Community mobilisation approaches to preventing adolescent multiple risk behaviour: a realist review

Abstract

Background

Adolescent multiple risk behaviour (MRB) is a global health issue. Most interventions have focused on the proximal causes of adolescent MRB such as peer or family influence, with systematic reviews reporting mixed evidence of effectiveness. There is increasing recognition that community mobilisation approaches could be beneficial for adolescent health. There are gaps in the current literature, theory and implementation that would benefit from a realist approach. We use a theory-driven evidence synthesis to assess how and why community mobilisation interventions work/do not work to prevent adolescent MRB and in what contexts.

Methods

This realist review used a six-stage iterative process, guided by the RAMESES framework. We systematically searched PubMed, MEDLINE, PsycINFO, Web of Science, CINAHL and Sociological Abstracts, from their inception to 2021. Studies were screened for relevance to the programme theory, assessed for rigour and included based on a priori criteria. Two independent reviewers selected, screened and extracted data from included studies. A realist logic of analysis was used to develop context-mechanism-outcome configurations that contributed to our programme theory.

Findings

We reviewed 35 documents describing 22 separate community mobilisation intervention studies. Most studies (n = 17) had a quality assessment score of three or four (out of four). We analysed the studies in relation to three middle range theories. To uphold our theory that these interventions work by creating a social environment where adolescents are less likely to engage in MRB, interventions should: (1) embed a framework of guiding principles throughout the community, (2) establish community readiness with population data and (3) ensure a diverse coalition with the support of intervention champions. Mechanisms such as empowerment through coalition ownership over the delivery of the intervention, cohesion across the community and motivation to work collaboratively to improve adolescent health are triggered to achieve social environment shifts. However, certain contexts (e.g. limited funding) restrict intervention success as these mechanisms are not fired.

Conclusions

For community mobilisation interventions to reduce adolescent MRB, the coalitions within them must seek to alter the social environment in which these behaviours occur. Mechanisms including empowerment, cohesion and motivation lead to this shift, but only under certain contexts.

Systematic review registration

PROSPERO CRD42020205342

Background

Health risk behaviours including hazardous alcohol consumption, tobacco smoking, antisocial behaviour, risky sexual behaviour and physical inactivity are global health issues and are often initiated and become habitual in adolescence [1, 2]. Adolescents who engage in one risk behaviour are likely to engage in others [3, 4], leading to an increased public health focus on multiple risk behaviour (MRB), which is defined by the co-occurrence of two or more risk behaviours directly or indirectly related to health [5, 6]. MRB can result in many deleterious health and social effects later in life, such as unemployment [7], poor educational attainment [8], obesity and mental health issues [9], cancers and premature mortality [8, 10]. This has, therefore, led to public health strategies that address multiple as opposed to single behaviours [11].

Most interventions addressing adolescent MRB have focused on the proximal causes such as peer or family influence, rather than targeting the wider environmental, social or structural context [12]. For instance, two Cochrane systematic reviews have assessed the impact of individual, family and school-level interventions on adolescent multiple risk behaviour [11, 13]. One of those reviews found mixed evidence, concluding that school-based universal interventions are potentially effective in ‘preventing engagement in tobacco use, alcohol use, illicit drug use and antisocial behaviour, and in improving physical activity among young people, but not in preventing other risk behaviours’ [13]. The authors highlighted that there was no strong evidence of benefit for family‐level or individual‐level interventions across the risk behaviour outcomes investigated [13]. The interventions included in this review were predominantly educational programmes. The effectiveness and equity of these ‘downstream’ interventions have been questioned [14] because health risk behaviours rarely have a single cause and occur in complex socio-cultural contexts [15]. As such, there is increasing recognition that structural changes that extend beyond individually focused educational programmes could be beneficial for adolescent health [15, 16].

Community mobilisation interventions

Recognition that decisions about health risk behaviours are made within a broad social context has led to the development and implementation of community-engagement interventions [17]. There is a range of community-engagement public health intervention types, which vary in the extent to which they emphasise community involvement in determining and delivering the programmes [17]. ‘Community mobilisation’ interventions are one such intervention type that work to engage community members to ‘take action towards achieving a common goal’ [18]. They have gained traction as a strategy for addressing complex and multifaceted problems [19]. Community mobilisation is a collaborative public health effort that is defined by the inclusion of a community coalition made up of diverse stakeholders (such as schools, businesses, residents, youth groups, emergency services and religious leaders) [12]. These stakeholders critically analyse the root causes of local problems, identify an array of potential solutions, develop multi-sector partnerships and implement multi-component strategies for creating local change and more health-promoting environments [15].

Community mobilisation efforts explicitly seek to affect community-level influences through changes in policies, practices, organisations and other features of the social or physical environment that may impact the health outcome or behaviour [20], signifying a shift away from individual behaviour change to a focus on the social determinants of health [21]. The premise being that if adolescents have a stronger affiliation with their community, and there are fewer opportunities to engage in risky behaviours and more opportunities for health promoting behaviours, then adolescent MRB within a given community will decrease [22]. These approaches may still include components which address individual behaviours (e.g. health promotion programmes within schools), but they seek to combine these with other structural factors (e.g. policy changes) as well as strong cross-community collaboration as part of a package of measures that are chosen and monitored by community stakeholders. The aim of the community mobilisation intervention is to be uniquely relevant to each community to see greater benefit for adolescent health, as the involved stakeholders understand what will work best to create changes in their community [23]. Figure 1 is a visual representation of one possible way community mobilisation interventions may be implemented and evaluated.

