Mapping review of interventions to reduce the use of restrictive practices in children and young people's institutional settings: The CONTRAST study

Abstract Restrictive practices are often used harmfully with children in institutional settings. Interventions to reduce their use do not appear to have been mapped systematically. Using environmental scanning, we conducted a broad‐scope mapping review of English language academic databases, websites and social media, using systematic methods. Included records (N = 121) were mostly from the United States and contained details of 82 different interventions. Children's participation was limited. Reporting quality was inconsistent, which undermined claims of effectiveness. Overall, despite a multitude of interventions, evidence is limited. Leaders should consider the evidence, including children's perspectives, before introducing poorly understood interventions into children's settings.


Strategies to address RP reduction
There is at the very least, a 'delicate balance' between restraint for the purposes of care, and causing preventable harm (Preisz & Preisz, 2019):1165. Previous research has explored strategies to reduce RP with adults in mental health (e.g. Bowers et al., 2015;National Association of State Mental Health Program Directors, 2006;Riley & Benson, 2018) and learning disabilities settings (Bowers et al., 2015;Deveau & McDonnell, 2009;Luiselli et al., 2004;Putkonen et al., 2013). There is limited empirical data, primarily based on case studies of single facility initiatives (Delaney, 2006;LeBel et al., 2010), that interventions effectively reduce RP use specifically with children in mental health services (Azeem et al., 2011;De Hert et al., 2011;LeBel et al., 2010;LeBel & Goldstein, 2005;Schreiner et al., 2004). Some of these interventions have been the subject of systematic reviews (e.g. Bowers et al., 2015), but the range of interventions implemented in practice does not appear to have been examined previously. Therefore, as a first step in understanding how restrictive practices may be reduced and/or applied without causing harm, this study aimed to identify and systematically map all available interventions seeking to reduce RP in children's institutional settings. It asked: What is known about interventions to reduce RP in children's institutional settings?

METHODS
The study design was a mapping review that used systematic methods (Bradbury-Jones et al., 2019;Carter et al., 2019;Clapton et al., 2009;Cooper, 2016;National Collaborating Centre for Mental Health, 2015;Perryman, 2016;Pham et al., 2014) and followed PRISMA reporting guidelines (Page et al., 2021). The protocol was registered online (National Institute for Health and Care Research, 2020).

Search strategy
It was known that there were numerous small-scale, standalone initiatives available for implementation in services, in addition to the small number of well-known interventions published in academic journals. Therefore, the search applied 'environmental scanning' (Parker et al., 2018), and included academic sources (ASSIA, BNI, CINAHL, CD and AS, CJA, Education Abstracts, EMBASE, ERIC, MEDLINE, PsycINFO, Scopus), grey literature and social media aimed at a global coverage. The method involved systematically searching, retrieving and reviewing all reports irrespective of effectiveness evidence, with a focus on ascertaining the range and characteristics of interventions.
An 'intervention' was any documented approach to reduce the use of RP, for example a RP training manual and a RP reduction programme described in an academic study would both be classed as interventions. Searches were developed for the following concepts: child or child behaviours; restraint practices or named programmes and a variety of institutional, healthcare and educational settings. Further detail of the search strategy is published separately (King et al., 2022).
The search was limited to English language reports dating from 1989 (Children Act, Stat, 1989). Searches were peer-reviewed and conducted June-August 2019, updated January 2020.
Additional information about interventions was obtained via email requests to authors and organisations. The full search strategy is accessible via: https://doi.org/10.5518/1077. Table 1 summarises the inclusion criteria. No restrictions regarding study design or quality were imposed. Ineligible interventions solely involved policy change or aimed to reduce the use of one type of RP by replacing it with another (Bradbury- Jones et al., 2019;Carter et al., 2019;Clapton et al., 2009;Graham et al., 2008;Hong et al., 2018;Pace et al., 2012;Perryman, 2016;Pham et al., 2014).

Data management and review
Records were managed within reference management software Endnote version X9 (Clarivate Analytics, 2018). Two reviewers (KB and KC) jointly screened titles/abstracts and full texts before independently assessing them against the inclusion criteria and then discussing and resolving any disagreements.

