Interventions that address institutional child maltreatment: An evidence and gap map

Abstract Background Child maltreatment has serious short and long‐term negative impacts for those experiencing it. Child maltreatment occurring in institutional settings has recently received substantial attention. However, evidence about the effectiveness of interventions that prevent, disclose, respond to, or treat maltreatment that has occurred in these environments is fragmented and can be difficult to access. This evidence and gap map (EGM) collates this research evidence. It was developed as a resource for stakeholders operating in the child health, welfare and protection sectors, including practitioners, organisational leaders, policy developers and researchers, wanting to access high quality evidence on interventions addressing institutional child maltreatment. Objectives The objectives of this EGM were twofold: (a) To provide a structured and accessible collection of existing evidence from finalised and ongoing overviews of systematic reviews, systematic reviews and effectiveness studies of interventions addressing institutional child maltreatment—for those who work to fund, develop, implement and evaluate interventions aimed at ensuring children's safety in institutional settings; (b) to identify gaps in the available evidence on interventions addressing institutional child maltreatment—thereby helping to inform the research agendas of funders and other organisations. Search Methods A comprehensive search strategy identified relevant studies from published and grey literature, comprising: (1) 10 electronic academic databases; (2) five trial and systematic review registries; (3) nine organisational websites; (4) websites and reference lists of inquiry reports associated with seven international inquiries into child abuse and (4) the lists of included studies within systematic reviews identified by the search strategy. Members of this EGM's Subject Matter Experts group were also invited to forward relevant unpublished studies or grey literature. Selection Criteria The selection criteria were developed to identify finalised and ongoing overviews of reviews, systematic reviews and primary studies that reported on the effectiveness of interventions addressing child maltreatment (including sexual abuse, physical abuse, neglect and emotional abuse) within institutional settings. Eligible effectiveness study designs included: randomised controlled trials (RCTs), nonrandomised trials, controlled before‐and‐after studies and quasi‐experimental studies. Reviews were eligible if they reported a systematic literature search strategy. Data Collection and Analysis All screening, data extraction, coding and critical appraisals were undertaken by two or more reviewers working independently, with discrepancies resolved via consensus or by a third reviewer. The titles and abstracts of studies identified by the search strategy were screened, and each full text of potentially relevant studies was further assessed for inclusion. Key data were extracted from all included studies and reviews. This included information about: publication details (e.g., year, author, country), inclusion/exclusion criteria (for reviews), study design, institutional setting, target population, type of maltreatment, intervention type and outcomes. Critical appraisal of included systematic reviews was achieved using the AMSTAR 2 tool, and completed RCTs were assessed using the updated Cochrane Risk of Bias 2.0 tool. Main Results Number of studies The electronic database search yielded 6318 citations, and a further 2375 records were identified from additional sources. Following deduplication and title/abstract screening, 256 studies remained for full text review. A total of 73 eligible studies (reported across 84 publications) met the inclusion criteria, including: 11 systematic reviews (plus, one update); 62 primary studies (including, three protocols for primary studies). Study characteristics The studies were conducted across 18 countries, however more than half (52%) were undertaken in the United States. Overall, most studies evaluated curriculum‐based interventions delivered in educational settings, primarily aimed at the prevention of sexual abuse. Institutional setting: Most studies evaluated interventions in school or early learning environments (n = 8 systematic reviews; n = 58 primary studies). Far fewer studies examined other organisational settings. Out of home care (including foster care, residential care and orphanages), and social service organisations servicing children were minimally represented. No studies were identified where the primary setting was sports clubs, churches/religious organisations, summer/vacation camps, detention centres/juvenile justice settings, or primary/secondary health care facilities. Target population: Most interventions targeted children rather than adults (n = 7 systematic reviews; n = 47 primary studies) from the general population. Fewer studies included populations known to be at an increased risk, or those already exposed to maltreatment. Just over a third of the primary studies conducted an analysis to ascertain differences in the effect of an intervention between the genders. Intervention type: Prevention interventions were the most studied (n = 5 systematic reviews; n = 57 primary studies), with additional studies including prevention approaches alongside other intervention types. Fewer studies evaluated interventions targeting disclosure, institutional responses, or treatment interventions. Type of maltreatment: The vast majority of the studies assessed interventions solely addressing the sexual abuse of children (n = 8 systematic reviews; n = 45 primary studies). The remaining studies addressed other forms of maltreatment, including physical and emotional abuse, or neglect, either in combination or as a sole focus. Outcomes: Primary reported outcomes reflected the bias toward child targeted interventions. Outcome measures captured child wellbeing and knowledge outcomes, including measures of mental health, children's knowledge retention and/or self‐protective skills. Measures of maltreatment disclosure or maltreatment occurrence/reoccurrence were less common, and all other outcome categories included in the EGM were minimally or not reported. A third of studies reported on some measure of implementation. Study quality The overall quality of the studies was low to moderate. Most systematic reviews were low‐quality (n = 10), with only one high quality review (and update) identified. Most completed RCTs had some concerns relating to the risk of bias (n = 30), and the remainder were considered to be at a high risk of bias (n = 19). Authors' Conclusions This EGM has highlighted a substantial need for more high quality studies that evaluate interventions across a broader range of institutional contexts and maltreatment types. The current evidence base does not represent countries with large populations and the greatest incidence of child maltreatment. Few studies focussed on perpetrators or the organisational environment. Further evidence gaps were identified for interventions relating to disclosure, organisational responses and treatment, and few studies assessed interventions targeting perpetrators' maltreatment behaviours, recidivism or desistence. Future studies should also include measure of programme implementation.