Fig. 1

Community mobilisation intervention example

Legend: A visual representation of the potential stages of implementing and evaluating a community mobilisation intervention. This is an example based on broad commonalities across interventions and is not based on one particular intervention programme

There is systematic review evidence suggesting that higher levels of community involvement in a public health intervention are linked to more beneficial effects and positive trends across a range of outcomes [24]. There is also some evidence to support the role of community mobilisation efforts in preventing health risk behaviours. For instance, such interventions have resulted in reductions in high-risk alcohol consumption and alcohol-related injuries [25]; alcohol-impaired driving [26]; uptake of smoking in young people [17] and youth violence [27]. Researchers have highlighted that with adequate resources and training, support from within the community and the adoption of evidence-based strategies, community mobilisation approaches show promise as an effective vehicle for addressing adolescent multiple risk behaviour [20]. Further, community-mobilisation efforts are also thought to be well suited to achieving health equity [28], due to ‘shared decision making’ [29] and the incorporation of ‘upstream’ or structural elements [30]. However, this has yet to be explored in relation to adolescent multiple risk behaviour interventions.

There are also significant challenges in implementing and evaluating such approaches, which is unsurprising given the dynamic set of social interactions and relational complexity one might expect in community-centred interventions [31]. These implementation challenges include a lack of community interest and long-term engagement, design inadequacies, inflated expectations, and weakness in planning and implementing the interventions [20, 32, 33]. Tensions and different expectations between scientists and community members as well as the practical difficulty in managing multiple components and stakeholder interests have also been cited as issues [12].

Evaluation is equally challenging [34], which is reflected in the lack of empirical evaluations of community mobilisation approaches compared to less complex (often education delivery) interventions focused at the individual level [15]. There is uncertainty around how long it might take to see an impact on behaviours, although it is expected to be a lengthy process. If changes in behaviours are found, the chain linking any effects on health risk behaviours to the coalition efforts is so long and complex that causal attributions become challenging [15]. These barriers have meant community mobilisation interventions have been missing from systematic reviews such as the aforementioned adolescent MRB review [13], as they are often evaluated through alternative methods to randomised controlled trials (RCTs) such as quasi-experimental studies. There has been one key review of the Communities that Care (CTC) community mobilisation process, including a narrative synthesis of its implementation globally [35]. Although not meeting the inclusion criteria for our document selection, this report proved to be useful during our broader realist review in mapping out all the iterations of CTC and providing reflections on why it was successful in some countries but not others.

Rationale for realist review

There is a strong rationale for an alternative review approach that moves beyond effectiveness measures and speaks to the complexities and challenges surrounding the delivery and evaluation of community mobilisation efforts [36]. The realist approach seeks to make sense of ‘what works, for whom, under what circumstances.’ It does this by developing and testing hypotheses about how contexts and mechanisms interact to produce outcomes [37]. Within realist approaches, ‘context’ refers to the physical, social, economic and cultural space within which an intervention is embedded [38]. ‘Mechanisms’ are defined as ‘the real but invisible forces that make the programmes work (or not). Mechanisms are not particular components of the programme but rather, they are the 'reactions the participants have to the resources the programme offers’ ([39], page 244, [40]). Outcomes can be intended health outcomes of which the intervention is targeted, or unintended consequences of the intervention. Interventions under the realist approach are viewed as producing outcomes not directly, but as a consequence of individuals engaging with these resources in a certain context to bring about change [37]. Context can influence the extent to which these resources trigger mechanisms (e.g. because of the structure, culture or social norms) and whether that is enough to produce observable outcomes [37]. In this review, we are concerned with examining configurations of contexts and mechanisms to determine how they bring about outcomes to articulate how community mobilisation interventions work for adolescent health. Therefore, a realist review was chosen as the most appropriate methodological approach.

There is a pre-specified protocol for our review which has been registered (PROSPERO registration number: CRD42020205342) and published [36]. This includes detail on the research aim, realist review approach and methodological stages, search strategy, inclusion criteria and plans for dissemination. We provide an outline of the methods here and note any modifications to the original protocol. The methodological steps are described below, using the RAMESES (Realist and Meta-narrative Evidence Synthesis: Evolving Standards) [40] to guide our approach.

Review aim

Our aim is to use a theory-driven evidence synthesis to assess how and why community mobilisation interventions work/do not work to prevent adolescent multiple risk behaviour and in what contexts. We are also interested in ‘who’ these interventions work for, so we can understand the impact of these types of interventions upon existing health inequalities. Although the focus of the review is adolescent multiple risk behaviour, we aim for our review to develop transferable learning about community mobilisation approaches more broadly in public health research and adolescent health interventions. The realist review was guided by the following sub-questions:

• (1) What are the mechanisms through which community mobilisation interventions produce outcomes?

• (2) What are the key contextual influences that determine whether the mechanisms produce outcomes?

Methods

Study design

This review was structured around the five review stages outlined by Pawson et al. [41] and has been informed by other realist review protocols in the field [42, 43]. Figure 2 is a diagram of the review process adapted from Power et al. [42].

