Non-pharmacological interventions to reduce restrictive practices in adult mental health inpatient settings: the COMPARE systematic mapping review

Introduction

Stage 1 focused on ascertaining the range and characteristics of interventions, irrespective of evidence of effectiveness, which involved systematically searching and reviewing all reports of interventions seeking to reduce the use of restrictive practices.

The search strategy approach drew on the increasingly utilised method of mapping52–56 to inform the purpose and output of the review. It differed from the method described in Bradbury-Jones et al. 52 because of the broad scope of the search and the inclusion of interventions in the current study. It was known that in addition to a small number of well-known interventions reported in the academic literature, there were numerous small-scale, standalone initiatives available for implementation in services. Not all of these would appear in a search restricted to the published research literature as they may be reported in unpublished literature or ‘non-research’ publications.

The search for relevant interventions records was informed by an ‘environmental scanning’57 approach suggested by Judy Wright, the project Information Specialist. Environmental scanning is a search methodology familiar in business contexts but relatively little used in health research. It permits the identification of more diverse information about an area than could be retrieved solely from published literature. In health-care settings, environmental scans have been used to inform future planning, to document evidence of current practice and to raise awareness. 57 It was therefore an appropriate choice for expanding the scope of the search strategy. Environmental scanning may involve a ‘passive’ approach that focuses on published and unpublished existing data or an ‘active’ approach where additional knowledge is generated through primary data collection. 57 In this study, a passive approach was used.

In keeping with objective 1 (to provide an overview of interventions aimed at reducing restrictive practices in adult mental health inpatient settings), the search criteria targeted diverse reports of non-pharmacological interventions aimed at changing the behaviour of inpatient adult mental health service staff to reduce restrictive practices. The scope of the searches was necessarily broad to include all records of an intervention, whether it was an evaluation or a descriptive report. 55 In order to include as many interventions as possible within the scope of the search, no quality threshold was imposed either indirectly (by restricting the search to high-impact journals)52 or directly via the search criteria or by screening. 54,56 Inclusion was not restricted by study design. 42 Interventions that solely involved policy change and those that aimed to reduce the use of one type of restrictive practice by replacing it with another were not eligible for inclusion.

In addition to interventions intended to reduce or eliminate restrictive practices, reports of interventions designed to improve quality or reduce or manage violence were included if their procedures and/or outcome measures addressed restrictive practices. Eligibility criteria are shown in Table 1.

The starting date of 1999 was decided by the date of introduction of the UK National Service Framework for Mental Health,58 which precipitated new quality standards and a significant shift in the orientation of services. Because of the research team’s prior knowledge of the paucity of the evidence base, there were no restrictions on study design and no quality threshold was imposed. Searches were conducted from February until June 2018, and repeated in April 2019.

Two main searches were developed to identify interventions to reduce the use of restrictive practice in adults with mental health disorders. The first search aimed to identify reports from the academic bibliographic databases. The second search aimed to identify unpublished reports, including those occurring in the grey literature, social media and other digital resources.

Academic bibliographic databases search

The first search was conducted in February 2018. A wide range of academic bibliographic databases were searched for published studies, including:

• British Nursing Index (BNI)

• Cumulative Index to Nursing and Allied Health Literature (CINAHL)

• Cochrane Central Register of Controlled Trials (CCTR)

• Cochrane Database of Systematic Reviews (CDSR)

• Database of Abstracts of Reviews of Effects (DARE)

• EMBASE

• Health Technology Assessment (HTA) Database

• HTA Canadian and International

• Ovid MEDLINE^®

• NHS Economic Evaluation Database (NHS EED)

• PsycInfo^®

• PubMed.

For full details of the databases, see Appendix 1.

The rationale for the academic databases was to select databases with good coverage of mental health studies and those covering studies of the nursing workforce, because the restraint reduction interventions are particularly important to this group that are dealing with aggressive and difficult situations on a ward. Two nursing databases (CINAHL and BNI), two of the largest medical databases (EMBASE and MEDLINE), the largest mental health database (PsycInfo), an evidence-based health database with good coverage of randomised controlled trials (RCTs) (Cochrane Library) and PubMed were searched to supplement the Ovid MEDLINE as they can contain some articles more recent than those included in Ovid MEDLINE. The team discussed and consented to this database selection, as proposed by the Information Specialist.

Search strategies were developed for the concepts: coercive interventions, mental health conditions and inpatients. Searches included subject headings and free-text words, identified by text analysis tools (PubReMiner), the Information Specialist (JW) and project team members. Further terms were identified and tested from known relevant papers. Searches were peer reviewed by a second Information Specialist. For full details of the search strategies, see Appendix 2 and Report Supplementary Material 1.

The search was updated and re-run in April 2019 in the same databases except for DARE and NHS EED, which were not searched as they had not received further content since the 2018 search. Owing to a change in database providers, BNI and HTA databases were searched via a different database host in 2018 rather than 2019. After checking index terms, two additional terms were added to the PsycInfo search: involuntary treatment/and psychiatric hospitalisation. All other searches remained the same.

Grey literature search

The second search was run from June 2018 to August 2018 to identify unpublished (grey) literature reports in databases, websites and social media sources. For full details, see Appendix 1.

The list of information resources to search was created collaboratively by the project team and information specialists. Websites for charities, government health departments, health-care organisations, health-care quality agencies, mental health organisations, professional societies/colleges and training providers were selected following an exercise to gather all potentially useful websites known by the project team, and those found by an information specialist scoping search. This large list was then organised into ‘types’ of organisations, such as health-care quality agencies, charities and government departments, and the team refined the list to include a set of 5–10 websites to search for each group that represented different countries/regions.

The team prioritised resources that were likely to provide relevant reports from North America, Australasia and Europe. Google (Google Inc., Mountain View, CA, USA) was used to search for interventions in 30 countries specifically identified in the 2016 Legatum Prosperity Index™ (a between-nations ranking system) as having the best health systems. A structured social media search incorporated YouTube (URL: www.youtube.com; YouTube, LLC, San Bruno, CA, USA), Facebook (URL: www.facebook.com; Facebook, Inc., Menlo Park, CA, USA) and Twitter (URL: www.twitter.com; Twitter, Inc., San Francisco, CA, USA).

The search strategies used in academic databases were adapted for use in grey literature databases, websites and social media sources. Website and social media searches used search terms similar to those used in the academic database searches, but fewer of them, and multiple short searches were run per resource, rather than one complex search. This ensured that the searches were consistent with the academic databases despite the limited ability of web and social media resources to process long strings of search terms or combine multiple searches. For further detail, see Appendix 2 and Report Supplementary Material 1.