Child maltreatment is a widespread phenomenon affecting millions of children, adults and communities around the globe. Child maltreatment includes sexual abuse, physical abuse, neglect and emotional abuse. It is a major social issue that has a negative impact on the physical, mental, spiritual, educational and interpersonal wellbeing of those experiencing and surviving itboth in the short-term and in the form of longterm consequences that reduce the quality of life into adulthood (Fang & Corso, 2007;Fang, Brown, Florence, & Mercy, 2012;Felitti et al., 1998;Jaffee et al., 2018;Lueger-Schuster et al., 2018;Maniglio, 2009;Moore et al., 2015;Teicher & Samson, 2016). In recent years, child maltreatment that occurs in institutional settings has received high public and policy recognition, and there is increasing interest in targeting this form of maltreatment.
Determining the prevalence of child maltreatment is considered difficult due to inconsistencies in measurement and suspected under-reporting (Finkelhor et al., 2014), resulting in considerable variability among estimates. Even so, estimates of the overall prevalence of child maltreatment are alarmingly high, and these may give some indication of the extent of this issue. A synthesis of existing meta-analyses from across the globe estimated overall prevalence at 127/1000 for sexual abuse, 226/1000 for physical abuse, 363/1000 for emotional abuse, 163/1000 for child neglect and 184/1000 for emotional neglect (Stoltenborgh et al., 2015). Prevalence rates are sensitive to a number of factors. There are both geographical and gender differences. For example, the Global Status report published by the World Health Organisation [WHO] (2014) reported the prevalence of child physical abuse in Swaziland to be 22%, whereas in countries including Kenya, Tanzania and Zimbabwe prevalence ranged between 53% and 76% with higher rates of abuse experienced by boys than girls. Rates can vary depending on whether incidences of maltreatment are self-reported or based on informants (Greger, Myhre, Lydersen, & Jozefiak, 2015;Moody et al., 2018), and can also vary with the identity of the perpetrator/s. The nature of the acts (how widely or narrowly different subtypes of maltreatment are defined and operationalised in studies) or how many items are used to measure prevalence can also impact on rate estimates. While there is some variability across estimates, it is clear that the occurrence of child maltreatment is unacceptably high. The overall economic cost of child maltreatment is also high, with average lifetime costs in the US upward of $200,000 per child, resulting in billions in cost burden from new cases each year (Fang et al., 2012;Letourneau et al., 2018). In short, child maltreatment is harmful, highly prevalent and costly.
Even less is known about the prevalence of child maltreatment that occurs in institutional contexts, such as schools, out-of-home care, youth/juvenile detention, sport clubs, recreational settings, religious organisations or other comparable child and youth serving organisations in which children live or spend time. Institutional child maltreatment encompasses any kind of maltreatment that occurs in institutional settings. In these settings, child maltreatment can encompass adults abusing children, children abusing other children or institutions enabling child maltreatment. Children may be more or less at risk, for reasons ranging from a lack of proper safeguarding in institutions (e.g. failing to respond to disclosures) (Australian Government, 2017;Lemaigre, Taylor, & Gittoes, 2017;Wurtele, 2012) to individual characteristics (e.g. age, developmental or other disabilities) Sullivan et al., 1992). Institutional child maltreatment as a field of empirical research is at an early stage (Blakemore, Hebert, Arney, & Parkinson, 2017;Proeve, Malvaso, & Delfabbro, 2016;Timmerman & Schreuder, 2014). It is not common for studies to differentiate between maltreatment occurring in institutional settings versus other maltreatment settings, and disentangling the impact of institutional maltreatment versus maltreatment that takes place in other contexts has not been routine. The empirical research to date has focussed primarily on sexual abuse within especially religious and out-of-home care institutions, whereas other types of maltreatment and settings have been less examined (Proeve et al., 2016). Recent studies conducted in Germany (Allroggen, Ohlert, Rau, & Fegert, 2018) and Norway (Greger et al., 2015) confirm that children placed in institutional care are at significantly higher risk of experiencing maltreatment, but less is known about maltreatment taking place in areas such as sports and exercise settings (Bjørnseth & Szabo, 2018). Regardless, child maltreatment taking place in these settings affects the lives of both victims, their families and their communitiesat times for generations.
Child maltreatment occurring in institutional settings has received substantial attention in recent years, both at the policy level, among practitioners and service agencies working with children in different capacities and roles, and also as part of the public discourse. The shift in attention and prioritisation of child maltreatment as a key concern of society is reflected in a broad range of official inquiries and associated reports conducted in recent years in especially high-income countries (Australian Government, 2017; German Inquiry into Child Sexual Abuse, 2020; Pennsylvania Attorney General, 2018; United Kingdom Independent Inquiry into Child Sexual Abuse, 2019). These inquiries have led to a prioritisation of child maltreatment within institutional settings, as both a specific and serious issue among policymakers, practitioners and service agencies working with children (Blakemore et al., 2017;Proeve et al., 2016). The inquiries have led to the production of multiple research reports examining the specific characteristics and consequences of institutional child maltreatment (Blakemore et al., 2017), how it can be prevented (Pitts, 2015;South et al., 2014South et al., , 2015, victims supported (Shlonsky, Albers, & Paterson, 2017), perpetrators and institutions held accountable and suitable responses implemented and maintained over time (Albers & Mildon, 2016;Parenting Research Centre, 2015). The problem has now rightly gained much wider recognition, being under the purview of the United Nations and the World Health Organisation, and gaining attention from parliaments, legislators, institutional governance and leadership, as well as the corporate and philanthropic sectors.
Evidence about the effectiveness of interventions aimed at preventing, disclosing, responding to or treating institutional child maltreatment is spread across multiple sources, and generally exists in the form of academic or grey literature. For institutions that wish to improve their practices and services in this area, it can be difficult and time consuming to find, access and interpret existing evidence. Therefore, there is still considerable confusion among sector stakeholders about what evidence exists for interventions developed to address institutional child maltreatment.
Evidence synthesis is a powerful tool that can bring together, integrate and interpret diverse knowledge sources using methods that are comprehensive, transparent and replicable (Littell & Shlonsky, 2010;Straus, Tetroe, & Graham, 2013). Evidence and gap maps (EGMs) are 'evidence collections' (Snilstveit, Vojtkova, Bhavsar, & Gaarder, 2013, p. 3) that provide a visual overview of the availability of evidence for a particular sectorin this case, interventions addressing institutional child maltreatment. They belong to a group of evidence synthesis products that aim to 'configure information' (Littell, 2018, p. 10). They do this by mapping out existing and ongoing systematic reviews and effectiveness studies, and by providing a graphical display of areas with strong, weak or non-existent evidence on the effect of interventions or initiatives. Evidence and gap maps therefore help to consolidate what evidence exists and what evidence does not currently exist about the effectiveness of interventions in a given area.
This EGM aims to provide a 'go to' knowledge base for stakeholders wanting to access high-level evidence on interventions addressing child maltreatment within institutions or organisations. To our knowledge, there are three existing EGMs related to issues of child maltreatment (Kornør et al., 2017;Pundir et al., 2019;Saran, Albright, Adona, & White, 2018). However, none have a specific focus on maltreatment that occurs in institutional settings. This EGM addresses this gap, making it an important resource for a wide range of stakeholders operating in child and youth serving organisations, such as kindergartens, schools, charities, churches, sports clubs, scouting associations, out-of-home care providers and the many other organisations that associate with children. Given the scale of interest in this issue, it is also expected to be an important resource more broadly.

Objectives
The objectives of this EGM were twofold: 1. To provide a structured and accessible collection of existing evidence from finalised and ongoing overviews of systematic reviews, systematic reviews and effectiveness studies of interventions addressing institutional child maltreatmentfor those who work to fund, develop, implement and evaluate interventions aimed at ensuring children's safety in institutional settings. 2. To identify gaps in the available evidence on interventions addressing institutional child maltreatmentthereby helping to inform the research agendas of funders and other organisations.

Protocol and Study Publication
The protocol for this EGM was prospectively registered (Albers et al., 2019; https://doi.org/10.1002/cl2.1039) and the final research report was published with Campbell Systematic Reviews (Finch et al., 2021;https://doi.org/10.1002/cl2.1139 following methods for producing a Campbell EGM (White et al., 2020). These documents should be referred to for additional information about this study, including methodology and results, and changes from the protocol.

Evidence and Gap Map framework
Evidence and gap maps include a visual representation of the evidence, usually presented in a tabular format that is predetermined prior to undertaking literature searches. This EGM has been vertically structured into interventions aimed at addressing institutional child maltreatment prevention, disclosure, institutional responses and treatment. The EGM's horizontal structure has been formed by outcome domains that relate to institutional safeguarding practices, maltreatment occurrence/reoccurrence, children's health and wellbeing, parent/caregiver behaviour, knowledge and attitudes and adult perpetrators of child maltreatment or child/youth offenders.