Fig. 2

Stages of realist review

Legend: Summary of stages of realist review adapted from Power et al. [ 42 ]. This depicts the steps for developing the initial programme theory, searching for evidence and synthesising the data with the input of key stakeholders (researchers and intervention delivery practitioners within the field of community mobilisation and adolescent health). Retroduction refers to inferences made through interpreting the data about the underlying causal mechanisms

Stages of realist review

Identifying working programme theories

We began our realist review by conducting scoping searches to identify potentially relevant documents that may help to explain how community mobilisation interventions work to address adolescent multiple risk behaviour. The search included academic databases (MEDLINE, PubMed, Web of Science), UK health websites and grey literature databases (OpenGrey, the King’s Fund, The Health Foundation), as well as Google Scholar. Broad search terms were used at this scoping stage (e.g. “community mobilisation”, “community coalition”, “youth”, “adolescence”, “health risk behaviour”, “substance use”, “antisocial behaviour”) and back and forth citation tracking was used until we developed a core set of documents to help build the initial programme theory framework [44]. This initial search was not designed to be exhaustive: this stage in the theory development is expected to be a ‘rough starting point’ that will be refined throughout the realist review process [45].

We reviewed the initial documents to develop a preliminary programme theory model (Additional file 1) and a long list of working theories that related to different elements in the model, with a description of the theories and how they explained how the intervention worked. For example, we had a theory specifically about the context of a geographical location that had social problems resulting from adolescent multiple risk behaviour and how that may trigger certain mechanisms within the intervention. We used this model and a preliminary theories document to engage in discussion with five key stakeholders (who remain anonymous) identified through our citation searching. Each stakeholder had experience or expertise in delivering community mobilisation approaches to adolescent multiple risk behaviour, with some being co-authors to the final sample of interventions and/or the preliminary set of documents. Several potential stakeholders were contacted through their organisational email address. Those that responded engaged in an online videocall with the lead reviewer to discuss the realist review. The stakeholders commented on the programme theory so far, indicated other studies that may be of use and gave insight into the experience (potential contexts and mechanisms) underlying the published work that then elicited alterations to the programme theory model. Changes were made to the programme theory model and the long list of different theories as a result of these conversations and during the data synthesis process. Additional detail on how the programme theory model evolved is in Additional file 1.

Search strategy

At this stage, we conducted more formal searches to identify literature and evidence. We searched the following databases: PubMed; MEDLINE; PsycINFO; Web of Science; CINAHL; and Sociological Abstracts, from their inception. We also searched grey literature on OpenGrey and external expert organisations and charity websites. We searched ProQuest for unpublished theses and dissertations. We used Google Scholar for citation searching and to check the reference lists of relevant papers.

We developed search terms and syntax from the initial background search in Stage 1, discussions within the research team and previous experience with search strategies used in two multiple risk behaviour systematic reviews [11, 13]. Search terms will include MeSH terms and free text related to “community mobilisation”, “adolescence” and a range of multiple health risk behaviours. No date restrictions will be used and only studies in the English language were assessed for eligibility. The search strategy can be found in Additional file 2.

While formalised and systematic, the sampling approach in realist reviews remains purposive to answer specific questions and develop theories [41]. Therefore, the process was iterative [41], with back-and-forth citation tracking remaining a key part of the iterative search strategy [46].

Study selection

Relevance and rigour assessment

Two independent reviewers (LT and CK) assessed each document to determine its relevance to our review, extracted information on a predetermined form and appraised the quality of each document. The form containing the relevant verification, data extraction and rigour appraisal can be found in Additional file 3. To be deemed relevant to our research aim, the study had to:

• Be written in the English language

• Describe a community mobilisation intervention (i.e. with a community coalition element)

• Describe an intervention targeting two or more adolescent health risk behaviours from a predetermined list, including substance use (tobacco, alcohol and drug use), violence and risky sexual behaviour. The behaviours did not necessarily have to be measured as outcomes for the document to be relevant, but needed to be the focus of the intervention to prevent them in young people aged 10–19 years. Adults may be included but the focus should be on adolescents.

• Include measurement of an outcome or result (qualitatively or quantitatively). Our outcome of primary interest, and what we based our programme theory on, is adolescent health risk behaviour outcomes. Other health, education or social outcomes or risk and protective factors (e.g. school connectedness, time spent with parents), community-level outcomes (e.g. changes in community crime levels) or process and implementation outcomes (e.g. the degree to which the intervention was successfully delivered or level of coalition functioning) are also of interest as they are intended to lead to reductions in population level adolescent MRB.

We included these types of outcomes because community mobilisation efforts often take many years to realise behavioural outcomes. There is a wealth of evidence on implementation outcomes relating to these interventions, which we deemed potentially useful in answering our question of how the interventions work in what circumstances. Some documents were not classified as relevant due to not meeting the ‘measuring outcomes’ criterion. They either described intervention challenges or outlined the implementation process, usually of interventions for which we did obtain documents presenting outcomes. We did not extract contexts, mechanisms and outcomes (CMOs) or data from these excluded documents, but some did contribute to theory development by adding contextual and mechanistic insight in a similar way to stakeholder input, which is acceptable within a realist review.

To appraise the documents’ quality (i.e. the rigour assessment), we used an adapted version of the Mixed Methods Appraisal Tool (MMAT) and the Critical Appraisal Skills Programme (CASP). We created a quality assessment form guided by Minian et al.’s [47] form adapted for their realist review on multiple health behaviour interventions for smoking cessation outcomes. Each study was assessed by two reviewers, with sets of questions for different study designs to derive a quality score: RCTs (eight questions), non-randomised studies (10 questions), qualitative studies (10 questions) and mixed method studies (six questions), all scored through a nominal scale (Yes/No/Not clear). An overall quality score was calculated for each study by taking the number of ‘Yes’ answers and dividing it by the number of questions. Scores were derived using the following descriptors: 0–25% (*), 26–50% (**), 51–75% (***) and 76%+ (****).

