Long-term impact of an educational antimicrobial stewardship programme in primary care on infections caused by extended-spectrum β-lactamase-producing Escherichia coli in the community: an interrupted time-series analysis
Summary
Background There is little evidence on the ecological effect and sustainability of antimicrobial stewardship programmes (ASPs) in primary-care settings. We aimed to determine whether a multimodal, educational ASP would be sustainable in the long-term and reduce the incidence of infections caused by extended-spectrum β-lactamase-producing Escherichia coli in the community by optimising antibiotic use.
Methods We did this quasi-experimental intervention study in 214 primary health centres of four primary health-care districts in Andalusia, Spain. Local multidisciplinary teams, comprised of general practitioners, paediatricians, primary-care pharmacists, and epidemiologists, were created in each district and implemented a multimodal, education-based ASP. The core activity of the programme consisted of regular one-to-one educational interviews between a reference interviewing physician and prescribing physicians from each centre on the appropriateness of their most recent (same or preceding day) antibiotic prescriptions based on a structured questionnaire. Appropriate prescribing was defined as compliance of all checklist items with the reference guidelines. An average of five educational interviews were scheduled per prescriber per study year. We did an interrupted time-series analysis to assess the effect of the intervention on quarterly antibiotic use (prescription and collection by the patient) and quality of prescriptions (as defined daily doses per 1000 inhabitants per day) and incidence per 1000 inhabitants of E coli producing extended-spectrum β-lactamase (ESBL) isolated from urine samples.
Findings The study was done between January, 2012, and December, 2017, in a pre-intervention period of 2012–13 and an intervention period of 2014–17. Throughout the study period, there were 1387 physicians (1116 general practicioners and 271 paediatricians) in the included health centres serving a mean population of 1 937 512 people (299 331 children and 1 638 181 adults). 24 150 educational interviews were done over the 4 years. Inappropriate antibiotic prescribing was identified in 1794 (36·5%) of 4917 educational interviews in 2014 compared with 1793 (26·9%) of 6665 in 2017 (p<0·0001). The intervention was associated with a sustained reduction in the use of ciprofloxacin (relative effect –15·9%, 95% CI –23·9 to –8·0) and cephalosporins (–22·6%, –35·9 to –9·2), and a sustained increase in the use of amoxicillin (22·2%, 6·4 to 38·0) and fosfomycin trometamol (6·1%, 2·6 to 9·6). The incidence density of ESBL-producing E coli decreased by –0·028 cases per 1000 inhabitants (95% CI –0·034 to –0·021) after the start of the programme, reversing the pre-intervention increase and leading to a relative reduction of –65·6% (–68·2 to –63·0) 4 years later. Interpretation Our data suggest that implementation of a multimodal ASP in primary care that is based on individual educational interviews improves the use of antibiotics and results in a sustained significant reduction of infections by ESBL-producing E coli in the community. This information should encourage the implementation of ASPs in primary care. Introduction The inappropriate use of antibiotics is one of the main factors responsible for antibiotic resistance.1 On We searched the Cochrane Database of Systematic Reviews and PubMed for studies on antimicrobial stewardship in the community setting published before March 1, 2019, using the search terms “antimicrobial stewardship”, “antibiotic stewardship”, or “stewardship program”, in addition to the terms “primary care”, “community care”, “ambulatory care”, “antimicrobial resistance”, and “extended-spectrum-beta- lactamase-producing Escherichia coli”, with no date or language restrictions. We reviewed guidelines on antimicrobial stewardship programmes (ASPs) published by relevant international scientific societies. We retrieved additional publications from the authors’ personal reference lists and the reference lists of included studies. We identified three systematic reviews pooling the evidence provided by these studies. In these studies, we found that the evidence on the efficacy of ASPs in primary care for improving antibiotic use was considered low to moderate. In addition, information about the association of key intervention components with improved prescribing and about the efficacy of educational interventions in changing clinicians’ prescribing behaviour was scarce, as were data on the sustainability and large-scale applicability of ASPs in health-care systems at the community level. Most studies focused exclusively on respiratory tract infections, with little information