Mobilising social support to improve mental health for children and adolescents: A systematic review using principles of realist synthesis

Social support is a well-recognised protective factor for children’s mental health. Whilst many interventions exist that seek to mobilise social support to improve children’s mental health, not much is known about how to best do this. We sought to generate knowledge about the ways in which social support can be mobilised to improve children’s mental health. We conducted a systematic review, which followed the principles of a realist synthesis. The following databases were searched: PubMed, CINAHL, Ovid MEDLINE, PsychINFO, EMBASE, Child and Adolescent Studies, EconLit and SocINDEX. Studies were included if the age of participants was between 0 and 18 years and they evaluated or described programme theories of interventions that sought to improve children’s mental health by mobilising social support. Relevance and quality of studies were assessed, and data were extracted and analysed narratively. Thirty-three articles were included. Studies varied substantially with regard to the detail in which they described the processes of mobilising social support and expected mechanisms to improve children’s mental health. Those that provided this detail showed the following: Intervention components included explaining the benefits of social support and relationships to families and modelling friendly relationships to improve social skills. Pathways to improved outcomes reflected bi-directional and dynamic relationships between social support and mental health, and complex and long-term processes of establishing relationship qualities such as trust and reciprocity. Parents’ ability to mobilise social support for themselves and on behalf of children was assumed to impact on their children’s mental health, and (future) ability to mobilise social support. Although interventions were considered affordable, some required substantial human and financial resources from existing systems. Mobilising social support for vulnerable children can be a complex process that requires careful planning, and theory-informed evaluations can have an important role in increasing knowledge about how to best address social support and loneliness in children.


Introduction
Social support refers to the extent to which an individual has access to, or perceives they have access to, assistance and resources provided by people in their social network [1]. It is concerned with the function of social relationships rather than their structural constellation, which makes it, to some extent, distinguishable from other concepts such as social capital or social connectedness, although the terms are overlapping and sometimes used interchangeably [2]. Whilst potential adverse effects of social support have been established, too, social support is generally regarded as an important protective factor for positive mental health at all ages, including during childhood and adolescence [2][3][4]. For children and adolescents, it can be associated with lower rates of depression, generalised anxiety and post-traumatic stress disorders [5][6][7][8][9][10], suicide [11], behavioural and school adjustment problems and risk behaviours [7,12,13]. Various studies that investigated the association between social support and protection from mental health problems found that sources of support (e.g. informal or formal) vary across the life span [8]. Findings from the youth literature suggest that sources and types or characteristics of social support might influence the magnitude of the protective (or sometimes adverse) effects of social support on mental health, but that important evidence gaps remain [14][15][16].
In the pursuit of realising potential mental health benefits for children and adolescents, the mobilisation of social support has been incorporated into the design of many interventions [17], either as one of several components, or as the only or main component. Researchers have highlighted the challenges of designing, implementing and evaluating what they call social support interventions due to the multi-dimensionality of the concept, which is defined and measured in many different ways [18,19].
Two main social support theories, the stress-buffering and main-effects models [20][21][22] have been leading the field for decades. Whilst the stress-buffering model suggests that social support reduces the impact of negative life events on a person's (mental) health, the main effects model hypotheses that there are (mental) health benefits inherent to social relationships irrespective of the stress experienced by a person. Based on those and additional theories, many different pathways and mechanisms have been proposed by which social support is expected to influence mental health [23][24][25]. They include: creating feelings of belonging, security and self-worth; developing trustful and intimate relationships; adoption of healthrelated behaviours through social networks; and improving access to resources and opportunities [22,26].
Overall, however, there is not much knowledge on how interventions should be designed to mobilise different types of social support in order to improve children's mental health [27]. This kind of knowledge, including about how different types and sources of social support influence mental health outcomes, which differ according to age group, is important in order to develop programme theories, and understand gaps in evidence [27]. By reviewing the intervention literature, we sought to understand: 1. Ways in which social support can be mobilised in order to improve the mental health of children and adolescents.
2. The mechanisms by which social support is expected to (or has been found to) lead to improved mental health for children and adolescents.
We hypothesised that the following areas would be important to investigate: sources and types of social support; metrics used for measuring social support and mental health; population characteristics. Finally, we wanted to understand resource inputs required for the delivery of interventions, and their potential role in influencing outcomes.

