Interventions for reducing violence against children in low‐ and middle‐income countries: An evidence and gap map

Abstract Background More than half of the children in the world experience some form of interpersonal violence every year. As compared with high‐income countries, policy responses in low‐ and middle‐income countries (LMICs) are limited due to resource constraints and paucity of evidence for effective interventions to reduce violence against children in their own contexts, amongst other factors. Objectives The aim of this evidence and gap map (EGM) is to provide an overview of the existing evidence available and to identify gaps in the evidence base on the effectiveness of interventions to reduce violence against children in LMICs. This report covers evidence published in English; a follow‐up study is under preparation focusing on evidence in five additional languages—Arabic, Chinese, French, Portuguese and Spanish. Methods The intervention‐outcome framework for this EGM is based on INSPIRE—Seven Strategies for Ending Violence against Children, published by WHO and other partners in 2016. The seven strategies include implementation and enforcement of laws; norms and values, safe environment; parent–child and caregiver support; income and economic strengthening; response and support services; education and life skills. The search included both academic and grey literature available online. We included impact evaluations and systematic reviews that assessed the effectiveness of interventions to reduce interpersonal violence against children (0–18 years) in LMICs (World Bank, 2018b). Interventions targeting subpopulation of parents, teachers and caregivers of 0–18 years’ age group were also included. A critical appraisal of all included studies was carried out using standardised tools. Results The map includes 152 studies published in English of which 55 are systematic reviews and 97 are impact evaluations. Most studies in the map are from Sub‐Saharan Africa. Education and life skills are the most widely populated intervention area of the map followed by income and economic strengthening interventions. Very few studies measure impact on economic and social outcomes, and few conduct cost‐analysis. Conclusion More studies focusing on low‐income and fragile and conflict‐affected settings (FCS) and studying and reporting on cost‐analysis are required to address gaps in the evidence. Most interventions covered in the literature focused on addressing a wide range of forms of violence and harm, which limited understanding of how and for whom the interventions work in a given context, for specific forms of violence. More impact evaluation studies are required that assess specific forms of violence, gendered effects of interventions and on diverse social groups in a given context, utilising mixed methods.


Results:
The map includes 152 studies published in English of which 55 are systematic reviews and 97 are impact evaluations. Most studies in the map are from Sub-Saharan Africa. Education and life skills are the most widely populated intervention area of the map followed by income and economic strengthening interventions. Very few studies measure impact on economic and social outcomes, and few conduct cost-analysis.
Conclusion: More studies focusing on low-income and fragile and conflict-affected settings (FCS) and studying and reporting on cost-analysis are required to address gaps in the evidence. Most interventions covered in the literature focused on addressing a wide range of forms of violence and harm, which limited understanding of how and for whom the interventions work in a given context, for specific forms of violence. More impact evaluation studies are required that assess specific forms of violence, gendered effects of interventions and on diverse social groups in a given context, utilising mixed methods.
1 | PLAIN LANGUAGE SUMMARY

| The extent of evidence on interventions to prevent violence against children (VAC) is unevenly distributed both geographically and by intervention
This Campbell-UNICEF evidence and gap map (EGM) includes interventions and outcomes, showing areas of evidence concentration that can be used to prepare evidence summaries to inform policy decisions, as well as identifying gaps in the evidence base which might benefit from a systematic review, research synthesis or additional impact evaluations.

| What is this EGM about?
The EGM provides a visual and interactive display of completed and ongoing studies structured around interventions and outcomes mapped in the INSPIRE framework. The framework includes these seven strategies: • Implementation and enforcement of laws This EGM includes studies on all types of VAC, that is, physical violence, sexual violence and emotional violence.
The report presents findings on interventions addressing specific forms of violence, including corporal punishment, peer violence including bullying, and intimate partner violence.

| What studies are included?
The map includes 152 studies: 55 systematic reviews and 97 impact evaluations.
Of the included systematic reviews, parent, child and caregiver support (21), norms and values (20), and response and support services (20) were the most commonly studied interventions.
Many of the systematic reviews were assessed to have methodological limitations. There are large numbers of reviews rated as being of low and medium confidence, particularly those related to parent, child and caregiver support and to norms and values interventions.
1.5 | What do the findings of the map mean? Impact evaluations and systematic reviews of interventions for reducing VAC have increased over the years, but there remain limitations that need to be addressed in research investment priorities and future studies. The evidence base is concentrated in Sub-Saharan Africa and South Asia, and select countries within these regions.
South Africa, Ethiopia and India are the most represented countries.
There should be more studies from low-income and conflictaffected settings, and more cost-analysis studies. Studies focusing on interventions linked to specific forms of violence, rather than multiple or unspecified forms of violence, could strengthen understanding of intervention effectiveness.
Overall, the EGM findings suggest the need to ensure increased investment in research to assess effectiveness of interventions, with specific attention to addressing thematic and geographical gaps.

| How up-to-date is this EGM?
The authors searched for studies published up to December 2019.

| Background
More than 1 billion children-over half the children in the world-report having experienced some form of violence in a previous year (Hillis, Mercy, Amobi, & Kress, 2016). VAC includes all forms of violence experienced by children aged 18 years and below, whether perpetrated by parents or other caregivers, peers, romantic partners, or strangers (WHO, 2018). As defined by UNICEF, the scope of violence includes, "all forms of physical or mental violence, injury or abuse, neglect or negligent treatment, maltreatment or exploitation, including sexual abuse" The Global Partnership to End Violence against Children, also launched in 2016, serves as a global platform aimed at "ending violence against children in every country, every community and every family" (Know Violence in Childhood, 2017) and advocates broadly for the use of INSPIRE to accelerate the implementation of violence prevention interventions. These developments, along with significant global commitments articulated in the sustainable development goals (SDGs), have provided greater impetus for global, regional and national level actions to end violence.
Although considerable research on VAC in high-income countries (HICs) is available, the same is not true for low-and middle-income countries (LMICs) as defined by World Bank country classification (World Bank, 2018b). The mapping of available evidence and especially the evidence required on effectiveness of interventions to reduce VAC is a priority area for policy and practice in LMICs (UNICEF, 2018b). This report summarises findings of phase 1 from an EGM of evidence published in English commissioned by UNICEF Office of Research-Innocenti and undertaken by the Campbell Collaboration. Phase 2 will include evidence available in five languages; Arabic, Chinese, French, Portuguese and Spanish.

