A systematic review of the literature on community-level child protection in low- and middle-income countries

ABSTRACT This systematic review synthesizes the current state of literature on community-level child protection in LMICs. The aim of the review is to present available evidence and effective strategies that implementing agencies can use to support community-level structures, practices, resources and processes. Multiple database searches were conducted, using search terms to capture community-level approaches to child protection in LMICs. The search identified 1,549 unique published articles and 1,745 grey literature resources. After a screening process based on pre-defined inclusion and exclusion criteria, 38 published articles and 204 grey literature resources were analyzed. The review highlights (1) a practice-research gap related to community-level approaches generally and more specifically in humanitarian settings; (2) the important role of different socio-ecological levels when implementing community-level interventions; and (3) a number of recommended strategies that implementing agencies can adopt in their work, such as ensuring ownership at community level increasing the potential of sustainability of an approach; linking with existing processes and structures; being inclusive in the approach; and carefully negotiating possible tension between traditional mechanisms and rights-based frameworks. Further studies, with a focus on outcomes for children, families and communities, are needed to broaden the current evidence-base and research on the applicability and effectiveness of community-level approaches in humanitarian settings is recommended to steer implementation in the sector.


Introduction
Violence against children remains a major global problem with devastating consequences (Norman et al., 2012;Pinheiro, 2006;Stark & Landis, 2016). Violence happens in families, educational settings, care institutions, places of work and in the community (Pinheiro, 2006;Wessells & Kostelny, 2013) and leads to long-term adverse outcomes including mental and physical health problems, under-performance at school, substance abuse, suicidal behavior, reproductive and sexual ill-health and involvement in violence and delinquency (Gilbert et al., 2009;Norman et al., 2012;Pinheiro, 2006).
Violence against children is a fundamental concern during humanitarian emergencies (United Nations, 1996;Rubenstein & Stark, 2017;Silverman & La Greca, 2002). While there is agreement that the first responsibility for the protection of children lies with state authorities, humanitarian aid and international development organizations are direct stakeholders when responding to protection risks .
In 2009, an inter-agency review of evidence on community-based child protection mechanisms was conducted, motivated by the large number of practitioners implementing these mechanisms without a strong supportive evidence base (Wessells, 2009). The review shows that community ownership is a key element, leading to sustainable child protection outcomes. Community-level action is well placed to support children facing risks; it is likely to reach large groups of children; and tends to be low cost and therefore more sustainable (Wessells, 2009). For most of the programs reviewed in the study, ownership remains fully or predominantly with external agencies (Wessells, 2009).
It has been over a decade since the initial review was conducted. Since then, the Inter-Agency Learning Initiative on Community-Based Child Protection Mechanisms and Systems conducted a number of studies and developed a Guidance and Toolkit for supporting a community-led approach to child protection (Child Resilience Alliance, 2018); the INSPIRE seven evidence-based strategies for ending violence against children were launched (WHO, 2016); and the Alliance for Child Protection in Humanitarian Action (2020) developed a field guide to strengthen community-level work. In the 2019 edition of the Minimum Standards for Child Protection in Humanitarian Action, a standard on community-level approaches was included, recognizing the significant role communities play in preventing and responding to risks children face (The Alliance for Child Protection in Humanitarian Action, 2019b). The term 'community-level' includes community-based mechanisms as described in the 2009 review (Wessells, 2009) and bottom-up initiatives commonly labeled as community-led and communitydriven (The Alliance for Child Protection in Humanitarian Action, 2019a, 2019b. This paper presents a synthesis of existing literature on community-level approaches to child protection in LMICs and effective strategies that humanitarian agencies can adopt to engage appropriately and effectively with community-level structures, practices, resources and processes.

Methods
To systematically review the literature, the steps outlined by Khan et al. (2003) were followed. The protocol followed for this review was registered on PROSPERO (CRD42018096954).