Quality appraisal
The purpose of quality appraisal was to understand the scope of the literature and not to exclude records. The Mixed Methods Appraisal Tool (MMAT; Pace et al., 2012) was used to categorise records and inform quality appraisal. The MMAT is suitable for appraising studies with diverse designs in complex systematic literature reviews, and has good validity . Comprehensiveness and consistency of reporting quality were appraised with reference to the WIDER tool reporting recommendations (Albrecht et al., 2013; see Table 3).

Data extraction and analysis
Available data were extracted regarding intervention, study participants, setting, outcome measures, costs, fidelity, acceptability and recommendations. Evaluations were identified by ascertaining whether a research question was described and whether the data required to answer the question had been collected (Hong et al., 2018); then allocated to one of the five MMAT study design categories: qualitative (QL); quantitative description (QTD); non-randomised (NR); randomised controlled trial (RCT); mixed methods (MM). Records that could not be classified by study design (i.e. were largely descriptive) were categorised as 'mapping records'. Available information about all interventions was subject to detailed analysis including intervention content, theoretical basis, population, outcomes and conclusions.

RESULTS
One hundred and twenty-one records (45 mapping records and 76 evaluations) were included in the review (see Figure 1; Table 2). Included records were diverse in format and reporting quality. The 45 mapping records described interventions without evaluating them. The 76 evaluation records comprised the following study designs: 41 NR; 23 QTD; 5 QL; 5 MM; 2 with insufficient detail of study design; 0 RCT. Evaluation design description was often unclear, though evaluation design could sometimes be inferred from other study details. Where reported, terminology was inconsistent.
All pre-2007 records (n = 23) were from the US. The geographical spread of publications increased from the mid-late 2000s. Seventy-nine records were from peer-reviewed sources. The remainder were from professional magazines, internal reports, training resources and blogs. Figure 2 summarises the pattern of publication over time. A sharp increase from the mid-2000s coincides with a US-wide policy response to newspaper reports highlighting deaths related to the use of restraint in facilities across the US (Huckshorn, 2010;Weiss, 1998).
All evaluations reported success, directly (e.g. reducing frequency, intensity or duration of seclusion and/or restraint) or indirectly (e.g. improvements to the social milieu).

Consistency and comprehensiveness of reporting
Reporting was poorly aligned with WIDER recommendations (Albrecht et al., 2013;See Table 3). Consistency and comprehensiveness were generally weak across all WIDER categories, and especially within the mapping records. Sampling strategies varied and included counts or rates of occurrences of RP (e.g. Azeem et al., 2011) and whole or part populations of children and/or staff (e.g. Nunno et al., 2003;Russell et al., 2009). Where provided, definitions of RP varied, for example seclusion only; restraint only (including mechanical methods); seclusion together with restraint. Type or intensity of physical hold was rarely detailed.
Demographic reporting about children in the setting was sporadic, typically describing age and/or gender and/or ethnicity. It was sometimes possible to extrapolate further information from the setting description; for instance Williams and Grossett (2011) describe a facility for individuals with intellectual disability. Restraints seem to be performed more frequently on children aged 5-11 than on their older peers (Ryan et al., 2007;Villani et al., 2012).
Little demographic information about staff samples was reported. One study described how some staff 'selected out' rather than engaging with a new culture (Elwyn et al., 2017). Shadili et al. (2012) speculated that the success of a restraint-reduction intervention may have been helped by the fact that most of the staff were female, and staff gender was acknowledged elsewhere as potentially relevant to intervention outcomes, for example (Glew, 2012; Singh  , 1999), but generally, staff groups were treated as homogenous for interpretation of study results.
Consent to participate was rarely mentioned and appeared to be mandatory in many evaluations, typically where staff training (Verret et al., 2019) or broad systemic change (Wisdom et al., 2015) were introduced.
Training interventions delivered directly to staff were commonly evaluated via data routinely collected when children were subject to RP (for instance see: Huckshorn, 2010;Kalogjera et al., 1989). Most evaluations did not report on delivery mode, intervention dose (e.g. duration or intensity of training), modifications or fidelity. See Table 3.

Interventions
The total number of distinct interventions identified within the 121 included studies was 82. Most (74/82) were applied once only, reflecting a common practice whereby individual settings developed tailored RP reduction initiatives. Table 4 lists those interventions that were applied more than once.

