescription (Supplementary Table 1) included the type of infection (p<0.01,), being assessed after hours (p=0.02), female gender (p<0.01), and age (p=0.07). These factors were adjusted for in comparisons of prescribing outcomes. Table 2 indicates the intervention effects on prescribing outcomes for the five selected infections combined (n=1,372 visits). There was an observed increase in the crude odds of receiving any antibiotic prescription at intervention clinics (crude odds ratio (OR)=1.72; 95% CI =1.00, 2.98). However, after adjustment for characteristics associated with antibiotic prescriptions, there was a 22% reduction in the odds of receiving an antibiotic prescription (adjusted OR = 0.78; 95% CI = 0.64, 0.96) at intervention clinics.

More prescriptions at intervention clinics were issued as delayed prescriptions (22.1%) than in control clinics (11.7%; adjusted OR = 2.29; 95% CI=1.37, 3.83). The odds that intervention physicians advised patients to start antibiotics immediately were reduced 37% compared to control clinics (adjusted OR= 0.63; 95% CI=0.45, 0.89). Antibiotic prescriptions for longer than 7 days duration were also decreased at interventions clinic (21.3%) compared to control clinics (29.3%; adjusted OR = 0.24; 95% CI = 0.13, 0.43). Use of recommended antibiotics were high in both control (84.9%) and intervention (92.2%) clinics (adjusted p=0.37).

There were no reductions in total antibiotic prescriptions for individual infections (Table 3). Prescriptions for URI cases were increased in intervention clinics (13.0% vs 7.7% control clinics, p<0.01), although most (56.5%) were delayed prescriptions. The odds of a delayed prescription was increased for acute sinusitis in unadjusted comparisons, and for acute cystitis in adjusted comparisons. The odds of a prescription for longer than 7 days was reduced for acute bronchitis only (adjusted OR = 0.46; 95% CI = 0.29, 0.73).

A post-hoc analysis of the effect of testing recommendations included in clinical decision aids found a throat swab or rapid strep test was ordered more frequently at intervention clinics (56/81, 69.1%) than control clinics (28/91, 30.8%) for sore throat presentations (adjusted OR = 4.63, 95% CI=2.81, 7.63). Chest x-rays were ordered more often at intervention clinics (19/49, 38.8%) than control clinics (13/126, 10.3%) for cases of acute bronchitis (adjusted OR = 3.85, 95% CI = 2.58, 5.76). Urinalysis and urine culture utilization was similar in both control (93.3%) and intervention (90.7%) clinics.

This community-based, primary care provider-focused antimicrobial stewardship intervention reduced total antibiotic prescriptions to adults with a respiratory or urinary tract infection, increased the proportion issued as delayed prescriptions, and reduced prescription durations for longer than 7 days.

Although the observed crude rates of total antibiotic prescriptions were not different between intervention and control clinics, there were a number of differences in baseline characteristics between the two groups of clinics that persisted after randomization. When clinic difference were adjusted for, there was a 22% reduction in the odds of prescribing an antibiotic at intervention clinics. In addition, these prescriptions were twice as likely to be issued as delayed prescriptions. This is relevant as only 33% of prescriptions with instructions to delay starting may ultimately be filled [28]. As a result, patients seen at intervention clinics were less likely to be advised to begin antibiotics immediately. Prolonged antibiotic treatment durations are also an issue in primary care [29]. Long duration prescriptions were substantially reduced at intervention clinics. In addition, the change in test utilization at intervention clinics suggests the decision aids were important in affecting clinical change, as testing recommendations were only included in these aids.

Other multi-faceted ASP intervention studies have varied in the intervention focus, types of infections, and measures of antibiotic use [10, 30, 31]. A Quebec-based trial of shared decision making reported a 50% reduction in patient-reported intent to start antibiotics after a respiratory infection [30]. However, delayed prescriptions were allowed, which may have been later filled, and changes in total or dispensed antibiotics was not reported. A British trial of physician education, decision support and audit and feedback reported a 12% relative reduction in antibiotic prescriptions, but for respiratory infections only [31]. An evaluation of a national AMS program (TARGET) where practices were unaware they were part of a trial found a small 2.7% relative reduction in dispensed antibiotics [10]. These studies highlight the challenges of multi-faceted primary care AMS interventions, and the need to consider broader prescribing characteristics other than total prescriptions to fully understand the potential impact of primary care stewardship on antimicrobial resistance.

Providers were compensated for participating in the current study. An American study similarly provided compensation to participants with a 70% participation rate [32]. Incentives have also been implemented outside of studies to promote optimal antimicrobial use. In Britain, a financial Quality Premium introduced in 2015 for reductions in antibiotic use was associated with an 8% absolute decline in antibiotic prescriptions [33]. The current study also provided logistical support for AMS activities in intervention clinics. Other jurisdictions have similarly utilized direct supports to clinics [34, 35]. A Scottish national AMS program utilized Antimicrobial Management Teams to support local prescribing leadership [34]. In Britain, Medicine Management Teams play a role in education of community providers [35]. The current study and this research suggest the provision of incentives and logistical support at the clinic level may be important in facilitating community AMS efforts.

A limitation of the study was not all clinicians at each clinic participated. Participating clinicians may have been more receptive to the AMS intervention than might be seen in the broader primary care community. Another limitation was the short study duration over one winter season. Whether the impact on prescriptions would be sustained is unclear. AMS effects have declined when interventions are removed [36, 37]. The study also relied on written prescriptions as an outcome rather than dispensed antibiotics. While reducing initial prescriptions and their durations are important to reducing dispensed antibiotics, patients may have sought out other providers and received antibiotics. However, as more providers are exposed to AMS practices, patients may receive consistent antibiotic recommendations from all prescribers. Finally, the safety of this program was not assessed. Other studies have found adverse events associated with reduced antibiotic use in primary care are uncommon and require large samples to detect [38]. However, the elderly may be one group where caution is needed [39]. Clear ‘safety-netting advice’ where antibiotics are not prescribed, appropriate use of delayed prescriptions and clinical follow up can help mitigate such risks and prevent a loss of provider confidence in stewardship efforts.

This study has demonstrated that clinically important changes in antibiotic utilization in primary care clinics are possible with local stewardship efforts involving provider-focused education, clinical decision aids, clinic support, ongoing audit and feedback, and compensation for ASP activities. Structured and supported community-based antimicrobial stewardship efforts, similar to those in hospital settings, warrant further study.