Eligibility Criteria
Studies that met the following eligibility criteria were included in the EGM.
Institutional settings. Institutional settings were defined as any public or private body, agency, association, club, institution, organisation or other entity or group of entities of any kind (whether incorporated or unincorporated), that also provides, or has at any time provided, activities, facilities, programs or services of any kind that provide the means through which adults have contact with children, including through their families (adapted from Australian Government, 2017). This included but was not limited to settings in education, health, residential care, detention, sports and recreation and religion. All geographic locations were included, and covered low-, middle-and high-income countries.
Target population. The key target population were children who were aged under 18 years, and living in or engaging in activities in institutional settings. However, although children were the key target group, study participants could also be adults. This EGM also included evidence on interventions that targeted perpetrators of institutional child maltreatment, as well as interventions aimed at improving the professional practice of staff and organisational standards of child and youth serving organisations. Target populations therefore included child victims, adult perpetrators, child/youth offenders and adult participants of eligible interventions.
Child maltreatment. Any type of maltreatment that may occur in an institutional setting, including sexual and physical violence/abuse, neglect and emotional abuse. This may encompass: adults abusing children; children abusing other children; or institutions enabling child maltreatment.
Interventions. Any intervention that aimed to prevent, support the disclosure of, respond to or treat the consequences of child maltreatment within institutional settings (see Table 1). These interventions could be targeted at the primary, secondary or tertiary level, with a focus on victims, perpetrators or institutions.
Outcomes. Studies were included where they evaluated and reported outcomes that could be grouped under the following six outcome categories within the scope of the EGM framework: institutional safeguarding practice; disclosure; child maltreatment occurrence/reoccurrence (child safety), child wellbeing, adult perpetrator; child or youth offender; and parent/caregiver outcomes. Each of these broad outcome groupings were further broken down into subcategories (see Table 2). Studies that reported adverse outcomes that were explicitly evaluated were also included in the EGM.
Study type. Primary effectiveness studies (primary studies) and systematic reviews were included. Primary studies were included if they used a randomised controlled trial design, or a quasi-experimental design with a control/comparison group. Reviews were eligible if they reported a systematic literature search strategy, and a subset of studies reported met the eligibility criteria of this EGM. The following study types were excluded: noncontrolled pre-post evaluations; case studies; cross-sectional studies; observational studies; opinion pieces, editorials; studies solely employing qualitative methods. Both finalised and ongoing studies were included. No limitations were placed on the year of publication. Studies reported in the following languages were included: English, German, French, Spanish, Italian, Portuguese, Dutch, Danish, Swedish and Norwegian. Studies in other languages were excluded due to a lack of further available translation resources.
Information sources. Comprehensive electronic database searches were undertaken in July 2019, using key words relating to 'institutional settings' and 'child maltreatment', and included the following 10 databases: Medline, PsycINFO, CINAHL, ERIC, Informit Families and Society Collection (Australian), Sociology Source Ultimate, Sociological Abstracts, Scopus, The Campbell Collaboration Library and Proquest-Dissertations and Theses. Further searches were undertaken to identify grey literature. This included searches of four trial and systematic review registries; nine organisational websites; and the websites and reference lists of inquiry reports from seven international inquiries into child abuse. The search for grey literature was also  expanded based on input from multiple subject matter experts and project stakeholders. All reference lists of included systematic reviews were also screened for additional studies. Further details of the complete search strategy, including individual database search syntax, can be found in the published report (Finch et al., 2021).
Screening and study selection. Seven reviewers were involved in the screening and study selection process. Each title/abstract identified by the search strategy was screened against the selection criteria, by at least two reviewers working independently. The full text of studies that were deemed potentially relevant at the title/abstract screening stage were further assessed by two reviewers working independently. Any discrepancies in the decisions made by reviewers were resolved by an additional reviewer, or by discussion/consensus. Authors who were involved in any of the identified studies did not take part in the screening and selection of those studies. The Covidence platform (Veritas Health Innovation, 2020) was used for literature screening. No automation or text-mining was used to identify studies.

Data Extraction, Coding and Data Management
Five reviewers took part in data extraction, coding and data management. Information within each of the included studies was extracted and coded by two reviewers working independently. Any discrepancies in the decisions made by the first two coders were resolved by an additional reviewer, or by discussion/ consensus. Study authors were contacted to obtain missing information. Authors who were involved in any of the identified studies did not take part in data extraction, coding or critical appraisal of those studies. Data extracted from the studies included the following information: the publication/study (e.g. year, first author and country undertaken), study design (e.g. randomisation and comparator groups), institutional setting, target population, type of maltreatment, intervention type and outcomes. The final version of the coding scheme, with all data items, can be found in the published report (Finch et al., 2021).

Quality Appraisal
Critical appraisal of included systematic reviews was achieved using the AMSTAR 2 tool (Shea et al., 2017) and completed RCTs were assessed using the updated Cochrane Risk of Bias 2 tool (Sterne et al., 2019). Five reviewers were involved in assessing the quality of the included studies. Two reviewers worked independently to assess each study, and any discrepancies were cleared via consensus or by an additional member of the review team working independently. Only RCTs were assessed for their risk of bias; neither protocols, nor primary studies which were not RCTs, were assessed. Systematic reviews that received a 'critically low' or 'low' assessment using the AMSTAR 2 tool were combined into a single 'low' category.

Presentation and Synthesis
The visual EGM was developed using the R Project for Statistical Computing (R Core Team, 2019). As per the EGM framework, the included studies are mapped in a table in which the rows are the intervention categories, and the columns are the outcome categories. A single study appears in several cells on the map if it reported on more than one intervention category and/or more than one outcome category. Study quality has been highlighted by colour (i.e. low risk of bias/high-quality = white, some concerns of bias/moderatequality = light grey, high risk of bias/low-quality = dark grey). The visual EGM has been supplemented by a narrative synthesis of the included studies, which encompasses a descriptive summary of the number of studies included in the EGM, and their distribution across different areas of interest, including institutional settings, target population, maltreatment and intervention type.

Included Studies
The search strategy identified 73 studies (across 84 publications): 62 primary effectiveness studies and 11 systematic reviews. Figure 1 shows the flow of studies that were identified from the search strategy, screened and finally included in the EGM. The academic electronic search strategy yielded 6318 citations, and an additional 2375 records from other sources were identified. After removing duplicates and screening titles and abstracts, 256 citations remained for fulltext review. A total of 84 eligible publications were identified after full-text review and these reported 73 unique studies. All included studies are listed in the reference list, denoted by an asterisk. The following information can be found in the published report (Finch et al., 2021): details of the source of each publication; a list of excluded studies alongside their primary reason for being excluded.
Of the total 84 included publications, 12 were completed systematic reviews. Three of these were scoping reviews that met our systematic review criteria. Two reviews were related: Walsh, Zwi, Woolfenden, & Shlonksy (2015) was an update of an earlier Cochrane review published in 2007 (Zwi et al., 2008). While both are included in the EGM, where the reported characteristics are identical for each (e.g. maltreatment type, institutional setting and target population), they have been represented collectively (i.e. counted as a single study). Where the reported characteristics differ, they have been represented separately (i.e. counted as two separate studies). Seventy-two publications of primary effectiveness studies (primary studies) were identified: 69 were completed studies, and three were ongoing (described in a published protocol where results had not yet been generated; Baker-Henningham, Vera-Hernandez, Alderman, & Walker, 2016;McElearney et al., 2018;Ssenyonga, Hermenau, Nkuba, & Hecker, 2018). Among the primary studies, five publications related to a study evaluating the Good School Toolkit (GST), and reported different outcomes from the same sample or a subset of the same sample (Devries et al., 2015;Merrill et al., 2018). Six publications reported results of the Bucharest Early Intervention Project (BEIP) (Bick et al., 2015;Humphreys et al., 2015;Johnson et al., 2010;Smyke et al., 2010;Troller-Renfree et al., 2015;Wade, Fox, Zeanah, & Nelson, 2018). These publications reported the same or different outcomes at various follow-up points from the same sample of children originally randomised for the BEIP. Two further publications reported results from a school-based prevention programme (Fryer, Kraizer, & Miyoshi, 1987;Kraizer, Fryer, & Miller, 1988). Though all of the publications reporting on these three studies are included in the EGM, where the reported characteristics are the same for more than one study (i.e. study design, maltreatment type, institutional setting, target population and country), they have been represented as a single study (i.e. the multiple publications are counted as a single study). Where the outcomes reported across the papers were different, the individual publications have been reported separately (i.e. counted as two separate studies).