Search strategy
Two search rounds were completed to locate relevant studies. First, a search was conducted to identify published articles, using the following databases: PubMed, PsycINFO, Web of Science, Sociological Abstracts and AnthroSource. Second, grey literature was searched through websites of networks relevant to the topic of the review. Finally, a request to share relevant resources was sent to members of the interagency Community-Level Child Protection Task Force (CCP TF) as part of the Alliance for Child Protection in Humanitarian Action. The original search was conducted in 2017, and an updated search of the published literature databases was conducted using the same search strategies in January 2022.
Search terms were defined in four categories: (1) LMICs following the World Bank Country and Lending Groups (1989Groups ( , 2003Groups ( and 2017. These were combined with (2) child protection search terms; (3) population search terms; and (4) community search terms.

Inclusion criteria
For inclusion in this review, studies had to (a) focus on LMICs defined by the World Bank as those with a gross national income per capita, calculated using the World Bank Atlas methods, of $12,695 or less (The World Bank, n.d..); (b) focus on community-level work on prevention and response to violence, abuse, neglect and exploitation; (c) be published in peer review journals including books and dissertations or be guidelines, manuals, training materials, evaluations and reports when identified in the grey literature search; (d) be written in English; and (e) no date restrictions were applied.

Screening
A total of 1,549 published articles were identified from the specified databases. Reviewing these involved a four-stage screening process. First, reviewers (AG, RE) screened for duplicates using bibliographic management software and hand search. A total of 638 duplicates were identified, leaving 911 articles to be considered for further review. Citations and abstracts of these articles were exported to Rayyan software and to Microsoft Excel sheets when databases were not compatible with Rayyan. Second, two reviewers (AG, RE) screened publications by title and abstract. A total of 751 articles were excluded at this stage for not meeting the inclusion criteria. Third, the full text of included articles was screened. Fifteen articles were not available in full text and of the remaining articles, 112 were excluded for not meeting the inclusion criteria. As per protocol, these studies were cross-referenced as a fourth step. The articles identified through cross-referencing and published articles identified through the grey literature search, required to go through the same screening process as has been described. A final dataset of 38 published articles was included for analysis.
The initial search for grey literature resulted in 1,745 resources. Reviewing these files started with a combined title/abstract and full-text screening as breaking these two steps up was not feasible due to the variety of identified documents. After removal of duplicates and exclusion based on the criteria listed in the protocol, 204 documents were entered in a Microsoft Excel overview for analysis and results are included in the paper. A flowchart of the search and selection process is added in figure 1.
Ten percent of included and 10% of excluded published articles were reviewed by two different coders to ensure consistency between reviewers. In cases of uncertainty in the screening process, the lead reviewer was consulted for a decision (RE).

Data extraction and analysis
The primary research question used in the review was: How can external actors engage appropriately and effectively with community-level groups, structures, systems and processes that are concerned with child protection? A Microsoft Excel data extraction tool was developed and used by the team to chart information. Three reviewers extracted information from published articles (MR, AG, RE) and an additional team of reviewers supported the extraction of relevant data from grey literature (including CV and GO). One additional reviewer extracted information from documents determined to be training materials. Information was extracted on topics included in primary and secondary review questions such as external actor engagement with community-level structures; risks effectively addressed by community structures; inclusivity of community participation and sustainability. In the data-extraction process, a check was done between reviewers (MR, AG, RE) to ensure consistency. Ten percent of the information extracted from grey literature and 20% of the information extracted from published articles were cross-checked. Three reviewers (MR, AG, RE) then analyzed the information extracted from published articles and two additional reviewers joined the analysis of information extracted from the grey literature dataset (CV, GO). During the review, the following definition for humanitarian settings was used: 'a serious disruption of the functioning of a community or a society involving widespread human, material, economic or environmental losses and impacts that exceeds the ability of the affected community or society to cope using its own resources and therefore requires urgent action' (Sphere Association, 2018, p. 5). The term can refer to slow-and rapid-onset situations, rural and urban environments and complex political emergencies in all countries (Sphere Association, 2018, p. 5). Development settings were defined as non-humanitarian settings, and following the UN Development Programme (UNDP)'s Human Development Index (HDI) (UNDP, 2019).
With a search focus on LMIC, the majority of articles focus on development contexts (n = 31) with considerably less studies focusing on humanitarian settings (n = 7). These seven articles detail: social reintegration of young mothers formerly associated with armed groups in Liberia, Sierra Leone and Uganda (McKay et al., 2011); levels of hope amongst children in Palestinian refugee camps in Lebanon ; functioning of existing community-based mechanisms in Aceh, Indonesia ; functions of child-friendly spaces (Wessells & Kostelny, 2013); community-based child protection mechanisms as the cornerstone of prevention efforts for war-affected children (Wessells, 2016); the role and experience of local faith leaders in promoting child protection in flood affected communities in Malawi (Eyber et al., 2018); and the results of an evaluation of school-based peace education and a community-based intervention to change harmful social norms and practices in Afghanistan (Corboz et al., 2019). In contrast to the published literature, an almost equal number of included grey literature documents refer to development settings (n = 82) compared to humanitarian settings (n = 84). Table 1 provides a summary of characteristics of the published articles that were reviewed.

