Early-career general practitioners’ antibiotic prescribing for acute infections: a systematic review

Abstract Background Antimicrobial resistance is a worldwide threat, exacerbated by inappropriate prescribing. Most antibiotic prescribing occurs in primary care. Early-career GPs are important for the future of antibiotic prescribing and curbing antimicrobial resistance. Objectives To determine antibiotic prescribing patterns by early-career GPs for common acute infections. Methods A systematic literature search was conducted using PubMed, Embase and Scopus. Two authors independently screened abstracts and full texts for inclusion. Primary outcomes were antibiotic prescribing rates for common acute infections by GPs with experience of 10 years or less. Secondary outcomes were any associations between working experience and antibiotic prescribing. Results Of 1483 records retrieved, we identified 41 relevant studies. Early-career GPs were less likely to prescribe antibiotics compared with their more experienced colleagues (OR range 0.23–0.67). Their antibiotic prescribing rates for ‘any respiratory condition’ ranged from 14.6% to 52%, and for upper respiratory tract infections from 13.5% to 33%. Prescribing for acute bronchitis varied by country, from 15.9% in Sweden to 26% in the USA and 63%–73% in Australia. Condition-specific data for all other included acute infections, such as sinusitis and acute otitis media, were limited to the Australian context. Conclusions Early-career GPs prescribe fewer antibiotics than later-career GPs. However, there are still significant improvements to be made for common acute conditions, as their prescribing is higher than recommended benchmarks. Addressing antimicrobial resistance requires an ongoing worldwide effort and early-career GPs should be the target for long-term change.


Introduction
Antimicrobial resistance is growing worldwide, exacerbated by inappropriate antibiotic prescribing. 1,2The majority of human antibiotic use occurs in general practice. 3,4One of the most significant determinants of antibiotic prescribing is the prescribing habits of individual GPs. 5,6][9] Evidence suggests that once prescribing habits are formed, they tend to remain stable over time. 8,9Training of GPs varies between countries, although typically following a similar process across nations.1][12] A few countries do not have mandatory specialist general practice training, allowing graduates to start working as a GP straight after graduation. 13esearch has shown that medical students consider resistance a public health concern, yet still have serious information gaps with regard to antibiotic usage. 14A qualitative study by Dallas et al. 15 found that GPs in vocational training in Australia are 'used to' prescribing antibiotics in the hospital setting where they regularly see serious infections. 15The transition from the hospital setting to general practice is a crucial point in a clinician's career. 16Therefore, investigating the antibiotic prescribing of GPs during the early-career period may inform the future stewardship of antimicrobial prescribing.
This group may be more receptive to interventions, given they have not yet formed these long-term prescribing habits.A systematic review examining interventions in junior doctors and medical students demonstrated prescribing behaviours can be altered. 17espite the importance of a GP's experiences early in their career for determining their ongoing antibiotic prescribing behaviour, there are currently no systematic reviews of studies on this topic.
We aimed to explore the antibiotic prescribing patterns of early-career GPs for acute infections, and if there is a relationship between antibiotic prescribing and working experience.

Exclusion criteria
Studies were excluded if they were: (i) examining complex or severe conditions, e.g.COPD, chronic bronchitis, community-acquired pneumonia, recurrent infection or severe infection; (ii) examining prescribing in complex patients, e.g.immune compromised, UTIs in males, or pregnant women; (iii) in non-general practice settings in primary care, e.g.residential aged care facilities, emergency departments and urgent care; (iv) prescribers who are not GPs but work in primary care, e.g.paediatricians or nurse practitioners; or (v) studies with fewer than five early-career GPs.

Search strategy
Databases searched were PubMed, Embase and Scopus.Articles were included if they were original research, and no limitations were placed on publication date.The last search date was 17 October 2022.
Articles not in English were examined separately and translated to English via Google Translate.Included full-text articles were then searched manually for additional records via citation searching, using Google Scholar.

Example search (PubMed)
Search terms used were 'primary care' OR 'general practice' OR 'general practitioner' OR 'family medicine' OR 'family practice' OR 'community care' AND 'early-career' OR 'trainee' OR 'registrar' OR 'resident' OR 'student' OR 'vocation' AND 'antibiotic' OR 'antibacterial agent' OR 'antibiotic resistance' OR 'anti-infective agent' OR 'antimicrobial stewardship' OR 'resistance' OR 'antimicrobials'.See Table S1 (available as Supplementary data at JAC Online) for our full search strategy.

Outcomes
Primary outcomes were antibiotic prescribing rates for common acute infections by early-career GPs.Secondary outcomes were any associations between working experience of the GP and antibiotic prescribing.

Data collection
Search results were downloaded into Covidence, and duplicates removed.Two authors (E.J.B. and G.M.) screened titles, abstracts and full texts independently using Covidence.Disagreements were discussed and, if required, resolved by a third author (M.L.V.D.).Data extracted into an Excel spreadsheet included study characteristics (setting, design, country), GP demographics (age, number, definition of early career, sex), patient population (age, number) and outcomes (condition, prescribing rates, working experience variable, statistical measures).