about the effectiveness of ASPs in other infections. In particular, urinary tract infections, which are commonly misdiagnosed and overtreated, were underrepresented. There was also a paucity of studies designed to assess the ecological effect of ASPs in the community, with a very low level of evidence on substantial or persistent reductions in the incidence of resistant bacteria, including E coli producing extended-spectrum β-lactamase. Added value of this study Our study supports the efficacy of ASPs in primary care to improve antibiotic use and associated ecological effects, such as reducing urinary tract infections caused by E coli producing extended-spectrum β-lactamases. It reinforces the value of educational interventions in increasing adherence to reference guidelines and change the prescribing behaviour of clinicians, and, to our knowledge, includes for the first time regular one-to-one educational interviews as the core activity of an ASP. The programme and its findings are sustainable over time and enforceable in health-care systems at the community level. Implications of all the available evidence Our findings suggest that multimodal educational ASPs based on peer-to-peer educational interviews can be useful to improve the use of antimicrobials and reduce bacterial resistance in the community, and that these changes can be sustained over time. This information reinforces the recommendations to implement ASPs in primary care and should encourage further studies to confirm the usefulness of individual educational interviews. Antimicrobial stewardship programmes (ASPs) are one of the key elements needed to achieve this goal.3 These interventions have proved useful in hospital settings,4,5 but there is little evidence supporting the long-term sustainability and ecological impact of ASPs in primary care, where most antibiotics used in humans are prescribedG,7 and up to a half of antibiotic prescriptions are inappropriate.8 In primary care, Escherichia coli is the most frequent cause of urinary tract infections (UTIs), which are one of the most common types of bacterial infections9 and are also the most frequently microbiologically diagnosed. Severe infections caused by Enterobacteriaceae producing extended-spectrum β-lactamase (ESBL) are associated with increased mortality rates.10 ESBL-producing bacteria frequently show co-resistance to other antibiotics such as fluoroquinolones.11 Antibiotic resistance in E coli is increasing throughout Europe, with the highest proportions of resistant isolates occurring in southern and south-eastern countries.12 Over the past 15 years, the prevalence of third-generation cephalosporin-resistant E coli, fluoroquinolone-resistant E coli, and ESBL-producing E coli has increased in Spain.13,14 Antibiotic consumption in this country is higher than the European average and has significantly increased in between 2012 and 201G.15 In this worrying national context, we developed a primary-care ASP and did a quasi-experimental study to assess the effect of its implementation on antibiotic use and the incidence of ESBL-producing E coli isolated from urine samples in a population of nearly 2 million people in Andalusia, Spain. Our hypothesis was that the programme would result in sustained improved antibiotic prescribing and, consequently, would reduce the incidence of ESBL-producing E coli in primary care in the long-term. Methods Study design This quasi-experimental, interrupted time-series study included doctors and all area residents from 214 primary health-care centres in four primary health-care districts in Spain (Aljarafe-Sevilla Norte, Huelva-Costa-Condado- Campiña, Sevilla, and Osuna), in coordination with microbiology laboratories at five reference hospitals (appendix p 1). The number of residents were recorded every quarter (appendix p 4). Patient data were anony- mised, so informed consent was waived by the study’s ethical committee, which was fully independent of the study. Study approval was granted by the ethics committee of the University Hospitals Virgen Macarena- Virgen del Rocío (19122014–03/2014) and valid for the Autonomous Community of Andalusia. Procedures The methodology has been described elsewhere.1G Briefly, the ASP had institutional support from the Ministry of Health of the Regional Government of Andalusia (Spain) from conception and consisted of multimodal edu- cational interventions led by local multidisciplinary teams set up at two levels. Level 1 comprised a reference doctor at each primary health-care centre and level 2 comprised a general practitioner, a paediatrician, a primary care pharmacist, and an epidemiologist within each primary health-care district, in addition to the medical director of each of these districts. These teams were coordinated by an infectious disease expert and a microbiologist at each reference