In addition, an information request for unpublished interventions was sent to mailing lists for the health and medical community, clinical librarians and mental health librarians. No suggestions of restrictive practice reduction interventions were received from the mailing list information request. Project team members forwarded relevant reports they saw on their own social media accounts and through personal contacts with experts. When contact details were available, authors of identified interventions aimed at reducing the use of restrictive practices were contacted for further data. A request for information was circulated around Restraint Reduction Network members.

Backward citation searching of cited references and forward citation searching using Google and PubMed were used in order to access fuller descriptions of interventions, including development, procedures and implementation to supplement records with minimal detail, such as conference and poster abstracts, Microsoft PowerPoint^® (Microsoft Corporation, Redmond, WA, USA) slides and some non-research reports. This procedure was also used to identify journal publications associated with a dissertation/thesis and published reports associated with unpublished records and non-research reports. These strategies were also supported with a Google search for authors. Individual journals were hand-searched; however, because of the disparate nature of journals reporting the study topic, no key journals were identified.

The results of the published and grey literature database and website searches were stored and de-duplicated in EndNote (Clarivate Analytics, Philadelphia, PA, USA) referencing software libraries. The results of the social media searches were stored in a Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) file.

Screening

Free online citation screening software (Abstrackr beta version, Center for Evidence Synthesis in Health, Brown University, Providence, RI, USA) was used to assist with screening. Abstrackr uses artificial intelligence to help reduce screening time by determining the relevance of papers, based on inclusion and exclusion terms entered by reviewers, and subsequently displays references in order of predicted relevance. 59

Retrieved references were imported into Abstrackr and the following settings were selected: a pilot phase of 100, double-screening, display-all (i.e. title, authors, abstract, keywords) and order by relevance. Two researchers (KC and KB) independently screened the first 100 references, documenting their decision-making. Terms were discussed and shared to ensure maximum efficiency and coherence after screening the first 100 references, again after screening 600 and again after screening 1000. In total, 55 terms indicating relevance were entered, including restrain, intervention, psychiatry, inpatient, and the names of specific interventions of interest. In addition, 78 terms indicating irrelevance were entered, including child, community, dementia and learning disability. The full list is provided in Appendix 3. Once 1500 references had been screened, no further references appeared to be relevant. Following the recommendation of Rathbone et al. ,60 references without an abstract were screened separately (n = 998) to avoid compromising Abstrackr’s predictions. Screening conflicts were discussed and resolved between KC and KB. This process generated a subset of full texts to retrieve for further screening.

Modification

Reports of modifications to the intervention protocol were recorded, including what was modified and how. 64 In this context, modification meant any planned deviation from the original intervention protocol.

Fidelity

When fidelity was recorded, it described reports of implementation as specified by the intervention protocol. 65

Theory

For the purposes of data extraction, theory was defined as a way of understanding, explaining and predicting behaviour, events and situations. Different scales of theory have been proposed:66 grand, mid and small.

Small theory or programme theory is how and why an intervention is proposed to work. It sets out the components of the intervention, the outcomes and how outcomes will be measured, often in the form of a logic model or driver diagram. 67 Although a 19-item measure for the assessment of the use of theory in interventions is available,48 it was apparent from screening and data extraction that few interventions made any use of theory, and most of the items in the measure would be recorded as ‘no’. Therefore, an adaptation was developed in which any interventions that explicitly referred to theory were examined for (1) whether or not they used theory to inform intervention design and implementation and (2) whether or not they related their findings back to the theory (this criterion was adapted from the theory coding scheme). 48 For example, a judgement would be made about whether or not a training intervention that was described as being informed by social learning theory had linked its training content and delivery back to the same theory, and subsequently whether or not the findings were discussed in relation to the theory.

Not reported

The term ‘not reported’ (NR) was used to indicate missing information, unless there was an explicit explanation of why the information was not provided, such as that costs were not recorded or a procedure was not followed. For example, if fidelity is not reported it can be assumed neither that fidelity was unmeasured nor that it was measured and unreported. The analysis and findings presented below therefore can reflect only what was reported.

Describing intervention content using the behaviour change technique taxonomy

Interventions to reduce restrictive practices use a variety of behaviour change techniques to change staff behaviour. The BCT taxonomy was used to describe and compare the content of the interventions identified. As described above, the BCT taxonomy can be retrospectively applied to completed interventions41,46 and to synthesise evidence. 41,50

Behaviour change technique coding

Two researchers (KB and KC) acted as coders and independently coded the documents using the BCTv139 as the basis for a coding manual for the data. Both researchers were trained in application of the BCT taxonomy and are experienced qualitative mental health services researchers. The analysis was also supported by NVivo 12 software (QSR International, Warrington, UK) for qualitative data analysis. NVivo provided an efficient means of storing, coding, checking and reviewing throughout the analysis. The software enables the generation of audit trails including spreadsheets with clear links to original data sets.

Twenty interventions of varying types were coded independently by both researchers before discussion about how BCTs had been identified and coded. The coding manual was developed as the two researchers (KB and KC) discussed and recorded details about how they had coded BCTs and for what reasons. Once the researchers were satisfied with the coding of this set of 20, the remainder were independently coded by one coder and then reviewed by the second. The researchers conferred when there was uncertainty and sought advice from author Ian Kellar, an expert in BCT, as required. Changes were made to the coding manual and coding to ensure consistency. Formal measures of agreement were not used because of the novelty of applying the taxonomy in this area. Appendix 4 provides further details.

Development of the coding manual

The starting point for the coding manual was the BCTv1 taxonomy and its definition and examples of each BCT. 39 However, most of the examples in the taxonomy referred to behaviour of health-care service users rather than health-care staff. In response to this discrepancy the current study applied the approach reported in Presseau et al. 47 For example, regarding ‘reframing’ in the context of staff–patient interactions, staff are supported to think of aggression as a response to trauma, that is the communication of distress. Further examples specific to the literature were developed to consider the content of the interventions to be described. As the coding progressed, additional examples and clarification based on areas of both discrepancy and consensus were added.

There were no intervention components that did not fit into the taxonomy. Some BCTs detected were aimed at health-care staff, for example instructions on how to perform a behaviour via training. Others were aimed at mental health service users, for example using distraction to reduce feelings of aggression and some were aimed at both groups, for example generating emotional and social support by encouraging socialising on mental health wards. Again, taking the approach of Presseau et al. ,47 these were treated separately. In line with the study aims, the focus was BCTs targeting staff behaviour.

