Peer-facilitated community-based interventions for adolescent health in low- and middle-income countries: A systematic review

Background Adolescents aged 10–19 represent one sixth of the world’s population and have a high burden of morbidity, particularly in low-resource settings. We know little about the potential of community-based peer facilitators to improve adolescent health in such contexts. Methods We did a systematic review of peer-facilitated community-based interventions for adolescent health in low- and middle-income countries (LMICs). We searched databases for randomised controlled trials of interventions featuring peer education, counselling, activism, and/or outreach facilitated by young people aged 10–24. We included trials with outcomes across key areas of adolescent health: infectious and vaccine preventable diseases, undernutrition, HIV/AIDS, sexual and reproductive health, unintentional injuries, violence, physical disorders, mental disorders and substance use. We summarised evidence from these trials narratively. PROSPERO registration: CRD42016039190. Results We found 20 studies (61,014 adolescents). Fourteen studies tested interventions linked to schools or colleges, and 12 had non-peer-facilitated components, e.g. health worker training. Four studies had HIV-related outcomes, but none reported reductions in HIV prevalence or incidence. Nine studies had clinical sexual and reproductive health outcomes, but only one reported a positive effect: a reduction in Herpes Simplex Virus-2 incidence. Three studies had violence-related outcomes, two of which reported reductions in physical violence by school staff and perpetration of physical violence by adolescents. Seven studies had mental health outcomes, four of which reported reductions in depressive symptoms. Finally, we found eight studies on substance use, four of which reported reductions in alcohol consumption and smoking or tobacco use. There were no studies on infectious and vaccine preventable diseases, undernutrition, or injuries. Conclusions There are few trials on the effects of peer-facilitated community-based interventions for adolescent health in LMICs. Existing trials have mixed results, with the most promising evidence supporting work with peer facilitators to improve adolescent mental health and reduce substance use and violence.


Results
We found 20 studies (61,014 adolescents). Fourteen studies tested interventions linked to schools or colleges, and 12 had non-peer-facilitated components, e.g. health worker training. Four studies had HIV-related outcomes, but none reported reductions in HIV prevalence or incidence. Nine studies had clinical sexual and reproductive health outcomes, but only one reported a positive effect: a reduction in Herpes Simplex Virus-2 incidence. Three studies had violence-related outcomes, two of which reported reductions in physical violence by school staff and perpetration of physical violence by adolescents. Seven studies had mental health outcomes, four of which reported reductions in depressive symptoms. Finally, we found eight studies on substance use, four of which reported reductions in alcohol PLOS

Inclusion criteria for the systematic review
We only included randomised controlled trials (RCTs) because these studies have a lower risk of bias compared to quasi-experimental studies. We included trials in which the majority (>50%) of participants were adolescents or participants with a mean or median age of [10][11][12][13][14][15][16][17][18][19]. Trials had to be located in the community (e.g. schools, youth clubs or primary health care centres) because this is where peer-facilitated interventions are commonly located. Trials also had to take place in LMICs (as defined by the World Bank [27]), and test an intervention delivered in whole or part by peer facilitators, defined here as persons or a majority of persons (>50%) with a mean or median age of 10-24 recruited from the group or community meant to benefit from interventions. We included trials of interventions involving peer education where peers sought to increase adolescents' knowledge or influence their attitudes, 'counselling', defined as peers providing support to help adolescents resolve personal or psychological problems, 'activism' involving peer-led campaigns to change health-related policy, and 'outreach' with peers engaging marginalised adolescents [28,29]. We included trials with primary or secondary outcomes relevant to areas of health need outlined in the report of the Lancet Commission on Adolescent Health and Wellbeing [2]: infectious and vaccine preventable diseases, undernutrition, HIV and AIDS, sexual and reproductive health, unintentional injuries, violence, physical disorders, mental disorders and substance use. We deliberately included interventions from across multiple adolescent health areas in order to compare effects across areas. For each area of health need, we included studies with outcomes related to the diseases and risk factors highlighted by the Lancet Commission Report, as well as diseases constituting the 10 main global causes of death or years lived with disability for [10][11][12][13][14][15][16][17][18][19] year olds [2,4]. These outcomes are shown in Table 1. We also included educational and employment marginalisation, which were considered key determinants of adolescent health. We did not include studies that were conducted in underprivileged populations in high-income countries. No date or language restrictions were applied. The review protocol is registered with PROSPERO (CRD42016039190). Our methods did not deviate from those specified in the protocol.