| Purpose
The purpose of the EGM is to provide an overview of the state of evidence, inform policy and programming and identify important gaps in evidence where the evidence base might benefit from systematic reviews, another type of research synthesis or new impact evaluation. The construction of an EGM is the first step towards building an evidence architecture to end VAC (White, 2019). The EGM will contribute to broadening the included interventions and outcomes in the INSPIRE framework to better reflect the state of evidence on VAC, and independently provide an updated overview of knowledge and evidence gaps for the field.

| Objective
The objective of this map is to provide an overview of the existing evidence base and gaps in evidence aimed at reducing VAC in LMICs using an intervention-outcome framework based on the INSPIRE framework-developed by the WHO, UNICEF and eight other international agencies and initiatives (WHO, 2016a(WHO, , 2016b.
The INSPIRE framework lists seven evidence-based strategies to end VAC: • Implementation and enforcement of laws Utilising this framework to code intervention categories, the EGM was developed to: i. Identify existing gaps in evidence to better inform future investment in research.
ii. Identify clusters of impact evaluations that offer opportunities for evidence synthesis.
iii. Identify, appraise and provide short summaries of existing evidence of the included systematic reviews and impact evaluations.

| What is an EGM?
An EGM is a presentation of the available, relevant evidence for a particular sector. It provides a visual display of completed and ongoing systematic reviews and impact evaluation structured around a framework. In the present map, the rows are intervention categories and the columns are indicator (outcome) categories. The present EGM provides an overview of all available evidence in English on the key outcome categories and interventions aimed at reducing VAC in LMICs. The EGM will be updated with evidence from other languages over 2020-2021.

| Scope
The scope of our EGM is defined by a framework of intervention and outcome categories and subcategories. We included impact evaluations and systematic reviews of effectiveness studies to reduce VAC.
Interventions were categorised by the seven INSPIRE categories, and assessed across seven outcome categories that they reported-(i) direct impact on violence, (ii) norms and values, (iii) economic and social factors, (iv) safety and risk factors for other harms, (v) health, (vi) education outcomes. Additionally, cost-analysis is included as a seventh outcome category. Violence against children: Violence against children includes all forms of violence against people under 18 years old, whether perpetrated by  parents or other caregivers, peers, romantic partners, or strangers (WHO, 2018).

Box 1 Key concept and definitions used in the EGM
What did this EGM include?
1. The EGM included studies on the effectiveness of interventions on both perpetration of violence against children as well as children victimized by violence. 2. The EGM included interventions aimed at addressing different types of violence studied included including physical, sexual and emotional violence. As there are different types and forms of violence against children, the report and maps are structured to reflect present findings of interventions that specifically addressed corporal punishment, intimate partner violence, and peer violence, as these were the most common forms of violence for which we found interventions were specifically tailored. 3. In addition, where interventions do not specify a type or form of violence, where they reportede more than one form of violence, where they covered broader harms experienced by children, or where interventions included attention to the consequences of exposure to violence, we have categorised them in an "unclassified" group.
Forms of violence included in the EGM:

Peer violence/Bullying (including cyber-bullying)
This is unwanted aggressive behaviour by another child or group of children who are neither sibling nor in a romantic relationship with the victim. It involves repeated physical, psychological or social harm, and often takes place in schools and other settings where children gather, and online (WHO, 2018).

Corporal Punishment:
Any punishment in which physical force is used and intended to cause some degree of pain or discomfort, however light. Most involves hitting ("smacking", "slapping", "spanking") children, with the hand or with an implement-a whip, stick, belt, shoe, wooden spoon, etc. But it can also involve, for example, kicking, shaking or throwing children, scratching, pinching, biting, pulling hair or boxing ears, forcing children to stay in uncomfortable positions, burning, scalding or forced ingestion (for example, washing children's mouths out with soap or forcing them to swallow hot spices). In addition, there are other non-physical forms of punishment that are also cruel and degrading and thus incompatible with the UN Convention on the Rights of the Child. These include, for example, punishment which belittles, humiliates, denigrates, scapegoats, threatens, scares or ridicules the child (UNCRC Committee, 2006). 3. Intimate partner violence (IPV) The World Health Organisation defines Intimate partner violence as "behaviour by an intimate partner or ex-partner that causes physical, sexual or psychological harm, including physical aggression, sexual coercion, psychological abuse and controlling behaviours" (WHO, 2014).