Methods
We carried out a systematic review of the literature, which followed principles of a realist synthesis [28,29]. Realist review or synthesis is an approach to reviewing evidence on complex social interventions which seeks to provide an explanatory analysis of how and why interventions work (or do not work) in particular contexts or settings and for particular populations. It combines theoretical understanding and empirical evidence, with a focus on explaining the relationship between the context in which an intervention is applied, the mechanisms by which the intervention works and the outcomes produced. Underlying this is an understanding that change is not just generated through the influence of interventions, but through resource inputs, human reaction processes and contextual factors. It is particularly suitable for the development of programme theories [28]. We used principles of realistic review in the inclusion of studies and when extracting data from studies.
We searched for studies concerned with the conceptualisation and evaluation of interventions that sought to mobilise social support to improve mental health of children and adolescents. We were interested in individuals of ages from zero to 18 years. We included infants in the review in order to capture interventions that seek to prevent mental health problems for children by focusing on early childhood.

Inclusion criteria
We included studies that examined interventions where the mean age was between 0 and 18 years. Studies were only deemed appropriate for inclusion if they described or evaluated interventions that had specific aims to increase social support as indicated by the inclusion of social support into the programme's or study's aims, as well as the inclusion of a measure of social support in the study design. We relied on authors' explicit descriptions of social support. For example, we would not infer from peer support intervention that the intervention was about social support unless the authors discussed social support explicitly. This approach has been used in a global review of active components present in interventions aimed to improve adolescent mental health [30]. Social support could refer to the child's or parent's social support as long as the intervention sought to mobilise social support in order to achieve improved children's mental health, which had to be an explicit goal. No (additional) restrictions were applied regarding type of settings. Primary outcomes were changes in children's mental health. Studies were included if they measured mental health or associated indicators or, for infants, predictors of mental health. This included studies that measured self-esteem, hope or coping for children, and studies that measured mother-infant attachment for infants. We accepted papers that reported on mental health outcomes in previous evaluations (if they were appropriately referred and cited in the paper). Studies also needed to include, as a secondary outcomes, a measure of social support. Outcomes for mental health and wellbeing and social support could use a standardised scale, a sub-domain of a scale, survey or activity data, or be evaluated qualitatively. Since we were interested in various evidence types (including conceptual papers reporting programme theories) we also accepted studies that did not specify outcome assessments but outlined the types of outcomes that could be included in evaluation studies.
Full texts of included studies needed to be in English language. There were no restrictions in terms of their study design; we included experimental, non-experimental, qualitative, and mixedmethod designs, evaluation protocols and conceptual papers reporting programme theories.

Exclusion criteria
We excluded studies of interventions that were seeking to improve parental behavioural outcomes but did not mention children's mental health in their programme goals. Consequently, we excluded studies of interventions that were only concerned with reducing child maltreatment. We excluded populations exposed to traumatic events or extreme adversities such as war, natural disasters, epidemics, and terrorist attacks. We also excluded studies that specifically targeted children with autism or severe communication needs.

Search strategy
Search terms that described the population, social support, and intervention were initially scoped on PubMed before a revised search strategy was developed for PubMed. The search strategy was adapted for each of the following databases: CINAHL, Ovid MEDLINE, Psy-chINFO, EMBASE, Child and Adolescent Studies, EconLit and SocINDEX. Searches identified studies between 01/01/2008 to 08/06/2018. An example of our search strategy is provided in the electronic material (S1 Box). Fig 1 shows the PRISMA flow chart of the screening process. Titles and abstracts were assessed by one reviewer (AB). Articles that clearly did not meet criteria were rejected at this stage. Full texts were retrieved for potentially relevant articles. The same reviewer (AB) screened studies based on full text. Studies where it was unclear whether inclusion or exclusion criteria were met were subject to a detailed screening process undertaken by four reviewers (AB, DP, JP, MS); this involved completing a screening tool, and various rounds of discussions.

