Educational Interventions to Reduce Prescription and Dispensing of Antibiotics in Primary Care: A Systematic Review of Economic Impact

Antibiotic resistance remains a crucial global public health problem with excessive and inappropriate antibiotic use representing an important driver of this issue. Strategies to improve antibiotic prescription and dispensing are required in primary health care settings. The main purpose of this review is to identify and synthesize available evidence on the economic impact of educational interventions to reduce prescription and dispensing of antibiotics among primary health care professionals. Information about the clinical impact resulting from the implementation of interventions was also gathered. PubMed, Scopus, Web of Science and EMBASE were the scientific databases used to search and identify relevant studies. Of the thirty-three selected articles, most consisted of a simple intervention, such as a guideline implementation, while the others involved multifaceted interventions, and differed regarding study populations, designs and settings. Main findings were grouped either into clinical or cost outcomes. Twenty of the thirty-three articles included studies reporting a reduction in outcome costs, namely in antibiotic cost and associated prescription costs, in part due to an overall improvement in the appropriateness of antibiotic use. The findings of this study show that the implementation of educational interventions is a cost-effective strategy to reduce antibiotic prescription and dispensing among primary healthcare providers.


Introduction
Antibiotic resistance is globally recognized as a serious hazard to global public health [1], associated with negative impacts on health outcomes and expenditure [1]. The major driver of antimicrobial resistance has been a huge increase in antibiotic use, which increased by about 91% worldwide, and by 165% in low-and middle-income countries over the last decades [2,3]. Estimates show that drug-resistant infections will continue to rise dramatically, and by 2050 it is expected that 10 million deaths will occur each year and incur economic losses of over USD 100 trillion unless adequate interventions to limit unnecessary antibiotic use are implemented [1,4].
The prescription and dispensing of most antibiotics occur in primary healthcare facilities [5,6] which positions the health professionals as crucial stakeholders and partners in antimicrobial stewardship efforts [7]. Additionally, 25 to 50% of all antibiotics prescribed in primary healthcare are proved to be unnecessary, with substantial geographical and prescriber variability [8][9][10].
However, summaries of evidence on the most effective strategies considering their economic impact in primary healthcare are still lacking in the literature. A proper cost analysis of the existent educational interventions on antibiotic use in primary healthcare might provide insight into both the effectiveness and cost outcomes of antibiotic use, namely direct (medical and non-medical) and indirect (productivity loss, hospitalizations, etc.) costs, with the potential to be used by health professionals and policymakers [1,31]. Moreover, this information can help to support investments in appropriate, less expensive, and more beneficial strategies to reduce antibiotic use and, consequently, restrict the growing antibiotic resistance [1,31].
Thus, this study aimed to identify and synthesize available evidence on the economic impact of educational interventions to reduce the prescription and dispensing of antibiotics among physicians and pharmacists in primary healthcare. Furthermore, this review also gathered information about the clinical impact of educational interventions on the use, prescription and dispensing of antibiotics in primary healthcare settings.

Data Sources and Search Strategy
Searches in the International Prospective Register of Systematic Reviews (PROSPERO) and PubMed were conducted preceding the design of the present systematic review to exclude the existence of reviews or protocols with the same purpose as that presented in this review. No similar studies were found, and the review protocol was registered and is available at PROSPERO (CRD42022311272) [32]. This systematic review was performed in accordance with the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) (the PRISMA checklist can be found in Table S1, Supplementary Material) [33]. A systematic literature search was performed by a researcher on 21 February 2022, on the following electronic databases: PubMed, Scopus, Web of Science and EMBASE. Monthly automatic updates from each database were activated to ensure the update of the evidence. Search terms were based on a combination of keywords and MeSH terms on the review topic adapted for each database: (Intervention OR education* OR program OR "health promotion" OR session OR workshop) AND (antibiotic* OR antimicrobial*) AND (economic* OR cost* OR spending OR expend*) AND ("primary care" OR "primary health care" OR "Community Health Services" OR "general practitioner" OR practitioner OR "Community Pharmacy Services" OR "Pharmaceutical services"). Full search expression for each database is available in Text S1 (Supplementary Material).
The search was limited to terms found in titles, abstracts, and keywords. Reference lists of the selected articles were also scanned for other potentially eligible studies. Authors were contacted to obtain full texts when needed.