Data extraction and synthesis

Data were extracted using a predesigned form collecting information about the study country and geographical area, the intervention activities, the coalition composition and the participants in the study. Data analysis was informed by the constructivist epistemological position, which is a paradigm aligned with the realist approach [39]. We structured our analysis around the recognition that knowledge is constructed by an individual’s perceptions. On a practical level, data analysis therefore began with reviewers coding information in the documents, coding any text that could relate to possible contexts, mechanisms and outcomes (CMOs) and including these in the data extraction form—being open to as many possibilities as we could think of. We met to discuss our long list of codes and searched for any similarities and conflicts in our initial coding.

Guided by realist review training, we then developed a diagrammatic tool using the initial codes, which we then systematically applied to each document independently. This process helped us build CMO configurations as it encouraged thinking about the causal pathways that resulted in the outcomes. In realist research, subjectively constructed knowledge is built on abductive and retroductive logic. Abduction is the creative inference required to imagine underlying causal mechanisms and retroduction is the theorising needed to develop a way of ‘testing’ whether these mechanisms exist [48]. Abduction has been described as a form of reasoning which examines evidence and makes inferences based on ‘educated guess work’ and ‘informed hunches’ about the causal factors linked to that evidence ([38], p.135). Therefore, displaying the possible contexts, mechanisms and outcomes of the interventions in a visual way allowed us to manoeuvre the different components and ‘test’ potential causal pathways. Each reviewer created their own set of CMO configuration diagrams, which were then used in multiple discussions until we agreed on the most salient set of CMOs for each document. An example of the diagram for one of the documents appears in.

Data synthesis involved assessing the documents and diagrams and compiling CMO configurations to map patterns of pathways of intervention functioning that help explicate programme theory [49]. Synthesis was driven by realist analytical approaches. For instance, coding involved deductive reasoning whereby we revisited our preliminary programme theory to look for alignment and conflicts within the data. We sought to be inductive in approach through grounding our reasoning in the data and being open to new and potentially challenging causal pathways. Finally, critical for realist synthesis, we practised retroduction and abduction in our synthesis, by selecting different combinations of interventions and meeting to discuss the similar or different causal pathways within each group. The CMO diagrams were used simultaneously with textual coding to explore commonalities and contrasts across the interventions and interpret potential causal mechanisms. For example, if particular mechanisms appeared to be meaningful within one intervention, this was applied to other interventions and the documents were revisited with this particular mechanism in mind. This interpretive and iterative process resulted in a set of middle range theories that contributed to the wider understanding of how the interventions work (or do not work), with evidence presented from the relevant studies.

Findings

Document characteristics

A total of 69 documents describing 22 different interventions were obtained for this review. Figure 3 shows the PRISMA flow diagram of how we reached that sample size. Where single interventions were described by multiple documents, the study team selected a core set of documents for data extraction, which resulted in a final focused sample of 35 documents. Decisions about these documents were made based on relevance to our research question and whether they were felt to add new insights about the intervention. Many documents repeated outcome measures at different time points and therefore added little extra detail to our analysis. We familiarised ourselves with the documents not selected for data extraction, being open to potential key contextual or mechanistic data; although data were not directly extracted from them, the information contained within them still informed our interpretations. Additional file 4 contains a table reporting the final sample of interventions, by the number of documents related to them and the type of outcome they report (e.g. implementation outcomes, community-level outcomes or individual adolescent health risk behaviour outcomes).

Fig. 3

PRISMA flow diagram of included studies

Legend: Flow diagram showing the stages of searching, screening and inclusion that resulted in our final sample to review

Table 1 presents the general characteristics of the sample documents selected for data extraction.

Table 1 Table of sample characteristics

Table 2 contains the relevant extracted data, example CMO configurations and quality assessment scores for each of the 22 interventions under review.

Table 2 Table of characteristics and extracted data

Main findings

Programme theory and middle range theories

As was expected due to the complexity and heterogeneity of approaches within community mobilisation interventions, it was a challenge to determine a single programme theory. Based on the previous literature explored in the background of this article and discussions with stakeholders, we determined a broad, overarching programme theory that for community mobilisation interventions to result in reductions in adolescent MRB: interventions must explicitly seek to affect community-level influences through coalition-led initiatives, by creating social environments that mean that adolescents are less likely to engage in risky behaviours.

Under the umbrella of our broad programme theory, we produced a synthesis of three middle range theories describing the various CMO configurations that explain elements of how these interventions work or do not work. Through critically reflecting on these middle range theories, which we have presented thematically below, we answer our research questions of which contextual and mechanistic factors are present within our sample to produce outcomes.

Our three middle range theories developed as contributing to our understanding (our overarching programme theory) of how the community mobilisation interventions ‘work’ are:

• (1) Community mobilisation interventions achieve positive adolescent MRB outcomes when supported by guiding principles and the focus is on collaborating across factions of society.

• (2) Community mobilisation interventions achieve positive adolescent MRB outcomes when the community readiness is established through understanding the community social norms and prevalence through population level data, which motivates the community to create an environment for prevention.

• (3) Community mobilisation interventions achieve positive adolescent MRB outcomes when the community coalition has diverse membership with expertise and community members, triggering empowerment, support and knowledge.