Communities' internal systems of care
The important role of families and existing community structures in the protection of children is presented in various studies. The quote 'it takes a village to raise a child' refers to societies in which primary care for children does not lay solely with caregivers but with families, larger kinship groups and communities (Seymour, 2013). A study in Papua New Guinea illustrates a context where ill health and neglect of children are collective concerns and are dealt with as such (Ikupu & Glover, 2004). Here, care for children is structured around the roles of older children, important adults and members of the extended family.
Children grow up and are protected within the environment of the community. The protective roles of the community, based on cultural norms and complex social relationships, are undermined when Eurocentric models are adopted (Ikupu & Glover, 2004). The importance of families is also noted by Roby and Shaw (2018) who recognize the significant position of grandparents as a safety net around children. Young people are also seen as actors in raising awareness on and assisting children at risk of harm (De Lange & Mitchell, 2014;Roche & Flynn, 2021;Stark et al., 2012).
The important role of cultural and traditional leaders, such as chiefs and clan leaders, is demonstrated in the review. Leaders raise awareness on harms to children and norms that predispose children to violence; they lend legitimacy to child protection efforts; and have the ability to encourage and take action to mobilize communities from within. A study in five West African countries shows that children and families rarely seek the involvement of formal protection services but that decision-making lays with chiefs (Krueger et al., 2014). The rapid ethnography conducted in Sierra Leone and in Kenya by Wessells et al. (2012), indicate that even in cases of rape, people prefer to use traditional processes rather than reporting to state authorities. Local families frequently resolve problems without going to Chiefs. The family of a sexually abused or impregnated girl 'compromises' with the family of the perpetrator. This kind of non-leader-based process is key in many indigenous systems of child protection . Hutchinson et al. (2015), Robinson and Hanmer (2014) and Eyber et al. (2018) report on the role of religion in child protection. Imams, for example, are described as 'counsellors who facilitate communication and give guidance, rather than enforcing change or interfering uninvited in family life' (Hutchinson et al., 2015, p. 402). Religious actors use religious teachings to raise awareness on social issues within the community Robinson & Hanmer, 2014). In addition to sharing child protection related messages in religious services, Robinson and Hanmer (2014) identify the following four additional roles of religion in the protection of children. First, the use and promotion of religious texts that respect the dignity of children; second, explain misconceptions of religious texts used to violate children's protection; third, condemn practices and traditions that put children at risk; and finally, speak out against those that use religion to justify violence against children. Robinson and Hanmer (2014) acknowledge the positive influence religion and spirituality can have on the daily lives of children, such as in enhancing resilience and mitigating the effects of violence.

External agency engagement with community structures, practices, resources and processes
Various studies report on the engagement of external agencies with communities regarding the protection of children. Muriuki et al. (2013) and Stuer et al. (2012) describe a community-based bereavement association, Idris, that collects fees and provides support services in the community. The role of the agency is to build capacity and provide technical assistance to a national organization. Subsequently, the national organization provides technical support and resources to the Idris, which provides direct support to targeted families (Muriuki et al., 2013). McKay et al. (2011) describe Participatory Action Research (PAR) with young mothers, formerly associated with armed groups. The mothers experience social isolation  Children received or accessed services at a higher level than children with no social worker and CHW support.
4 African countries

Literature review and interviews
Impact of cultural factors on child protection at community level.
Understanding of cultural practices is essential for the success of child protection responses in communities. O'Leary et al.

Refugees in Lebanon
Administering hope scale and case studies Impact of a communitybased child protection project.
Children who received a social work intervention were shown to have significantly improved hope. Petersen et al. The intervention showed increased knowledge of, and willingness to discuss, child sexual abuse.