hospital. We implemented five educational interventions: (1) central (at regional level) and local (at the level of the hospital or primary-care district) explanation and dissemination of the programme by the scientific committee and local teams;1G (2) open online health courses created for the programme to train the local teams on the appropriate use of antibiotics for the most common infectious diseases (eg, respiratory tract infections, urinary tract infections, skin and soft tissue infections, and sexually transmitted infections);17 (3) regular clinical sessions at each centre to update the local protocols for management of the most common infections in the community, on the basis of both the ecological data provided by each reference hospital’s microbiological laboratory and on regional antimicrobial reference guidelines;18 (4) educational interviews, the core activity of the programme, which consisted of regular one-to-one interventions; and (5) provision of quarterly reports to all participating centres, with an analysis of results for all outcomes obtained for feedback and benchmarking. The programme was disseminated to doctors and other health-care professionals in the participating centres using a web-based platform, training courses, seminars, and face-to-face sessions.1G The scientific committee comprised members of scientific societies actively involved in the prevention, diagnosis, and treatment of infections, and was responsible for creating, managing, coordinating, and disseminating the programme.1G For the educational interviews, the reference doctor from each centre selected a patient at random who had been prescribed antimicrobial treatment at the centre on the same or the preceding day. The reference doctor then held an educational interview with the prescribing doctor, reviewing key aspects of diagnosis and antimicrobial therapy on the basis of the best evidence available in the reference Aljarafe Area Antimicrobial Guidelines.18 The educational interviews followed a structured questionnaire (appendix p 2). As the approach was strictly educational, for the purposes of furthering knowledge and appropriate prescribing practice, no changes in the prescription were required. All doctors from the participating sites were scheduled to receive an average of five educational interviews per year. After every educational interview, the prescriber was invited to complete a voluntary anonymous satisfaction survey (appendix p 3) to assess the level of acceptance of the programme. The Ministry of Health of the Regional Government of Andalusia covered the cost of the training courses and meetings, allocated members with time dedicated to the educational activities, and included the objectives of the ASP in the annual agreement signed between each centre and the Andalusian Health Service. Outcomes We assessed inappropriate antibiotic prescribing through the educational interviews. Prescriptions were deemed appropriate when all questions of the questionnaire were assessed positively; if one or more of the items was answered negatively, the prescription was deemed ina- ppropriate. We assessed antibiotic prescriptions recorded every 3 months (quarterly) through the computerised pharmacy records of reimbursed and dispensed drugs, as a proxy for antibiotic consumption, of the resident population counted for the corresponding quarter. The records database is hosted by the Pharmacy Service of the Andalusian Health Service; it is a computerised system that collects data for all dispensed and reimbursed antibiotics prescribed electronically across the Health Service. Data were expressed as defined daily doses per 1000 inhabitants in the patient study population per day (DID) for WHO’s Anatomical Therapeutic Chemical Classification System group J01 (antibacterials for systemic use) and for the following antibiotics prescribed in the ambulatory setting: amoxicillin–clavulanic acid, cipro- floxacin, levofloxacin, cefuroxime, third-generation cepha- losporins, amoxicillin, and fosfomycin trometamol. We also estimated quarterly incidence density of ESBL- producing E coli isolated from outpatients’ urine cultures per 1000 inhabitants per day and as resistance proportion (ie, the number of ESBL-producing E coli isolates divided by the number of all E coli isolates). Similar to antimicrobial consumption, this assessment was also done in the number of inhabitants for each quarter. Strains were defined as susceptible or resistant according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) recommendations.19 ESBL production was determined by MicroScan ID/AST panels (Beckman Coulter; Brea, CA, USA) and confirmed by combination disk test with a cephalosporin alone (cefotaxime or ceftazidime) and in combination with clavulanic acid on the basis of EUCAST criteria.20 The level of acceptance of the