Assessment of relevance and quality
Following guidance for realist reviews [29], studies were appraised as to their relevance as well as their rigour. The relevance of the study was assessed based on the extent to which the study defined, conceptualised and measured social support, and explained how it was mobilised and expected to improve children's mental health outcomes. Using the latest version of the Mixed Methods Appraisal Tool [31] the study rigour was assessed in relation to choice of study design, sample size, data collection methods, and outcomes. Following the guidance and algorithm provided by the tool, we applied 'low', 'high' and, where information was insufficient to rate the criterion, 'can't tell' ratings. The algorithm provides quality criteria as well as examples of how to apply those for qualitative, quantitative randomised controlled trails, quantitative non-randomised controlled trials, quantitative descriptive studies and mixed-method studies. Studies were not excluded based on relevance or rigour. Instead, the rating informed the interpretation of findings.

Data extraction, analysis and synthesis
Data were extracted from all sections of papers using bespoke forms and analysed narratively using headings of a realist synthesis and categorised into age groups of children. Age categories included infants aged 0 to 2 years, children aged 3 to 9 years, and adolescents aged 10 to 18 years. For studies, where the age range fell between two categories, they landed in the category that captured more years; e.g. if the inclusion was 5 to 12 years, the study would land in the 3 to 9 years category. By identifying data patterns, a realist synthesis seeks to derive information about relationships between resource inputs, human reaction processes, and contextual factors for interventions or intervention components, and how those lead to particular outcomes. In this paper, our main interest was to understand how social support was conceptualised, e.g. with regard to types of social support, which changes in human interaction processes were assumed to be required in order to improve children's mental health outcomes, and how those were modelled into the intervention design. We used the above-mentioned dimensions (intervention components, context, mechanisms, and outcomes) for the synthesis of the data and we present findings by research questions.

Results
Thirty-three studies were included . S1-S5 Tables in the supporting information provide details of the studies including the details of how assessments of study relevance and quality were derived. In many studies, social support was not well-conceptualised, and many studies were weak in explaining how social support was mobilised or expected to lead to improved children's mental health. Most studies did not specify the types or sources of social support they sought to address or the rationale for doing so. Social support sometimes only referred to a single source of support such as health professionals, peers or mentors, parents or school staff. Interventions most commonly mobilised the social support of parents, followed by those studies that were about mobilising social support of children. Only a few were about increasing social support of the family as a whole.
More than a third of the interventions were mentoring, peer support, or a combination of the two. Other interventions included parenting education, training or support (covered in

PLOS ONE
seven studies), complex family support involving goal-setting and support-planning, linkage projects with schools and other public services, community capacity-building and service integration approaches, and psychoeducation or mental health literacy training. More than half of the programmes were delivered in the form of individual (family) support, and the rest were delivered in the community in group sessions or in mixed formats. Two studies were delivered via telephone or internet. Most interventions were provided by volunteers, community workers or psychological therapists. Only four studies [43,49,51,56] were conducted in middleincome countries (South Africa, India and Pakistan), whilst the rest were from high-income countries (North America, Europe, Australia, Japan). Tables 1-3 provide details about studies in relation to their programme components, contextual factors, population, mechanisms for improving children's mental health, and findings on outcomes.