Search strategy
KR-C used customised search strategies (S1 Text) to search for studies that met the inclusion criteria in Medline, Embase, Cochrane Library, CINAHL, African Index Medicus, Web of Science, Psycinfo and ERIC up to 9 th March 2017. The search was later updated to 22 nd June 2018. We identified ongoing studies by contacting adolescent health experts and searching the International Clinical Trials Registry Platform. We found further studies by searching relevant reviews. Fig 1 summarises the study selection process. KR-C or AB screened the title and abstract of each article to identify and exclude those that were irrelevant. KR-C and AB or AP then independently screened the full text of all remaining articles for relevance. Any discrepancies were discussed and resolved by the review team and/or by contacting authors. S2 Text outlines reasons for excluding articles at full text screening. S1 Table describes details of ongoing studies. We used Covidence and EndNote reference manager software to manage articles retrieved by the search [30]. For each study that met the inclusion criteria, KR-C and AB or AP independently extracted data on general study details, trial design, participant characteristics, sample size, intervention, control condition, outcomes and summary measures, for example a risk ratio (RR), odds ratio (OR), or linear regression coefficient (β). We noted whether interventions involved education, counselling, activism and/or outreach strategies. We extracted data from the first outcome assessment post-intervention based on a hierarchy of clinical outcomes first (e.g. HSV-2 participants and personnel blinding, outcome assessment blinding, incomplete outcome data, selective outcome reporting and other bias [31].

Data synthesis
We mapped the evidence using a narrative summary of intervention characteristics by area of health need. Within each area of health need, we also considered how complementary intervention activities, setting, type of facilitator and participant age could influence intervention effects. Although we initially planned to do a statistical meta-analysis, this was not possible because of the wide variation in types of interventions and outcomes.

Results
We found 43 articles that described 20 relevant randomised controlled trials with a total of 61,014 participants at baseline. S2 Table summarises the characteristics of these studies. Six were conducted in low-income countries, seven in lower-middle income countries and seven in upper middle-income countries. Fourteen interventions were linked to schools or a college. Twelve interventions had additional non-peer-facilitated components, for example health worker or teacher training, and dissemination of educational materials. These other non-peerfacilitated components are described in Table 2. Table 2 describes the characteristics of peer-facilitated intervention strategies, including the selection, training and supervision of peers. Interventions were diverse: peer facilitators conducted education, counselling, outreach and activism. Nineteen of the 20 studies featured peer education activities. Peers ran group-based sessions for classmates and other students [32][33][34][35][36], facilitated groups in the community, [37][38][39][40][41] performed street plays or created dramas [37,[42][43][44], ran workshops with parents [45], and distributed educational materials [45,46]. Nine of the 20 studies incorporated peer counselling strategies. These ranged from low intensity approaches where peers encouraged their classmates not to give or accept cigarettes [34], to higher intensity approaches where peers led manualised interpersonal psychotherapy groups [41]. Peer activism was used in five studies to develop and enforce anti-smoking/tobacco policies [34,43], work with community leaders to provide opportunities for adolescents [45] and run a 'student court' to manage school discipline issues [42]. Peer outreach was used in four of the 19 studies. For example, in Thailand, peers used communication skills to convey risk reduction messages to drug users in their social networks [20]. As part of the CERCA (Community-Embedded Reproductive Health Care for Adolescents) intervention in Nicaragua, peers mentored adolescents to help them build decision-making competence related to sexual and reproductive health, and referred and accompanied them to health services when needed [45].