"Unclassified"
Here we include studies where interventions do not specify a type or form of violence, where they report more than one form of violence, where they cover broader harms experienced by children, or where interventions include attention to the consequences of exposure to violence, we have categorised them in an "unclassified" group.
On the outcomes, we included studies that reported any of the above outcome categories, provided the intervention is intended to reduce VAC and its risk factors.

| Interventions
The search included interventions aimed at addressing different types of violence including physical, sexual and emotional violence.
As there are different types and forms of VAC, the report and maps are structured to present findings of interventions that specifically address corporal punishment, intimate partner violence, and peer violence, as the three most common forms of violence for which we found interventions were specifically tailored.
In addition, where interventions do not specify a type or form of violence, where they report more than one form of violence, where they cover broader harms experienced by children, or where interventions include attention to the consequences of exposure to violence, we have categorised them in an "unclassified" group.
Boxes 1 and 2 present the key working definitions that we use throughout the report.

| Outcomes
Included studies were those that reported interventions to reduce different types and forms of VAC. Studies were categorised by seven outcome categories on which they reported: direct impact on violence, norms and values, economic and social factors, safety and risk factors for other harms, health and education outcomes (Box 2). Additionally, the report includes cost-analysis as a seventh outcome category.

| Search method
The EGM search method was developed to scan literature in two Any indicators measuring changes in both public safety as well as individual or community risk factors associated with early marriage, FGM and child labour or incidence of these forms of violence? Economic and social Any measures related to impact on social discrimination and social inclusion. These also include impact on related "drivers" of violence against children, i.e., poverty and food security. Education These include measures of factors relating to school environment e.g. gender roles and life skills, teacher engagement. These also include measures of factors affecting school performance, as well as overall school performance reports including truancy/exclusion levels, school enrolment. Cost-analysis These outcome measures the economic cost of violence prevention interventions and highlights economic impact of implementations.

| 5 of 37
Phase 2 is underway and extends the searches to Arabic, Chinese, French, Portuguese and Spanish.

| Selection criteria
The primary population of interest for this map is children and adolescents in the age group of less than or equal to 18 years from LMICs (World Bank, 2018b).
Studies, with multiple populations were included in the map if they include a portion of children under 18 years, had an intervention to reduce VAC and/or its risk factors such as poverty, child marriage, child labour and the study conducted in LMICs.

| Screening, data extraction and confidence appraisal
The screening was conducted using the eligibility criteria defined in Supporting Information Appendix 3.
Two independent reviewers screened titles and abstracts and a third reviewer resolved any discrepancies. Two independent  reviewers coded the included studies from full text. The coded information includes bibliographic details for the study, the interventions and outcomes from the framework that the study measured and other relevant aspects such as population, region and countries.
Two independent reviewers appraised the confidence of all included studies using standardised tools for systematic reviews, AMSTAR 2 (Shea et al., 2017) and impact evaluations (Modified risk of bias tool). Studies were confidence rated as high, medium or low confidence based on critical appraisal tool findings.
We did NOT exclude any study based on study confidence.

| Total number of studies and their type
The map includes 152 studies of which 55 are systematic reviews and 97 are impact evaluations. The impact evaluations predominantly include randomised controlled trials (RCTs)-51 of the impact evaluation are RCTs. The other study designs were quasi-experimental-before and after design (35), matching/propensity score matching methods (4), difference in difference (4) and regression discontinuity design (3).

| Distribution of impact evaluations across intervention and outcome categories of INSPIRE
The studies are unevenly distributed across intervention and outcome areas. Education and life skills is the most widely populated area of the map. Nearly 43 out of 97 impact evaluations have interventions related to education and life skills. Income and economic strengthening is the next most populated area with 37 studies followed by parent, child and caregiver support with 29 studies, and norms and value with 23 studies.
The most common outcome measure is direct impact on violence (n = 92), followed by norms and values (46) and safety and risk factors (45). Very few studies measured impact on economic and social outcomes (18) and only two studies conducted cost-analysis.

| Distribution of systematic reviews across intervention and outcome categories of INSPIRE
Parent-child and caregiver support and norms and values are most widely populated with 21 reviews each, closely followed by education and life skills (20) and response and support services (20) and income and economic strengthening with 15 reviews. Limited evidence was identified on safe environment (4) and laws, crime and justice (2). The most common outcome measure is direct impact on violence (53), followed by norms and values (23), health (19) and safety and risk factors (15). Very few studies measured impact on economic and social outcomes (3).

| Confidence in study findings
The systematic reviews and impact evaluations included were assessed for the level of confidence that could be placed in their findings. Only 17% of the 55 systematic reviews included are highconfidence systematic reviews, which means that more than three quarters (83%) are either low or medium confidence systematic reviews (18 and 25, respectively). A similar picture emerges for impact evaluations, where only 30% of the 97 included impact evaluations were rated as high confidence, which means that 70% were rated as either low of medium confidence (47 and 20, respectively).

| Geographic distribution
The EGM provides us with the opportunity to analyse the distribution of studies within regions. For instance, the distribution of impact evaluations is uneven across regions, with a concentration in sub-Saharan Africa (59), followed by South Asia (13)

| Conclusion
Our study reveals that rigorous evidence synthesis on the interventions for VAC remains limited in many areas. There is concentration of impact evaluations in education and life skills and income and economic strengthening, specifically on safe environment where the systematic reviews are scarce. Also, most of the evidence is found to be concentrated in few countries in Sub-Saharan Africa and South Asia with South Africa, Ethiopia and India being the major countries of coverage. More studies focusing on low-income and conflictaffected settings are needed because violence affecting children is anticipated to be more prevalent in such settings. More impact evaluations are needed to assess the impact of more than one kind of strategy, evaluate effects for vulnerable populations, as well as report on cost-analysis of interventions. The focus of majority of studies has been on impact of interventions on victims of violence or their parent/caregivers or teachers; however, there seemed to be gaps in studies assessing the impact of interventions on perpetration of violence. There is a need to conduct high-confidence impact evaluations and systematic reviews. More than 1 billion children between two to seventeen yearshalf the children in the world-report having experienced some form of interpersonal violence in a previous year (Hillis et al., 2016) Violence affects children in LMICs as well as in HICs; however, the burden and types of violence vary substantially between economies.
Violence, exploitation and abuse against children occur in the homes, families, schools, care and other institutions, justice systems, workplaces and communities across all contexts, including online, and including as a result of conflict and natural disasters (UNICEF, 2011).
Globally, three in four children aged two to four worldwide experience violent disciplines by their caregivers on a regular basis; six in ten children are punished by physical means (UNICEF, 2017). Violent discipline also takes place in schools: one in two school-age children Second, much of the violence experienced by children is at the hands of adults who are typically to be found within a circle of trust and caregiving-parents, teachers, neighbours and authority figures.
Third, there is also a striking rise in peer violence as children grow older and violence spills into peer relationships including through bullying (offline and online), dating and intimate partner violence, as well as gang violence (Know Violence in Childhood, 2017). The gendered experience of violence also manifests as children grow older and transition to adolescence.
Vulnerability to violence is also experienced by children who face other violations of their rights such as early marriage or child labour.
For instance, girls who marry before 18 are more likely to experience intimate partner violence than their peers who marry later (UNICEF, 2005a).