Data analysis
Data were tabulated and narrative analysis was undertaken.Subgroup synthesis of primary outcomes was by condition, with prescribing rates and 95% CIs where available.Secondary outcomes were presented by the nature of the relationship (direction and magnitude) between prescribing and experience, and variable used.

Risk of bias in individual studies
The Newcastle-Ottawa scale was used for cohort and casecontrol studies, and adapted for cross-sectional studies; see Table S2. 22Control arms of randomized controlled trials (RCTs) were assessed using the Cochrane Risk of Bias tool. 23E.J.B. performed the risk of bias and G.M. checked a randomly selected number of studies.

Search results and study characteristics
Of the 1483 records identified, 376 duplicates were removed, leaving 1107 records for title and abstract screening.There were 128 records eligible for full-text screening; see PRISMA diagram (Figure 1). 18Thirty studies were included, and after citation searching was performed, 41 studies were included in the review, the characteristics of which are presented in Table S3.][26] Studies examining training versus non-training practices without reporting career stage of the prescriber were excluded (n = 20).

Continued
Systematic review used to measure the association between antibiotic prescribing and experience of the GP.

Secondary outcomes
Of the 27 studies examining secondary outcomes, 17 found a statistically significant relationship between experience and reduced prescribing, 8 found no statistically significant difference, and 2 found more experience resulted in less prescribing (Tables 2 and 3).

Studies concluding less-experienced GPs prescribed fewer antibiotics
Of the studies with a statistically significant relationship (n = 17), 7 found that early-career GPs have decreased odds of prescribing antibiotics compared with later-career GPs, with OR ranging from 0.25 to 0.68. 19,30,31,35,57,60,61The other 10 found that working experience significantly influenced antibiotic prescribing.Walsh et al., 57 examining antibiotic prescribing for 'non-indicated conditions', reported the lowest OR of 0.25 (95% CI 0.18-0.36). 57Akkerman et al. 39 concluded that 'years in practice' was the most important factor explaining variation in antibiotic prescribing, accounting for 29% of prescribing variability. 39ainous et al. 55 found that lower antibiotic prescribers (25 th percentile and below) compared with high prescribers (75 th percentile and above) had significantly fewer years since graduation. 55wo studies found partially significant results, depending on the country or the antibiotic class prescribed. 59,63Safaeian et al. 59 examined 3372 GPs' prescribing of different antibiotic classes, and found that early-career GPs were less likely to prescribe cephalosporins, macrolides and quinolones, but more likely to prescribe an aminoglycoside.Neither study found a statistically significance difference for penicillins, sulphonamides and tetracyclines, compared with later-career GPs. 59ordoba et al. 63 examined prescribing for sore throat across six countries; Lithuania was the only country with a statistically significant relationship between years in practice (OR 0.05; 95% CI 0.01-0.3). 63However, all countries had very low sample sizes of early-career GPs (11-63), overall small sample sizes and high variability in prescribing between GPs. 63

Studies that did not identify a relationship between experience and antibiotic prescribing
Of the eight studies that found no statistically significant relationship, three compared early-career with late-career GPs' antibiotic prescribing, 6,34,58 and five examined the influence of years in practice on antibiotic prescribing. 5,29,33,36,66Seven of the eight studies did not report the number of early-career GPs included, or the range of years in practice.The one study that did report this included eight GPs with 5-10 years' experience, and no GPs with experience of <5 years. 5wo studies provided additional information regarding GPs in training. 29,56Pynnonen et al. 56 found that 'having a GP trainee present during a patient visit' reduced the likelihood of prescribing (OR 0.36; 95% CI 0.2-0.65). 56In the study by Petrovic et al., 29

Studies finding more-experienced GPs prescribed fewer antibiotics
Di Martino et al. 27 found that, with an increase in increments of 5 years' experience, the odds of prescribing reduced (OR 0.97; 95% CI 0.96-0.99). 27They examined all patients aged 6-13 years in a region of Italy, including 5097 physicians, 15% of which were paeditricians. 27egnan et al. 53 found that prescribers who were boardcertified before 1997 had a lower rate of antibiotic prescribing compared with those registered more recently (63% versus 76%, P = 0.02).Those in teaching practices in this study prescribed 22% fewer antibiotics (73% versus 51%, P ≤ 0.01). 53

Risk of bias
The 39 observational studies were generally considered at low risk of bias, with only 3 having serious risk of bias using the Newcastle-Ottawa Scale (Table 4).The main concerns were sample size (n = 10), selection bias (n = 8) or confounding (n = 10).Selection bias was due to either: not being representative of GPs in their country (n = 6); excluding low antibiotic prescribing GPs (n = 1); or GPs were aware of the study aims (n = 1).The majority of records controlled both patient and GP factors, although some may have been subject to confounding, either by only focusing on GP factors/not controlling patient factors, or by controlling only a small number of confounders.Outcome measurement across almost all studies was appropriate, with most using record-linked data (n = 37; 95%).
One of the two RCTs had a high risk of bias; participants were aware of their allocation and 5/40 GPs in the control group dropped out after randomization. 35Reason for declining to participate was not reported; however, participants may have declined after randomization as they would not receive the C-reactive protein testing kits given to the intervention group (not readily available in Latvian general practice). 35Participants recorded their own prescribing, and were not required to record all consultations for infections.The other RCT had a lower risk of bias; although it did not report if participants were aware of the intervention, the intervention was embedded in their regular GP training (unlikely to cause performance bias). 50