Peer-facilitated strategies
The duration of peer-facilitated components ranged from three weeks [47] to four years [39]. Training duration and intensity ranged from a one hour information session [35] to a four-week programme [46]. Peer facilitators were school students in nine of the 20 studies, and school graduates in six. Five studies did not provide information on the education level of facilitators.
Study quality was variable (Table 3): three studies were at low risk of bias across all seven domains [32,40,42]; 15 did not report methods used for allocation concealment; eight did not report methods for random sequence generation. One study was at high risk of bias because it had a small number of clusters and results were not adjusted for clustering or confounders [37]. In another, schools refused to participate after the baseline survey and it was not clear whether data were missing because of this or for other reasons [48]. Two studies encountered unexpectedly high rates of adolescent out-migration and were forced to change their study design substantially with implications for the statistical power of the study [39,45]. In one study in Nicaragua, loss to follow up was 76%, with important differences between resurveyed adolescents and those lost to follow up [45].

Study outcomes and intervention effects
We did not identify any studies focusing on infectious and vaccine preventable diseases, undernutrition or unintentional injuries. More studies measured outcomes related to sexual and reproductive health (nine studies), substance use (eight studies) and mental disorders (seven studies) than any other area of health need. Below, and in Table 2, we present intervention details and findings by area of health need. HIV and AIDS. Four studies reported HIV/AIDS-related outcomes [20,39,40,44]. All involved a community component and peer education. Two examined the effects of combining peer facilitation with programmes for parents, community stakeholders and health worker training [39,44]. None of the four studies reported a positive effect of the interventions.
Sexual and reproductive health. Only one study [40] found an effect of peer-facilitated interventions on clinical sexual and reproductive health outcomes: Jewkes et al tested the effects of a structured curriculum of peer-facilitated group education on sex and love, contraception and sexually transmitted diseases among adolescent boys and girls in South Africa, and reported a reduction in Herpes Simplex Virus-2 infection (HSV-2) (RR 0.67 CI 0.47-0.97) Peer-facilitated interventions for adolescent health in low-and middle-income countries: A systematic review [40]. However, other studies found negative results: one study from Tanzania reported an increased prevalence of gonorrhoea among young women (RR 1.93 CI 1.01-3.71) following school-based reproductive health education led by teachers followed by scripted dramas by peer educators [44]. Another study from Thailand used a curriculum of group education and role-play sessions to help young men and women reduce their use of metamphetamines and sexual risk-taking, and to communicate with others in their social networks about these risks [20]. The study found an increased incidence rate of gonorrhoea in the intervention group compared to the control group (4.69 per 100 person years vs. 0.43 per 100 person years, p<0.05). Self-reported symptoms of sexually transmitted diseases (STDs) were reduced in two studies from India and Nigeria [37,46]. In Balaji et al.'s Indian study, complaints of vaginal symptoms and penile discharge only decreased significantly in urban areas (OR: 0.49, 95% CI: 0.26-0.93 and OR: 0.36, 95% CI: 0.24-0.55, respectively), where peer facilitators were linked to schools [37]. Peers were also trained and supported within schools in the study by Okonofua et al, which reported a reduction in self-reported symptoms of STIs in Nigeria (OR 0.63 CI 0.43-0.91) [46]. A trial of peer-led after-school life skills training sessions reported an increase in condom use among boys (β 0.217 p = 0.004) in Tanzania. A South African trial of peer-led interpersonal psychotherapy groups to help adolescents learn how to resolve distress and access emotional support also led to girls reporting more condom use among their partners (β 0.21 p = 0.02 [36,41]. Conversely, one trial of peer mentors helping adolescents build competence in making deliberate choices and referring them to health facilities reported reduced condom use (β -2.66 p = 0.039) [45].