| Risk factors and context
Data suggest that risk factors for exposure to or experience of violence manifest differently across gender, age groups and context. WHO estimates that the highest child homicide rates occur in adolescents, especially boys, aged 15-17 years and among 0-4year-old children (WHO, 2004). A 2011 review estimated the global lifetime prevalence of childhood sexual abuse to be about 18% for girls compared with 8% for boys (Stoltenborgh, Van IJzendoorn, Euser, & Bakerman-Kranenburg, 2011). Girls all over the world are victims of child marriage, forced pregnancy, and female genital mutilation, particularly in LMICs. Despite wide variation by country, physical punishment is common in LMICs

| Impact of VAC
The severity of the effects of violence experienced by children is magnified because of the serious intergenerational impacts on the future wellbeing of children as they transition to adulthood. Experience of violence during childhood increases the risk of becoming victims or perpetrators of violence during adulthood (WHO, 2016b).
Violence can negatively affect physical, mental, sexual, and reproductive health, and may increase the risk of acquiring HIV in some settings (WHO, 2017). VAC is associated with poor educational outcomes, economic insecurity including food insecurity, parental unemployment, inadequate housing and other basic necessities for children and families in LMICs (Peterman, Neijhoft, Cook, & Palermo, 2017). The global costs related to physical, psychological and sexual VAC have been estimated to be between 3% and 8% of global GDP (Pereznieto, Montes, Routier, & Langston, 2014).

What does this EGM include?
This EGM focuses specifically on violence experienced by children whether perpetrated by adults or peers.

The EGM is focused on effectiveness studies of interventions
where the primary aim is the reduction of VAC.

Approach
The INSPIRE framework has been identified as the basis for the intervention-outcome framework for the present EGM (WHO, is designed to help governments and agencies to monitor progress

Box 3 SDG agenda 2030 targets to end violence against children
The SDG 2030 Agenda includes a specific target to end all forms of violence against children (16.2). Abuse, neglect and exploitation of children are also mainstreamed across several other targets. The following goals and targets are particularly relevant for ending violence against children: Goal 5: Achieve gender confidence and empower all women and girls 5.2 Eliminate all forms of violence against all women and girls in public and in private spheres, including trafficking and sexual and other types of exploitation. 5.3 Eliminate all harmful practices, such as child, early and forced marriage, and female genital mutilation Goal 8: Promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all 8.7 Take immediate and effective measures to eradicate forced labour, end modern slavery and human trafficking and secure the prohibition and elimination of the worst forms of child labour, including recruitment and use of child soldiers, and by 2025 end child labour in all forms. There are additional related ongoing maps and published maps.
There is a map on child maltreatment and neglect which is restricted to HICs (Axelsdottir, Biedilae, & Albers, 2018). A second map for LMICs focuses on child neglect (Sinha, Radhika, Jha, & John, 2018) whereas the present map focuses on VAC (we elaborate on this distinction below).
The third map (Albers, et al., 2019) is examining institutional responses to child maltreatment, not focusing on any specific region.
A fourth evidence map has been published in 2019 by the Institute for Security Studies (ISS) in partnership with the University of Johannesburg and Witwatersrand University to address violence against women and children in South Africa (Amisi, et al., 2019).
There is no existing EGM for interventions to address VAC in LMICs.

| Objectives
The objectives of this EGM are to: i. Identify existing gaps in evidence to better inform future investment in research on interventions for VAC.
ii. Identify clusters of impact evaluation that offer opportunities for evidence synthesis on VAC.
iii. Identify, appraise and provide short summaries of existing evidence from systematic reviews and impact evaluations of the effects of interventions to reduce VAC.

| Scope
The scope of this EGM is to capture available effectiveness studies on reducing VAC in LMICs. We included studies assessing effectiveness of interventions to reduce victimisation as well studies assessing impacts of interventions on perpetration of VAC.
This map excludes self-directed violence and child neglect, as this is the scope of an ongoing EGM on child abuse and neglect that includes self-directed injuries in LMICs (Sinha et al., 2018). We also excluded structural and collective violence such as genocide, wars, political violence because the nature of effective interventions to combat collective violence are likely to be dependent on sociopolitical context and factors and are thus considered to be outside the scope of the present EGM.