Age-group specific findings: Infants (0 to 2 years)
How is social support mobilised, to which populations, and in which context? Four studies [33][34][35][36] were concerned with providing or mobilising social support for parents of infants to improve children's mental health (S3 Table). Two of the interventions were linking parents with their community through a trusted lay person, who would connect the parent with informal and formal support [34,35]. One intervention was linking mothers with health professionals [36]. Two of the interventions included teaching skills such as mother-infant interaction [33] or broader social skills that would allow mothers to build relationships [35]. In the study by Mitchell et al [35], a mentoring mother modelled friendly relationships and helped to create opportunities for the mothers to practice newly gained skills together in the community. All four interventions addressed some form of informational support in regards to parenting; one intervention addressed informational support only [36], whilst one specifically addressed all types of social support (i.e. emotional, practical and informational support) [34]. The study by Stubbs and Achat (2016) [34] was the only one that targeted the whole families, whilst the others targeted mothers. All four took place in high-income countries. They targeted parents at risk of social isolation, stress and mental health problems (Table 1). One intervention targeted women with postpartum depression [33]. In three studies [33,35,36], mothers received the intervention alongside professional (mental) health services (Table 1).
Does an increase in social support lead to improved children's mental health, and what are the mechanisms by which this is (expected to be) achieved? Three studies [34][35][36] reported increased social support, which in two studies referred to social support from health professionals, the community or formal services not measured with standardised scales ( Table 1). None of the studies reported an increase in support from partners, parents or friends (even though all four studies measured this). One moderate-quality study [33] that captured perceived social support using a standardised scale reported findings in relation to social support that favoured the control group. Authors explained this as follows: mothers in the control group formed their own networks that were more sustainable than the relationships formed by mothers in the intervention group with volunteering peers. In additionthe matching of volunteers to mothers was considered inadequate and the teaching component of the intervention might have hindered the development of equal and trusting relationships between peers and mothers.
Studies described how social support was expected to protect against negative impacts of depressive symptoms and stress, improve mother-infant interactions, parental self-efficacy, agency, and self-esteem (Table 1). In turn, those would allow parents to form new relationships, and this would improve child cognitive and social development, children's ability to form social relationships in the future as well as improve child behaviour and wellbeing. Infant outcomes measured in studies included infant-attachment, and socio-emotional and cognitive development (S3 Table). Only one low-quality study [35] reported positive effects on motherinfant attachment (and improved parenting skills). Two studies [34,36] reported that design errors might have explained the lack of evidence on infant outcomes.

Age-group specific findings: Children (3 to 9 years)
How is social support mobilised, to which populations and in which context? We identified thirteen studies [32,[37][38][39][40][41][42][43][44][45][46][47][48] in this category (S4 Table). This included two studies [37,39] that did not specify the age, but where we inferred from the background information that they referred primarily to children in early or mid-development ages. Most interventions aimed to increase parents' social support by: directly providing social support, increasing access to services, reinforcing to parents the importance of social relationships and teaching relationship or help-seeking skills ( Table 2). For a few interventions, this specifically referred to improving relationships with childcare institutions or schools. Three interventions sought to change capacities of social networks and whole service systems (including schools) to mobilise social support for parents [40,47,54]. Some interventions focused on increasing positive emotions such as hope and self-esteem, which were expected to lead to development of new relationships. Studies focusing on changing the perception of parents about social support   described how interventions were increasing parents' satisfaction and trust with public institutions by providing a trusted person, who would facilitate those links (see for example Drummond et al. [40]). Some studies assumed that children of parents with increased social support would acquire new socio-emotional skills, thus allowing them to build their own social support systems in the future, highlighting the intergenerational effects of social support (see for example Doty et al. [39]). All but one study [54] targeted children and their families experiencing socio-economic disadvantage, including children of parents with mental illness from migrant, black or ethnic minority backgrounds. The study by Hauken et al [54] targeted children whose parents were living with cancer. One study [43] took place in a low-income country with high rates of HIV, substance abuse, and violence, whilst all other studies took place in high-income countries. Studies described families' social isolation and lack of social support, which could include their alienation from school and public services, due to distrust towards government, based on their own past, or intergenerational experiences as a community (Table 2). Studies described problems experienced by children, which included behavioural and health problems, poor literacy and numeracy skills, low self-esteem, lack of bonding with parents, and academic underachievement. Two studies referred specifically to families involved with the child welfare system.
Does an increase in social support lead to improved children's mental health, and what are the mechanisms by which this is (expected to be) achieved? The vast majority of studies reported increases in parents' social support, which referred most commonly to improved family relationships, and to a lesser extent, to other parents, and improved relationships between families and schools ( Table 2). Only one study [41] referred to social support as mobilised by children directly, whilst all other studies referred to social support as mobilised by parents (and teachers) on behalf of the child. In some studies, social support was reported as an outcome of the implementation of the intervention, referring for example to mentoring or peer support, whilst in other studies it was reported as a primary or secondary outcome.
The majority of studies reported improved child behaviour, cognitive and social development outcomes, alongside improved school performance or attendance, as well as improved coping, psychological functioning or help-seeking (Table 2). Some studies reported that effects were only small, and two studies [43,62], including a high-quality one, reported negative effects on child behaviour, emotional problems or school adjustment. One study explained this as short-term negative emotions when opening up about painful experiences, whilst the  Table 3. Information about programme theories and findings of studies concerned with adolescents (10-18 years).