Discussion
In the majority of studies identified, early-career GPs prescribed fewer antibiotics than later-career GPs, across a variety of conditions and countries.Although highly heterogeneous in variables used to describe the outcome and measurement of the variables, most studies found more years in practice was associated with higher likelihood of antibiotics prescribed.This may be encouraging for future antibiotic stewardship if these lower antibiotic prescribing rates of newer generations of GPs reflect increased awareness of the importance of antimicrobial resistance over the past decade. 2,67However, it is also reasonable to speculate that GPs begin their career with more evidence-based prescribing, but this may deteriorate with time in practice, due to financial, time and patient Systematic review across a comprehensive range of geographical and socioeconomic Australian settings, and has limited biases. 49However, ReCEnT captures data only in the first 2 years of clinical experience in vocational training and we do not know if prescribing habits persist.
In addition, it makes our review Australia-centric, particularly with regard to prescribing rates, limiting generalizability of the findings.
The search strategy was narrowed to include early-career GP terminology: this may have excluded studies from a wider range of GP career stages that didn't separately present data of earlycareer GPs.There may also be other terms for GPs in training used in non-English-speaking countries that were not identified in our search string for 'early-career'.To mitigate this potential selection bias, a comprehensive citation search was performed.
Most of the literature was current (70% from the past 10 years); however, three of the studies were published prior to 2000. 28,51,55Two of these found more-experienced GPs prescribed more, and the other reported prescribing rate data. 28,55ome of the included studies had small sample sizes, or were from single regions, which may not be representative of GPs in their respective countries.Many studies examined the association between prescriber age and antibiotic prescribing but were excluded as working experience may vary across GP ages.
We intended to perform meta-analysis, but this was not appropriate due to methodological and clinical heterogeneity of the included studies.The PROSPERO protocol stated that secondary outcomes were 'appropriateness of antibiotic (first line, second line, specified with regard to authoritative prescribing guideline in country)'.This was changed, because first-line antibiotic treatment choice differs across international guidelines, and therefore was not comparable.This was revised to be 'any associations between working experience of the GP and antibiotic prescribing', after the search was performed, but before full-text screening.The change of the secondary outcome was needed and added important information to the narrative about antibiotic prescribing of early-career GPs.

Comparison with existing literature
A systematic review by Hawkins et al. 69 comparing Australia, Sweden and the UK found that neither antibiotic consumption nor community knowledge has changed significantly in Australia and the UK since 2011.In line with the Hawkins review, 69 we also found the lowest antibiotic prescribing rates in Sweden, a country with established low antibiotic prescribing. 70 previous qualitative review determined that key driving factors of unnecessary prescribing included diagnostic uncertainty, time pressure and patient pressure. 68A qualitative study of GPs in vocational training presented similar themes, but also pointed to the inexperience of the GP, and the influence of the supervisor. 15GPs in training viewed guidelines favourably, and following them was deemed desirable. 15Conversely, in the literature review, lack of adherence to guidelines/continuing professional education was noted by 10 of the 17 studies. 68Less-experienced GPs' preference to use guidelines, in addition to their recent medical education, may explain why their prescribing is lower than more-experienced GPs. 15,682][73] Although data from training practices includes early-career GPs, supervisors who may be of varying experience were also included in 'training practices' as their unit of analysis.Training practices' antibiotic prescribing may be of interest for further review, as knowledge in this area could inform medical education.Data on early-career GPs, particularly after vocational/specialist training, are still lacking.

Implications for research and/or practice
We found limited international data on early-career GPs' antibiotic prescribing, and this varied by country.Antimicrobials continue to be overprescribed, even by early-career GPs, who have had recent medical education.Antimicrobial resistance is a global problem, and it is important to achieve a greater understanding of early-career GPs' prescribing in a wider range of settings.A previous non-randomized trial of education targeting GP trainees demonstrated a (short-term) substantive decrease in antibiotic prescribing for acute bronchitis. 50Further interventions targeting early-career GPs could examine effects on antibiotic prescribing for other conditions (and assess longer-term sustained changes), so contributing to future antimicrobial stewardship.

g
Family medicine resident: first 3 years in practice (USA).h GP residents: first 5 years in practice (Sweden).

Table 1 .
Early-career GPs' antibiotic prescribing rates for various acute infections

Table 2 .
Secondary outcomes-ORs comparing early-career GPs with later-career GPs on the prescription of antibiotics

Table 3 .
Measurements of the association between GPs' working experience and antibiotic prescribing P = 0.016) compared with those who practice without specialist training (in Serbia, one can practice as a GP without postgraduate GP training).29