| Types of evidence
This EGM includes systematic reviews and impact evaluations on the effectiveness of interventions to prevent VAC.
Systematic reviews are eligible to be included if they contain one or more studies from LMICs. The key characteristics for a review to be included as a "systematic review" 1. A clearly stated set of objectives with pre-defined eligibility criteria for studies.
3. A systematic search that attempts to identify studies that would meet the eligibility criteria. 4. A systematic presentation, and synthesis, of the characteristics and findings of the included studies.
Among impact evaluation, we included RCTs; non-experimental evaluations (controlled before-after studies, uncontrolled before-after and interrupted time-series); regression discontinuity design (RDD); difference-in-difference (DID); and studies employing matching techniques such as propensity score matching (PSM). Relevant ongoing studies are also included. The detailed eligibility criteria are given in Supporting Information Appendix 3. We did not include qualitative research studies as per the scope of this EGM.

| Types of outcome categories
The seven main outcome categories are listed in Table 3.

| Types of location/situation
The primary population of interest for this map is children and adolescents from LMICs. LMICs are defined by World Bank as: • low-income economies-those with a Gross National Income (GNI) less than $995; • lower-middle-income economies-those with a GNI per capita between $996 and $3,895; and • upper-middle-income economies-those with a GNI per capita between $3,896 and $12,055 (World Bank, 2018b).

| Types of settings
The EGM includes all types of settings where interventions for VAC were implemented. Based on the framework, the settings include school, home, centre or facility, community, etc. We did not exclude any study based on setting.

| Search methods and sources
The search for the EGM was conducted online in two stages:

| Description of outcome categories
Included studies were those that reported interventions for different types and forms of VAC. Studies were categorised by seven outcome categories on which they reported: direct impact on violence, norms and values, economic and social factors, safety and risk factors for other harms, health and education outcomes (Box 4). Additionally, the report assesses the availability of information on cost-analysis as a seventh outcome category.
The outcome categories and subcategories were mapped across the intervention categories and subcategories in the present map.
The seven main outcome categories are listed in Table 3.

| Description of population
Children irrespective of their sex in the age group of less than or equal to 18 years were included in the EGM.

| Description of geographic location
Studies from low-income countries, LMICs and upper-middle-income countries are included in the present map.
The category LMIC is defined by (World Bank, 2018b) as: • low-income economies-those with a Gross National Income (GNI) less than $995; • lower-middle-income economies-those with a GNI per capita between $996 and $3,895; and • upper-middle-income economies-those with a GNI per capita between $3,896 and $12,055.
There are 47 countries in the LMICs criteria defined by The World Bank globally.
Global systematic reviews were included if they contained at least one impact evaluation conducted in LMICs. Impact evaluation and systematic reviews from HICs alone were excluded.
6.5 | Analysis and presentation

| Presentation
The EGM has traditionally two primary dimensions intervention as rows and outcomes as columns. In the online interactive map, we used the additional dimensions as filters:

Box 4 Distribution of studies from Africa
South Africa: Five (5) studies from South Africa relate to parenting interventions to reduce child abuse. The other studies from South Africa include microfinance interventions (2) and gender transformative approaches to reduce violence (2) and an assessment of the effectiveness of cash transfers on child labor (1). Uganda: The majority of the studies from Uganda are on skills training and microfinance to reduce VAC. The other studies from Uganda include interventions such as gender-transformative approaches (1), parenting interventions (1) and interventions for teachers to reduce violence in schools (1). Kenya: The impact evaluations from Kenya, the country with the third highest number of intervention studies in Sub-Saharan Africa, include interventions with gender transformative approaches to reduction of all forms of violence against children (2), economic transfers and parenting interventions (1), reduction of female genital cutting (1) and child marriage (1).
which report more than one form of violence, or which cover broader harms experienced by children, or which include attention to the consequences of exposure to violence, such as, for example, an intervention studying impact of counselling to empower sexually abused girls through cognitive behavioural group therapy training which is categorised here) For this map we will present three forms of visualisation of evidence (Snapshots for each added as annex) 1. Interventions as rows and outcomes as columns.
2. Interventions as rows and additional filters; forms of violence and perpetrator of violence as columns.
3. Interventions as rows and additional filter; regions as columns

| Dependency
Each entry in the map is a systematic review or impact evaluation of effectiveness. Each item in the map reports one study, as we did not come across any review of reviews. The readers must be aware that though the representation of studies is separate (for systematic reviews and impact evaluations), there is an overlap of evidence as one study can be represented in more than one cell, if the study dealt with multiple interventions or outcomes.

| Data extraction and management
We used a standardised data extraction form (presented in Supporting Information Appendix 5) to extract descriptive data from all studies meeting the eligibility criteria.
Each included study was coded independently by a team of two coders (J. A. and P.P.) using the coding tool (Supporting Information Appendix 5) covering study characteristics, population, interventions, outcomes, region, countries and type of violence. A third reviewer (A. S.) resolved the conflicts. We contacted the authors for the articles that we could not retrieve. If we did not receive the reply within 15 days, we kept such studies under "awaiting classification". A reference list of the included studies and systematic reviews were compiled by A. S., P. P., and J. A.
Two authors (H. W. and R. S.) validated coding and reporting through random coding checks for a portion of studies.
7.3 | Tools for assessing risk of bias/critical appraisal of included reviews

| Critical appraisal
We critically appraised both systematic reviews and impact evaluations to indicate confidence in study findings.

Systematic reviews
For systematic reviews we scored each study using the 16-item checklist, AMSTAR 2-"Assessing the Methodological Confidence of Systematic Reviews" (Shea, et al., 2017)
The 16 items cover: (1) population, interventions and outcomes in inclusion criteria, (2) ex ante protocol, (3) rationale for included study designs, Seven of these criteria have been applied as relevant in our critical assessment. Items 2, 4, 7, 9, 11, 13 and 15 are considered "critical". Study confidence is rated "high" if there is no more than one non-critical weakness, and medium if there is no critical weakness but more than one non-critical weakness. Studies with one or more critical weaknesses are rated "low" confidence.