The taxonomy deals with BCTs concerning both behaviour and outcomes of behaviour. Outcomes can be the stopping of a behaviour (e.g. stopping smoking and improving health) or the commencement of a behaviour (e.g. exercising and reducing weight). These were distinguished by treating incidents of restrictive practices as ‘outcomes’ and ‘behaviour’ as the efforts made to reduce these (e.g. de-escalation). The interventions contained more focus on outcomes than on behaviour as these are easier to record and report; however, some interventions did encourage the examination of near-misses and successes (i.e. where restrictive practices had been avoided), perhaps through team meetings, and these were seen as examples of monitoring of behaviour rather than outcomes.

One problematic aspect of the taxonomy is its use of ‘self’ in terms of ‘self-monitoring’, ‘self-reward’, ‘self as role model’, ‘valued self-identity’ and ‘self-talk’. The initial screening of the literature had revealed that there was very little reference to individual health-care staff at all and no self-monitoring from individuals. Therefore ‘self’ was interpreted in a collective sense and was applied in instances in which the ward team were, for example, self-monitoring rather than being monitored at arm’s length via management. This interpretation was validated during coding, since no examples of individuals (rather than ward teams) were detected, and, in addition, many ward-based initiatives had been generated from the ward staff, rather than from management or at a broader policy level. Therefore, when it was reported that a ward recorded its own incident data, this was classified as ‘self-monitoring of outcomes’. When data were recorded centrally, this was recorded as ‘monitoring of outcomes’. The other opportunity for self-monitoring was in interventions that used debriefing after an incident. If it was specified that staff were encouraged to reflect on their role in the incident, this was coded as ‘self-monitoring of behaviour and/or outcomes’.

One coding decision made in relation to staff training was that, if this was mentioned at all, it was coded as BCT 4.1 ‘instruction on how to perform the behaviour’, regardless of the level of detail (following Presseau et al. 47). Sometimes, interventions referred only to staff training in de-escalation, with no further detail; it was agreed that, in this circumstance, an assumption could be made that, at the minimum, there would be instruction involved. This is not in keeping with the specified instructions for coding BCTs, but had the rationale that, if only those interventions that had gone on to describe the actual content of this training had been coded, the presence of training within the interventions would have been severely under-reported.

Several of the BCTs refer to prompts or cues. It was agreed that, where a checklist or tool had been implemented on a ward (e.g. during a risk assessment on admission), this would be coded as a ‘prompt’, as it prompted staff to carry out behaviour that had the intention of avoiding restrictive practices. Care planning or risk assessment were treated within the context of ‘goals and planning’. This was because, although they were focused on the service user, ‘goals and planning’ can also refer to agreement on how staff will respond to service users’ needs. ‘Problem-solving, goals and planning’ could also be identified in post-incident debriefing, depending on how the debriefing was described.

The difference between ward and service level in monitoring of outcomes was also seen in a number of other BCTs that could be applied at the individual staff/service user level, ward level, organisation level and policy level. For example, ‘goal-setting (outcome)’ was detected at all of these levels (see Appendix 6 for further details).

Descriptive statistics were used to count which BCTs featured in each intervention to provide an overall frequency of the most commonly occurring BCTs across the data set, and what clusters they were from. It was further established which BCTs were used in interventions with particular components, for example training, or audit and review, or service user involvement.

Describing intervention outcomes and relating back to behaviour change technique content

The outcomes of evaluations were extracted and described. These outcomes were then related back to the BCTs contained in the intervention subject to evaluation. The BCTs in evaluations with both positive and negative findings were identified and described.

Assessment of study quality using the Mixed Methods Appraisal Tool

In stage 1, the primary objective was to identify and document all interventions in records that met the eligibility criteria. The quality of the records was of interest; however, although it was anticipated that the records would be diverse in quality, it was also expected that the data set would contain valuable information about the range of interventions being considered for use in practice. Therefore, no records were excluded on the basis of quality.

In the current study the MMAT51 was used at two levels: to identify records of interventions that had been evaluated and to assess the quality of the evaluation reports. To get a sense of the quality of the evidence, the screening questions of the MMAT51 were used during data extraction to establish whether or not the intervention had been subject to an evaluation. The MMAT was again used for further examination of evaluations during data analysis.

The MMAT was designed for use in complex systematic literature reviews that include quantitative, qualitative and mixed-methods studies. The MMAT was developed from theory and a literature review, and has been found to have good validity. 51,68 Quantitative and qualitative studies are judged against four criteria and mixed-methods studies are judged against three. Scores are between 0% and 100%, although caution is advised against relying solely on the score, and reviewers are encouraged to provide a narrative description of the study features that lead to that score. The quantitative domain is split into three subdomains: randomised controlled, non-randomised and descriptive. The characteristics of the MMAT meant that it was selected as the most suitable tool with which to judge study quality in the context of wide-ranging research methods.

The distinction between records, standalone interventions and intervention families

The 175 records that were identified reported a total of 150 unique interventions (see Figure 1). Of these, 121 records reported standalone interventions. The remaining 54 records contained 100 references to 29 intervention families. Intervention families consisted of interventions with multiple records, sometimes in different formats and some of which have been implemented (and evaluated) multiple times (see Appendix 7). The intervention for which the most records were identified was Six Core Strategies (n = 18), including research reports, intervention reports, service reports and tools. With the exception of four records that pertained to two studies respectively,23,99 all other records were unrelated to each other – there were no follow-up or replication studies. Similarly, 10 records were identified for Safewards, which were unrelated research reports except for two that reported the same study in different formats. 88,100 Just three interventions had replication studies: City Nurse,101,102 Patient Focused Nursing103,104 and Review. 105,106 Two interventions had follow-up studies: Initiatives to Reduce Seclusion and Restraint107,108 and Open Door Policy. 109–111 One intervention (Brøset Violence Checklist112–114) had been evaluated in a pilot and a subsequent study. In the case of six interventions (Beacon Project,115,116 Recovery Based Principles,6,117 Early Recognition Method,118,119 REsTRAIN Yourself,22,85,87,120,121 Scottish Patient Safety Programme for Mental Health91,122–124 and Talk First83,86,125,126), the multiple records related to the same application or study of that intervention, often in different formats.