Visual Evidence and Gap Map
The visual EGM was based on the EGM framework (see Figure 2). The cells within the map show the number of studies for each study type (RCT, quasi-experimental (QED) and systematic review). A second visual EGM was developed to further highlight the institutional settings addressed by the included studies: early childhood settings, out-of-home care, school, youth services organisations, multiple settings (see Figure 3). Study quality has been highlighted by colour (i.e. low risk of bias/high-quality = white, some concerns of bias/ moderate-quality = light grey, high risk of bias/low-quality = dark grey).

Study Characteristics
Detailed information about the study characteristics of each individual study can be found in the published report (Finch et al., 2021).
Language, publication year, country. One primary study was published in German (Feldmann, Storck, & Pfeffer, 2018) and one primary study was in Spanish (del Campo Sánchez & Sánchez, 2006), with the remaining studies published in English. The earliest primary study included in the EGM was published in 1985. No more than four studies were published per year before or during 2011. Of the total studies, 54% were published between 2012 and 2020, with the peak number of completed primary studies published in 2018 (n = 10). The first systematic review was published in 1994, 9 years after the first primary study was published, with the four most recent reviews published in 2017. Just over half of the primary studies were conducted in the United States (n = 32). Canada produced four studies, three studies each came from Germany and the United Kingdom (one from Northern Ireland and two from Scotland), six countries produced two studies each (Turkey, Ireland, China, Spain, The Netherlands and Uganda) and a further eight countries contributed one study each (Australia, Ecuador, Indonesia, Jamaica, Malaysia, Romania, Taiwan and Tanzania).
Institutional settings. Most primary studies (n = 48) were conducted in school settings, including: primary/elementary school (n = 40), middle school (n = 1), secondary/high school (n = 7), kindergarten to year 12 (K-12) educational settings (i.e. schools inclusive of all years) (n = 1) and out-ofschool-hours care programmes (n = 2). Two studies included more than one of these school settings (Barron & Topping, 2013;Ssenyonga et al., 2018). Eleven primary studies were conducted in early childhood settings (e.g. kindergarten, preschool and day-care), and three of these also included primary/elementary school settings (Fryer et al., 1987;Kraizer et al., 1988;Kraizer, 1991;Wurtele et al., 1986). Two primary studies were conducted across multiple settings, which included: health, school and social services agencies who respond to child maltreatment (Cerezo & Pons-Salvador, 2004), and organisations delivering services that children access or attend (e.g. schools, day-care and church organisations) (Rheingold et al., 2014). Three were conducted in out-ofhome care, including foster care and orphanages (the BEIP study and associated publications), a residential school for the deaf (Sullivan et al., 1992) and group homes (Van Lieshout et al., 2019). No studies were identified where the primary setting was sports clubs, religious organisations, summer camps, detention centres or primary and secondary health care facilities.
Most of the 11 systematic reviews reported on studies conducted exclusively in school and/or early childhood settings (i.e. kindergartens, preschool and day-care) (n = 5). Of these, one systematic review (plus, one update) included studies reporting both primary/elementary and secondary/high school settings (Walsh et al., 2015;Zwi et al., 2008), one systematic review included early childhood and primary/ elementary settings (Heidotting, 1996), one included middle and secondary/high school settings (Ricardo, Eads, & Barker, 2011), one systematic review included several settings (across early childhood, primary/elementary and secondary/high school) (Topping & Barron, 2009) and one included only early childhood settings (Pitts, 2015). Of the remaining five systematic reviews, four focused on residential care (e.g. orphanages and out-of-home care) (Hermenau et al., 2017;McKibbin, 2017;Sherr, Roberts, & Gandhi, 2017;South et al., 2015). Two systematic reviews included studies conducted across various settings (Quadara, Nagy, Higgins, & Siegel, 2015;Radford, Foster, Barter, & Stanley, 2017), including school and early childhood settings, voluntary and faith-based organisations and sports clubs (coded as 'multiple settings').
Most reviews (n = 7) examined interventions solely for children, one included evaluations of interventions targeting only institutional staff and/or adult care providers (e.g. teachers) (Hermenau et al., 2017), and four reviews included studies assessing interventions for either or both of these populations (Quadara et al., 2015;Radford et al., 2017;Sherr et al., 2017;South et al., 2015).
Types of maltreatment. Most completed and ongoing primary studies included interventions that had a singular focus on sexual abuse (n = 46), with 10 additional studies focussing on sexual abuse alongside other maltreatment types (total sexual abuse: n = 57). Four primary studies assessed interventions specifically addressing physical abuse, and a further 10 incorporated physical abuse alongside other maltreatment types (total physical abuse: n = 14). Child neglect was the primary focus of the BEIP study, and three other studies also addressed neglect alongside other maltreatment types (total neglect: n = 4). No study focussed exclusively on emotional abuse, but emotional abuse was considered in the GST study and four others (total emotional abuse: n = 5).
Of the 11 systematic reviews, eight (plus one update) included studies that reported on interventions relating solely to sexual abuse. The other three systematic reviews included primary studies that reported on one or more types of child maltreatment. Hermenau et al. (2017) and Sherr et al. (2017) included studies assessing physical and emotional abuse, as well as neglect, and Ricardo et al. (2011) included studies reporting on sexual, physical and emotional abuse.

Quality Appraisal
Of the 49 reports of completed RCTs, all were assessed to have either a 'high risk' of bias (n = 18) or 'some concerns' (n = 31) using the Cochrane Risk of Bias 2 tool (See Figure 3). No study received an overall assessment of low risk. Refer to the published report for the overall risk of bias assessment and assessments for individual studies (Finch et al., 2021). Overall, most systematic reviews (n = 10) were assessed as being of low quality (i.e. low confidence in the reported results) using the AMSTAR 2 checklist (Heidotting, 1996;Hermenau et al., 2017;McKibbin, 2017;Pitts, 2015;Quadara et al., 2015;Radford et al., 2017;Ricardo et al., 2011;Sherr et al., 2017;South et al., 2015;Topping & Barron, 2009). Two received a high-quality rating (i.e. high confidence in the reported results) (Zwi et al., 2008;and update Walsh et al., 2015).