Impact Evaluations
We used a modified risk of bias tool (Supporting Information Appendix 7) to rate the confidence of impact evaluations.
This tool includes six criteria that are appropriate for the as- However, coding a mention of power calculations signals the importance of both conducting and reporting power calculations.
3. Attrition or losses to follow-up: can be a major source of bias in studies, especially if there is differential attrition between the treatment and comparison group (called the control group in the case of RCTs) so that the two may no longer be balanced in pre- reported and there is a significant difference of 10 or more than 10%, medium confidence if imbalance is between 5% and 10%, and high if an RCT or if imbalance is 5 or less than 5% ( Figure 1).
F I G U R E 1 Graphical representation of critical appraisal tool for impact evaluations PUNDIR ET AL.
| 17 of 37 8. Overall assessment: The overall assessment uses a "weakest link in the chain" principle so that the overall assessment is the lowest of assessment given to any of the relevant items. As mentioned above, not all items are used in this assessment. So the overall assessment, is the lowest of the assessments for Items 1, 4, 6 and 7.

Ethical considerations
Many The nine-item tool had five critical and four non-critical items.
The scoring identified "Strong ethical standards", "Moderate ethical standards" and "Weak ethical standards". The complete tool with scoring can be found in the Supporting Information Appendix 8. We did not disqualify any study based on the adequacy of ethical standards.
We note that a uniformly used and standardised ethical appraisal tool does not exist for VAC studies that explicitly mention points on privacy, referral, adverse event protocol and informed consent.

| Description of included studies
There are many more entries in the map than there are studies.
This is because the studies, especially systematic reviews, were coded and mapped for all the intervention categories and subcategories that they included. The results present the findings from the 152 included studies (55 systematic reviews and 97 impact evaluations).
F I G U R E 2 PRISMA for VAC evidence and gap map. VAC, violence against children Table 4 shows the aggregate maps of interventions and outcomes. • The impact of parent, child and caregiver support interventions on violence (48) and norms and values (26).

| Laws, crime and justice
• This is the least populated sector with only five studies identified.
• Most common outcomes reported were direct impact on VAC (5) and impact on changing norms and values (4).

| Norms and values
• Thirty-five (35) out of 43 studies for norms and values interventions focused on community mobilisation programmes.
• Fourteen (14) studies were identified on media campaign.
• Limited evidence was found on the effectiveness of bystander interventions (4).
• On the outcomes front, 26 studies reported impact on changing norms and values such as change in gender roles (21), empowerment (5) and belief in parenting practices (3).
• There are other outcomes reported as well, such as safety and risk factors for other harms (25), direct reduction of violence (47) and impact on health (19).

| Safe environment
• This sector has limited evidence with only 15 studies, of which nine studies were identified on interventions for creating safe places and six on making existing environment safe.
• On the outcomes front, there are 13 studies reporting direct impact on violence, 11 studies reporting impact on safety and risk factors for other harms and 7 on changing norms and values.

| Parent, child and caregiver support
• Parent, child and caregiver support includes 48 studies, 39 of which focused on parent training and education and 10 on peerrelationship and training.
• On the outcomes front, we identified only 4 studies that assessed the impact on maternal and paternal mental health.
• Reduction in VAC (53) is the most commonly reported outcome, followed by change in norms and values (29), impact on health outcomes (23) such as, sexual and reproductive health (12), child mental health (11), mental/paternal mental health (7) and substance abuse (4).

| Income and economic strengthening
• The intervention area income and economic strengthening has a total of 53 studies. Majority of studies were found assessing the effectiveness of economic transfers (40), followed by income generation, savings or credit interventions (16).
• Insurance and welfare schemes is least populated with only five (5) studies.
• On the outcomes front, safety and risk factors for other harms (46) is the most commonly reported outcome after reduction in violence (49). Safety and risk factor for other harms includes outcome like reduction in child labour and trafficking (29); female genital mutilation and child marriage (18). • Second common outcome reported is the impact on economic and social outcomes (19) such as employment and labour force participation (7), savings and credit (9) and poverty and food security (5).

| Response and support services
• Thirty-six studies were identified related to the intervention area response and support services, of which 21 related to screening and training and 17 to counselling and therapeutic approaches.
• There are stark gaps in the evidence identified on the effectiveness of interventions for children in care (3) and dissemination of information by media and communication (2).
• The most common outcomes reported as a result of response and support interventions are impact on violence (44) and health outcomes (23).

| Education and life skills
• The intervention area education and life skills-which maps the studies of effect of educational and school-based interventions on VAC is the most heavily concentrated (60) ( Table 4).
• Under education interventions, gender transformative approaches and life and social skills are almost equally concentrated with 41 and 32 studies, respectively.
• On the outcomes front, 24 studies reported impact on educational outcomes as a result of the interventions.
• Most commonly reported outcomes as result of interventions in this area were found influencing norms and values (37), followed by impact on health (26) and safety and risk factors for other harms (26).
8.4 | Synthesis of included studies by primary dimension (intervention-outcome)

| Number of studies by intervention categories
Systematic reviews Figure 3 shows the number of systematic reviews by intervention categories. Laws, crime and justice (2) and safe environment (4) are the least populated areas of the map. We identified an almost equal number of systematic reviews across the categories of parent, child and caregiver support (21); norms and values (21), response and support services (20) and education and life skills (20).

Impact evaluation
As with systematic reviews, laws, crime and justice (3)

Impact evaluations by intervention and study design
RCTs account for close to half of the impact evaluations (51 RCTs out of 91) being particularly prominent in the intervention areas education and life skills (19), norms and values (16) and parent, child and caregiver support (10) ( Figure 5).