Study Selection and Eligibility Criteria
The screening process occurred in three steps and was conducted independently by two researchers (VR and ME): First, articles were excluded based on title, abstract and keywords. In step 2, full texts of the articles were evaluated to determine eligibility based on previously defined criteria. Then, in step 3, the selected articles were re-evaluated to assess their adequacy for data extraction. During the whole screening process, the researchers consecutively applied the following eligibility criteria: Studies were included if they: (i) were written in English, Portuguese, French, or Spanish; (ii) were experimental, quasi-experimental or observational studies; (iii) described original educational interventions such as communication and education activities, stewardship programs, treatment algorithms, delayed treatment, peer or community oversight, medication reviews, or any other framework on antibiotic use in primary healthcare (strategies that were merely administrative, applied incentives or coercion were excluded from this study); (iv) included as target population physicians (general practitioners and all other specialists) and/or pharmacists; (v) described at least one economic effect measure of the educational interventions on the prescription behavior of physicians and/or dispensing behavior of pharmacists in primary healthcare facilities. Studies not conducted in humans (e.g., in vitro or animal studies), review articles, qualitative studies, magazines, news, research protocols, thesis, reports, dissertations, abstracts, communications, posters, letters to the editor, unpublished work, editorials, commentaries, books, book chapters without original data, guidelines, statements, position papers and case studies were excluded.

Data Extraction and Quality Assessment
Data extraction retrieved information on authors, year, country, study design, study population, disease, type of intervention, a brief description of the implemented educational interventions, antibiotic-related outcomes (cost-related and non-cost related) measured, study time period translated in months and the perspective assessed.
The risk of bias in each study was assessed independently by two researchers (VR and VN) using the Joanna Briggs Institute (JBI) critical appraisal checklist for economic evaluations [34,35], since this scale might be applied to studies including economic impact measures independently of their study design. Disagreements were resolved by consulting a third author (ME). Additionally, the inter-rater agreement of the quality assessment performed by the two reviewers was evaluated using Cohen's kappa [36].

Data Syntheses and Analysis
The primary outcome was antibiotic cost. The results of the studies were summarized qualitatively and quantitatively. The decision matrix for the economic effects of the educational interventions had three possible outcomes adapted from a previous study [37]: (i) reject intervention: the educational intervention resulted in higher antibiotic cost, or had a higher cost and similar effectiveness, or similar costs and lower effectiveness, or higher cost and lower effectiveness; (ii) unclear: neither increase nor decrease in antibiotic costs were reported, or the intervention had lower cost and lower effectiveness, similar cost and similar effectiveness, or higher cost and higher effectiveness; (iii) favor intervention: the educational intervention resulted in savings in antibiotic cost, or had lower cost and similar effectiveness, similar costs and higher effectiveness, or lower cost and higher effectiveness. Figure 1 presents the literature search flow diagram. The systematic database search identified 4119 publications. After removing 1689 duplicates, the title, abstract and keywords were screened for 2430 papers. From these, 74 papers were full text screened. A total of 33 studies were included. Results of the quality assessment showed that only seven papers [38][39][40][41][42][43][44] presented more than two quality criteria classified as compliant and/or clear (Table S2 of Supplementary Material). The agreement between the two reviewers was substantial (k = 0.61, 95% CI 0.49-0.74, p < 0.001) and a final consensus was reached.