Not all studies contribute to our thinking around the middle range theories and therefore not every CMO configuration contains every study, if there was no evidence within the study. Further, there are times where interventions report positive outcomes, but we have interpreted that a mechanism in question has not been ‘triggered’. Therefore, in the analysis tables where studies are listed under ‘Mechanisms not triggered’, this is not to say that they are not potentially successful and useful interventions, but simply that we did not interpret them to be adhering to that particular CMO configuration. Throughout the text (C), (M) and (O) labels appear to highlight aspects we interpreted as contexts, mechanisms and outcomes, to add clarity and enhance transparency of our analysis.

Middle range theory 1—Community mobilisation interventions achieve positive adolescent MRB outcomes (O) when supported by guiding principles (C) and the focus is on collaborating across factions of society (M)

While community mobilisation interventions aim to address the community-level causes of adolescent multiple risk behaviour, only some of the included studies appeared to be committed to that goal. The community mobilisation interventions that had the greatest success in terms of ongoing implementation and prevention of adolescent MRB tended to focus on embedding the intervention ‘principles’ throughout multiple factions of the community (e.g. schools, family, wider community/neighbourhood) (C). In these cases [20, 23, 51, 52, 63, 74, 84], the intervention is a framework, model or way of thinking, framed by ‘guiding principles’ as opposed to one-off programmes that need to be delivered in a certain way. This approach allows support and a way to structure collaborative efforts, without being too prescriptive, instilling empowerment for coalition members (M) to choose how to deliver the intervention within a proven framework [22].

Moreover, by taking the focus away from individual behaviour change or delivering school-based educational programmes and instead focusing on encouraging coalition and community cohesion through establishing strong collaboration and strengthening connections between family, schools and the wider community (C) [50], these seven interventions were able to deliver multi-level strategies and all work towards the same goal of social environmental change. This approach, over time, resulted in reductions in adolescent MRB in their community (O). The reason they were successful, we believe, is that they were more closely aligned with an environmental and systems way of thinking, instead of trying to teach young people not to engage in MRB, which is increasingly known to be largely ineffective and potentially increasing of inequalities [85]. If the intervention’s ‘guiding principles’ are woven throughout schools, policy areas including health and education, communities and families (C), then the societal shift needed to prevent adolescent MRB will come from the ‘bottom up’ and is more likely to be directly relevant to the community trying to be changed. It also brings different areas of the community in even if they do not sit directly on the coalition (e.g. school leaders, church leaders), as changing the social environment requires collaborative participation of a wide range of community members (M) [22]. Details of ‘outreach’ activities used to instil the guiding principles and encourage collaboration appeared infrequently within the data beyond note of ‘informalised collaborative activities’ [20], but some included meeting with city and parks personnel and making phone calls to elected officials [74] and supporting grassroots organisations, and partnerships with police [23].

Mechanistically, having a set of ‘guiding principles’ (C) encourages empowerment (M) among coalition and community members (where an instructive intervention pack would not), through the ability to design and deliver relevant programmes. The coalition can constantly revisit the principles when making decisions [22]. Some studies contradicted this theory in that they still managed to result in positive health risk behaviour outcomes in the absence of this context and mechanism combination. For example, interventions in the USA such as Communities that Care (CTC) [55], PROSPER [76] and Community Prevention and Wellness Initiative (CPWI) [53] appeared to be focused on a core component of the compulsory delivery of evidence-based (usually school-based) programmes with less apparent focus on the establishing collaborative networks (C). Although these studies do begin with a framework for intervention delivery, the essential element of collaboration and creating environmental shifts appeared to be less important than selecting from a ‘menu’ of evidence-based programmes (EBPs) [57, 76]. In contrast, the CTC programme in Australia had stronger focus on the collaborative approach and systems way of thinking (M) and less on which specific types of programmes were delivered [63]. For this reason, we determined that some studies strongly aligned with this middle range theory [20, 23, 51, 52, 63, 74, 84], but not others [53, 55, 59, 72, 76, 83], despite all seeing reductions in health risk behaviour outcomes.

We found that other, less successful, interventions were either focused more on simply choosing and delivering programmes as opposed to establishing long-term connections (M) [67,68,69] or were embedded within in societal contexts that made it difficult to strive for a systems approach (C) [54, 79]. These interventions appeared to be less successful in creating that social environmental shift needed for achieving positive intermediate or health risk behaviour outcomes (O). It was a challenge to fully elucidate the causal pathway with this part of the theory as most studies did not go into detail about the extent to which they adhered to the guiding principles and how exactly they applied the principles to their own decision-making (Table 3).

Table 3 CMO configuration 1

We further interpreted, unsurprisingly, that in order to facilitate system-wide prevention of adolescent MRB, there needs to be a strong and supportive funding infrastructure (C). We found this to be a crucial context important for triggering mechanisms such as collaboration, flexibility and motivation (M) from the coalition members and wider community (participants) to continue with the intervention. This CMO configuration was supported by evidence from 12 studies, which reported positive implementation or coalition functioning [55, 59, 69, 72, 76] and/or health risk behaviour outcomes [20, 22, 23, 51,52,53, 55, 59, 71, 72, 76, 83, 84]. It is not enough to simply have the ‘model of guiding principles’ context, but this must be coupled with long-term, flexible funding that accounts for both technical staff assistance and differing strategies the coalition may implement [50]. Further, the environmental strategies likely to have the most success (e.g. paying for young people to engage in health promoting activities [22], scholarship programmes for at risk youth [74], policy changes around alcohol or tobacco smoking [23]) are also likely to be the most expensive comparative to school-based education programmes. This contextual condition was further highlighted through examples of interventions that had been successful in one context but then had failed to work in the same way in other contexts/countries. Namely, Communities that Care and the Icelandic Prevention Model were both adapted for other contexts (e.g. CTC in the UK [67] and IPM in Chile [51, 52]) and all highlighted funding longevity as a significant barrier to success. Further, the Connect to Protect programme in Thailand [54] did report a commitment to shifting the social environment, but the context they were working in (e.g. significant methamphetamine problems within the community, funding issues, legal approach to drug use, limited community involvement) had contextual factors that made it extremely difficult to deliver the intervention [54]. Our second core result is therefore that context is crucially important to achieving seeing reductions in adolescent MRB from community mobilisation interventions and a condition on this middle range theory (Table 4).