Clinical setting

Sixty-seven interventions (45%) did not provide any detail about the clinical setting other than it being adult inpatient (see Appendix 9). A further 27 interventions (18%) did not report clinical setting as they were not reporting implementation, for example training links. The intervention that has been applied in the widest clinical settings is Six Core Strategies. Six standalone interventions127–132 had been implemented in multiple settings within the same intervention/study, as had four intervention families (i.e. interventions with multiple records): Six Core Strategies (in acute and secure wards); Brøset Violence Checklist [in acute wards and psychiatric intensive care units (PICUs)]; Tower Hamlets Violence Reduction Collaborative (in multiple acute settings); and Initiatives to Reduce Seclusion and Restraint (in multiple, unreported settings). A further five interventions had been applied across various clinical settings: the Scottish Patient Safety Programme for Mental Health had been implemented in acute, PICU and forensic settings, the Dynamic Appraisal of Situation Aggression – Inpatient Version (DASA-IV) had been implemented in acute and high-dependency wards, the Novel Seclusion Reduction Program in acute and forensic wards, and Sensory Modulation and Sensory Rooms had both been implemented in acute wards and PICUs.

The most common setting for implementing interventions to reduce restraint and seclusion was acute wards (n = 40/150; 27%). Nevertheless, interventions had also been implemented on PICUs (n = 11), and on forensic (n = 10), secure (n = 8) and specialist geriatric (n = 6) wards. The least common settings were admission wards (n = 1) and high-dependency units (HDUs) (n = 1).

Geographical setting

Just five interventions (3%) had been implemented in different countries, with Six Core Strategies23,75,133 having the widest geographical spread covering six countries (i.e. Canada, England, Finland, New Zealand, Spain and the USA). No Force First24,134 had been applied in three countries (i.e. Australia, England and the USA), as had Sensory Modulation135–137 (Australia, Denmark and New Zealand) and Sensory Rooms138 (Australia, England and the USA). The Brøset Violence Checklist114 (Canada and Switzerland) and Patient Focused Nursing104,139 (Australia and the USA) had each been applied in two countries. Safewards16 was applied in three countries (Australia, Denmark and England). Overall, the countries applying the widest range of interventions were the USA (n = 60) and the UK (n = 59). The origins of two interventions were not reported (see Appendix 10).

Intervention aims and targets

Three intervention aims were identified to (1) reduce, eliminate or prevent, (2) improve and (3) manage or monitor.

One hundred and five of the 150 interventions (60%) had a single aim, 35 (23%) had two aims and five (3%) had three aims88,131,133,140–142 (Table 4). The remaining 14% did not report an aim or target. 77

As seen in Table 4, 81 interventions reported a single target, 29 had multiple targets and the remainder did not report a target. The most common target was seclusion and restraint (n = 53 interventions), followed by restraint only (mechanical, physical or prone restraint) (n = 25 interventions), seclusion [including long-term segregation (n = 21 interventions)] and generic terms [e.g. ‘restrictive practice’, ‘conflict and containment’, ‘coercive measure’ (n = 29 interventions)]. Another common target was service user behaviour [e.g. aggression, violence, ‘problem behaviour’ (n = 26)]. Just 11 interventions included pro re nata (PRN) medication or chemical restraint as a target: in six it was the sole target. One of these interventions specified eliminating unsupervised PRN medication (and reducing PRN medication overall),143 another aimed to replace PRN medication with ‘other clinical strategies’ (no further explanation provided),131 and another to reduce restraint associated with PRN medication and security involvement. 69 None of the interventions explicitly reported targeting the use of rapid tranquillisation, although Beckett et al. 130 referred to reviewing its use (in their procedures) and Sarkar78 examined the impact of their intervention on their use of rapid tranquillisation (reported in their outcome measures). Some interventions specified the type of restrictive practice they wanted to address, whereas others specified frequency115,131,144 or duration. 115,145,146

The most common aim for improvement was quality (e.g. environment, ward functioning, staff presence, service user to staff ratios, quality of care, ‘communication’) (n = 11). Other common targets for improvement included safety (n = 18), service user behaviour (‘dangerous’, ‘disruptive’, ‘risk’, ‘challenging’, ‘problem’, ‘aggression’) (n = 26),20,78,112–114,118,129,136,137,141,147–156 and staff skills and attitudes (n = 14). Three interventions targeted staff injury,129,157 anxiety141 and burnout,70 whereas 14 targeted service user outcomes and experiences, including service user harm,114 the service user experience (e.g. feeling of safety158), experience of care79,159 and service user outcomes. 160 Others targeting harm or safety did not specify service user or staff (e.g. Bell and Gallagher161) or included both (e.g. Lo74). In addition to those targeting quality of care (and implicitly targeting staff behaviour), five specifically targeted staff behaviour in terms of staff attitudes and perceptions,162 knowledge and efficacy,21,141,163 and culture. 132

Reference to theory

Mention of theory was absent from many interventions. Three interventions referred to having a ‘theory of change’85,153 but provided no further detail about what this was, how it had been developed and how it was tested and refined. Many of the ‘quality improvement’ interventions used a plan, do, study, act (PDSA) cycle: a mechanism to repeat and adjust interventions until they achieve the desired effect. 12,16,17

Some interventions130,137 made explicit cited reference to programme-level theories that had informed their intervention procedures, such as Sensory Modulation or Trauma-Informed Care. Other programme-level theories cited sought to explain staff behaviour, service user behaviour, therapeutic relationships and organisational change. These studies often sought to test or modify not the actual theory but rather the impact of using interventions based on them in relation to the reduction of restrictive practices.

The most frequently cited theory related to staff behaviour was social learning theory,71–73,152,164 which was used to support training interventions that sought to improve the self-efficacy of individual staff and staff teams.

Bonner’s theoretical model for debriefing and post-incident review165 informed one intervention166 and the general aggression model167 informed another that sought to reduce aggression via Sensory Modulation. 152 Kernberg’s theory of personality organisations and transference-focused psychotherapy168,169 informed clinical guidelines that aimed to reduce restraint. 170 Other theories mentioned included those seeking to explain care giving processes: Peplau’s interpersonal relations171 and Watson’s caring theory. 172 Both Safewards173 and City Nurse102 were based on the theoretical work of Bowers et al. 101 regarding conflict and containment, and the interaction between service users and staff.

Five interventions were informed by broad organisational change theories. Kanter’s structural empowerment theory sees organisational empowerment and potential for change as being influenced by the individuals within the organisation having access to information, resources and opportunities to learn. 174 The transtheoretical model of change175 is generally applied to individuals, explaining how people prepare for and enact change, although Colton176 applies it to organisations to structure a tool kit to prepare to reduce restrictive practices. Schein’s model of organisational culture177 informed the development of the Six Core Strategies. 178 This focuses on publicly espoused values and assumptions. Senge179 focuses on the capacity of organisations and, by extension, the individuals within them to learn and focuses on a common goal. 133 The Iowa Model of Evidence Based Practice was used to guide the implementation of a rapid response team. 74 The full analysis of behaviour change techniques used in the interventions can be found in Chapter 5.