Settings and locations
Most records (87/121) reported US-based studies. A further 21 were Europe-based (UK n = 18; Finland n = 1; Netherlands n = 1; France n = 1), and the remainder in Canada (n = 4), Australasia (Australia n = 3; New Zealand n = 1), Singapore (n = 1) or in more than one country (n = 1). Three records did not report geographical location. Regardless of study origin, all reported interventions had been delivered in the US, with some additionally delivered elsewhere.
As seen in Table 4, the number of times an intervention was reported could differ from the number of times it was used and/or evaluated; for example 6Cs were delivered on 12 separate occasions and evaluated five times.
Just under half of the records (60/121) related to mental health settings. Other service settings were health and social care (n = 23 records); criminal justice (n = 11); education (n = 10) or multi-functional services, for example healthcare and education (Shield, 2006).

Intervention focus
Children's participation was identified in 6 out of 82 interventions and was typically low-level, for example community meeting attendance (Azeem et al., 2011;National Association of State Mental Health Program Directors, 2006;Padhi et al., 2019); limited influence over treatment (Miller et al., 2006;Wisdom et al., 2015); or contribution to a consumer satisfaction survey (Azeem et al., 2011;Winnipeg, 2015).
All interventions included staff training, though it was not necessarily made explicit how this would affect RP use. Training included: goal setting with staff (Azeem et al., 2011) and/or children (Holstead et al., 2010); RP data review (Campbell, 2004;HM Inspectorate of Prisons, 2015;Rettmann, 2019); introduction to a new resource, for example a sensory modulation room (Carter et al., 2008;Seckman et al., 2017) Training length varied from 1 to 35 hours. Details regarding: length; intensity; content; training provider; mode of delivery; numbers, profile and post-training assessment of staff were often not provided.
An example of a multi-strand intervention is 'Six Core Strategies (6Cs)

Outcomes evaluation
In total, 228 measures were used across all interventions (mean 3; range 0-9). Twenty-two were standardised measures (Table 5), and they were found in 14 evaluations. Non-standardised measures, identified by the absence of supporting references, were generally study-specific, reporting simple counts and various rates and proportion calculations, for example the average number of incidents per child over a given period.
Reported outcomes were in four broad categories: use of RP; staff development and activity; resource implications and child progression and satisfaction. The most common outcome measures were as follows: number of restraints (n = 63 records); duration of restraints (n = 9); number of seclusions (n = 36); duration of seclusions (n = 7); injuries (n = 8); incidents (n = 11) and number of restrictive interventions (n = 8 records).
Most evaluations reported only pre/post descriptive data without statistical or control group comparison. All reported favourable outcomes. A small number reported mixed results, for example across settings (Winnipeg, 2015) or time points (Nunno et al., 2015). Many studies did not report timeframes or time points. Typically, the targeted RP reduced over time post-intervention, though improvement could be uneven (e.g. Campbell, 2004;Deveau & McDonnell, 2009). There was no reporting of unhelpful interventions.
Multi-strand interventions or those involving gradual change could confound attempts to clarify cause and effect (Martin et al., 2008;Pollastri et al., 2016;Reynolds et al., 2016); for example McGlinn (2006) observed: …psychiatrists at the study facility changed the manner in which they medicated clients between the two study periods. The number of staff involved in an incident was not reported at all, nor was psychological harm. Four records reported the number of injuries to staff, and eight reported total injuries to staff and children combined, but no record reported both, suggesting, significantly, a lack of focus on injuries, especially to children.

Assumed change process and design principles
There was limited discussion of underpinning theory. Many quality improvement interventions used 'Plan, Do, Study, Act' (PSDA), a mechanism that repeats and adjusts interventions to achieve the desired effect. Some interventions cited programme-level theories informing intervention procedures, for example sensory modulation or trauma-informed care. The most frequently cited theory relating to staff behaviour was social learning theory, used to improve staff individual and team self-efficacy.

Costs reported
Twelve evaluations reported financial costs. Financial analysis was diverse in terms of cost unit, study/intervention period and accounting period (e.g. financial year, calendar year, part year).