Interventions
Following the EGM framework and coding, interventions were categorised as prevention, disclosure, response or treatment approaches. For a further detailed synthesis and description of the interventions reported within the studies, refer to the extended report (Finch et al., 2021).
Prevention. The overwhelming majority of studies assessed the effectiveness of prevention interventions (n = 58 primary studies; n = 5 systematic reviews), and a smaller number included prevention approaches alongside other intervention types (n = 2 primary studies; n = 5 systematic reviews). Prevention interventions were defined as any intervention where the primary aim was to decrease the likelihood or risk of child maltreatment occurring or recurring in the future. This encompassed universal interventions for any child or adult, as well as interventions aimed at specific populations. Examples of the types of prevention interventions that could be included were school-based safety programmes, organisational guidelines/practices and interventions to reduce perpetrator reoffending.
Most prevention interventions were delivered in an educational setting and were curriculum-based, with a focus on increasing child awareness and understanding of sexual abuse, and teaching self-protection skills (n = 54). For most (n = 43), the main intervention involved workshops or lessons, alongside written, audio-visual or other resources (e.g. films or plays, images, activity books and parent resources), and was delivered directly to children in groups via an external agency or existing trained institutional staff or students. The intensity of these sessions varied from brief standalone educational programs involving single sessions (Daigneault, Hebert, McDuff, & Frappier, 2012;Pulido et al., 2015); delivery of between two to eight lessons over the course of 1-2 weeks (Cecen-Erogul & Hasirci, 2013;Conte, 1985;Dake, Price, & Murnan, 2003;Fryer et al., 1987;Jin, Chen, Jiang, & Yu, 2017;White et al., 2018;Wurtele, Gillispie, et al., 1992), and more intense delivery with multiple lessons delivered over longer periods ranging from 5 to 10 weeks (Citak Tunc et  Twelve primary studies reported on interventions that aimed to improve the knowledge, attitudes and practices of the organisation's staff via trainingsome with and some without follow-up supportin educational settings (n = 9), and multiple youth service organisations (n = 2). Among these interventions, training for staff ranged from a brief 1 hour session (Gushwa et al., 2018), up to 5 days (Ssenyonga et al., 2018), with inclusion of follow-up support strategies such as in-school coaching (Baker-Henningham et al., 2016;Dryden et al., 2014), performance feedback and text messaging (Baker-Henningham et al., 2016), and supervision and peer networks (Ssenyonga et al., 2018). Two of these interventions involved online or web-based delivery (Rheingold et al., 2014;Gushwa et al., 2018).
Four school-based prevention interventions used more comprehensive approaches, seeking to embed the programme across the broader school community, and included multiple strategies (combined with curriculum approaches) delivered over a longer timeframe (from two terms, up to a year) (Baker-Henningham et al., 2016;GST study;McElearney et al., 2018;Ratto & Bogat, 1990). For example, the GST was aimed at multiple levels within the schools including head teachers, administration, classroom teachers and students with multilayered training, processes and school-led activities for each level.
Disclosure. Disclosure interventions were defined as any intervention that aimed to facilitate, support or promote the disclosure of child maltreatment. This encompassed a range of universal interventions, such as traditional or social media campaigns, or child helplines, as well as therapeutic interventions for children that aimed to promote disclosure (e.g. play therapy). It included tertiary interventions relating to perpetrators, such as mandatory reporting, and also included any intervention that aimed to promote disclosure within an organisational context (e.g. staff training and organisational guidelines).
No primary studies were evaluations of interventions aiming solely to facilitate disclosure of child maltreatment. However, multiple prevention interventions included components that aimed to provide children with knowledge and/or skills to disclose maltreatment to a trusted adult. Nine studies evaluating these interventions included participant rates of disclosure either during or directly after participation (Czerwinski, Fine, Alfes, & Kolip, 2018;Daigneault et al., 2015;Devries et al., 2015;Barron & Topping, 2013;del Campo Sánchez & Sánchez, 2006;Hazzard et al., 1991;Kolko, 1987;Kolko et al., 1989;Oldfield, Hays, & Megel, 1996), and two studies specifically assessed disclosures rates of children currently suspected of experiencing abuse, who had at some point in the past taken part in one of these programs (MacIntyre & Carr, 1999b;Elfriech et al., 2020). Two reviews searched for primary studies assessing interventions aiming to increase disclosure (Quadara et al., 2015;Radford et al., 2017). However, these reviews either did not identify primary studies, or did not identify primary studies that met our inclusion criteria for disclosure interventions.
Response. Response interventions were defined as any intervention that aimed to improve institutional responses to the occurrence of child maltreatment in relation to each of the target populations. Response interventions included legal or regulatory mechanisms aimed at introducing new procedures for institutions to follow, organisational guidelines and/or practices (e.g. response framework), support for the victim and/or family, working with child protection agencies and providing training and/or crisis support to staff within organisations. Response interventions were evaluated by fewer studies (n = 2 primary studies; n = 5 systematic reviews), and for all bar one systematic review (Hermenau et al., 2017), these were reported alongside, or combined with, preventionfocused interventions.
Cerezo and Pons-Salvador (2004) used a quasiexperimental approach to assess a largescale intervention that aimed to increase detection of child maltreatment across a single territory in Spain. The intervention involved professional training based on motivational interviewing approaches and support. It was delivered in multiple settings to professionals from all frontline health and social services agencies, and schools, in the territory. An RCT reported by Rheingold et al. (2014) compared a web-based and in-person training versions of an intervention with a dual focus on preventing and responding to child sexual abuse among children (ranging from 0 to 18 years). It was delivered to staff from youth service organisations including schools, churches, day-care, extracurricular activity agencies, state agencies, group home/ residential settings and healthcare settings. The programme included education about child sexual abuse, ways to minimise child sexual abuse, how to recognise the signs and how to respond appropriately when a child discloses (Rheingold et al., 2014).
Only one of the five reviews focussed solely on interventions relating to institutional responses to child maltreatment (Hermenau et al., 2017). This review investigated the effects of structural interventions and caregiver trainings on child development, for children living full time in institutional care environments across the world (e.g. orphanages and residential care). It included interventions that aimed to change the organisational structure and culture of the institutions, as well as the ways in which caregivers interact with children. The four remaining reviews included studies assessing response interventions alongside other intervention types (Quadara et al., 2015;Radford et al., 2017;Sherr et al., 2017;South et al., 2015).
Treatment. Treatment interventions were assessed by fewer studies still (n = 2 primary studies; n = 2 systematic reviews). Treatment interventions were defined as any intervention that aimed to provide a therapeutic response to a target population. This included therapeutic interventions provided to children who experienced child maltreatment in institutions, and interventions targeted at perpetrators of institutional child abuse.
Two primary studies assessed the effectiveness of treatment interventions, including the six reports describing the BEIP. The BEIP study randomly assigned children in Romanian orphanages to remain in institutional care or be removed and placed in high-quality foster care (the treatment intervention). Sullivan et al. (1992) used a quasi-experimental approach to assess the effectiveness of a treatment intervention for hearing-impaired children between the ages of 12 and 16 years, who had been sexually abused while attending a residential school for the deaf. The psychotherapeutic intervention was offered to the children by the school and involved 2 h of individual therapy per week for 36 weeks, delivered by a clinical psychologist and a supervising psychiatrist with expertise in the psychology of deafness and fluency in sign language (Sullivan et al., 1992).
A scoping review by McKibbin (2017) examined treatment interventions focused on harmful sexual behaviour and child sexual exploitation among children and young people living in out-of-home care. This review included 17 papers describing interventions, including treatment interventions, for young people who display harmful sexual behaviour. The authors concluded that evidence about the elements of a successful tertiary prevention response, including trauma-informed therapeutic treatment was well-developed particularly in the UK. The review by Quadara et al. (2015) examined prevention, early intervention and therapeutic responses to child sexual abuse and described one study comparing children in institutions with home-based care as a form of treatment.