Violence. Three studies reported violence-related outcomes, two of which found reductions in violence. Both of these successful interventions involved activities for teachers and adolescents and both used a combination of peer education, counseling and activism strategies. Devries et al evaluated the Good School Toolkit in Ugandan primary schools: students took part in intervention-implementing committees to reduce violence, create dramas and facilitate a student court to handle school discipline issues. They found reductions in past week and past term physical violence perpetrated by school staff, reported by students (past week: OR 0.39 CI 0.25-0.62; past term: OR 0.31 CI 0.18-0.53) [42]. They also reported a reduction in violence from peers, and a reduction in violence by school staff against adolescents who had functional difficulties and/or a disability [49]. Balaji [32]. An evaluation of a multicomponent school-based intervention to improve adolescent health and nutrition in India-judged to be at high risk of bias-measured no effect on BMI [50].
Mental disorders. Interventions for mental disorders were diverse and included peer outreach, counselling and education interventions that addressed determinants of mental health such as violence and substance use. Four of the seven studies with mental health outcomes reported improvements in depressive symptoms [19,36,46,50]. These four interventions were from diverse locations (Uganda, Philippines, India and Thailand) and involved a range of peer-facilitated strategies (education, outreach and counselling). Only one [47] of the four positive studies focused on an actual mental disorder, and reported a reduction in the severity of depression. Three of these four successful interventions were linked to schools or colleges [37,47,51].
Substance use. Four out of eight studies reporting substance use outcomes found positive effects. Interventions reduced alcohol drinking among young men (OR 0.68 CI 0.49-0.94) [40] and the risk of non-smokers becoming regular smokers (OR 2.23 CI 1.20-3.85) [35]. One study in urban schools in India tested Project MYTRI, a multi-component intervention with classroom curricula, a poster campaign and peer-led activism. The study found betweengroup differences in the rate of growth of cigarette smoking (p = 0.05), bidi smoking (p<0.01), and any tobacco use (p = 0.04) among students [48]. Among urban adolescents in India, Balaji et al reported a reduction in use of tobacco, cigarettes and alcohol (OR 0.63 CI 0.45-0.89) [37]. Three [35,37,48] of the four studies reporting positive effects were linked to schools, including two where school students acted as peer facilitators [35,48].
Educational and employment marginalisation. Only two studies measured effects on educational and employment marginalisation [38,51]. In Tanzania, the Young Citizens Programme aimed to develop adolescents' individual and collective efficacy to raise awareness of HIV [38]. One outcome in this trial was academic self-efficacy (e.g. "I have learned how hard work helps me in math"), but there were no improvements in this outcome. The Suubi intervention in Uganda was aimed at AIDS-orphaned adolescents and involved a microfinance intervention, financial education and mentorship by older peers aged 17-23. Evaluation of the programme showed an increase in the number of adolescents saying they planned to go to secondary school and that they were more certain they could accomplish their education goals [51].

