Study selection process
The database search identified 253 articles, of which 58 were duplicates. An additional 16 articles were identified through manual searches. The titles and abstracts of 211 articles were screened, with 81 articles excluded due to not meeting the eligibility criteria. Full-text analysis was performed for 130 studies, of which 79 were excluded to yield a final total of 51 articles for inclusion in this review. Reasons of exclusion include irrelevancy to AMS, AMR or antimicrobials in general as well as studies conducted in hospital settings. Figure 1 illustrates the search and selection process.
Correlation of antimicrobial resistance with antibiotic usage and/or volume of antibiotic use
The link between antimicrobial use and AMR continues from the hospital into the PHC setting, as evidenced by various studies across the globe. An ecologic analysis of data from 20 countries demonstrated that antibiotic use exerted selection pressure that has directly increased the resistance of multiple known pathogenic types of Streptococcus sp. [22]. Similarly, a chart review of Canadian Indigenous communities found methicillin-resistant Staphylococcus aureus (MRSA) to be the organism responsible for over 40% of skin and soft tissue infections as well as a high prevalence of antibiotic use in these communities, although direct correlations between the two phenomena could not be made [23]. Further evidence can be found in a systematic review by Hansen et al. [24], where immediately following macrolide exposure, resistant bacteria were identified in participants with greater frequency. Although later trends were subsequently met with inconsistency, resistance is regarded as a major adverse effect of macrolide treatment [24]. This is substantiated by a study by Doan et al. [25] involving children below 5 years, in which mass azithromycin administration resulted in up to 7.5 times (95% confidence interval [CI] 3.8–23.1) higher genetic determinants of resistance to macrolides at 48 months. Costelloe et al. [26] demonstrated that antibiotic resistance is highest in the month following treatment, with effects lasting up to a year. This systematic review presented pooled odds ratio (OR) for resistance among urinary tract bacteria of 2.5 (95% CI 2.1–2.9) and 1.33 (95% CI 1.2–1.5), and among respiratory tract bacteria of 2.4 (95% CI 1.4–3.9) and 2.4 (95% CI 1.3–4.5) within 2 and 12 months of antibiotic treatment, respectively [26].
Prescribing practices in PHC play a major influencing role in AMR rates. Prescription, dispensation and use of penicillin, especially classes susceptible to beta-lactamase, were associated with increased risks of nasal carriage of resistant S. aureus, with ORs of 1.18 (95% CI 1.04–1.35) and 1.09 (95% CI 1.00–1.18) according to two large cross-sectional studies [27, 28]. Azithromycin use was likewise correlated with nasal carriage of S. pneumoniae and S. aureus strains which are resistant to macrolides [29, 30]. Hare et al. also discovered a dose–response relationship between azithromycin use and carriage of macrolide-resistant S. pneumoniae and S. aureus strains, which were found in 21% (95% CI 14–28) and 1% (95% CI 0–5) respectively when azithromycin was not used, 28% (95% CI 17–38) and 14% (95% CI 6–22) when azithromycin was used infrequently, and 38% (95% CI 26–49) and 23% (95% CI 14–33) when azithromycin was used frequently (nptrend P = 0.002 and P < 0.001 respectively) [29]. The systematic review by Costelloe et al. [26] accentuated this relationship by drawing associations between multiple or longer durations of antibiotic courses with higher AMR rates, especially for amoxicillin and trimethoprim. However, a randomised controlled trial (RCT) that included 520 children with acute otitis media found that a shorter amoxicillin–clavulanic acid antibiotic regimen of 5 days did not affect AMR rates (P = 0.58) and instead, produced less favourable health outcomes with greater likelihood of clinical failure and worse mean symptom scores (P = 0.001) compared to the standard 10-day regimen [31]. Nevertheless, this study has limited generalisability due to the age group and disease type [31].
Risks for AMR development are not limited to the specific antimicrobial agent used to treat a particular infection, as suggested in a recent revision of a Cochrane systematic review on the mass treatment of trachoma using antibiotics [32]. Here, evidence shows with high certainty that S. pneumoniae, S. aureus and Escherichia coli exhibited approximately fivefold increase in resistance at 12 months not only to agents used to treat trachoma, namely azithromycin and tetracycline, but also to clindamycin despite having no role in trachoma management [32]. Likewise, Doan et al. [25] found that beta-lactam resistance determinants were 2.1 times (95% CI 1.2–4.0) higher after mass azithromycin distribution. Cross-resistance is a genuine cause for concern, as it expedites the development of multidrug-resistant pathogens, thereby rendering progressively more lines of treatment ineffective.