programme was assessed by satisfaction surveys. Statistical analysis We used an interrupted time-series analysis, which is a robust quasi-experimental design to evaluate the effectiveness of population-level health interventions.21 To assess the effect of the ASP on antibiotic prescribing and incidence of resistance, we used a longitudinal segmented regression with a generalised least squares approach and an autoregressive moving-average model to account for autocorrelation. We could thus estimate changes in level or trend of prescription and resistance after inception of the programme, adjusting for seasonality and identifying potential outliers. We used the Akaike Information Criterion to find the most parsimonious models, checked for normality of residuals, and validated the autocorrelation structures by means of likelihood ratio tests to select the final model for each outcome. To identify whether changes in outcomes were attributable to the ASP, we calculated absolute and relative differences between observed changes and estimated values that would have been expected in the absence of the intervention for the last quarter of 2017, 4 years after the inception of the programme. Proportions of inappropriate prescribing outcomes were compared between years with the χ test with the Yates correction. All statistical analyses were done using R software, version 3.5.0. We additionally did a joinpoint regression analysis22 to assess the robustness of the interrupted time-series analysis findings. A p value lower than 0·05 (two-tailed) was considered significant. We used the Joinpoint Regression Program, version 4.G.0.0, to detect significant contemporaneous or lagged changes in resistance trends related to the intervention. Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results The mean study population comprised 1 937 512 individuals (299 331 paediatric patients and 1 G38 181 adult patients; appendix p 4) seen by 1387 prescribing physicians (111G general practitioners and 271 paediatricians). There were 1 941 119 patients (304 814 children and 1 G3G 305 adults) in the area in the quarter before the start of the programme and 1 94G 548 in the final quarter (289 555 children and 1 G5G 993 adults). The study was done from Jan 1, 2012, to Dec 31, 2017, spanning 24 calendar quarters. The pre- intervention period ran from 2012 to 2013 and the intervention period ran from 2014 to 2017. 24 150 educational interviews were done with the 1387 prescribers from 214 primary care centres during the intervention period. 4917 educational interviews were held in 2014, followed by a mean 11% (SD 10·1) average annual increase, reaching GGG5 educational interviews in 2017. The mean number of educational interviews per prescriber was 4·4 (0·G) per year. The prescribers from 52G8 (21·8%) education interviews completed the anonymous satisfaction survey that they were invited to after every educational interview, and gave a positive assessment of the interview in 4952 (94·0%) instances. The rate of inappropriate antibiotic prescription showed an annual reduction of –3·2% (r² 0·998) from the 2014 to 2017 (p=0·001), decreasing from 1794 (3G·5%) pres- criptions identified from 4917 educational interviews in 2014 to 1793 (2G·9%) prescriptions from GGG5 interviews in 2017 (figure 1). The most common causes of a prescrip- tion’s inappropriateness were the choice of antibiotic (2800 [3G·9%] of 7579 inappropriate prescriptions) and the duration of treatment (2G19 [34·5%]; appendix p 5). Total antibiotic prescriptions were stable during the study period, with no significant changes detected. Compared with the expected antibiotic consumption based on the baseline trend, long-term reductions in DID, with significant absolute and relative effects (difference from the pre-intervention trend), were observed at the end of the 4-year intervention period for cefuroxime, but not for third-generation cephalosporins, and ciprofloxacin (table, appendix p G). Similarly, significant sustained increases in DID were found for amoxicillin and fosfomycin trome- tamol (table, appendix p G). No sustained changes were detected between baseline and intervention trends for levofloxacin or amoxicillin–clavulanic acid (table). Time-series plots for individual antibiotics (appendix p 7) show seasonal prescription patterns, with peaks in the first quarter of each year, closely associated with influenza epidemics, and minimum values during the summer periods. A significant outlier was detected, coinciding with the 2014–15 influenza epidemic, which produced an increase in the overall antibiotic use of 2·7G5 DID (95% CI 1·404–4·12G, p=0·0008) in the first quarter of 2015 (appendix p 8). Susceptibility testing was done on all of the G7 428 E coli clinical isolates collected during the G years of the study. The incidence density of ESBL-producing E