Study Selection and Quality Assessment
Antibiotics 2022, 11, x FOR PEER REVIEW 4 of 22 Figure 1 presents the literature search flow diagram. The systematic database search identified 4119 publications. After removing 1689 duplicates, the title, abstract and keywords were screened for 2430 papers. From these, 74 papers were full text screened. A total of 33 studies were included. Results of the quality assessment showed that only seven papers [38][39][40][41][42][43][44] presented more than two quality criteria classified as compliant and/or clear (Table S2 of Supplementary Material). The agreement between the two reviewers was substantial (k = 0.61, 95%CI 0.49-0.74, p < 0.001) and a final consensus was reached.
Three of the reviewed studies [52,57,62] specified that the improvement in appropriate behavior emerges from a reduction of broad-spectrum antibiotics use. Furst et al. [42], who assessed antibiotic consumption between 1999 and 2012, reported a significantly decreased consumption of antibiotics, except for β-lactamase-resistant penicillins [42] and also in restricted and non-restricted antibiotics consumption. Chazan and colleagues [44] showed a reduction in general antibiotic consumption, more specifically broad-spectrum antibiotic use in comparison with narrow-spectrum antibiotics. Concordantly, two additional studies [68] reported a significant decrease in broad-spectrum antibiotics along with an increase or stable use of narrow-spectrum antibiotics.
Armstrong et al. [46], who exclusively recruited bladder and kidney infection patients, revealed an increase in the administration of first-line trimethoprim-sulfamethoxazole or nitrofurantoin as initial antibiotic therapy accompanied by a decreasing consumption of fluoroquinolones [52].
Finally, Wei et al. [62] stated that to avoid patient complaints and the difficulty in differentiating between viral and bacterial infections, doctors tended to prescribe antibiotics even when not supported by the proper guidelines [65,68].
Specifically, the study of Aksoy and colleagues [45] conducted in Turkey over a period of eight years including family physicians reported decreased overall APR, mainly explained by a reduction in the number of boxes, items, and costs of antibiotic prescriptions. Another study [2] revealed that primary healthcare physicians who received a letter alerting the detection of their high prescription trends, tended to prescribe shorter-term prescriptions. Butler and colleagues [56] showed a significant reduction in terms of oral antibiotic dispensing, specifically for macrolides and phenoxymethylpenicillins (penicillin V), for all age groups and health conditions after implementation of the Stemming the Tide of Antibiotic Resistance (STAR) educational program when compared to the usual care. Two additional studies referred a reduction in prescription rates in specific subset of antibiotics, such as first-line antibiotics [64] and in broad-spectrum antibiotics [62].
Studies by Cals and colleagues [51] and Oppong et al. [54] reported that the reduction in APR was more pronounced among GPs that received an intervention combining the use of C-reactive protein (CRP) tests and training aimed at enhancing communication skills.
Walker and colleagues [50] stated that the increasing utilization of generic first-line antibiotics was the main cause behind the reduction in cost per claim of antibiotics. Likewise, the study of March-López et al. [68] reported that the total spending on antibiotics showed a significant reduction, after the implementation of a multifaceted antimicrobial stewardship intervention, principally due to reductions in total spending of quinolones and amoxicillin/clavulanic acid.
Within the twenty-nine articles that analyzed antibiotic costs, five of them focused their cost analysis on the route of antibiotic administration, including oral [42,44,56,64,69] or, simultaneously, oral and intravenous [69] antibiotics.

Antibiotic Treatment Success and Quality-Adjusted Life Years
The success of treatment using antibiotics was evaluated in five studies [39,46,49,52,68], with only one study [39] reporting positive and significant results.
While the study by Armstrong and colleagues [46] focused their analysis only on bladder and kidney infections, three other studies analyzed a wider set of pathological conditions, namely urinary tract infections [68], pharyngotonsillitis [68], and acute respiratory infections [39,49,68]. Armstrong et al. [46] demonstrated that the overall treatment success rate of kidney and bladder infections remained essentially unchanged; however, after the treatment guideline implementation, the highest success rate results were achieved using antibiotics such as fluoroquinolone and doxycycline groups, respectively. Another study [39] showed that only 1% of the attending patients with URTI progressed to pneumonia.
Quality-adjusted life-years (QALYs) were measured in two studies [47,54]. Gong et al. [47] reported that the three intervention arms (i.e., suggested alternatives; accountable justifications; and peer comparison) yielded more QALYs at a lower cost when compared with the control group.