Impact evaluations by intervention area and type of violence
Physical violence is the most common type of violence studied across the various intervention categories ( Figure 6).

Fine-grained analysis by intervention subcategory
Further fine-grained analysis reveals that though there is a fair amount of evidence on parent, child and caregiver support, evidence is limited on children in care (3) and media and communication interventions (2). There are only five (5) studies on substance abuse and five (5) studies on morbidity and mortality as outcomes.

F I G U R E 3 Number of systematic reviews by intervention categories
There are also limited studies on insurance and welfare schemes assessing their impact on the reduction of violence in children (5).
Empty reviews are systematic reviews with no included eligible studies in a particular sector. There were two empty reviews included in the present EGM: the first, Parker & Turner (2014)

| Number of studies by outcome category
Systematic reviews Figure 7 shows systematic reviews by outcome area. Direct impact on VAC is the most studied outcome (54) followed by influence on norms and values (22) and impact on health (19).

Impact evaluations
As with systematic reviews, direct impact on VAC was the most reported outcome (91); however, we also found studies reporting impact on changing norms and values (47), safety and risk factors for other harms (45), education (36) and health (34). We found few impact evaluations conducted in relation to cost-analysis (2), and relatively few studies assessed economic and social outcomes (19) (Figure 8).

Number of impact evaluations by types of violence
Physical violence is the most studied type of violence (60) with equal number of studies on sexual (37) and emotional violence (36) (Figure 9).

Systematic reviews
This pattern remains similar for systematic reviews with a concentration in Sub-Saharan Africa (36), South Asia (27), East Asia and Pacific (28) and Latin America and Caribbean (24) as shown in Figure 11 and Figure 12. Table 3 gives the list of countries and the frequency of studies from these countries. Twelve impact evaluation studies in the EGM are from South Africa and 10 are from India. Uganda and Kenya also recorded a high number of impact evaluations (Table 5).

By study design
In terms of methodology, a fair number of impact evaluations (45) had evidence from RCTs, a common impact evaluation method. As shown in Table 5, other common methodologies were quasiexperimental-before and after design (35), matching/PSM methods (4 studies), 3 used DIDs and 3 used RDDs given in Table 6.
Out of the 35 quasi-experimental studies, 10 studies used uncontrolled before-after design while the 25 studies used a controlled before-after design. Controlled before-after study designs are more reliable in terms of measuring effectiveness of interventions. As uncontrolled before-after designs are used in the absence of a control group, they are considered to provide lower confidence evidence on the effectiveness of interventions.

By forms of violence
Twenty-one impact (21) evaluations assessed the impact of interventions specifically on intimate partner violence, followed by corporal punishment (10), and peer violence (7). Remaining studies could not be specifically classified by form of violence addressed, and are hence presented as a general, "unclassified" category.

Conflict-affected and fragile countries
Analysis was undertaken according to a list of conflict-affected and fragile regions and countries (WHO, 2018a) (the list available when this EGM was conducted). Studies from high-, medium-and lowfragility countries and their neighbours are given in Table 7.
F I G U R E 1 1 Geographic heat map of systematic reviews included in the EGM. EGM, evidence and gap map F I G U R E 1 2 Countries with the largest number of combined studies (impact evaluations and systematic reviews) The greatest number of studies from high fragility contexts were from Pakistan (13) followed by Nigeria (11). For medium-fragility contexts, the greatest number of studies were from Iran (11) followed by Egypt (9). For low-fragility contexts, the greatest number of studies were from Ethiopia (26) followed by Bangladesh (12).
Overall, the number of impact evaluations from conflict-affected countries was low, and the same studies were included in many systematic reviews.

Funding agencies
More Although there was no restriction of year on screening, the systematic reviews, included in the EGM, were conducted between the years 2009 and 2019.
The confidence of systematic reviews, as appraised for this EGM, remained consistent over the years with mostly medium-confidence studies, followed by low-confidence and then high-confidence studies in number as shown in Figure 16. Our overall observation, however, is that there is a need to develop more tailored ethical adequacy appraisal tools better suited to the specific requirements of research on violence.

| Ethical adequacy
T A B L E 5 Number of impact evaluation (n) by countries (World Bank, 2018b)

Low-income countries
Low and middle income
The reason for a high number of low-confidence systematic reviews lies in factors such as reviews have been produced by organisations or individuals that have not complied with international checklists and standards which are readily available for production of systematic reviews. Also, the search and screening procedures were not as per standard guidelines, and studies were insufficiently transparent with respect to methodology and did not provide the list and reasons for excluded studies. Figure 19 shows the results of the critical appraisal of the included 97 impact evaluations. Study confidence was rated high, medium or low for each of the criteria, applying the standards as stated in the critical appraisal tool given in the report (given in Supporting Information Appendix 7).

| Impact evaluations
Only 30% of the 97 included impact evaluations were high confidence, which means that 70% were either low of medium confidence (47 and 20, respectively). The large number of lowconfidence studies is largely driven by concerns related to attrition.
Sixty-six percent (66% of the impact evaluations) had no mention of power calculation. Attrition was not reported in nearly half (40)