Mode of delivery: intervention procedures

The extraction process highlighted the procedures used by each intervention to address restrictive practices. A total of 15 unique procedures were identified from the analyses and these were organised into six themes (Figure 4).

FIGURE 4.

Intervention procedures by theme. PPI, patient and public involvement.

Staff-focused procedures

Staff-focused procedures were those that were aimed at and undertaken solely by staff with a view to influencing staff use of restrictive practices. One procedure was training, which could cover, among other topics, de-escalation (e.g. Laker et al.,21 Sullivan et al. ,104 Lee et al. 127 and Jonikas et al. 180) or crisis management (e.g. Steinert et al. 20 and Melin69). Another staff-focused procedure was role modelling, which could involve supervision or mentoring (e.g. Fletcher et al. 181 and Noorthoorn et al. 182) and identifying champions, experts or specialists (e.g. Bowers et al. ,102 Tully et al. 90 and Lombardo et al. 159) to set an example to ward staff.

Service user-focused procedures

Service user-focused procedures were those that focused on and sometimes involved service users but always involved staff. One of these procedures was risk assessment – whereby service users’ triggers would be recorded – which was often undertaken using a tool designed for this purpose (e.g. Brøset Violence Checklist114 or the Early Detection Plan118). Another related procedure was care planning, in which service users and/or staff planned appropriate and preferred strategies to prevent or respond to distress and/or incidents. These sometimes involved service users in identifying their own triggers and forming their own plans, for example a ‘personal safety plan’132 and also included PDSA. 161

Alternative approaches

Two further procedures were classified as alternative approaches because they proposed alternative ways of either preventing or responding to service users’ distress. Sensory approaches included Sensory Modulation via the installation of sensory or Comfort Rooms (e.g. Novak et al. ,183 Barton et al. 184) and/or the availability of sensory equipment (e.g. Lee et al. 127) and/or use of Sensory Modulation techniques (e.g. Yakov et al. 185).

Incident-focused procedures

Other procedures were incident focused, that is, they were responses to incidents of restrictive practices. These included incident review procedures, in which organisations (staff and managers) conducted retrospective chart audits (e.g. Qurashi et al. 186) or collected and monitored their incident data (e.g. Donat106 and Friedman et al. 131) to establish baseline and progress rates, or to identify patterns for targeted intervention. In contrast to this whole-system review, debriefing was conducted immediately or soon after an incident, and included the staff and service users involved and possibly others who witnessed the incident (e.g. Duxbury et al. 120). The final procedure was rapid response, where specially trained rapid response teams were formed to respond to and provide support to incidents when they happened [e.g. psychiatric emergency response teams (PERTs); see Smith et al. 142 and Prescott et al. 128].

Organisation-focused procedures

We also identified several organisation-focused procedures. These were system-wide structural and cultural changes including making changes to staffing levels (e.g. Parasurum et al. 77), increased one-to-one nursing (e.g. Jungfer et al. 109) and/or staff availability to/contact with service users (e.g. Beezhold et al. 79 and Lewis et al. 132). Another procedure involved changing nursing approaches [e.g. such as implementing the City Nurse model,101 the Bergen model,129 a recovery approach (e.g. Repique et al. 187) or a trauma-informed approach (e.g. Madan et al. 108)]. This theme also included improvements to communication (e.g. Stead et al. 188), community meetings (e.g. Mistral et al. 189), de-escalation (e.g. Cowin et al. 163) and safety huddles (e.g. Taylor-Watt et al. 153 and Stead et al. 188). Another procedure involved policy change (e.g. Short et al. 157 and Sullivan et al. 190). Finally, leadership-related procedures involved senior management being involved in meetings, making statements of commitment.

Service user involvement in interventions

Forty-eight interventions involved service users in some way, but the type and extent of involvement varied greatly. In some cases, service users were involved in multiple ways, whereas in others they had limited roles. A number of interventions involved service users in a consultation or advisory role, for example as committee representatives,133 participants of project teams158 (e.g. working group on medicines and rapid tranquilisation130), service user panels,160 or advisory committees. 88,115,140,163 Service users were consulted for their views16 and feedback on rules,116 sensory rooms116,140 and research. 17,140 Others described involving service users in the design or co-production of parts of the intervention, such as safety plans,116 information leaflets,130 comfort rooms,23,191 training16,115,163 and the selection of intervention components. 16,116 Service users were involved in intervention delivery, for example in ward/community meetings,17,23,186,191 delivering activities,116 and training,88,116,137,163,182,192,193 displaying their positive messages for current service users,181,192,194,195 and as peer advocates,142 counsellors196 and support workers. 99,193 Only two interventions specified that service users were paid for their involvement. 24,197 The remainder that reported service user involvement provided no further detail.

Seventy-one per cent (n = 158) of interventions involved multiple procedures, which ranged from 2 to 10 procedures (Figure 5).

FIGURE 5.

Number of procedures per intervention.

The most common procedures were training/education (n = 151), changes to nursing approaches (n = 72) (e.g. implementing Trauma-Informed Care or the Recovery Approach) and reviewing incident data (n = 80). Least mentioned were rapid response teams (n = 11) and activities (n = 14) (see Appendix 11).

As illustrated in Figure 6, although most interventions involved staff training, few reported the content, mode of delivery or training provider in any detail. The documentation reporting 89 interventions did not report any detail at all. The most common mode of delivery was group training or workshops (n = 37). Six reported using e-learning or online training and a further five reported multimedia components (e.g. video, PowerPoint) to their training. Four interventions reported using a train-the-trainer model. Others described training as one to one (n = 2), face to face (n = 3) or on the job (n = 1). Two interventions mentioned using champions and exchange visits respectively. Training was specified as provided in-house in 64 interventions and, of these, only seven specified the provider (these included quality improvement team, occupational therapists, unit manager or researcher). Training was delivered by external providers in 24 interventions, and 20 of these specified the provider. Providers mentioned included Bergen Model Representatives, Centre for Creative Leadership, City Nurse, Crisis Prevention Institute (CPI), ePsychNurse, Omega, Safewards, Substance Abuse and Mental Health Services Administration (SAMSHA), National Enforcement Training Institute, JKM Training, AQUA and Recovery Innovations. The remaining interventions did not report any information about the training provider.