Economic Effects of the Educational Interventions
The implementation of educational interventions was associated with savings in antibiotic cost, or lower cost in 30 of the 33 included studies.
All costs mentioned across the included articles were acquired retrospectively, except in two studies [48,53] that obtained costs from the literature and three other studies [47,49,54] that collected this data.
The main direct medical costs assessed in the included studies were expenses associated with healthcare providers (clinician salaries, out-of-hours care costs, telephone calls), hospitalization (e.g., physician consultations, laboratory tests and material charges) and drugs (e.g., treatment, medication, and prescription expenditures). The costs involved in the intervention design and implementation (organizing and administrative intervention costs, seminars/workshops/online training courses costs, educational materials costs, travel costs, outreach visits costs, and staff costs) were also reported in several studies [38,43,[47][48][49][51][52][53][54][56][57][58]60,64,69]. The study of Gillespie and colleagues [58] also highlighted some additional costs related to electronic software needed to support interventions.

Discussion
This study systematically reviewed the economic impact of educational interventions implemented to significantly improve or reduce antibiotic prescription and dispensing among physicians and pharmacists in primary healthcare settings.

Antibiotic Consumption and Prescription
Overall, educational interventions were demonstrated to have a positive impact in both the conscientious consumption of antibiotics and appropriateness of antibiotic prescription and dispensing in primary healthcare provider prescriptions. Evidence from the thirtythree included articles showed that one of the major aims of educational interventions is to guide antibiotic prescription, which predominantly culminate in favorable outcomes, either by decreasing the overall amount of antibiotic prescriptions or by improving prescription quality. This positive impact on the appropriateness of antibiotic prescriptions, in primary healthcare settings, is in accordance with the existing literature [70,71]. The appropriate antibiotic prescription was associated with short-and long-term cost reduction, since it promoted cost reductions associated with less hospitalizations, second-line inpatient antibiotic use and non-antibiotic drug costs (i.e., equipment, workload, etc.), for instance, resulting in improvements in morbidity and mortality [57].
Nevertheless, the inappropriate use and prescription of antibiotics are not always easy to reverse since, in many cases both clinicians' and patients' attitudes may present barriers to the implementation of good practices. The absence of patient awareness on the hazards caused by antibiotic-resistance, lack and/or ineffective communication between prescribers, patients and pharmacists, the pressure by patients towards physicians to prescribe antibiotics, the clinicians' fear of patients worsening, and the expectations of practitioners and patients may represent barriers to provision of the correct prescription and dispensing of antibiotics [2,60,62,70,72]. The study of Wei et al. [62] reported examples of reasons provided by doctors to maintain their antibiotic prescription trends, namely difficulty in differentiating between viral and bacterial infections and to avoid patient complaints.
Generally, antibiotic use is higher in more deprived areas, explained by factors such as the lack of regulations to prevent the over-the-counter sale of antibiotics, the limited availability of essential diagnostic procedures, the inadequate training of healthcare professionals, and the high burden of illness and comorbidities [73]. However, the literature also showed extremely high values of antibiotic prescription and dispensing in high-income countries, namely in primary healthcare facilities where most antibiotics are supplied [72]. In the study of Aksoy and colleagues [45], a reduction or maintenance in antibiotic prescription behavior was described; however, the overall drug prescription raised significantly. In comparative terms, it would be interestingly to verify this trend (i.e., reduction in antibiotic prescription and simultaneous increase in prescription of the overall drugs), in the other included studies; however, this analysis was not possible since none of the other included studies evaluated the same outcomes.
As mentioned before, the antibiotic prescription quality showed to be a crucial factor in the improvement of antibiotic prescription practice. However, prescribers tended to favor long-term prescriptions, therapies combining multiple antibiotics with similar pharmacological characteristics, as well as prescription of broad-spectrum antibiotics, all factors negatively associated with poor prescription quality, and, indirectly, with antibioticresistance [6,74]. Three studies [52,57,62] out of the included papers stressed the importance of reducing the use of broad-spectrum antibiotics for improving quality of antibiotic prescription. Concordantly, two additional studies [44,68] reported a significant decrease in the use of broad-spectrum antibiotics in favor of narrow-spectrum antibiotics, which [43,44,68] showed to be an appropriate alternative [43,44,68], since therapy with narrow-spectrum antibiotics is associated with a lower risk of drug-related adverse effects and a higher health-related quality of life [74,75]. Other concerns associated with antibiotic misuse were prescription without clinical indication and improper consumption associated with the choice of a suitable molecule and dosage of therapy according to patients' characteristics and location of infection [73]. To sum up, despite the efforts to increase appropriate prescription attitudes among primary physicians using educational interventions, several studies demonstrated that additional changes/interventions are still required since prescription values remain above the desirable goals [38,42,43,50,53,60,64,65,67].