| Areas of evidence concentration
Most of the impact evaluations measured impact of "education and life skills" and "income and economic strengthening" in reducing all forms of VAC. The intervention category "Parent, child and caregiver support" had a fair concentration of studies as well.
In terms of outcomes, impacts on "changing norms and values" and "health" impacts were the most studied outcomes apart from direct reduction of VAC.
The distribution of studies (systematic reviews and impact evaluations) across regions is uneven. As regions, Sub-Saharan Africa and An emerging insight from our EGM is that the evidence seems to be concentrated on reducing victimisation and there seem to be gaps in evidence on the effectiveness of interventions for perpetration of violence.
There seems to be differences in understanding and defining diverse forms of VAC, as our findings show that a large proportion of the studies reported interventions for "unclassified" types and forms of violence. It was rare for studies to include information on cost-analysis (including cost-effectiveness and cost-benefit analysis). There were only two studies that conducted cost-analysis.
Evidence is scarce on assessing the effectiveness of "maternal and paternal mental health", "children in care" and "media and communication" interventions. Only five studies noted the effectiveness of "insurance and welfare schemes" interventions in reduction of VAC and they are noted to be of medium to high confidence.
There are also evidence gaps in studies assessing the effectiveness of interventions for reduction of VAC of less than 3 years of age and of children belonging to ethnic minorities or children with disabilities.
There are major gaps in studies assessing the effectiveness of parent-child and caregiver support on safety and risk factors for other harms and economic and social outcomes. Evidence was lacking in terms of the impact of the intervention area "response and support services" on economic and social outcomes.
There is limited evidence on the studies assessing effectiveness of interventions related to specific forms of violence such as intimate partner violence amongst children and adolescent, peer violence and corporal punishment.
There are very few studies from low-income countries as compared with middle-income countries (low and upper-middle). This is of concern because many low-income countries from the regions report high violence prevalence against children and the need for effective interventions is critical.
The EGM analysis showed that there are many studies from Sub-Saharan Africa, but it is mainly concentrated in a few countries,

| Limitations of the EGM
Violence is as complex issue, with a range of intersecting forms, drivers and risk factors operating at the level of the child, family, community and wider institutions. This EGM has been based on an as expansive a framework as possible, though limitations may remain in the scope and approach. We trust and expect that other researchers will address limitations going forward, contributing to more robust evidence architecture to inform this field.

| Recommendations
Main recommendations from the findings of the EGM: 1. The EGM should be the basis to identify gaps in evidence such as on key intervention areas and forms of violence, so that investments in research can be better targeted to address gaps.
2. The EGM finds that there is a need to improve the confidence of impact evaluations. This includes standardising definitions of violence used, study design, implementation and also reporting. 5. The requirement to report on the application of ethical standards must be made mandatory as part of funding applications, as well as research publication. There is a need for a specific, tailored tool for ethical appraisal of research on violence, and ethical considerations need to be included in confidence appraisals as well.
6. The EGM should be made a living review and expanded to include qualitative literature, to provide a database for researchers to update with relevant literature, as well as for policy makers and other stakeholders to have easily accessible summaries of relevant literature for their interest.
7. More high-confidence impact evaluations should be funded and implemented. One way to ensure confidence can be by adhering to the standardised international checklists for relevant study design (available for impact evaluations as well as for systematic reviews) and making training available and adaptable to context. 8. In the areas of high evidence concentration such as education and life skills and income and economic strengthening, evidence summaries should be generated gain an idea of the extent of the developing country literature, and also to develop taxonomy of approaches relevant in these contexts.
9. As a next step, this EGM can provide a preliminary content for the web based evidence portal for VAC in LMICs. Wider stakeholder consultations with experts in the field is needed.
9.6 | Stakeholder engagement throughout the EGM process 1. A consultative process was followed for refining the EGM con-  ➣ More studies are needed from low-income and conflict-affected settings.
➣ More high-confidence systematic reviews and impact evaluations need to be commissioned with standardised guidelines and tailored guidance to ensure robustness, as most of them included in the map were found to be of low confidence.
➣ More impact evaluations are needed that report outcomes for vulnerable groups of children, requiring larger sample sizes in studies and better disaggregation in study design, methodology and reporting.
➣ More impact evaluations are needed that assess gendered effects of interventions and on diverse social groups in a given context, utilising mixed methods to better understand for whom the interventions work, and how.
➣ Ethical standards need to be better-defined and made a manda- Jill has attended training workshops on evidence synthesis provided by both 3ie and Campbell. Prachi Pundir has experience in systematic reviews and has previously worked on systematic reviews and metaanalysis with Public Health Evidence South Asia, Manipal, and all authors are proficient in carrying out the various processes in an EGM, such as eligibility screening, confidence assessment and coding.
Statistical analysis: Ashrita Saran and Prachi Pundir have training in data management and using statistical softwares for quantitative data analysis.
Information retrieval: Ashrita Saran and Prachi Pundir have training in designing and implementing search strategies. P. P, A. S., H. W., J. A. and R. S. contributed to writing and revising this report. The search strategy was developed, validated and piloted by P. P. and A. S. Search was conducted by P. P., A. S. and J. A. P. P.
will be responsible for updating this EGM.

CONFLICT OF INTERESTS
One of the co-authors is an Innocenti staff member who commissioned the study, helped shape the search strategy, and worked with the Campbell Collaboration team to coordinate internal and external review processes and to contextualise findings and recommendations. The main search, screening, coding and analysis were independently conducted by Campbell Collaboration staff independently.

PLANS FOR UPDATING THE EGM
Once completed, the EGM is planned to be updated yearly, provided we have availability of funds. The lead author and/or the corresponding authors will be responsible for updating the EGM.

DIFFERENCES BETWEEN PROTOCOL AND MAP
In the course of finalising this map, there have been some deviations from the protocol, informed by practical considerations: 1. "Intimate partner violence", "peer violence" and "corporal punishment" was added as forms of violence.
2. A typology of violence used in the protocol on "community violence and collective violence" was removed as it was found to be beyond the scope of this map. The focus for this map was to capture studies on interpersonal VAC.
3. We added "perpetration of violence" as a filter.

SOURCES OF SUPPORT
Financial support for this research was provided by UNICEF Office of Research-Innocenti. Non-financial support was not received for the production of this map.