FIGURE 6.

Distribution of procedures.

Intervention dose: duration and intensity

Many evaluations did not report details about the duration and intensity of the intervention. Partial details, such as overall duration of the intervention or of an individual component (usually training), were sometimes, but not always, provided. Often, the evaluation period and the duration of intervention implementation were not distinguishable. Similarly, the duration of individual intervention components was often not reported. With this proviso, interventions ranged in length from 10 months to 5 years. Some interventions described providing standalone training sessions, whereas others were conducted over a short period of time (e.g. 1 week) or longer (e.g. several months). One evaluation described offering refresher sessions. 152

Intervention materials

Interventions reported using various materials in the implementation of the intervention, including training materials, guidelines, multimedia resources, tools, posters, slides and policies. Some referred to materials that are publicly available on the internet (i.e. the Six Core Strategies, Safewards, Brøset Violence Checklist).

Costs and funding

Eighteen interventions referred to the cost of implementing the intervention or its financial impact. Costs were reported in the currency relevant to the study setting. Several studies provided details of the costs of one or more elements of the intervention, for example: US$10,000 for a Snoezelen room,198 US$11,456.98 for two Sensory Rooms,191 US$4000–5000 for sensory equipment,136,198 £70,000189 and £2000154 for environmental improvements, US$600 for PERT including office supplies and digital pagers,74 £69,285.25 for staff de-escalation and restraint training including replacement costs and overheads21 and US$20,000 for consulting fees. 198 Others referred to costs incurred but did not specify the amount, for example sensory equipment;127,196 camera, television monitor, three two-way radios;149 or staff training. 23,196 Putkonen et al. 17 reported the costs of their intervention as the equivalent to two person-years per year. Bell and Gallacher161 stated that their interventions incurred no costs.

Seven interventions reported who funded the intervention. Three reported receiving funding from the hospital where the intervention was implemented,23,74,127 whereas Mistral et al. 189 received funding from the mental health care trust. Putkonen et al. 17 specified that funding came via the hospital performance improvement project from research funding from the National Institutes of Health and Welfare. McEvedy et al. 137 reported receiving funding from the Victorian state government-funded programme. Lloyd et al. 136 reported that they received funding but did not specify the funder.

Nine interventions reported some cost–benefit analysis. Mistral et al. 189 reported a 62% reduction in time lost to staff short-term illness. Laker et al. 21 recommended further analysis, having been unable to draw conclusions from insufficient data regarding the costs of incidents, specifically damage to property, staff or service users and injury-related absence. Brown et al. 154 reported savings of 49% associated with reduced staff absence and injury (from £119,988 prior to implementation to £61,376 post implementation). Short et al. 157 reported a decrease of 77% in lost work days due to staff injury (90% of which were attributable to physical interventions). Lo74 argued that because the PERT intervention did not require any additional staffing resources it was likely to bring economic benefits. Finally, Putkonen et al. 17 reported 75% reduction in the number of sick days in the information period and 65% reduction in the intervention period, compared with the previous year; and in addition, reported 80–82% shorter duration of sick days.

Evaluations of interventions

Evaluations were identified using the screening questions of the MMAT: the presence of a research question and the collection of data required to answer that question. Those reports that passed the screening were then appraised using the MMAT to be given a score in the form of a percentage.

Of the 109 evaluations that we identified, 106 were research reports and there was one intervention report,92 one service report156 and one practice example. 159 Six theses and five abstracts were included. Most evaluations (n = 95) were published in 42 peer-reviewed journals spanning mental health, nursing, psychiatry and quality. The most common publication titles were Psychiatric Services (n = 12), Journal of Psychiatric and Mental Health Nursing (n = 7), Journal of Psychosocial Nursing and Mental Health Services (n = 5), International Journal of Mental Health Nursing (n = 6) and Psychiatric Quarterly (n = 7). The remaining journals featured between one and four records.

Controlled trials

There were 18 evaluation studies that used a control for comparison purposes (see Appendix 12) and six RCTs (see Table 5). Fifteen reported significant findings in relation to the decrease in use of a restrictive practice. However, they did not all use the same definition for the restrictive practices, with two studies combining all restrictive practices. 16,113 Only one study reported an increase in restrictive practices but this was confounded by major service changes during the intervention. Of those studies that reported decreased use of restrictive practices, eight reported restraint, eight measured seclusion, two measured PRN medication and two measured observation.

Qualitative studies

Cabral and Carthy15 examined the staff experience of implementation of Safewards, including its perceived benefits and challenges. They reported that staff perceived benefits, but there was some resistance, and they also identified a knowledge and skills deficit of the model. McEvedy et al. 137 explored services’ feedback on the effectiveness of a train-the-trainer intervention on knowledge transfer and translation of this knowledge into practice. They reported that knowledge transfer to staff was achieved, with most services subsequently training further staff (in some cases adapting the train-the-trainer materials). However, they found little evidence of translation of Sensory Modulation or Trauma-Informed Care into practice. Goulet et al. 166 examined the staff and service users’ perspectives of the impact of the intervention, including barriers to and facilitators of its implementation. Staff reported that the post-incident review gave them the opportunity to discuss the service users’ feelings and restore the therapeutic relationship, while staff were able to learn from the experience. Smith and Jones140 explored staff and service users’ experiences of using the sensory room, reporting that they considered it to be a positive contribution to the unit, that it had been incorporated into care plans and improved communication. Staff perceived that there had been a reduction in the use of seclusion but it is unclear whether or not service users were asked. Repique et al. 187 explored nurses’ views of Recovery-Based Principles training and reported that they perceived the training materials to be good but lacking specific detail and examples, and stated a preference for a live presenter with the opportunity to ask questions. Nurses also expressed a desire for more direction from leadership and service user feedback.

Chandler201 reported that seclusion and restraint were reduced ‘dramatically’ (although this is not quantified) and explored how the unit achieved this reduction and promoted safety. The author described how opportunities, information, support, resources and relationships were the factors that gave staff the leadership to feel empowered and, in turn, they empowered service users. These factors were key to reducing restrictive practices and promoting safety. For example, staff were able to learn about the impact of trauma, service users’ sensorimotor needs and stress responses, and the establishment of respectful and trusting relationships. Staff were also able to reflect on their responses to service user stress.

Interventions subject to multiple evaluations

Two follow-up studies (n = 2), one pilot plus main study (n = 1) and two replication studies (n = 2) were identified.