Antibiotic Cost and Costs of Antibiotic Prescription
The cost of the prescribed medicines is one of the most important drug utilization indicators allowing the assessment of rational use of drug performance by clinicians [76]. Within the scope of this review, two of the major outcomes analyzed in twenty-nine [2,[38][39][40][41][42][43][44][45][48][49][50][52][53][54][55][56][57][58][59][61][62][63][64][65][66][67][68][69] of the included studies were the changes in antibiotic cost and costs of antibiotic prescription and dispensing. To reduce antibiotic costs and associated costs, one of the strategies currently adopted is the availability of generic first-line antibiotics that resulted in significant cost savings over their branded and broader-spectrum counterparts [62]. This finding is also supported by the study of March-Lopez et al. [68] who reported a significant reduction in the total spending on antibiotics due to reductions in total spending on broad-spectrum antibiotics, and by Walker and colleagues [50] who stated that the reduction in the cost per claim of antibiotics was mainly triggered by the increasing utilization of generic first-line antibiotics. This is in line with previous literature reporting that the use of broad-spectrum antibiotics, which are commonly more expensive, and alteration in clinicians' prescription practices, namely the prescription of second-line antibiotics, are considered some of the main drivers behind extra costs in healthcare [77].
In addition to monetary effects, directly related to antibiotics and their associated costs, inappropriate antibiotic prescriptions and dispensing also have negative effects on productivity [77,78]. Productivity losses are related to the amount of work time lost, due to lack and/or non-productivity as a result of reduced concentration, as well as the expected number of additional hours needed to conclude the regular amount of work [53,77]. The studies of Calls et al. [51] and Dekker et al. [53] were unique in that they considered productivity loss, with Dekker and colleagues reporting that the productivity loss of parents represented the highest costs regarding the intervention group, whereas Calls et al. did not observe any significant difference. These societal benefits of interventions are crucial and should thus not be underestimated. Nevertheless, only three studies [47,48,53] assessed data from a societal perspective.

Economic Effects of the Educational Interventions
Educational interventions showed to be of most relevance, especially for policymakers, regarding enhancement of prescription and dispensing, among physicians and pharmacists in primary healthcare settings, and for promoting cost savings [57].
Despite being stated in the literature that CEA is one of the most reliable tools of process and economic analysis [80], more than half of the included studies [2,[38][39][40][41][42][43][44][45]49,50,52,[55][56][57]59,[64][65][66][67][68][69] implemented a CA. CEA is an extremely useful method in terms of direct comparison of different interventions with identical out-comes [80,81]. When compared to other methods of analysis it presents the benefit of bringing into focus the relative advantages and disadvantages of implementing several interventions from both cost and clinical perspectives, since it allows, simultaneously, the identification of the intervention that confers more benefits to patients, and that provides more cost savings for health systems to help inform policy decisions [81].
As seen within the papers that performed a CEA, there is ample variation in the reported outcomes measured (ICER, WTP, NMB) throughout the included economic evaluation studies, that in conjunction with other methodological discrepancies are responsible for impeding, in a certain mode, the utility of data on the effectiveness and costs in the ongoing practice in primary healthcare facilities. Return on investment, another measure of cost outcome and a method of cost-benefit analysis, allows the measurement of educational intervention costs and the financial recovery of these interventions estimated as a net benefit, i.e., the total benefit minus the total cost, over total cost [82]. Figueiras et al.'s study [57] was the only that reported the return on investment as a measure of cost outcomes. Nevertheless, this type of economic analysis frequently neglects the patients' health, since it is based on temporary recovery. Thus, to surpass this limitation, it is crucial to adopt other outcome measures, that consider the quality of life in patients who experienced the clinical outcomes, enabling comparisons between economic evaluations [82]. QALY is an example of one of those effectiveness-standardized outcome measures that were implemented by Oppong et al. [54] and Gong et al. [47].
Resistance patterns, which may diverge according to the geographical regions, and the financial difficulties of healthcare systems represent huge challenges to the implementation of cost-effectiveness interventions among countries and to apply knowledge translation to current clinical practice [82]. Furthermore, evidence for the cost-effectiveness of implementing educational interventions still requires further exploration since although numerous studies [39,42,52,53,[56][57][58][59]62,66,69] have been performed with the application of a follow-up analysis to check the middle-and long-term effects of the intervention, the evidence is not yet consistent on this topic.
Cost-effectiveness presents an example of an economic health evaluation that is crucial for policymakers and healthcare providers to formulate the best decisions regarding alternative methods of action, since in addition to accessing health benefits, i.e., effectiveness, it also allows the evaluation of relevant information about the costs needed to implement educational interventions [37]. Considering the three possible outcomes, favor intervention, unclear and reject intervention, previously identified in the study of Munn et al. [37], three studies [40,41,61] did not report favoring of interventions. Possible explanations for this finding were the selection of unsuitable control arms (e.g., historical controls), the reduced impact, as less than expected of the implemented interventions, the inadequate practice settings where interventions were implemented and the lack of sensitive prescribing habits regarding cost information.
The findings of this study did not allow us to identify the most cost-effective strategy for improving or reducing antibiotic prescription and dispensing in primary healthcare. However, overall, the implemented educational interventions seemed to result in significantly positive cost and clinical outcomes via the adoption of appropriate antibiotic prescription and dispensing practices in primary healthcare settings.