DISCUSSION
This appears to be the first review using systematic methods to map RP reduction interventions for children's institutional settings. Environmental scanning (Graham et al., 2008) was novel in the context, identifying resources that might otherwise have been overlooked.
The review highlighted a lack of evidence to clarify which interventions are effective in reducing RP. Evidently, many service providers develop their own interventions or adapt or applying existing ones without reporting useful levels of detail about intervention or study procedures. How children's beliefs, circumstances, expectations, experiences, identities, resources or values may interact with RP reduction interventions remain unclear.
It remains unclear why staff training received particular attention. While the health sector literature demonstrates widespread enthusiasm for using staff training to improve service user outcomes (Ameh et al., 2019;Hatfield et al., 2020), evidence of effectiveness is inconsistent (Bosco et al., 2019;Hassiotis et al., 2018;Knotter et al., 2018). It may be useful to consider alternatives, such as attention to staffing levels (Baker & Pryjmachuk, 2016), team reflexivity (Lines et al., 2021) or organisational change theory (Hussain et al., 2018).
Problematic reporting supports Purtle (2020) in suggesting an underdeveloped evidence-base around trauma-informed interventions in children's settings. Relationships between aims, intervention and results were often unclear, perhaps untested. For example where RP reductions followed a staff education intervention, simple chronological associations could be conflated with cause and effect, with little consideration of fidelity or confounders.
Interventions are not necessarily designed to produce evidence (Girelli, 2004;Wilson et al., 2015). For complex issues, practitioners may prefer multi-strand interventions (The Australian Psychological Society Ltd, 2011). Setting-specific interventions may not contribute to the broader body of evidence, though better reporting of multi-strand interventions could clarify whether these are especially beneficial (Duncan et al., 2020).
Incident numbers were frequently used as effectiveness evidence. However, there was little reporting of factors such as number of children involved or injuries sustained. Broad, collapsed data of this type may not easily portray practice realities, limiting its potential to inform decision-making.
Comparisons across the dataset were complicated by diverse study outcomes. Although the most common measures were RP incidents, the numbers were calculated differently, for example counts or rates. Potentially, a brief, low-intensity restriction could count the same as a lengthy, damaging, complex and high-intensity incident. This reflects results from a comparable review of RP reduction interventions in adult mental health settings (Baker et al., 2021).
The limited evidence may reflect values affecting progress in this field of research (Lineham, 2018). A disenfranchised and silenced population (children in institutions) can scarcely influence the allocation of research monies (Archard & Skivenes, 2009;Care Council For Wales, 2016;Lansdown, 2011); whereas the increase in records from the 2000s coincides with media reports (Busch & Shore, 2000;Weiss, 1998) that stimulated US-wide support for RP reduction (National Association of State Mental Health Program Directors, 2006).
Most studies reported some positive outcomes around reducing RP and none reported unhelpful interventions. However, no RCTs were identified and only around a third of records reported quantitative data. Contributory factors may include marginalisation of studies that do not demonstrate large effect sizes (e.g. exploratory or preventive research; Mavridis & Salanti, 2014); suppression of unwanted outcomes because of funding issues (Morrow, 2022) and potentially, inherent difficulties in developing ethical RCTs in this context.

Strengths and limitations
No previous reviews have systematically mapped evaluated and unevaluated interventions to reduce RP in children's institutional settings. The study findings are transferable to any institutions that have children in their care; however because non-English language records were ineligible, results were skewed towards the Global North-specifically, most evidence was from the US.
Environmental scanning enabled the inclusion of wide-ranging interventions in diverse formats. The absence of quality inclusion criteria contrasted fundamentally with conventional systematic reviews (Agency for Healthcare Research and Quality, 2020;Fajardo et al., 2019;Parker et al., 2018). This restricted options for producing systematised results, but arguably generated a more realistic picture of practice.
Children in institutions may lack voice, power and opportunities to protect themselves from RP (Kiraly & Humphreys, 2013). The omission of children's perspectives reflects poor respect for children's rights and opinions (United Nations, 1989). Staff diversity was also overlooked. More attention to study design and reporting could help understand differential implications of interventions in relation to children and staff abilities, beliefs, background, gender, geography, identity, race, religion and values.