Goals and planning

Behaviour change techniques in the cluster ‘goals and planning’ were those identified most commonly in the interventions, with just over one-fifth (22%) of identified BCTs contained within this cluster (Figure 10). All the BCTs in this cluster were identified: ‘discrepancy between current behaviour and goal’ (BCT 1.6), ‘behavioural contract’ (BCT 1.8), ‘review for behaviour’ (BCT 1.5), ‘commitment’ (BCT 1.9), ‘review for outcomes’ (BCT 1.5), ‘goal-setting (behaviour)’ (BCT 1.1), ‘goal-setting (outcome)’ (BCT 1.3), ‘action-planning’ (BCT 1.4) and ‘problem-solving’ (BCT 1.2).

FIGURE 10.

Behaviour change technique cluster 1: goals and planning.

Problem-solving refers to when the individual or team is prompted to analyse what is influencing behaviour (unsuccessful de-escalation resulting in increased restrictive practice) and find solutions that attempt to overcome the problems or increase the likelihood of it happening. The ‘problem-solving’ described by interventions occurred in response to a problem of high rates of restrictive practices, or the resulting staff or service user injury. These problems could be identified at different levels: regional level, leading to a change in policy, within a service or in the review of individual service users. These reviews might involve the service user in the problem-solving process by taking place during individual care-planning meetings, debriefing after incidents or communal service user meetings. Problem-solving not involving service users took place during nursing handovers (e.g. McEwan et al. 123), staff clinical supervision (e.g. Prescott et al. 128) and in response to routine data review:

In weekly clinical supervision sessions, relevant data on seclusions and the circumstances leading up to its use were systematically discussed, in order to find ways to prevent future occurrence.

Mann-Poll et al. 212

Sometimes tools were used to support problem-solving, risk assessments and care plans:

A five-item recovery rounds checklist was developed by professional practice to prompt reflection and problem solving.

Hernandez et al. 99

A daily nursing assessment was also initiated by nursing staff. This included a brief mental state component . . . and assessment of risk of violence or harm to self or others, [resulting in an] individual service plan for the following 24 hours. This tool was important as its aided communication between nurses.

Sullivan et al. 104

‘Action-planning’ and ‘goal-setting’ followed a similar pattern, often as a result of problem-solving: planning with individual service users to try to prevent incidents that could result in restrictive practices, for example McEwan et al. ,123 or with staff teams when a ward had a target to reduce the number of episodes of restrictive practices, for example Lo. 74 There was also an indication of strategic action-planning and goal-setting, often through the formation of project steering groups to formulate institution-wide action plans to meet and review goals, for example Huckshorn. 133 Mention of reviewing goals was usually with reference to an individual service user and the goals documented in their care plan, for example Riemer. 156

Fewer interventions described ‘goal-setting’ than ‘action-planning’. Those that did consistently specified the reduction or elimination of restrictive practices, most commonly restraint and seclusion. 89 Other goals included improving staff cohesion and service culture:

The goals of the initiative included further reductions in S/R [seclusion and restraint] use and continued culture change to make the psychiatric inpatient and emergency services more patient centered and trauma informed.

Wale et al. 196

There were no examples of goal-setting with individual staff to reduce their use of restrictive practices.

Eight interventions24,176,190,201,213–216 described the use of a public ‘commitment’ (BCT 1.9) at strategic level to reduce restrictive practices:

Management has articulated (verbally and in writing) its intention of reducing the use of seclusion and restraint and/or to eliminate their use entirely.

Colton176

One of the more profound policy transformations was initiated by a declaration from the president and chief executive officer of RI [Recovery Interventions, Inc.], Gene Johnson, who mandated that seclusion and restraint practices would no longer be used and that the NFF [‘no force first’] policy would be implemented companywide, including in its crisis centers.

Ashcraft et al. 24

Three interventions described behavioural contracts to support individual staff commitment to reducing restrictive practices:16,137,217

[Trainees confirmed] that they would deliver the training when they returned to their service. The intention was that trainees should train end-users in all modules using the training materials they themselves had been trained with, in order to provide consistent training and maintain the integrity of the material.

McEvedy et al. 137

One other intervention made reference to a behaviour contract but it was unclear whether it referred to service user or staff behaviour. 103

Very few interventions discussed any discrepancy between the goal of the intervention and staff behaviour. Safewards has ‘clear mutual expectations’ that acknowledge this possibility, for instance to ‘. . . encourage the patients to refer to these expectations with staff, when they fail to uphold them’213 whereas Clark et al. 217 describe the commitment that staff make to refraining from using restrictive practices, via care plans with individual service users. Clark et al. 217 also report that patients are supported to challenge staff:

For example, if a tertiary intervention is used without any attempt at secondary intervention strategies, the patient is well within their rights to state that staff have not fulfilled their side of the contract and that the restrictive practices used may not have been justifiable.

Clark et al. 217

In only two instances was a discrepancy between current behaviour and intervention goal detected. 213,217

Shaping knowledge

The cluster ‘shaping knowledge’ contained 16% of the overall BCTs and included those that capture the imparted information about ‘antecedents’ (BCT 4.2) (40%; n = 60) and ‘instruction on performing the behaviour’ (BCT 4.1) (90%; n = 136), half of the four BCTs within this cluster (Figure 11). ‘Re-attribution’ (BCT 4.3) and ‘behavioural experiments’ (BCT 4.4) were not detected. Information about antecedents was detected as being provided at the theoretical, service, ward and individual service user levels.

FIGURE 11.

Behaviour change technique cluster 4: shaping knowledge.

Information about antecedents for individual service users was sought either proactively on admission145,212 or on a regular basis (daily153 or weekly200) and either more broadly in a care plan or as part of a specific safety plan or risk assessment. Very few interventions reported the involvement of family in this process:145,182,212

Implementation of Safety Plans, a collaborative document completed by the patient with the staff that recorded stressors and triggers.

Maguire et al. 116

This information could also be sought via debriefing after an episode of restrictive practice:

(d) identify triggers and antecedent behaviors that may have resulted in the use of restraint; (e) discuss alternative behavior and healthy coping strategies that may effectively minimize the future use of restraint should similar situations reoccur.

Riahi et al. , 201623

Information about antecedents could be informed by review of aggregate service data:153

Change ideas were developed through sharing theories about why violence was occurring and what would help to mitigate this. A range of stakeholders contributed to this theory-building in a facilitated workshop, including staff of all levels of seniority and different professional backgrounds, service users and the police liaison officer.

Taylor-Watt et al. 153

Staff training often included more general information about antecedents:

. . . staff were trained in early recognition of warning signs, thereby improving their risk-assessment skills, and their ability to prevent and manage aggression in early stages.