Outcomes
This section describes the outcomes of interest to the EGM that were measured and reported across the included studies. For a further detailed synthesis and description of the outcomes reported across the studies, refer to the extended report (Finch et al., 2021).
Two recently published systematic reviews included interventions delivered in out-of-home care settings (Hermenau et al., 2017;McKibbin, 2017). Hermenau et al. (2017) included studies that assessed interventions aimed at improving the quality of care in institutional environments, reporting a broad range of outcome measures and measurement instruments used to assess changes in caregiving and institutional quality and attachment. They included institutional safeguarding practice outcomes relating to both operational practice (e.g. measures assessing changes in caregiving quality, child-caregiver ratios) and the institutional environment (e.g. measures of environmental quality and structural changes to the institutional environment). A scoping review by McKibbin (2017) included studies reporting on interventions addressing harmful sexual behaviour and child sexual exploitation for children and young people living in residential care. The reported institutional safeguarding practice outcomes were related to operational practice, and included outcomes measuring staff members' knowledge about, and skills relating to, recognising children's problematic sexual behaviour (McKibbin, 2017).
Child maltreatment disclosure. Eleven reports of primary studies reported outcomes relating to child maltreatment disclosure (Czerwinski et al., 2018;Elfreich et al., 2020;Daigneault et al., 2015;Devries et al., 2015;Barron & Topping, 2013;del Campo Sánchez & Sánchez, 2006;MacIntyre & Carr, 1999b;Hazzard et al., 1991;Oldfield et al., 1996;Kolko et al., 1987Kolko et al., , 1989. All of these studies evaluated school-based interventions aiming to prevent child maltreatment. Outcome measures included: participant, teacher and/or parent reported disclosure of sexual abuse over the course of the intervention and evaluation (Barron & Topping, 2013;del Campo Sánchez & Sánchez, 2006;Hazzard et al., 1991;Kolko et al., 1987;1989;Oldfield et al., 1996); child reported courses of action in response to hypothetical scenarios, including: possible disclosure options (Czerwinski et al., 2018), childreported likelihood of future disclosure (Kolko et al., 1989), youth recognition of sexual assault and response to a hypothetical disclosure of sexual assault (Daigneault et al., 2015), and students' self-reports of physical violence from school staff (assessed in a follow-up survey) (Devries et al., 2015). Two studies (Elfreich et al., 2020;MacIntyre & Carr, 1999b) specifically assessed disclosure rates of children who were suspected of experiencing maltreatment and who had at some point in the past taken part in a school-based prevention programme. MacIntyre and Carr (1999b) reported children's disclosure of sexual abuse after they had been referred to a sexual abuse assessment unit, and Elfreich et al. (2020) assessed child disclosure of abuse during forensic interviews.
Three systematic reviews included studies reporting on interventions' impact on disclosure-related outcomes. A highquality review by Walsh et al. (2015) (an update of Zwi et al., 2008) included school-based sexual abuse programmes, and reported on disclosure of sexual abuse by child or adolescent participants during or after undertaking a programme. Pitts (2015) included studies that reported on the disclosure of child sexual abuse. Radford et al. (2017) also included studies that reported on measures of safe disclosure (e.g. rates of disclosure) to peers, adults, institutions and services, including disclosure of nonrecent abuse.
Child safetymaltreatment occurrence or reoccurrence. Thirteen completed and ongoing primary studies reported/will report outcomes related to child maltreatment occurrence/ reoccurrence (GST; Nkuba et al., 2018;Taylor et al., 2010;Baker-Henningham et al., 2016;McElearney et al., 2018;Ssenyonga et al., 2018). Eleven studies evaluated interventions focused on preventing maltreatment in educational settings (e.g. schools and day-care), with most addressing physical violence. Outcome measures included: student selfreported violence perpetrated by staff (Good School Kit); teacher and student reports of emotional and physical violence ; and student-reported exposure to violence (Ssenyonga et al., 2018); and teacher-reported use of violent disciplinary methods (Ssenyonga et al., 2018). A further two QED studies used a child-report questionnaire to determine childrens' experiences of inappropriate touching involving an uncomfortable or potentially abusive interaction (Kolko et al., 1987(Kolko et al., , 1989. Three systematic reviews included interventions delivered in residential care settings that reported on child maltreatment occurrence/reoccurrence (Hermenau et al., 2017;Sherr et al., 2017;South et al., 2015). Outcomes examined included: selfreports or observations of maltreatment from staff/adults (physical/emotional), as well as peer to peer violence in institutional care (Sherr et al., 2017), sexual abuse in out-ofhome care (South et al., 2015), exposure to violence of children living in a child care institution (Hermenau et al., 2017), and documented abuse in official records (Sherr et al., 2017).
Child wellbeing. Child wellbeing outcomes were coded into five subcategories: knowledge and awareness, mental health, cognitive functioning, social functioning, and health and development.
Across the child wellbeing outcome subcategories, more primary studies reported outcomes relating to knowledge and awareness (n = 51) than primary studies reporting mental health outcomes (e.g. internalising and externalising behaviours, self-esteem and emotional intelligence; n = 23), outcomes relating to child cognitive functioning (e.g. education scores, memory and executive functioning; n = 5), social functioning (e.g. communication, self-control and social connections and relationships; n = 6) or physical health and development (e.g. brain development; n = 2). All 51 studies reporting child knowledge and awareness outcomes evaluated curriculum-based prevention interventions delivered in educational settings, with most focussing on child sexual abuse. Fewer studies assessed changes in participant knowledge and awareness about other child maltreatment types, such as physical or emotional abuse (n = 7) (Barron & Topping, 2013;Dake et al., 2003;Dhooper & Schneider, 1995;Dryden et al., 2014;Edwards et al., 2019;Kraizer, 1991;Wolfe, MacPherson, Blount, & Wolfe, 1986).
Nine systematic reviews included studies that assessed interventions' impact on child knowledge and awareness (e.g. knowledge of child sexual abuse and protective behaviours, attitudes towards violence and knowledge of safe sexual relationships; Walsh et al., 2015;Heidotting, 1996;Pitts, 2015;Quadara et al., 2015;Radford et al., 2017;Topping & Barron, 2009;Sherr et al., 2017;McKibbin, 2017;Ricardo et al., 2011). Six systematic reviews included studies that assessed an interventions' impact on child mental health outcomes (e.g. externalising and internalising symptoms, post-traumatic stress symptoms, self-esteem and aggression; Walsh et al., 2015;Pitts, 2015;Topping & Barron, 2009;Sherr et al., 2017;Hermenau et al., 2017). One systematic review that evaluated interventions delivered in institutionalised care, included studies that reported child cognitive functioning outcomes, including child mental development, language development and intelligence (Hermenau et al., 2017). Hermenau et al. (2017) was the only systematic review that included studies evaluating interventions' impact on child social functioning, including outcomes relating to children's social-emotional competencies and skills, as well as attachment and caregiver relationships. This was also the only review that reported on child physical development and health outcomes, including psychomotor development specifically, as well as general development (including cognitive, language, social-emotional development).
Outcomes related to adult perpetrators or child/youth offenders. Two studies reporting outcomes relating to adult perpetrators and child/youth offenders. Baker-Henningham et al. (2016) reported on an ongoing RCT for a prevention focused study that plans to include observations of teachers' use of violence against children in day-care settings in Jamaica. Edwards et al. (2019) evaluated a bystander-focused interpersonal violence prevention programme with high school students in the United States, reporting self-reported youth offender outcomes including sexual harassment, sexual assault and stalking victimisation and perpetration among high school students.
No systematic review reported outcomes for adult perpetrators or child/youth offenders that specifically related to child maltreatment that occurred in an institutional setting.
Parent or caregiver behaviour, knowledge or attitudes. Five studies reported parent or caregiver behaviour, knowledge or attitudes (Merrill et al., 2018;Wurtele, Gillispie, et al., 1992;McElearney et al., 2018;Kolko et al., 1987). Wurtele, Gillispie, et al. (1992) compared teachers and parents as instructors of a personal safety programme delivered to preschool children and assessed parents' perceptions of their child's understanding of protective behaviour concepts, and their application of those behaviours. Merrill et al. (2018) assessed changes in parental normative beliefs relating to school-based physical discipline when assessing the GST programme (Merrill et al., 2018). In their evaluation of a multicomponent 'whole-school' programme designed to teach 4-11 year olds how to keep safe from all forms of maltreatment, McElearney et al. (2018) will assess parents' confidence in talking to their children about keeping safe. Kolko et al. (1987) reported changes to parental knowledge about sexual abuse when evaluating a school-based sexual abuse prevention intervention.
No systematic review reported outcomes related to parent or caregiver behaviour, knowledge or attitudes.
Intervention implementation. Of the primary studies, 23 reported outcomes relating to the implementation of the intervention, including one ongoing study (Ssenyonga et al., 2018). Outcomes representing feasibility (i.e. the utility, fit or practicality of the implemented programme), adoption (i.e. uptake or utilisation of the intervention), fidelity (i.e. the degree to which an intervention was implemented as it was intended), acceptability (i.e. perception among implementation stakeholders that an intervention is satisfactory in relation to content, complexity, comfort, delivery and credibility) and penetration (i.e. reach, spread and institutionalisation) (Proctor et al., 2011), were reported across these studies. Aspects of fidelity were assessed in 15 studies, acceptability was reported in 11, five studies reported aspects of penetration (Devries et al., 2017;Nkuba et al., 2018;Ssenyonga et al., 2018;White et al., 2018), and feasibility Ssenyonga et al., 2018) and adoption (Devries et al., 2017; were each reported by two studies. Adverse outcomes. An EGM is not designed to report on the direction of findings in relation to the reported outcomes. Therefore, we cannot report adverse effects on outcomes where the intervention had a negative effect but was hoped to have a positive effect (e.g. knowledge of sexual abuse). However, some studies included specific outcomes that aimed to capture adverse effects. For example, measures of anxiety, and fear and touch aversion, were commonly used to assess whether education-based prevention programmes targeting the sexual abuse of children had a negative effect on their wellbeing. For the most part, these programmes did not appear to adversely impact children. A single study (Taylor et al., 2010) reported that an intervention addressing gender violence and sexual harassment, delivered to sixth and seventh graders, reduced peer violence victimisation and perpetration, but may have increased dating violence perpetration, or at least the reporting of it.