Robinson and
Hanmer (2014) Not specified A discussion paper Working with religious communities in addressing child abuse and neglect.
Partnering with religious communities is essential but the communities can also use religion to justify harms.

Roby and
Shaw (2018) Uganda Questionnaires A community-based program for families raising vulnerable children.
Households' need decreased significantly. Children's senses of belonging appeared promising.

Roche and
Flynn (2021) Philippines Exploratory, interpretive case study design.
Explores local child protection practices, actors and functions.
Indonesia Qualitative study Functioning of existing community-based mechanisms.
Child protection systems that include community level learning are better-able to promote the health and safety of children. ) and stigmatization, resulting in community reintegration challenges and the rejection of their children. Through PAR, the young mothers formed a group in which they identify their problems, develop and implement actions to address these, and evaluate their actions (McKay et al., 2011, p. 112). In this example, agency staff spent significant time in organizing the groups of mothers to identify capacity building needs and provide training with the ultimate purpose of shifting leadership to the mothers over time (McKay et al., 2011). A notable, recurrent problem was that staff tended to direct or 'facilitate excessively', so they had to be encouraged to step back to create space for the young mothers to take the key decisions and guide their own actions. Reid et al. (2014) outline an approach that focusses on social mobilization and awareness raising on child sexual abuse in Trinidad and Tobago, following a PAR approach. After an initial community rejection of the agency's centralized approach, a decentralization step was taken with the purpose to reach more people through content shaped by the community.
While some agencies use a strength based approach when engaging with communities (Cook et al., 2017; L. Thompson et al., 2019;Wessells, 2016), deficiencies of the formal system and family support may be reasons for agencies to divert to a community approach (Lei et al., 2019). Thurman et al. (2008) point to a risk of targeted agency support to vulnerable groups in a community which may unintentionally lessen the level of support they receive from the community. Several factors are identified as contributing to positive effect of external engagement, such as the development of links between agencies and existing community structures; time spent in the community to gain trust and confidence in the collaboration; development of relationships with interested individuals and voluntary groups; and advocacy with local authorities to address causes of abuse. Lack of linkages between formal and informal systems, and top-down approaches are seen as preventing positive engagement. Only two reports mention external resources as an important element of agency support, whereas mobilization of internal resources right from the start is seen as the preferred strategy in other reports.
When reported, the duration of external engagement of agencies with communitylevel groups on child protection is relatively long, varying from 3 to 9 years in three studies (Krueger et al., 2014;McKay et al., 2011;Shadowen et al., 2017) and 2 to 10 years in 27 grey literature documents. Only two documents in the grey literature report on external engagement below the duration of 1 year.