Social support outcomes Child outcomes
Asghar et al 2018 [49] Building life skills of girls and training their caregivers and service providers in supporting girls Displaced and hostcommunity adolescent girls; some living in camps or with restricted movement in public; exposed to genderbased discrimination and stigma Social support networks together with self-esteem and hope (human assets) and physical assets hypothesised to protect from future violence Relationship to a mentor (trusted adult) expected to lead to greater resilience  other explained this as insufficiently skilled staff, who did not have child development knowledge. Two studies [32,37] did not report child outcomes but reported improvements in parental agency or parent-child relationships. Mechanisms by which social support was expected to improve children's mental health referred primarily to an increase in parents' social support (Table 2). A few interventions were specifically designed to teach parents to ask for and utilise social support, which in turn was expected to improve parental capacity, improve family interaction and reduce child behaviours problems. In some studies, increased access to informational support, better links to schools and other services were considered to lead to improved child development and wellbeing. One study [39] explained this link with children's ability to develop socio-emotional skills that would support their psychological adjustment and ensure access to social support networks in the future. Social support was seen as providing opportunities for experiences that would allow children to build and engage in social relationships, for example by engaging in recreational activities.

PLOS ONE
relationships and social skills; offering safe spaces or opportunities for young people to practice their social skills; encouraging youth to seek help for social support; or changing perceptions of the benefits of social support. Interventions sought to provide various types of social support including informational (e.g. how to seek a job), material (e.g. borrowing a car), or emotional (e.g. how to leave an unhealthy relationship). A few interventions referred to providing social support to parents or whole families, or supporting them in developing social support networks. This included providing information about social support, reinforcing the importance of social support, or removing access barriers to social support (Table 3). Other intervention characteristics were primarily educational, e.g. in the form of psychoeducation or self-management. Most interventions applied empowerment and strengths-based approaches towards education. Twelve studies referred to youth exposed to a number of risk factors such as living in poverty, in single-parent households and being treated unequally because of ethnicity, sexual orientation, or gender (Table 3). Youth had low literacy skills, dropped out of or had been excluded from school, had been or were at risk of being removed from their families, or experienced mental health problems, discrimination or abuse. Two studies were about universal preventative interventions, which addressed transition to employment and mental health of young male athletes. Three studies [49,51,56] took place in low-and middle-income countries.
Does an increase in social support lead to improved children's mental health, and what are the mechanisms by which this is (expected to be) achieved? More than half the studies reported evidence of an increase in social support, referring mainly to perceived social support measured with standardised scales. This commonly referred to specific types or sources of support such as by families, peers or mentors. Two studies reported no effects on perceived social support [46,50] and one study [53] reported negative effects among boys when the intervention (mentoring) resulted in relationship break-ups between mentor and mentee.
Most studies that reported positive effects on social support also reported positive effects on depression, anxiety or behaviour, or on indicators of mental health such as self-esteem, selfefficacy, coping, hope or resilience (Table 3). Often, those changes were reported alongside improved school attendance, performance or functioning. However, a number of studies reported mixed findings (that is, some mental health outcomes improved, but others did not) or small effects. One study [43] found that child behaviour problems could become worse, which they attributed to insufficiently skilled staff. Eleven of the sixteen studies were of either moderate or high quality.
A range of expected mechanisms for adolescent mental health referred to protective or buffering effects of social support, whilst others referred to social-cognitive effects of social support, such as sense of belonging, identity, self-esteem, self-control and self-regulation, or to relationship aspects such as trust or sense of connection (Table 3). One study [57] described in detail the types of social support provided by different sources of support and hypothesised that peers were more appropriate for providing emotional support, and mentors more appropriate for providing advice and guidance (which was also supported by their findings). Mechanisms for interventions that targeted parents' (rather than youth's) social support included changes in parenting attitudes, behaviours, knowledge and skills, as well as an increased perceived benefit of social support and ability to navigate services for the young person One study [60] hypothesised that the intervention stimulated positive effects of social support on mental health, because of an increased social stimulation to care for one self.
Resource inputs to deliver interventions. Studies varied substantially in the detail reported on resource inputs, costs of programmes, or resource implications. Overall, there was not enough information to carry out systematic data extraction and analysis. However, we identified some relevant information and common themes. A third of the studies explicitly highlighted the affordability and scalability of interventions, with several studies even expecting a positive return-on-investment, for example because of expected reductions in criminal justice costs linked to improvements in delinquency and criminal behaviour. Whilst most interventions were provided by volunteers or low-skilled staff, and only included a couple of days of training, thus suggesting low cost of programme delivery, there was also evidence that interventions could require high levels of (unplanned) resources, including substantial time inputs from staff employed by public sector agencies [38,42,47]. Implementation challenges were potentially driving up costs, in particular in areas in which prevention had low priority and staff were sceptical of the value of the intervention [44,45]. One study reported that, in order to remove access barriers for parents and families, additional investment was required to fund travelling costs and childcare [55]. A few studies highlighted that the interventions should not replace existing support for vulnerable populations, but be provided alongside professional support [35]. Only one study reported intervention costs, and those were USD 10,000 to 12,000 per child per year [41]. Some papers discussed whether interventions could be effectively provided at low cost [33,43].