coli decreased significantly during the intervention period. The observed increase in the pre-intervention period (0·004 cases per 1000 inhabitants per quarter, 95% CI 0·002 to 0·005, p<0·0001) showed a change of –0·028 cases per 1000 inhabitants (–0·034 to –0·021, p<0·0001) after implementation of the ASP, with a change in slope of –0·00G cases per 1000 inhabitants per quarter (–0·007 to –0·005, p<0·0001; figure 2). Compared with the expected resistance incidence based on the pre-intervention trend, an estimated relative reduction of –G5·G% (95% CI –G8·2 to –G3·0) was obtained 4 years after the start of the programme, accounting for an absolute reduction of –0·121 cases per 1000 inhabitants (–0·148 to –0·095). The joinpoint regression analysis done in addition to the interrupted time- series analysis supported these findings, detecting a significant point of change in the 8th quarter of the study (95% CI G to 10, p<0·05), coinciding with the start of the intervention (appendix p 9). The proportion of ESBL-producing E coli decreased from 7·1% (893 of 12 575) to 5·5% (539 of 97G7) from the year before the intervention to the final year of the intervention (p=0·0001; appendix p 10). The upward pre- intervention trend (slope 0·12, 95% CI 0·09 to 0·15, p<0·0001) showed a decrease of –1·3% (95% CI –1·4 to –1·2, p<0·0001) after the start of the programme, followed by a downward trend (change in slope –0·18, 95% CI –0·21 to –0·15, p<0·0001), with an estimated absolute effect of –4·1% (95% CI –4·7 to –3·5) by the end of the study period (appendix p 11). Discussion Our results show a positive long-term effect on antibiotic prescribing of an educational-based ASP, designed for the primary care setting, which had an excellent level of acceptance by clinicians. The ASP improved prescribing appropriateness and led to a reduction in the use of ciprofloxacin and cephalosporins, to an increase in the use of amoxicillin and fosfomycin trometamol, and to a decrease in the incidence of ESBL-producing E coli in the community across a health-care system. Attention to antimicrobial stewardship is increasing. In 2011, the EU urged all member states to implement action plans against antimicrobial resistance, promoting the prudent use of antibiotics.23 However, the population- weighted mean consumption of antibiotics for systemic use in the EU and the European Economic Area did not show any significant change during the period 2013–17 (22·3 DID in 2013 vs 21·8 DID in 2017), nor did the consumption of penicillins (11·1 vs 11·5), cephalosporins (2·2 vs 2·0), or quinolones (1·8 vs 1·G).24 These data highlight the value of our intervention that led to a change in prescribing behaviour by reducing the use of ciprofloxacin and cephalosporins and increasing the use of amoxicillin and fosfomycin trometamol. Multifaceted interventions, such as those in our study, are more effective than single interventions in changing antibiotic prescribing behaviour.7 Additionally, studies have found that stewardship interventions such as audit and feedback improve professional practice, with greater effects when they are increasingly frequent, and have highlighted the importance of continuity to sustain positive effects in the long term.25 On the basis of this evidence, we designed and implemented an intervention to improve antibiotic prescribing behaviour that lasted for 4 consecutive years and provided quarterly feedback. Whereas most studies have focused exclusively on respiratory tract infections,G,7,25 our ASP focused on all types of infections using a comprehensive educational approach, in line with expert recommendations.G,7 Our ASP did not include any restrictive interventions. Although restrictive interventions, such as formulary restrictions, pre-authorisations, therapeutic substitution, and automatic stop orders, can have rapid effects on the use of targeted antibiotics and subsequently reduce the resistance of bacteria linked to these antibiotics, the capacity of these interventions to change antibiotic prescribing behaviour in the long term is not clear, as prescribers might consider any restrictions to be a limitation to their clinical freedom.4,25 An important achievement of the programme was the optimisation of antibiotic therapy. Overall antibiotic prescribing did not decrease significantly, probably because of the high increase of 2.77 DID (95% CI 1·40 to 4·13) in consumption that took place in the 13th quarter of the study (appendix pp G, 7). This increase coincided with a influenza epidemic in the first quarter of 2015, which had the highest overall cumulative age-adjusted infection incidence since the 2004–05 influenza season.2G Influenza incidence was 21% higher in 2015 than in 2014 (30G·5 cases vs 253·G cases per 1 000 000 inhabitants) and the duration of the influenza season was 14% longer (8 vs 7 weeks).2G This