Table 4 CMO configuration 2

Middle range theory 2—Community mobilisation interventions achieve positive adolescent MRB outcomes (O) when the community readiness is established through understanding the community social norms and prevalence through population level data (C), which motivates the coalition and community to create an environment for prevention (M)

“Community readiness” repeatedly arose as an important concept [53, 54, 57, 72] and is widely discussed in the community prevention literature [58, 87]. Community readiness refers to a community’s ‘levels of awareness of the problems which they need to address, their perceptions, expectations, and interests related to the problem, their judgments and necessary decision-making processes as well as capabilities, the necessary human and non-human resources to implement effective prevention interventions, and their readiness to initiate prevention activities and to institutionalise and continue them if and when necessary.’ ([87], p.1084). Put simply, if the social environment of the community is going to be changed to support adolescent health (O), the coalition need to know which communities are in most need, which risk and protective factors require the greatest attention and whether the community will accept programmes and in what form (C) [22].

An important part of community readiness in terms of MRB interventions was that the local community had social norms and values compatible with prevention interventions as well as a concern for adolescent health (C) [50]. For example, for the intervention to be successful, the local community and coalition members would need to (1) be concerned about adolescent MRB within their community (e.g. through experiencing high local substance use or violence prevalence rates) and (2) have community and societal social values amenable to a prevention approach to these problems. It is only through these contextual ‘community readiness’ factors that the motivation mechanism would be fired among both the coalition members driving intervention delivery and community members receiving the intervention (M) (Table 5).

Table 5 CMO configuration 3

Through this CMO, we see that it is not only contextually important that adolescent health risk behaviours are prevalent, but that their framing and discursive representation is also crucial. For example, if adolescent MRB is not viewed as a problem for the community to try and tackle, or if it is viewed as a legal issue (e.g. drug use and violence) (C), then there will be little interest in the community or local stakeholders in pursuing a prevention approach (M). This particular example was highlighted in a included studies in Thailand [54] and Mexico [79], which referenced the established legal frameworks for tackling adolescent substance use as a major challenge to getting the community and stakeholders unified within the intervention, with substantial distrust of government agencies (C). Therefore, the wider context surrounding adolescent health risk behaviours is important for triggering the level of commitment (M) from community members to take part in the intervention. In turn, if the community are committed to the issue and the coalition can galvanise the community, it is expected that positive implementation and health risk behaviour outcomes would follow (O).

Using population data was critical for both establishing community readiness through assessing the size and nature of the problem of adolescent MRB and allowing the coalition and community to be actively involved in data collection/analysis (C). Population data was also crucial for tracking the intervention over a long time period, something mentioned by some of our stakeholders and highlighted in a number of studies [51,52,53, 59, 63, 64, 69, 84]. To be clear, this is beyond simply evaluating the intervention using survey data or evaluation methods (which all studies did), but is about using population and/or epidemiological data as a the ‘foundation’ of the community mobilisation effort (C) [69]. This approach is also needed for guiding the coalition in establishing what kinds of programmes, at what levels, they may choose to implement [22].

Studies that did not use initial data analysis of adolescent MRB and other related risk and protective factors as its foundation, had challenges doing so or did not mention it as a major part of ‘readiness’ (C) [23, 54, 67, 68, 76, 82], we interpreted, were less able to design strategies uniquely matched to the needs of that community (M). These studies saw the potential value of community mobilisation interventions and delivering events and programmes for the community (and some mentioned the goal of environmental change [23, 54]), but did not take the time to understand how these programmes might bring about long-term change. Without meaningful population-level data at initiation and a commitment to tracking that data throughout, mechanisms including flexibility (e.g. the ability to react to evidence coming from the community) (M) can lose sight of the intervention aims [22]. This issue was raised by one study, highlighting that while they were aiming to empower community members by giving them flexibility (M), in practice it meant that ‘activities were only indirectly related to youth health behaviours’ (O) [80], whereas greater engagement with population data would have resulted in strategies more directly connected to the intervention objectives (Table 6).

Table 6 CMO configuration 4

Middle range theory 3—Community mobilisation interventions achieve positive adolescent MRB outcomes (O) when the community coalition has diverse membership with expertise and community members (C), triggering empowerment and support (M)

We identified that the coalition composition was an important resource (C) that determined whether certain reactive mechanisms were triggered in particular contexts. Across the documents, we found there to be limited detail about the makeup of the coalitions involved in the interventions. Where we were able to extract data, however, we inferred from the data that the coalitions most likely to create an environment where adolescents are less likely to engage in risky behaviours were those that achieved a balance of expertise (e.g. civil servants, policymakers, social scientists) and community relevance (e.g. community members, parents, young people) through coalition membership. We determined that focusing on getting the right configuration of the coalition (C) enacted the mechanisms of empowerment, knowledge and ability (e.g. ‘political leverage’ [20]) (M) needed to alter the environment (O).