Overall, the emergence of AMR in the PHC setting has been associated with high-volume antibiotic prescribing and use. Therefore, sustained, large-scale efforts are required to regulate such practices.
Appropriateness of antimicrobial prescribing in rural and remote primary health care
Inappropriate prescribing is generally defined as the prescription of antimicrobial agents that do not adhere to authorised guidelines in terms of type of antimicrobial chosen, dose and/or duration, or are deemed unnecessary [33]. In rural and remote areas, inappropriate prescribing was reported to be highly prevalent, with 18–88% of antimicrobial prescriptions deemed inappropriate depending on the country [34,35,36,37,38]. Antibiotics were the antimicrobial agents most frequently incorrectly prescribed [35,36,37,38,39].
Clinicians in rural and remote PHC services tended to be high-volume prescribers of antimicrobial agents [40,41,42,43]. Antibiotics were prescribed more frequently in rural regions compared to urban areas, 85% versus 68–80% respectively [40,41,42]. Primary health care services in rural and remote areas face additional challenges in the provision of health care, including limited resource allocations and a different patient population compared to urban clinics, which may contribute to differences in prescribing.
An Australian study showed no significant difference in prescribing behaviour and adherence to guidelines between rural- and urban-based physicians [44]. While this study portrayed a positive outlook in Australia, the results cannot be generalised as the focus was on early-career general practitioners and urinary tract infections only, and rural practices constituted only 16% of the study locations [40]. An audit of antimicrobial use in 15 remote PHC clinics across three states in northern Australia found that appropriateness was high compared to general practices in urban settings [45]. Approximately 91%, 82% and 65% of antimicrobial use in Western Australia, Northern Territory and Queensland adhered to clinical guidelines, endorsed by experts, or constituted agents with the narrowest targets. This appreciably high level of appropriateness was demonstrated by 86% of nurses and 73% of doctors in this region [45]. The high rates of appropriateness in this study may be a result of government endorsed clinical guidelines that the nurses are legally bound to comply with in order to prescribe.
Measures of prescribing appropriateness are highly dependent on specific diseases and clinical guidelines used. For example, a study on childhood diarrhoea by Rhee et al. [35] designated the presence of dysentery as the standard for antibiotic prescription, with over-prescription defined as antibiotics given for non-dysentery cases and under-prescription as absence of antibiotics for children with dysentery. In some instances, prescribing appropriateness were inferred directly from antimicrobial use, particularly antibiotics prescriptions for infections not commonly suspected to be of bacterial origin such as acute upper respiratory tract infections [46]. Often, an antimicrobial prescription is deemed appropriate if it is in accordance with local or international therapeutic guidelines designed for that disease entity [35, 44]. Nevertheless, an Australian study assessed both prescribing appropriateness and guideline adherence, which accounted for prescriptions deviating from guidelines for a justified clinical reason [45].
Risk factors associated with inappropriate use/prescribing of antibiotics
Rurality is often found to be an independent risk factor for inappropriate prescribing [30, 39,40,41, 47]. Rural and remote areas have widely distributed populations who generally experience poorer health and who are serviced by fewer doctors per population [39]. This invariably translates to increased physician fatigue due to high workloads, which is associated with excessive and error-prone prescribing [46, 48]. Despite having AMR awareness, rural clinicians were less likely to adopt steps to reduce antibiotic prescriptions, as a survey by Salm et al. [49] suggested.
Aside from the remoteness of practice mentioned above, several other factors contribute to the overuse and misuse of antimicrobials, which ultimately play a major role in the emergence of AMR. Chiefly, physician knowledge and behaviour exert a strong influence on antimicrobial use in the population. Numerous studies have found insufficient awareness of the importance of judicious prescribing and lack of knowledge on local resistance patterns to be major drivers of AMR [38, 47, 50,51,52,53]. Most frequently, physicians prescribe antibiotics for viral infections, indicating a substantial need for clinician education [39, 50]. This problem is not unique among clinicians in rural and remote settings, as those in urban primary care practices share similar knowledge, attitudes and practices [50, 51, 53].