unexpected increase might have also adversely affected the post-intervention slopes of antibiotics commonly used for respiratory tract infections, such as amoxicillin–clavulanic acid, levofloxacin, and cepha- losporins, attenuating the effect of the intervention. Dispensed and reimbursed ciprofloxacin and cepha- losporins decreased significantly, whereas prescriptions for amoxicillin and fosfomycin trometamol increased during the intervention period, in line with the recommendations of the programme guidelines.18 The increase in third-generation cephalosporins after the start of the ASP was probably driven by improved adherence to the reference guidelines, although the use of these antibiotics in primary care remained low in our study area, compared with EU data for 2014–17 (appendix p G). Our findings show the effect of the intervention on antibiotic use, consistent with other studies.G However, when ASP interventions are discontinued and do not change prescribing behaviour, antimicrobial prescribing tends to revert to baseline levels.27 Very few studies have reported persistent effects beyond 12 months.25,27,28 To avoid the negative effects of discontinuation, our ASP has remained operational without interruption. 4 years after the inception of our programme, the incidence density of ESBL-producing E coli was G5·G% lower than expected on the basis of the pre- intervention trend. Several findings suggest that our ASP—specifically, reduced prescriptions for ciprofloxacin and cephalosporins after the start of the intervention—was the main reason behind this sustained ecological effect. First, cephalosporins and fluoroquinolones have been identified as independent risk factors for community- onset bacteraemia due to ESBL-producing E coli.29 Second, two studies30,31 have shown that decreasing antibiotic consumption has a positive ecological effect in both hospital and primary care settings by reducing the incidence of multidrug-resistant Enterobacteriaceae, including ESBL-producing E coli. However, those studies did not implement ASPs but assessed the effect of nationwide interventions restricting the use of cipro- floxacin; in one of the studies, once the restriction was lifted, quinolone use and E coli susceptibility returned to baseline levels.30 Third, according to data from the European Antimicrobial Resistance Surveillance Network, from 2014 to 2017, which is the time period of our intervention, there was a significant increase in the population-weighted mean percentages of E coli isolates with resistance to third-generation cephalosporins in the EU and European Economic Area.32 Fourth, to ensure that the effect on incidence density was not biased because of potential increases or decreases in the number of urine samples submitted, resistance proportion was assessed. The trend analysis showed changes in resistance proportion that were consistent with the incidence density model. Fifth, to our knowledge, no other competing interventions occurred contemporaneously with the programme, there were no changes in study techniques, protocols, and guidelines, and the composition of the study population remained unchanged before and after the intervention. Finally, a limitation of our study was that the interrupted time-series analysis did not include a control group, but we also did a joinpoint regression analysis to increase the strength of inference and to support our findings. This analysis detected significant inflection points in the incidence density of ESBL- producing E coli, consistent with the findings of the interrupted time-series analysis, with a steep decrease in the first quarters after the inception of our ASP, followed by a significant, less intense sustained decrease until the end of the study period. This effect was probably because the possibilities of an improvement in the results were higher at the beginning of the programme than later on. Therefore, it seems plausible that the reduction in the incidence of ESBL-producing E coli in the community was attributable to the intervention. It is unclear whether the increase in the use of fosfomycin trometamol in our study was associated with the decrease in incidence of ESBL-producing E coli. The relationship between fosfomycin trometamol and ESBL-producing E coli is unknown, but fosfomycin has good in-vitro activity against common uropathogens, including ESBL-producing E coli.33 Furthermore, a single dose of fosfomycin trometamol, as recommended in our ASP, had similar clinical and bacteriological efficacy against E coli to longer courses of ciprofloxacin, norfloxacin, co-trimoxazole, nitrofurantoin, cefuroxime, ceftibuten, and amoxicillin– clavulanic acid in people with lower UTI and pregnant women with asymptomatic bacteriuria.33 Studies are needed to investigate the effect of fosfomycin trometamol on ESBL-producing E coli.