Discussion
Our systematic review is the first to summarise results from trials of peer-facilitated interventions for all areas of adolescent health in LMICs: to our knowledge, the only other review of peer-facilitated interventions to assess effects for multiple health outcomes was conducted in 1999 and mainly included studies from high-income countries [28]. We found 20 trials focused on six of the nine areas identified by the Lancet Commission for Adolescent Health and Wellbeing: sexual and reproductive health, HIV/AIDS, physical disorders, mental health, violence, and substance use. There was some evidence that interventions improved mental health and reduced violence and substance use, but the diversity of components and outcomes prevented us from making definitive statements about effectiveness. We found no trials with positive effects on HIV-related outcomes, heterogenous results for physical disorders and sexual and reproductive health outcomes, and no trials on infectious and vaccine preventable diseases, undernutrition, or injuries.
Our review has three main limitations. The diversity of interventions and outcomes prevented us from meta-analysing the data within or across adolescent health areas. It also prevented us from understanding the extent to which facilitator characteristics, other intervention components and locations (e.g. school vs. non-school components) might explain heterogenous results within areas. To remedy this, future studies could provide more accurate descriptions of the content of interventions, and use comparable outcome measures within areas of adolescent health need. Further reviews could also focus on individual adolescent health areas and examine a broader range of study designs and methods.
A second limitation was our inability to assess publication bias. Although we contacted authors for clarifications, many articles screened lacked information about facilitator age, and we may not have identified all eligible studies [43,45]. Risk of bias was variable across studies, with no specific pattern within and across areas.
Finally, several trials only included our outcomes of interest as secondary indicators. For example, some were powered to detect differences in sexual and reproductive health outcomes but also included outcomes related to violence and mental health [39]. Such trials may have been under-powered to detect significant differences between intervention and control arms for secondary indicators, and prone to false positive (Type I errors) due to multiple testing.
In line with previous systematic reviews, we found heterogeneous effects of peer-facilitated interventions on sexual and reproductive health, suggesting that peer facilitation alone is unlikely to be the solution to improving this area of health [15,52,53]. This is unsurprising given the breadth and strength of socio-political factors affecting sexuality and access to services for sexual and reproductive health.
We found more promising evidence for peer-facilitated interventions to improve adolescents' mental health and reduce violence and substance use, but too much heterogeneity in interventions and outcomes to make definitive conclusions. Effects on mental health, violence and substance use have some plausibility: peer-facilitated interventions can strengthen peer networks, increase social support, change social norms and improve school environments [10,54].
Fourteen out of 20 studies in our review examined interventions with a school-or collegebased component, including three out of four studies with positive effects on depressive symptoms, and all positive studies on violence. There are many potential benefits to locating interventions in schools: there may be pre-existing support systems for peer facilitators, and facilitators have a 'captive audience' of participants in a classroom setting [55]. Potential disadvantages of working in schools include the potential for hierarchies between teachers, peer facilitators and participants to hinder communication, a lack of engagement with out-ofschool adolescents, and the risk of entire schools dropping out of the intervention [28]. Previous studies have shown that using peer facilitators rather than teachers to deliver health education does not necessarily make an intervention more effective [28]. This may be because peer facilitation often involves implementing interventions developed by older adults. The benefits of such interventions could be lost if adolescents feel the intervention is no longer relevant or that they cannot relate to peer facilitators. Successful school-based interventions in this review were largely devised by research teams, though half consulted with young people during intervention design or implementation phases [37,42,46,48]. More formalised involvement of adolescents in the development of peer-facilitated interventions is likely to be beneficial [24,28,52].
We identified two peer-facilitated interventions that engaged adolescents in peer leadership roles, and focused on capacity building rather than knowledge transfer [20,38,56]. These interventions had positive outcomes for mental health and self-efficacy (deliberative and communicative self-efficacy and emotional control). Interventions that engage a higher proportion of peer leaders may be more sustainable in populations with high rates of adolescent mobility, where retaining peer facilitators may be challenging. Interventions that engage peer facilitators in mobilising communities of young people have been successful in non-school settings [20,38]. Reaching young people who are not in school is important to ensure equity. Offering them leadership opportunities through participatory interventions might help to achieve this.
Critically, twelve of the studies in this review involved interventions with additional, nonpeer-facilitated components, with evidence of positive effects on mental disorders, violence and substance use. The enthusiasm for multi-component interventions-while challenging from the point of view of attribution-reflects the widespread acceptance that adolescent vulnerabilities are influenced by factors at multiple, interacting socio-ecological levels. Reviews of interventions for the prevention of violence have highlighted that interventions with multiple components that address these multiple layers are more likely to succeed than interventions that only address one [28]. These multi-component interventions require evaluations that theorise and assess the interaction between peer and non-peer-facilitated components, or the environment within which interventions are delivered as complex system [57,58].
In conclusion, peer-facilitated community-based interventions show promise to improve mental health and reduce violence and substance use in LMICs, though further robust studies are needed to strengthen the evidence base. Future research should focus on theorising and assessing the contribution of peer-facilitated interventions and their interactions with nonpeer-facilitated components in these areas of adolescent health.