Although many rural-based physicians commonly recognised the negative consequences of over-prescription, some nevertheless continue to prescribe antimicrobials liberally to patients [38, 51, 54]. This can be due to both internal and external reasons. Doctors cited apprehension of medical complications from undertreatment, perceived patient expectations and intentions to preserve good relations, and the notion that selected prescribed antibiotics would not contribute to resistance as grounds for discrepancy in personal knowledge and clinical practice [38, 51, 52, 54,55,56,57]. Concurrently, external pressures including financial considerations in line with incentivising drug sales and patient retention may compel physicians into prescribing inappropriately [38, 52, 54]. System factors such as insufficient accessibility to follow-up as well as the fear of litigation or related medico-legal issues were also strong drivers of excessive antibiotic prescribing [51, 54].
Furthermore, attitudes of physicians may influence the development of AMR in rural and remote settings. It is possible that some PHC physicians are complacent, prefer to prescribe by habit, or are reluctant to keep up with new recommendations in prescribing [47, 53]. Other health care professionals such as pharmacists and informal health providers may contribute to AMR [50, 55] particularly in countries and regions where there is relatively unrestricted access to antimicrobials.
Patient factors are also identified as playing a role in the development of AMR. In particular, non-compliance to prescribed antimicrobial treatments, especially poor adherence to the full course of antibiotics, facilitates the development of AMR in the community [50, 58]. Treatment decisions are influenced by patient demands, and often driven by unfounded beliefs of obtaining the ‘fastest cure’ [55]. In some countries and regions where antimicrobial dispensing is not strictly regulated, members of the public may self-prescribe antimicrobial agents for ailments [50, 58,59,60]. The inappropriate use of antibiotics and demands on clinicians to provide antibiotics may result from a poor understanding of the implications of their misuse, limited access to appropriately trained doctors, scarce resources and limited health literacy [58, 61].
Patient attributes associated with inappropriate prescribing and higher AMR rates include patients of a younger age group, female gender, with other co-morbidities and higher socioeconomic status [28, 39]. While further studies are required to elucidate the mechanisms and processes behind these correlations, it is postulated that they could be explained by greater anxiety levels in these populations. For example, parents desire a rapid relief of symptoms for their children, and patients of a higher socioeconomic strata can have greater demands [40, 41]. Patient expectations in relation to antibiotic treatment was identified by physicians, as placing undue pressure on the prescription of antimicrobials to satisfy patients [57].
Antimicrobial stewardship strategies in remote and rural primary health care settings and their effectiveness
A total of 16 interventions are included in this section under four distinct themes according to their intervention focus, with further detail outlined in Table 4. Although the measured outcomes exhibited a high degree of heterogeneity among the studies, a few main groupings can be elucidated from the pooled data, namely rates and appropriateness of antimicrobial prescribing. Nine studies had antimicrobial prescription rates as their primary outcome measure [62,63,64,65,66,67,68,69,70]. Of these studies, seven had significantly lowered prescription rates post intervention [62,63,64,65,66, 68,69,70], in which results ranged from an 11% decrease when the United Kingdom incentivised reduced prescribing of specific antibiotics [63] to a 49% drop when a multi-dimensional clinician-focussed intervention was implemented in China [66]. A total of four studies reported on appropriateness of prescribing [64, 67, 71, 72], and all but one—the MIKSTRA study [67]—demonstrated meaningful improvements after intervention.
Health care provider and patient education
Three publications focused on the provision of education to health care providers, where clinicians and supporting staff received training on antimicrobial prescribing, including implementation of antimicrobial stewardship guidelines and strategies such as delayed prescriptions [65, 67, 73]. Although immediate antibiotic prescription reduced symptom severity and duration of acute otitis media in children by about a day, these benefits were marginal as symptoms were already improving after 24 hours [73]. This study also found that doctors’ perception of patient expectations were often overestimated—most parents were satisfied with delayed antibiotic prescribing for their child’s illness [73]. A large RCT in rural China implemented a comprehensive clinician-oriented intervention which consisted of clinical guideline implementation/enforcement, monthly reviews on physicians’ prescribing rates, training on communication skills and brief education for caregivers [65]. Besides proving to be highly cost-effective [74], this multifaceted education strategy successfully decreased antibiotic prescribing rates for childhood upper respiratory tract infections by 29% (P = 0.0002) at 6 months [65] and up to 36% (P < 0.0001) at 18 months by virtue of improved physician knowledge and communication with patients [66].