Although there were limited details regarding the coalition configurations in most studies, from what we could glean, the data suggested that most interventions attempted to form a coalition that had a mixture of expert stakeholders and community members [20, 23, 53, 54, 59, 63, 64, 67, 74, 76, 79, 84, 82]. The study that had purely civil servants on the coalition [69] and the study that was made up entirely of community members [72] both faced challenges in implementation and achieving prevention outcomes, which they partially attributed to the coalition functioning. Critically, several studies tried to embrace the mixture of expertise and community members (C) yet had significant issues during implementation (M) or did not report community reductions of adolescent MRB (O) [54, 64, 69, 79, 80] (See Table 7). For example, for the Communities that Care in the Netherlands [64], there was a challenge in engaging the right mix of people, with the coalitions being made of up of representatives from local institutions working with children, who often left when they moved job role. They noted a distinct lack of participation from ‘students, business leaders and volunteers [community members]’ [64] (C), which they found to be a barrier to achieving the right knowledge and community attachment (M) needed to make environmental changes for adolescent health (O). Further, Brown et al. noted that lower community support and lack of collective identity (C) within the coalitions in the Red de Coaliciones Comunitarias de Mexico study [79], led to ‘lack of momentum necessary to take action’ (M).

Table 7 CMO configuration 5

We determined that for this middle range theory to be upheld and result in positive outcomes, additional components are needed so that the coalition part of the intervention can function successfully. Intervention champions were interpreted as an important contextual feature (C) within the IPM [50], Communities that Care (CYDS) [59], New Directions [20], Project Freedom (Lawrence) [72] and Project Freedom (Wichita) [74] and was mentioned by a stakeholder describing the implementation of Communities that Care Australia [63]. These champions speak about the intervention as much as possible, as a formal part of their responsibilities. It becomes a major part of their role to advocate for the intervention, so it remains on the agenda for public health prevention, schools and other agencies. These champions are a particularly crucial for triggering collaboration and establishing connections across different domains of the community (M). Crucially, two studies [79, 80] highlighted that not having community champions (C) meant they had a reduced ability to create the collaborative networks (M) needed to improve the environment for adolescent health (O) (Table 8).

Table 8 CMO configuration 6

Although only highlighted by five studies, we found that intervention champions helped establish community-wide understanding and trust (M). They also kept strategic focus on the intervention and acted as a catalyst from cross-community collaboration [59]. Intervention champions would not necessarily be needed for single component educational programmes. However, for community mobilisation interventions with the goal of joining up different areas of community and policy, they are an essential glue that allows social environmental shifts. They can communicate what is and is not working across the communities and work to continually embed the principles for reducing adolescent MRB (i.e. creating health promoting environments) across schools, policy and neighbourhoods. We present this evidence within this middle range theory as we found that intervention champions essentially bolster the coalition. Intervention champions alone cannot achieve the reductions in adolescent MRB without a strong and diverse group of experts and community members to support them (C). Finally, something missing from the data was mention of involving young people in the coalition (C). This might be happening, but not reported, as one of our stakeholders claimed this was a core element of CTC Australia but there was not a description in the document [63]. However, we had expected to see greater presence of young people within the community mobilisation interventions but given the lack of data on this element this is not something we have been able to explore using CMOs in the present synthesis.

Discussion

This realist review synthesis identified important contextual and mechanistic factors that are critical to the successful implementation and effectiveness of community mobilisation interventions for adolescent multiple risk behaviour. The central finding of our review is that for community mobilisation interventions to ‘work’, they must be committed to evoking a set of ‘guiding principles’ that coalitions can use to deliver the intervention their own way. One framework that has been particularly successful in this regard is Icelandic Prevention Model (IPM) [84], with this model becoming increasingly popular and several countries either planning to implement it or exploring it as an option [88,89,90]. However, we should note the critiques surrounding IPM due to the lack of clarity on successful intervention components and few studies being able to establish causal links to health outcomes [51, 91]. This was reflected in our sample as only three documents measured outcomes [51, 52, 84] despite the intervention having been run for many years across the world [51]. Therefore, although we inferred success in these interventions and attributed that partially to the adherence to guiding principles and social environmental change, we present this finding with caution due to the lack of outcomes papers and the challenge in establishing whether it was the intervention or some other environmental factor (e.g. Covid-19 pandemic [51], policy or societal changes [84, 92]) that resulted in the outcomes.

We further found that the context of the delivery of these interventions is crucially important. The importance of context is illuminated by commentary documents relating to the delivery of the Communities that Care (CTC) model in countries outside the USA. For instance, Basic [87] describes the community readiness-related barriers to intervention success in Croatia. Low awareness among stakeholders of the problems of adolescent health risk behaviours, as well as a lack of realistic assessment of available and accessible resources for initiating and sustaining changes, hindered the progress of CTC. Perez-Gomez et al. [93] saw promise in CTC for middle- and low-income countries in Latin America, but their pilot in Colombia conveyed that time and funding must be spent to establish community readiness among residents but also local authority personnel. There was often initial interest in the intervention from local authorities, but a reluctance to make financial commitments to the programmes and at times a lack of motivation to drive the intervention [93]. Further, for IPM, a commentary by Koning et al. [91] highlighted how the context of Iceland is particularly suited to the intervention and warned against other countries trying to replicate without strong consideration of context.