Child protection issues effectively addressed by community-level structures, practices, resources and processes
The review provides limited insight into which child protection risks are (most) effectively addressed at community level. Over 30 types of different child protection risks are reported in the included grey literature covering a broad range of threats to children's protection, development and well-being. The most frequently cited addressed risks were as follows: child labor, child marriage, trafficking and corporal punishment. Project evaluations are mostly describing process and output indicators rather than outcomes at the child, family and community level.
Sexual abuse, violence and exploitation are frequently reported as especially challenging to address. Shame and stigma are seen as factors that prevent survivors from seeking help (McKay et al., 2011) and from reporting, which is specifically the case when the violations occur in the family or when the perpetrator is otherwise known to the child survivor. Perceptions of what constitutes violence in need of a response from a system perspective are likely to influence community members' willingness to disclose harms to children (Child et al., 2014). Existing socio-cultural traditions and taboos may discourage open discussion of sexual matters, maintaining secrecy around such violations. Hutchinson et al. (2015) noted distrust amongst religious leaders towards external child protection actors and unwillingness to work with them on particularly sensitive issues such as sexual abuse. Support from community leadership can be seen as a supporting factor for community members to assist survivors of sexual abuse (Kisanga et al., 2011).
Some issues, such as domestic violence, are considered within the realm of family responsibility, and social barriers and shame make it difficult for people to seek help from within the community . Community structures face challenges as well when addressing risks rooted in traditional norms and cultural practices, such as corporal punishment and early marriage.
Besides these challenges, successful examples are shared where community processes were key in addressing child protection issues. Stark et al. (2012) describe the role of peers as a central source of psychosocial support. Improved psychosocial well-being, decreased involvement in transactional sex and increased economic independence amongst young mothers formerly associated with armed groups is described by McKay et al. (2011) as a result of community-level programming in West and East Africa. A reduction in abuse, mistreatment, stigma and discrimination are identified outcomes of a care and protection program for orphans and vulnerable children implemented by traditional burial societies in Ethiopia (Stuer et al., 2012). The involvement of traditional burial societies in the protection of children affected by HIV and AIDS led to increased awareness and enlarged community support for those in need (Stuer et al., 2012). Wessells (2015) describes how teenage pregnancy was identified by Sierra Leonean communities as a priority risk that was addressed effectively. A community-led approach was used which emphasized dialogue, reflection, community decision-making and action. The approach also facilitated linkages between communities and the formal system and enabled the community to begin to question other related harms, such as early marriage. Other risks reported to be addressed with positive outcome by community structures in grey literature include child labor via parents and children's groups, HIV/AIDS via peer groups, unaccompanied and separated children via kinship care, and impacts of natural disaster.
The review also highlights cultural practices that address child protection risks with positive effect. In Sierra Leone, the shame and exclusion of girls previously associated with armed forces and groups returning to their communities was addressed by traditional cleansing ceremonies that both addressed the shame the girls experienced and led families to accept girls back as 'pure' again (Robinson & Hanmer, 2014). A traditional healing and justice/reconciliation practice addressed traumatic stress reactions, such as bedwetting and nightmares, in children reintegrated into their communities in northern Uganda. Participation in these ceremonies also demonstrated the community's desire for acceptance and forgiveness of reintegrating children (Robinson & Hanmer, 2014).
Though not linked specifically to the ability to effectively address particular child protection risks, the review found several overall factors that are associated with better outcomes than others. Community-level work is identified to achieve better outcomes when building on existing processes and groups; when allowing members to pool their technical knowledge, skills and professional networks; when mandates, roles and responsibilities are clear; when they can negotiate possible tensions between traditional mechanisms and rights-based frameworks that favor formal justice mechanisms; and when regular monitoring, ongoing capacity development and mentoring support are received. Interventions should be implemented in line with broader community strategies through which protective social norms are developed (Petersen et al., 2005). Positive outcomes were generally linked to higher levels of community ownership, when the community itself identifies and addresses challenges and achieves collective goals (McKay et al., 2011;Muriuki et al., 2013;Wessells, 2015;Wessells & Kostelny, 2013), and when the used approach was inclusive (McKay et al., 2011;Wessells, 2015). When programs create space for meaningful participation and collaboration with participants, they can be empowered to identify solutions to risks impacting children in their families and communities (Lackovich Van Gorp, 2017). Cultural beliefs and practices determine people's perceptions on child abuse and protection. These perceptions therefore play an important role in the success of a child protection intervention (Murove et al., 2010).
Several resources cited constraints to positive effects of community structures and processes in addressing child protection risks. Among reported limitations are the lack of follow-up to reported child protection cases; the inability to address the scale of identified problems; lack of appreciation for and recognition of community volunteers; minimal access to funds; and lack of similarity between formal system priorities and community concepts of child well-being and protection.
While not reported extensively, the review suggests that community structures and processes are particularly successful in prevention through awareness-raising and community dialogue where formal systems and top-down approaches are better suited for response.

Meaningful participation of children
The involvement of children in community structures and processes strengthens their ability to more effectively understand and respond to child protection risks. McKay et al. (2011) demonstrate that the meaning of the concept of 'meaningful participation' varies among different child protection agencies. In some cases, it is at the 'low-end' of participation, focused on providing children with information, involving them in consultation or membership of a community-level structure or gathering information through agency-directed focus groups or questionnaires. Wessells (2015) notes that good facilitation can concurrently support community planning processes which are inclusive and promote children's agency to help guide community child protection action. McKay et al. (2011) mention the importance of training and mobilizing adults to respect children's views and support meaningful child participation in the community in age-appropriate ways, including through the opportunity to engage in decisionmaking processes. Skovdal et al. (2008) also highlight the importance of including children in decision-making to ensure the responsiveness and greater results of community-level approaches. In Sri Lanka, children who experience neglect are identified as agents of change in delivering interventions to their mothers (Rathnayake et al., 2021).