On the other hand, modest improvements to antibiotic prescribing were observed in the Finnish MIKSTRA study which promoted the use of first-line treatment for the treatment of acute maxillary sinusitis according to national guidelines recommendations [67]. An education programme using two training methods facilitated by local general practitioners were employed, which were: problem-based learning (PBL) based on group work and an academic detailing (AD) process involving various information sources, feedback and external visits [67]. The RCT only managed to yield minor positive effects on antibiotic prescribing with use of the first-line drug amoxicillin and appropriate duration of antibiotic courses, with the former increasing from 33 to 45% and 39 to 48% while the latter increased from 32 to 47% and 34 to 40% in PBL and AD centres, respectively. These shortcomings were likely due to various resource issues throughout the trial, which included physician shortages in rural PHCs and improper adherence to guidelines brought about by education methods which did not adequately address the local context [67].
Patient education was investigated by a quasi-experiment conducted among Lao villagers by Haenssgen et al. [75] in which a one-off education session involving interactive health-based activities was implemented with mixed results, especially on behavioural change towards antibiotic use. There were noticeable improvements in awareness and understanding of antibiotic resistance, but not attitudes. A combined approach involving education of both the physician and the patient using presentations and printed materials on relevant health information was explored by three studies, all of which showed significant reductions in undesirable outcomes [62, 70, 71]. Chiswell et al. and Belongia et al. decreased antibiotic prescription rates in the rural community, with the former even lowering immediate antibiotic prescriptions from 31.1 to 13.5% (P < 0.05) [62, 70]. Cummings et al. reduced the proportion of antibiotic-inappropriate prescriptions by 14.9% (95% CI − 20.30% to − 9.05%; P < 0.0001) via distribution of education materials as well as clinician peer comparisons of prescribing behaviour [71].
Physician support systems
The effectiveness of a clinical-decision support system (CDSS) was researched in two RCTs in rural areas of the United States: Samore et al. [64] and Gonzales et al. [69]. Different implementations of CDSS—based on paper and electronic tools—were employed and compared, and an extension study was performed in the former focussing on a CDSS system based on personal digital assistants [72]. These CDSS-based interventions were effective at mitigating prescriptions for conditions not requiring antibiotics by 32% in one study [60] and between 11.7 and 13.3% in the other [69]. CDSS also resulted in an increase of 2.7% (P = 0.016) in provider adherence to guideline recommendations [72].
A cohort study by Madaras-Kelly et al. [76] described an intervention carried out by rural community pharmacists in which patients were interviewed and the prescribing clinicians contacted before dispensing of a broad-spectrum antibiotic. This intervention did not attain favourable responses from the general public [76]; whereby only 7% (n = 4) of participants consented to the pharmacist contacting the prescribing physician to discuss alternative therapies, leading to challenges in performing the intervention to its full extent [76].
Surveillance
An important component of AMS programmes is comprehensive surveillance of antimicrobial prescribing and resistance which provides data to focus stewardship efforts. Three studies reported on different surveillance tools, which were: a modified General Practice version of the National Antimicrobial Prescribing Survey (GP NAPS) to assess the clinical appropriateness of antimicrobial use according to local Australian guidelines [45]; a mathematical modelling of molecular testing for gonorrhoea resistance surveillance [77]; and the IQVIA Xponent antibiotic database to identify high prescribing physicians [78]. By nature, these studies did not actively involve participants, but rather served to collate relevant data through retrospective data analysis or simulation. All three tools were validated and deemed fit for integration into clinical practice to assist AMS efforts in the future [45, 77, 78].
National policies
Achieving AMS on a nationwide scale necessitates upstream policies and regulations, often in regard to financial aspects of prescribing. Two studies by Hammond et al. and Yin et al. provided an overview of the effects of national health policies in the United Kingdom and China, respectively [63, 68]. In 2014/15, reduced prescribing of certain antibiotics in PHC was incentivised in the UK [63], while China implemented a zero mark-up policy and a national strategy for antibiotic use [68]. In the UK, although overall antibiotic dispensation decreased 11%, the national AMS policy produced mixed results with reduced resistance of community-acquired urinary E. coli to amoxicillin, ciprofloxacin and trimethoprim, but increased resistance to cefalexin and co-amoxiclav [63]. China’s strategy lowered the overall antibiotic consumption by 22% and 8% for parenteral and oral antibiotics, respectively [68]. The two countries’ administrative decisions affected both urban and rural PHC centres, with differences in results between these settings delineated in the Chinese study where antibiotic consumption rates in rural PHCs fell from its peak after in 2014 whereas urban PHCs saw a steady increase in antibiotic use even after policy implementation [63, 68].