These examples reiterate that even with a proven community mobilisation intervention like CTC or IPM, the wider context that supports intervention delivery (e.g. funding, local authority buy-in, community understanding of the problem) affects the trust, motivation and ability of potential coalition members, which are essential mechanisms for sustaining the intervention to achieve positive outcomes. Although this finding is in some ways generic and not necessarily directly unique to adolescent MRB outcomes, we nevertheless see it as important due to our programme theory that states that social environmental change is needed to reduce MRB, which is more complex, long-term and expensive than educational programmes [22].

Further, we were unable to identify particular programmes within the community mobilisation interventions that were noted as crucial (e.g. a school-based initiative or community component). While this was not an objective of our study due to our focus being on the wider community mobilisation and coalition effort, it would have no doubt been informative to communities hoping to implement these strategies. This absence of specific recommendations around programmes or components is partly due to the lack of description of said individual programmes, but also if we were to recommend specific programmes, we would be contradicting the impetus of community mobilisation interventions. That is, several studies suggested that success does not lie within the single components of the programmes delivered [20], but instead within the shift within society towards a way of thinking that fosters adolescent MRB prevention. This shift is more probable with comprehensive approaches, with a greater number of prevention activities at different levels [20]. However, we do recognise critiques that suggest that without identifying what the successful intervention components are, it is difficult to establish whether the intervention has had a positive impact on outcomes [91].

The composition of the coalition was highlighted as a key factor. Those studies with the greatest success strongly recommend a mixture of experts, policy stakeholders and community members, with ‘intervention champions’ being employed to maintain those connections [20, 22, 59, 63, 72, 74]. We identified that most of the coalitions were made up of adults only or did not report ways in which young people input into the intervention. The Australian version of Communities that Care adapted the original CTC model and created an active youth board who were involved in the ongoing development of the intervention [63]. This might be an indication of the time period of the studies, with many studies being designed before 2001. It was around this time that there was an acceleration in calls for participatory approaches to be adopted in public health research and practice [94]. The value of public involvement in public health intervention research is now well recognised [95]. Further, involving youth in the design and delivery of youth-based interventions is a matter of equity, social justice and reducing power imbalances [96]. The involvement of the target group, adolescents, has also been shown to be critical for intervention effectiveness [97]. We would anticipate that if community mobilisation interventions are to be successful in the future, the studies included in this review would need to be revised to incorporate greater youth involvement.

We were unable to identify CMOs related to the ‘for whom’ question in our synthesis. One document [83] reported gender differences in intervention effects, but could only offer a speculative explanation as to why there were greater reductions in boys’ health risk behaviours than girls. Across documents, there was often a lack of detail about the young people and community members who were the target of the intervention, with a greater focus on coalition members. Some studies described the intervention setting, in a general sense, in that the focus was on youth of colour [80] or from deprived areas [54]. Without a subgroup analysis or narrative reflection on how aspects of the intervention affected such groups differently, we were unable to draw any conclusions on any dimension of inequality. This finding is unsurprising given the period during which many of the studies were undertaken (1980s–2000s). The two previous systematic reviews on individual, family and school-level interventions for adolescent multiple health risk behaviour similarly found a lack of reporting of the demographic characteristics needed to assess differential intervention effects, particularly for older studies [6].

Strengths and limitations

Realist reviews, like traditional systematic reviews, are ultimately limited by the data they can extract. The kind of information we needed to understand the contexts and mechanisms, such as process evaluation data, researcher reflections and challenges and successes are less likely to be published in peer-reviewed journals, if at all. The stakeholder engagement sessions provided some insight into the mechanisms and contexts surrounding some interventions. However, given the period that many of the interventions took place and how complex many of them were, we were limited in how much detail we could gather. Documents were also limited in the detail they provided around the composition of the coalition, the strategies they implemented and the population they reached. These data limitations also meant we were unable to draw any conclusions about health inequalities. Further, we know of several community mobilisation interventions that are planned or ongoing [87, 88, 93], but as they did not publish a report of outcomes, we could not include them in the analysis—although they were used to understand the patterns found within the included studies.

There may have been some interventions that were not picked up in our searches. Multiple risk behaviour is a construct referring to the co-occurrence of two or more health risk behaviours. However, this term is not widely used within the vast field of community mobilisation approaches. Given that our search terms included health risk behaviour, we may have missed some interventions that do not overtly mention health risk behaviour but measured relevant outcomes. Many interventions appeared relevant but were excluded because they only reported one health risk behaviour outcome. Further, the majority of programmes in our sample were delivered in the USA. Our searching style and syntax choices (e.g. choosing to restrict to ‘coalition-based interventions’) will have likely had an impact on our results. For instance, ‘coalition’ as a term may not be used commonly in contemporary research or outside of the USA—as is the case within the Icelandic example, which proved to be a key study not picked up by initial searches. Despite engagement with stakeholders outside of the USA, many studies related to the Communities that Care programme. This undoubtedly narrowed our analysis, particularly in the early middle range theory development stages.

Conclusions

For community mobilisation interventions to ‘work’ and reduce adolescent multiple risk behaviour within the community, the coalitions within them must seek to alter the social environment in which these behaviours are likely to occur. Mechanisms including empowerment, motivation and knowledge lead to this success, but only under certain contexts. In particular, we found successful interventions that aligned with our middle range theories tended to (1) be guided by principles that were interwoven across the community, (2) establish community readiness and use population level data to make coalition decisions and (3) consider carefully the composition of the coalition so that the right balance of expertise and local knowledge were included to achieve social environmental change.