Strengths and Limitations
This systematic review was associated with some limitations that merit discussion. Firstly, publication bias might have affected our findings. As previously stated, four scientific databases (PubMed, Scopus, Web of Science, EMBASE) were selected for searching potential selection of studies. Thus, grey literature, that includes a wide range of documents difficult to search and retrieve, such as studies with null findings, abstracts, and other unpublished documents were not included. However, the defined inclusion criteria allowed the coverage of a wide range of studies with different settings and designs.
Secondly, our findings were limited by the level of heterogeneity within the included studies, observed in the different study designs, data collection methods, statistical cost analysis methodologies and ways of presenting cost values. To overcome this limitation, we transformed all costs into Purchasing Power Parities (ppp) and this allowed us to equalize the purchasing power of different currencies, by eliminating the differences in price levels across countries. The high heterogeneity also hindered us from drawing closer comparisons between costs and performing a meta-analysis. Additionally, some common issues observed across the analyzed papers were the lack of randomization of the study population and/or the absence of a control group, resulting in a reduced ability to determine whether the results were due to the educational intervention or external factors. The imbalance between participants, i.e., more primary care physicians than pharmacists, may have also limited generalization of results. Another limitation is associated with the timeline of some of these studies, since twenty-two of the revised papers, did not perform or report follow-up results, so it is not possible to conclude whether the interventions were effective over middle-and long-term periods. Finally, most of the included studies did not report sociodemographic characteristics of the physicians/pharmacists and the few [2,40,41,44,52,[55][56][57]59,[61][62][63]67] that reported did not explore correlations with the trends of antibiotic prescription, dispensing and cost.
Despite these limitations, we may highlight the extensive, rigorous, and systematic search across the four distinct databases (PubMed, Scopus, Web of Science, EMBASE) following well-established and updated guidelines for conducting systematic reviews [33,83], which might be considered the main strength of this study.

Conclusions
Our results suggest that educational interventions were associated with improvements in the overall prescription rate, dispensing, and consumption of antibiotics as well as significant reduction in antibiotic costs. These results support the need for public health actions to qualify primary healthcare providers, principally in low-and middle-income countries, through the implementation of cost-effective educational interventions to reduce the prescription and dispensing of antibiotics in primary healthcare settings and a consequent decrease in the associated healthcare costs.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antibiotics11091186/s1, Table S1: PRISMA checklist; Table S2: Quality assessment of the studies included using the Joanna Briggs Institute critical appraisal checklist for economic evaluations; Table S3: Characteristics of studies selected for full data extraction (n = 33). Text S1: Full search expression for each database.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.