Sustainability
The sustainability of an intervention is often founded on voluntary, community-led action with strong motivation and reliance on local resources (Wessells, 2015). Community-level groups have local ownership when communities see them as their own and when they are helping fulfil their responsibility to support children (Wessells, 2016). Muriuki et al. (2013) note that communities that take ownership of child protection issues are more likely to have better scope, depth and coordination of services and be sustainable. Although investing in community mobilization may initially cost more and take longer to show results than when direct services are provided, not doing so will prolong dependency on external agencies (Thurman et al., 2008). Agency-led community-level groups usually only have low-to-moderate levels of community ownership as these are often driven by external technical experts who control the resources, power and decision-making, rather than communities themselves (Wessells, 2016). Local informal structures and organizations (CBOs) are well placed to identify and target children at risk (Kidman & Heymann, 2009) and can provide continued support after an intervention has finished (Muriuki et al., 2013).

Discussion
The aim of the present study was to synthesize the existing literature on community-level approaches to child protection in LMIC and to identify effective strategies that humanitarian agencies can adopt to engage appropriately and effectively with community-level structures, practices, resources and processes.
The main findings of the review include (1) a practice-research gap related to community-level interventions generally and more specifically in the humanitarian sector, with a limited number of evaluations focusing on outcomes for children, families and communities; (2) the important role of the different socio-ecological levels when implementing community-level interventions; and (3) several recommended strategies that implementing agencies can adopt to engage with communities, such as ensuring ownership at community level increasing the potential of sustainability of an approach; linking with existing processes and structures; being inclusive in the approach; and carefully negotiating possible tension between traditional mechanisms and rights-based frameworks.

Practice-research gap
There appears to be a practice-research gap related to community-level child protection interventions overall and more specifically in the humanitarian sector. The evidence base in the child protection sector and on community-level child protection approaches needs strengthening (The Alliance for Child Protection in Humanitarian Action, 2018; Thompson, 2015;Wessells, 2015). Despite the growing number of studies that focus on work with communities on child protection, the current state of the literature provides limited information about effective approaches and mainly focusses on process and output indicators rather than evaluation of outcomes at child, family and community level. A significant number of published papers focus on development contexts (78%) instead of humanitarian settings (19%). Interestingly, this division is nearly equal within grey literature. This demonstrates that community-level approaches are implemented in humanitarian contexts, but publications on evaluations are lagging. Important to note is the increasingly blurred humanitarian-development line, recognizing that working in 'silos' of humanitarian and development sectors may not be appropriate in protracted and repeated crisis (Barakat & Milton, 2020; Center on International Cooperation, 2019; Kaga & Nakache, 2021). Even though the duration of external engagement with community structures is not mentioned in all included resources, only two grey literature reports mention a duration of engagement below 1 year. All other reports extend duration of engagement from 1 year and above. In the three published papers that mention duration of engagement, the reported duration is between 3 and 9 years. Following the humanitarian program cycle (Inter-Agency Standing Committee IASC, 2015), it is likely that community-level work focusses on protracted crisis and development contexts rather than sudden onset emergencies of which no examples were presented in the included literature.