Discussion
This review synthesised knowledge about how social support can be mobilised through interventions that seek to improve children's mental health outcomes. It is hoped that this knowledge will be useful for practitioners or researchers who seek to develop, implement or evaluate interventions in this area.

Discussion of main findings across age groups
Our review found that social support was not well-conceptualised in intervention studies, and studies were generally weak in explaining how social support was mobilised and expected to lead to improved mental health for children. Most studies did not specify the types or sources of social support they sought to address or the rationale for doing so. These limitations have been identified previously [19,20]. Studies that did have a more detailed programme theory in relation to social support were describing the process of mobilising social support as complex, dynamic and long-term. They described various components of this process, such as educating children or parents about the benefits of social support, offering repeated opportunities for practising social skills and for experiencing the benefits of positive relationships through reciprocity and trust-building. The importance of such processes has been confirmed in studies which found that relationship satisfaction and reciprocity of relationships are important contributors to improved mental health [65] and reduced loneliness [66]. Some of the identified studies theorised a complex interaction between social support and mental health, in which social support could be a means to positive mental health, as well as the outcome of processes in which aspects of mental health (e.g. self-esteem) were improved, and this led to a capacity to engage further in social relationships. In the field of social neuroscience, underlying cognitive or biochemical processes have been found that seek to explain this bidirectional relationship [67][68][69]. They suggest that certain mental capacities or cognitive abilities are required in order for a person to see the value of, and engage in, social relationships and in collective actions (socalled 'we intentions') [68]. Those are likely to be diminished for people experiencing prolonged lack of social support and loneliness due to changes in the nervous systems and in gene expression [70], which can trigger fear-based responses to situations, thus leading to erosion of trust in relationships and further isolation [67,71]. Our review also found that most interventions specifically targeted families from low socio-economic backgrounds, but studies did refer to potentially different mechanisms between social support and mental health for this population as identified in the literature [22].