Discussion and Applications to Practice
Overall, there were more evidence gaps across the EGM than areas with high-quality evidence, and most of the studies included in the EGM were low to moderate quality. Therefore, any assessment of effectiveness of the interventions on the reported outcomes should be cautiously interpreted. This, combined with the fact that most studies were published in the last 5 years, suggests that empirical research on the effectiveness of interventions addressing child maltreatment in institutions is very much at an early stage and highlights a considerable need for further research.
Most studies focused on children, on prevention and on sexual abuse specifically. This is not proportionate to the prevalence of different maltreatment types. For example, current estimates suggest that physical abuse is at least as prevalent as sexual abuse (WHO, 2014). Most studies targeted universal child populations, with far fewer targeting children who are high-risk or who had already been exposed to maltreatment. That most of the studies evaluate interventions for children raises some concerns that could reasonably be framed as 'unintended harm'. While children have rights to provision, protection and participation in areas that affect them, relying solely on interventions focussing on children potentially places the burden of responsibility of prevention and disclosure of child maltreatment on children, rather than on perpetrators of abuse or on the organisations that serve them. Of further concern is that, by focussing on children in this way, the child may feel responsible or may be blamed if maltreatment occurs. Sadly, there is a long history of blaming the victim, especially with respect to violent sexual offenses such as rape. Asking the question, 'What can I do differently to prevent becoming a victim?' can easily translate into selfblame if maltreatment occurs.
There is a clear need for more high-quality evaluations of interventions that cover the whole spectrum of players that this issue concerns, including children, perpetrators, adults in institutional environments, as well as the institutional environment itself. In some cases, this kind of research can be unpopular and difficult to promote or fund, such as research on offenders. However, in this instance the onus of responsibility should be on governments, funding agencies, criminal justice systems and the institutions themselves, who have been unwilling or unable to fund offender prevention, response, and treatment interventions now and in the past.
Education and early childhood settings were by far the most well-studied. This is perhaps not surprising, given that most children have more contact with schools than any other institution and studying children in schools is relatively easy. In contrast, evidence assessing the effectiveness of interventions across other institutions, such as out-of-home care (e.g. foster care, residential care and orphanages), was very limited. For many types of institutions within our scope, there were no specific studies at all. For instance, there were no studies specifically targeting religious organisations, sports clubs or other recreational settings.
There are several potential explanations for this. While it is certainly the case that many institutional settings have not adequately addressed child maltreatment, there are also instances where interventions have been implemented, but have not yet been evaluated or have not been evaluated in an institutional context. The EGM's eligibility criteria excluded studies that did not explicitly define an institutional setting. However, there are also evaluations that either focus on maltreatment in family settings, or do not specify where the maltreatment occurred. Taking this into account, it is possible that existing evidence-based interventions targeting general populations, or specific populations outside of an institutional setting, may also be effective, or may be adapted and be effectively used within an institutional context. For instance, interventions targeting sexual abuse perpetrators could possibly be adapted to specifically target people who perpetrated sexual abuse in an institutional setting. Or, interventions targeting staff in schools may be adapted to target staff in other organisational contexts. However, this approach has limitations: institutional environments are diverse, and one-size-fitsall interventions are unlikely to be effective without at least some modifications. There are also differences in risk factors for perpetrators and victims, as well as differences in the experience, perpetration and response to maltreatment both across different institutional settings, and when compared to other settings where maltreatment occurs (Quadara et al., 2015;Radford et al., 2017). These factors would need to be considered when adapting existing interventions to (other) institutional environments.
Though the studies look at many countries, the evidence is dominated by studies undertaken in the United States and Europe. It is clear therefore, that the available research does not currently represent countries with the largest populations (and, potentially the greatest incidence of child maltreatment), nor does it represent countries with the highest estimated prevalence of child maltreatment (WHO, 2014).
Most of the included studies assessed education or skillsbased interventions delivered to children. Fewer were delivered to institutional staff, teachers and/or adult care providers. There was a lack of evidence targeting adult perpetrators and only one study of youth offenders, and where evidence was identified, the focus was primarily on children who display sexually aggressive behaviour toward other children. Most interventions focussed on sexual abuseand specifically on preventing sexual abuse. Though we did identify a cluster of studies focused on addressing physical violence in schools (including harsh discipline), far fewer interventions targeted other maltreatment types.
A further major gap was identified in relation to studies evaluating interventions that specifically aimed to improve disclosure. Interventions with a particular focus on disclosure were not studied in any of the primary studies that were found and included in only one systematic review. However, there were a number of studies reporting on disclosure outcomes relating to prevention programs which included disclosure components. Evidence supporting the effectiveness of organisational response-based approaches was lacking in both breadth and quality. Of the small number of studies, only one was an RCT, which evaluated a very brief staff training intervention. Studies that assessed treatment interventions that addressed child maltreatment experienced or perpetrated in institutional settings were also extremely limited and solely focused on out-of-home care settings. Prevention-based interventions were by far the most highly represented group of interventions. Of these, most reported on school-based interventions that primarily aimed to provide children with knowledge and skills to better protect themselves from maltreatment, often with elements geared toward normalising and promoting help-seeking.
The predominance of curriculum-based interventions in education settings targeted toward children is also reflected in the outcomes presented in the EGM. Across all the included studies, outcomes relating to child knowledge and awareness were reported more than any other type of outcome. Child mental health and maltreatment occurrence outcomes were also reported in a substantial number of studies. It is perhaps not surprising, given the nature of child maltreatment and its measurement in institutional contexts, that these studies mostly focused on short-term, self-report risk indicators for maltreatment rather than measurements of whether maltreatment actually occurred. Overall, reported outcomes tended to focus on children, and not perpetrators. Direct measures of perpetrator maltreatment behaviours, recidivism and desistence were included in only two primary studies.
Despite lowering our inclusion criteria for primary studies well below the RCT threshold, there was scarce evidence reporting outcomes relating to institutional safeguarding practices that may better support the prevention, disclosure, and organisational responses to child maltreatment. Unfortunately, these gaps may be due to a lack of concerted, rigorous efforts at evaluation within institutional settings. Though the reasons for this are unknown, it is potentially associated with a reluctance to look closely at institutional failures and to evaluate them in a way that builds the knowledge base for prevention work in this area. The past has seen a larger research focus on maltreatment in family/home settings than in institutions. This is only now being challenged as victims of child sexual and physical abuse recount their experiences, seeking justice and restitution, sparking numerous inquiries across the world. Hopefully, this level of scrutiny and a demand for a meaningful response will translate into a growing number of safeguarding approaches that are rigorously evaluated.
Finally, only a third of the studies reported one or more outcomes that related to implementation. These included measures of feasibility, adoption, fidelity, acceptability, and intervention penetration. Implementation outcomes are 'the effects of deliberate and purposive actions to implement new treatments, practices, and services' and describe the result of intentional actions to deliver a policy or an intervention (see Proctor et al., 2011). Measuring implementation is important in determining how or whether an intervention was delivered as intended, information that is essential to ascertaining its effectiveness. Moreover, the effectiveness of an intervention may be compromised by insufficient attention to implementation. Measures of implementation also provide information about whether an intervention is acceptable to participants, and/or whether it is likely to be successfully adopted in real life contexts. The fact that most studies in the EGM did not report on measures of implementation is concerning, given that many studies reported on interventions which were delivered by multiple individuals (e.g. practitioners and trained staff) and across multiple study sites. This creates ample scope for variation in what gets delivered, which may impact the reliability of a study's findings.
Overall, the evidence included in the EGM is sparse and of low to moderate quality. There is much need for further highquality research, specifically: evaluating interventions in a broader range of institutions; in countries with the largest populations, in which the greatest prevalence of child maltreatment in institutions is likely to occur; assessing interventions that focus on perpetrators and the organisational environment (as well as children); assessing interventions addressing a broader range of maltreatment types (i.e. not only sexual abuse); assessing interventions focussed on disclosure, organisational responses, and treatment (both victim/survivor and offender); assessing interventions targeting perpetrators, maltreatment behaviours, recidivism, and desistence; and assessment and reporting of implementation outcomes.
Given that the evidence-base for interventions specifically addressing institutional child maltreatment is sparse, it is difficult to assess whether an intervention which achieved some result in one location or setting would achieve that same result elsewhere. For instance, could school-based education and skills training interventions be appropriately translated to other institutions and/or other populations? For example, could the kind of child trainings which have been studied schools be delivered at Scouts? Could training for teachers be delivered to clergy? Could effective institutional safeguarding practices or policies be adapted to other organisational contexts and/or personnel? Are treatment principles for children who experienced maltreatment in other settings appropriate and effective for children who experienced maltreatment in an institutional environment? Or, are interventions for perpetrators as effective with populations of perpetrators who abused within an institutional setting?
Clearly, interventions that are moved from one type of setting to another may not work as well there. This highlights the importance of continuing to evaluate an intervention when it is delivered somewhere other than the setting it has been shown to be effective in. For example, if an education and training intervention which has been effective when delivered in schools, is used in a sports or recreational setting, it should be further evaluated there. Similarly, an education and training intervention for school staff to prevent child physical and emotional abuse may be transferrable to coaches and mentors in sport and recreation contexts but would need to be evaluated with those personnel in their contexts.