CONCLUSIONS AND RECOMMENDATIONS
RP reduction in children's institutional settings should be a priority for practice, policy and research. Key recommendations concern the linked issues of intervention development, evaluation and reporting. Without clarity about current RP use, interventions evaluation will remain unsuitable for informing evidence-based practice guidance. Above all, the near absence of children's voices seems to be a critical failing in this field.
The interventions identified in this review seem numerous and wide-ranging. The focus on training for staff is without clear justification. The limited geographical scope of most interventions indicates a need for insights beyond the Global North. A better understanding of demographic trends, institution type and governance could inform the adaptation of interventions to reduce RP for diverse groups.
Most interventions are multi-strand and evaluation design tends to be bespoke for the setting. Resultant difficulties in comparing results across studies suggest an urgent need to streamline intervention reporting. Accessible guidelines for a core outcome set that is feasible for researchers and practitioners to use in real-world settings, would be a valuable step towards improving practice.
Policy makers, commissioners and practitioners could avoid further investment in interventions whose outcomes are not known. Intervention reporting frequently lacks detail, consistency and comprehensiveness, combined with an over-simplification of cause and effect. Robust evaluation methodologies appropriate for multi-strand interventions, combined with adherence to reporting conventions, could help develop an evidence base to support policy and practice.
Despite numerous enquiries and recommendations, concern about the use of RP in children's institutional settings is ongoing. The impact of RP on children and staff's psychological and physical welfare, and the potential for harm, and even death, should not be underestimated. Children worldwide will continue to face malpractice and their care will remain sub-optimal without a sustained focus on RP reduction.
A better understanding of interventions may lead to discernible improvements in service delivery. It will inform decision-making about staff training, which in turn could influence everyday professional practices, promoting therapeutic relationships and staff well-being. Most importantly, vulnerable children in institutional settings could be protected from trauma, injury and deaths, thus benefiting wider society.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in University of Leeds Open Access data repository at [https://doi.org/10.5518/1077]. #10;Additional queries to corresponding author Professor John Baker j.baker@leeds.ac.uk Krysia Canvin is a qualitative methodologist with 25 years' experience of research focusing on vulnerable groups' health and mental health and is particularly interested in coercion, secure settings, and health inequalities. She has authored almost 100 publications and presentations on these topics. Stella Branthonne-Foster is a young person with lived experience of restrictive practice. She has worked with a number of organisations, including NHS England and the National Collaborating Centre for Mental Health, on issues surrounding children and young people's health.

Tim
McDougall is a Director of Nursing and Quality at Lancashire and South Cumbria NHS Foundation Trust, and is also National Professional Advisor (Children and Young People's Mental Health) for the Care Quality Commission. He spent 15 years as a CAMHS Nurse Consultant and several as a Clinical Director in CAMHS and Children's Services. He was formerly Nurse Advisor for CAMHS at the Department of Health in England and has been a member of several National Advisory Councils. He is interested in quality improvement and service transformation. Barry Goldson is PhD FAcSS, is Professor Emeritus, Department of Sociology, Social Policy and Criminology, University of Liverpool where he was previously Professor of Criminology andSocial Policy (2006-2009) and Charles Booth Chair of Social Science (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). He is best known for his work in the fields of youth criminology and youth/juvenile justice studies within which he has earned significant visibility and international standing. During 2017-2019 he was a member of the Expert Advisory Board that supported the United Nations Global Study on Children Deprived of Liberty.
Ian Kellar is an Associate Professor of health psychology with expertise in developing and evaluating interventions that use behaviour change techniques for implementation both in the UK and in LMICs. He attained CPsychol (Health) in 2006. He is a member of the College of Experts to support the DHSC/UKRI Global Effect on COVID-19 Health Research.
Judy Wright is a senior information specialist and qualified librarian leading the information specialist team in Leeds Institute of Health Sciences (LIHS) and supporting the management of the Academic Unit of Health Economics (AUHE). She manage a portfolio of information specialist work including custom-made literature searching, reference management and search methodology advice for systematic reviews, realist reviews, evidence syntheses, research proposal scoping, economic evaluations and cost-effectiveness models.
Joy Duxbury is Professor of Mental Health at Manchester Metropolitan University. She worked on numerous funded projects pertaining to mental health, inclusion and participation. Research on the exploration of stakeholder perspectives and implementation has been a strong feature of my work to date. She have a background in forensic mental health and more recently my national and international focus has been upon minimising restrictive practices such as physical restraint and coercion across varied settings.