Socio-ecological framework
The review highlights the important role of the different socio-ecological levels when implementing community-level interventions. It is widely accepted that child protection prevention and response must be looked at beyond the influence of caregivers, including the broader neighborhood, community and cultural contexts that are outlined in Bronfenbrenner's ecological model (The Alliance for Child Protection in Humanitarian Action, 2019b; Ben-David & Nel, 2013; Daro & Dodge, 2009). A specific focus on the social ecology is needed in humanitarian settings where needs may have changed and the supportive roles of communitylevel structures, practices, resources and processes may have been affected (Vindevogel et al., 2012). The literature included in the review recognizes the significant role extended families and communities play in the care for children (Ikupu & Glover, 2004;Roby & Shaw, 2018;Seymour, 2013). Child protection issues are perceived as collective concerns and are responded to as such. An important role is also attributed to cultural and traditional leaders in raising awareness and validating actions to prevent and respond to maltreatment (Krueger et al., 2014;Wessells et al., 2012). Families and children regularly favor involvement of community leaders over formal child protection services. Religion and religious actors can play a meaningful role in the protection of children by providing guidance; raising awareness; and condemning practices in relation to harm to children and endorsing protective practices (Eyber et al., 2018;Hutchinson et al., 2015;Robinson & Hanmer, 2014).
The significant role ascribed to families, communities, local leaders and religion, aligns with the localization agenda established during the World Humanitarian Summit in 2016 (Schmalenbach et al., 2019). During the summit, commitments were endorsed outlining the required changes that enable local and national actors to take on a leading role in humanitarian action (Charter for Change, 2019; OCHA, 2019; Schmalenbach et al., 2019). Ongoing dialogue and equal partnership between international, national and local actors allow for a process in which affected people are at the center of a response (Grand Bargain Participation Revolution work stream, 2017; CHS Alliance, 2018). Based on the typology of approaches of external agency engagement with communities (Benham, 2008), Wessells (2009Wessells ( , 2015 reflects on community engagement and ownership of community-level child protection processes and concludes that most frequently, external agencies lead a response and communities are being invited to participate in the action. Continuous engagement in and learning from the global localization agenda and concrete practices can be beneficial for the child protection sector taking steps to increase, where relevant, community ownership of protection action in humanitarian responses.

Recommended implementation strategies
A number of strategies are recommended in the published and grey literature that can be used by implementing agencies when engaging with community-level structures, practices, resources and processes, i.e. (a) develop links with existing processes and structures; (b) be inclusive in the approach; (c) carefully negotiate possible tension between traditional mechanisms and rights-based frameworks; (d) advocate with local authorities to address causes of violence; and (e) build on internal resources from the start. The role of external agencies varies from setting up and strengthening community-level child protection groups, to building capacity and providing technical assistance. Humanitarian agencies may prefer top-down approaches, and these can be effective in contexts such as rapid onset emergencies. However, such approaches limit full participation of communities in child protection initiatives (Kostelny et al., 2020). The review indicates that community ownership is an important element for effective and sustainable communitylevel work (McKay et al., 2011;Muriuki et al., 2013;Wessells, 2015;Wessells & Kostelny, 2013). Ownership is established when the community identifies and addresses its own challenges and achieves collective goals. It is the role of agencies to allow space for that. When community processes are inclusive, and when children are included in decisionmaking, the responsiveness and results of community-level approaches increase.

Limitations
This systematic review has some limitations that should be considered. First, the review included English language literature only. Second, in contrast with the protocol, the reviewers did not search individual websites of members of the Child Protection Area of Responsibility and Alliance for Child Protection in Humanitarian Action. This decision was taken because of time constraints, the large amount of resources already identified, the general request to members of the CCP TF to share resources and the diminishing identification of unique resources. The authors consider minimal effect of this deviation. Third, due to the limited number of studies available, the diversity of study types and challenges in using study quality to present results, a decision was taken to include all studies to find promising strategies that agencies can adopt to support community-level approaches. Last, researchers published results mainly on process and output indicators and rarely on outcomes at child, family and community level. The review unearthed the paucity of research on community-level approaches.

Conclusion
This systematic review on community-level approaches to child protection demonstrated a practice-research gap; highlighted the importance of different socio-ecological levels when working with communities; and recommended a number of strategies that implementing agencies can use in their work with community-level structures, practices, resources and processes. For example, ensuring ownership at community level increasing the potential of sustainability of an approach; linking with existing processes and structures; being inclusive in the approach; and carefully negotiating possible tension between traditional mechanisms and rights-based frameworks. Although grey literature shows increasing attention to -and use of -community-level approaches, including in humanitarian settings, the limited number of published evaluations prevents us from drawing conclusions on effective approaches. More studies, with a specific focus on outcomes for children, families and communities, are needed to broaden the current evidence. Topics such as community self-mobilization around children's issues, community ownership and inclusivity of community-level approaches should be reviewed with regard to effectiveness and sustainability. Furthermore, research on applicability and effectiveness of the use of community-level approaches in humanitarian settings is urgently needed. This information will be essential to steer implementation in the sector.