Limitations
The EGM involved an extensive and rigorous search for peer reviewed and grey literature and examined over 8000 citations. We also sought relevant studies from contact with experts in the field. Despite this, it is possible that some studies relating to institutional responses to child maltreatment were missed. When screening at the title or abstract level, we may have incorrectly excluded some studies where information provided did not clearly reveal relevance to the setting or topic. Similarly, some relevant studies with crossover to settings outside the scope of our EGM, including health or clinical settings, may have been excluded based on the setting criteria. Snowballing techniques were not used for screening primary studies, and though we screened the primary studies included in the included systematic reviews, we did not screen all the studies in their reference lists.
Though the search terms were carefully designed, and piloted, relevant studies could still have been missed because of our included terms or because of variations in database indexing. There may have been studies in other languages that were not picked up by our search strategy, or studies that used different language/terms to describe institutional settings or child maltreatment. We will further assess the appropriateness of search terms in future updates to ensure that the search terms include relevant terminology.
Finally, due to unclear reporting, it was at times difficult to categorise intervention type, and identify the exact institutional setting where the intervention was delivered or where the abuse took place. As a result, we categorised the information based on what was available, and at times, some assumptions were necessary.

Conclusion
This EGM shows a need for more high-quality studies that assess interventions across a broad range of institutional contexts and maltreatment types. The evidence gaps are particularly evident for countries with large populations, and therefore the greatest number of children affected by child maltreatment. Few studies focussed on perpetrators or organisational environments. Evidence gaps were also identified for interventions relating to disclosure, organisational responses and treatment, and few studies were identified that assessed an intervention's impact on perpetrators' maltreatment behaviours, recidivism and desistence. There is also need for more studies to measure and report on implementation.

Author Note
Porticus commissioned this review to support its own and others' ongoing work to enhance organisational safeguarding. To ensure that all standards to produce a Campbell Collaboration EGM were met, Porticus was not involved in any technical steps taken to produce the EGM, including information retrieval, data analysis and reporting of findings.
Prof. Aron Shlonsky, Dr. Robyn Mildon and Dr. Bianca Albers have co-authored publications commissioned by the Australian Royal Commission into Institutional Responses to Child Sexual Abuse.