Discussion of findings by age groups
Infants (0 to 2 years). Interventions in this category tended to be provided alongside (mental) health services to mothers at risk of stress or mental health problems during the perinatal period. They sought to address information needs alongside other support needs, and to help mothers engage with and improve their relationships with (health) professionals. Authors of these studies expected that increased social support for mothers (and fathers) would improve child social and emotional competence either through social-cognitive (e.g. parental self-efficacy) or stress-buffering mechanisms. Small et al (2011) [72] found a lack of impact of social support interventions in this area, which they explained with their focus on information support (i.e. parenting education) rather than companionship, emotional and appraisal support. Similarly, Milgrom et al (2019) [73] highlight the importance of providing different types of social support at different time-points during pregnancy and after birth. As suggested by a high-quality study in our review [33], professional-like advice was potentially crowding-out feelings of trust and self-worth, and naturally evolving relationships, suggesting therefore the challenge of mobilising social support sustainably. Evaluation challenges prevented us from deriving conclusions about whether social support provided to parents during the perinatal period improved children's mental health.
Children (3 to 9 years). Interventions in this category described the social isolation of families, who had very limited formal support from public institutions such as childcare facilities or schools. Several interventions focused on rebuilding such relationships and transforming them from one based on power imbalance to one that was reciprocal and built on trust. Authors of studies expected that by improving those relationships, parents would start engaging in and enjoying child-centred activities, thereby leveraging social capital for the benefit of their children, which in turn would improve children's long-term wellbeing. Another set of interventions focused instead on social support as a protective factor for improved parenting practices and capacities, which in turn was expected to improve family functioning and contribute to improved child development.
In this review two intervention types had the potential to achieve positive child behaviour. One focused on changing bi-or multi-directional relationships involving families and professionals (and sometimes wider communities). The other focused on parents' behaviour., It has been argued that only the first follows a truly ecological model of shared child responsibility supported by international legislation of child rights [74].
Adolescents (10 to 18 years). Interventions included in this category sought to reduce major risks for vulnerable groups, in particular with regard to school failure and risky life choices. Vulnerabilities of youth related to sexual orientation, mental health, and their exposure to discrimination, violence and abuse. Social support was mobilised by providing opportunities for learning and practising social skills in healthy relationships and safe environments. Developing trust, identity and confidence were important mechanisms for improved mental health. Most interventions focused on the young person's own social support network. The importance of supporting young persons' social networks in order to help them develop skills they require in adulthood has been highlighted as a priority matter in global youth policy [75]. The importance of developing adolescents' social skills and enabling them to improve interpersonal relationships has been identified a central ingredient towards improving their mental health [30,76].

Strengths and limitations
To our knowledge, this is the first review of social support interventions specifically looking at children's mental health. We applied realist review principles thoroughly and consistently throughout the research with the aim of generating findings that can guide theoretical thinking around developing programme theories, logic models, and evaluation designs. As with many psychosocial phenomena, there other concepts closely related to social support (such as social connectedness, social capital, loneliness). Investigating one concept but not others will naturally have limitations. For example, it means that we excluded studies in which interventions mobilised or altered social relationships and improved social skills, but did not specifically investigate this from a social support perspective [77]. As typical for realist reviews, the application of inclusion and exclusion criteria was complex. It was difficult to decide whether studies sufficiently conceptualised or measured social support and children's mental health to justify their inclusion. Whilst we sought to address this challenge by adding an additional screening step, we cannot rule out a certain lack of consistency.

Implications for policy, practice and research
Loneliness and social isolation attract major interest as contributors to poor mental health [78], with young people experiencing loneliness with greatest frequency or intensity of all age groups [79]. Increasing perceived social support, which is considered to be equivalent to reducing loneliness [80,81], might help prevent or reduce mental health problems in young people [82,83]. Few children or young people approach health professionals for help with their mental health problems [84,85] and are instead much more likely to seek help from existing networks of formal or informal supports, such as from teachers and friends [86]. Therefore, interventions seeking to mobilise such networks might have an important role in promoting mental health in this population. However, findings from this review also suggest that, in order for interventions to be effective, they might need to be population-and context-specific, and consider the complex nature of social support. Especially for vulnerable populations who might experience discrimination, lack skills and trust to engage in social relationships, approaches might need to involve changing attitudes towards social support, motivations to engage in social support, and skills to do so. Achieving those changes involves time and resources. As highlighted in a recent review of interventions to reduce loneliness among people with mental health problems [82], it is often unclear whose responsibility it is to invest their time and resources. Social care and community organisations, community (mental) health services and schools are potentially well-placed to actively foster development of informal and formal networks [74,87,88]. However, it also requires policies, strategies and investments that support this kind of systems change. A requirement for a wider roll-out of most interventions includes the knowledge about who should be targeted. Findings from Cacioppo et al (2009) [89] suggest that targeting individuals at the periphery of social networks might have positive knock-on effects for whole communities. Future research and practice developments might be needed to explore how best to identify such children or families at risk of social isolation.
Noticeably, the majority of programme theories in studies identified by our review mobilised parents' social support and focused on improving children's behaviour problems. Less consideration was given to the impact of interventions that mobilise social support to improve child emotional problems, as well as those that mobilise social support networks from the perspective of the child. Additionally, our review only identified one study that included online support. Digital technologies might potentially play important roles in providing social support [90,91]. However, their programme theories are often not detailed in studies highlighting the need for more development work [92].