Our programme aimed to teach general practitioners and paediatricians how to improve their antibiotic- prescribing behaviour without compulsory measures, such as antibiotic restrictions or changes to any prescription, to avoid interference with their autonomy. We applied the concept of peer-teaching34 through regular one-to-one discussions. The acceptable workload involved in, and the high level of acceptance of, the intervention helped to integrate the ASP into the routine practice of every centre. In our opinion, the effectiveness
of our ASP derives from its educational design, with the educational interview being the core activity of the programme that was embedded in a bundle of non- restrictive educational actions. This approach has previously shown success in hospitals.5 Our study shows that a sustainable ASP based on regular one-to-one peer- educational interventions can also be successfully implemented in the primary care setting across a whole health-care system.
A strength of the study is the interrupted time-series analysis, which is the strongest quasi-experimental research design to evaluate the effect of a health intervention when a randomised trial is not feasible or unethical.21 There is a risk in these studies of confounding from other interventions happening at the same time, but no other stewardship programmes were in operation in primary care in the region before the implementation of our ASP. Pre-existing local antimicrobial guidelines were used on a discretionary basis in some health-care centres of Andalusia before the onset of our study, and some primary care physicians who were not part of any stewardship programme might have attended workshops on antibiotic use, although these were carried out before our multimodal education-based ASP. Moreover, before our intervention, the annual agreement, signed by the Andalusian Health Service and the primary care centres, included antimicrobial-prescribing objectives. Therefore, it is possible that those interventions could have positively affected the use of antibiotics in some centres during the pre-intervention period.
The fact that our programme did not aim to educate patients and that prescribers were not trained in communication skills might constitute other weaknesses of the programme, as those interventions have been associated with improved prescribing outcomes when combined with clinical education of physicians.G,7 Including such a mixed strategy in an ASP would be recommended. We did not analyse the clinical effect of the intervention because we did not find a widely accepted indicator for that purpose. We have instead used a process indicator, the quality of antibiotic prescribing, to measure the effect of the programme. Further efforts to define clinical indicators in outpatient ASPs are needed. Although we did not do a cost- effectiveness analysis, the health outcomes of the programme have contributed to its sustainability. The ASP activities are ongoing in all public primary health- care centres of Andalusia, and since 2017 it has been expanded to include physicians, pharmacists, and dentists working in private practice in primary care. However, this research was done in a specific public health-care system, thus it might require adaptation to the characteristics of other settings.
In conclusion, our findings suggest that this multimodal educational ASP can improve the use of antibiotics and reduce the incidence of infections caused by ESBL-producing E coli in the community, and that these changes can be sustained over time. This information reinforces the recommendations to implement ASPs in primary care and should encourage further studies to confirm the usefulness of individual educational interviews to improve antibiotic stewardship and reduce the spread and prevalence of bacterial resistance.