Georgieva et al. 150

‘Instruction on how to perform the behaviour’ (BCT 4.1) was one of the most frequently coded BCTs, present in 91% (n = 136) of interventions. Interventions varied as to how much detail they provided. Some merely reported that staff were trained in de-escalation, whereas others provided more detail. Specific instructions were provided, for example regarding how to resolve conflict and de-escalate situations or use Sensory Modulation strategies. Training often included Trauma-Informed Care (see BCT 13.2: ‘framing/reframing’) as well as avoiding restrictive practices through use of risk assessment, care planning and respectful communication. Many interventions also provided instruction for when de-escalation had failed with elements about managing violence, restraint skills and post-incident care, as well as legal and ethics issues.

Some training involved service users116 and when this was the case this had influenced the content of the training:

PMVA [Prevention Management of Violence and Aggression] training was revamped with much greater focus on communication, de-escalation and building therapeutic alliance.

Lombardo et al. 159

Very little mention was made of training supporting staff to regulate their own emotional responses; where it was mentioned it included raising awareness of and control of feelings, especially fear and anger. 15,129

Antecedents

The BCT cluster ‘antecedents’, involving factors that might influence whether or not restrictive practices can be avoided, was a theme throughout many interventions, typically in terms of preventing situations where service users might become distressed and conflict occur. ‘Antecedents’ was the third most populated cluster, containing 15% of BCTs (Figure 12). Half of the six BCTs constituting the antecedents cluster were identified: ‘restructuring the physical environment’ (BCT 12.1), ‘adding objects to the environment’ (BCT 12.5) and ‘restructuring the social environment’ (BCT 12.2). ‘Avoidance/reducing exposure to cues’ (BCT 12.3), ‘distraction’ (BCT 12.4) and ‘body changes’ (BCT 12.6) were not detected. Broader ward- and organisation-level changes to the physical and social environment described here as BCTs that aim to reduce antecedents by addressing individual service user needs are described elsewhere under ‘prompts’.

FIGURE 12.

Behaviour change technique cluster 12: antecedents.

Feedback and monitoring

Behaviour change technique cluster 2, ‘feedback and monitoring’, accounted for the monitoring of ward data and, if they were fed back to the ward, in what ways this feedback took place. Eleven per cent of BCTs were in this cluster (BCTs 2.1–2.7) (Figure 13). Both feedback and monitoring related primarily to outcomes (reduced restrictive practices) of the behaviour (de-escalation) although there was some evidence of monitoring of behaviour. Monitoring was either self-monitoring (including ward-level monitoring) or by others, for example at system level through incident reports. ‘Biofeedback’ (BCT 2.6) was not detected.

FIGURE 13.

Behaviour change technique cluster 2: feedback and monitoring.

Identity

The cluster ‘identity’ referred to the identity of the individual staff members and 9% of the BCTs identified were in this cluster. Two of the five potential BCTs in this cluster were detected: ‘framing or reframing’ (BCT 13.2) and the ‘identification of the self as a role model’ (BCT 13.1). The BCTs ‘incompatible beliefs’ (BCT 13.3), ‘valued self-identity’ (BCT 13.4) and ‘identity associated with changed behaviour’ (BCT 13.5) were not detected (Figure 14).

FIGURE 14.

Behaviour change technique cluster 13: identity.

Social support

The cluster ‘social support’ (Figure 15) accounted for 7% of the BCTs in the interventions and all three possible BCTs were identified as taking place within staff teams at ward level, more broadly within the organisation, or being shared with external agencies. Social support was either ‘unspecified’ (BCT 3.1), or further defined as either ‘emotional social support’ (BCT 3.3) or ‘practical social support’ (BCT 3.2).

FIGURE 15.

Behaviour change technique cluster 3: social support.

Practical support included ‘hands-on’ support in terms of sharing ideas for good practice in reducing restrictive practices. Taxis227 reported that at the outset a ‘weekly RN [registered nurse] discussion group’ was formed, with the goal of addressing practice issues, by raising awareness of the restraint and seclusion incidents, and to ‘build a consensus’ for increased use of less restrictive alternatives. Rather than providing ready-made answers for all situations, the goal was to ‘form a collegial environment in which these matters could be discussed’. This collegiality would serve as an important element as the programme progressed.

Melin69 reported that the meeting was intended to be a forum in which staff involved in treatment could have a positive and constructive dialogue about the interventions and the behaviour of colleagues. Guidelines included making sure that treatment staff were back in control before the meeting started, clarifying what happened, reviewing how staff responded and looking for ways to improve and strengthen responses in the future. Treatment staff were to be supported and encouraged, and trust in colleagues was to be expressed.

Safety huddles were coded as involving practical social support. The Northumberland, Tyne and Wear NHS Foundation Trust83 reported that the Positive Safe Team provide ongoing support including training, consultancy and opportunities for sharing good practice.

Some training providers offered ongoing practical social support for participants228 and, depending on the nature of training, it can offer practical social support in terms of feedback on role play.

The use of rapid response de-escalation teams, which featured in 12 interventions, was classed as providing practical social support.

Associations

As illustrated in Figure 16, associations were detected in 5% of BCTs in 57 interventions, although only one BCT ‘prompts or cues’ (BCT 7.1). None of the other BCTs in this cluster were detected: ‘cue signalling reward’ (BCT 7.2), ‘reduce prompts or cues’ (BCT 7.3), ‘remove access to the reward’ (BCT 7.4), ‘remove aversive stimulus’ (BCT 7.5), ‘satiation’ (BCT 7.6), ‘exposure’ (BCT 7.7) or ‘associative learning’ (BCT 7.8). Associations were predominantly prompts or cues (BCT 7.1) in the form of standardised assessments for service users on wards relating to risk of violence. 127 Mersey Care NHS Foundation Trust229 reported training staff in the PICU in the use of a structured risk assessment tool: the DASA-IV. DASA-IV scores were incorporated into daily care-planning. DASA-IV is a seven-item scale used for daily assessment of inpatients. Higher scores indicate a possible need for increased attention over the following 24 hours to reduce the risk of a serious violent incident.

FIGURE 16.

Behaviour change technique cluster 7: associations.

Other prompts included posters181,213 displayed in offices or wards and flow charts reminding staff of less restrictive practices (e.g. Alberta Health Services). 230

Comparison of behaviour

As illustrated in Figure 17, cluster 6, ‘comparison of behaviour’, comprises three BCTs: ‘demonstration of behaviour’ (BCT 6.1), ‘social comparison’ (BCT 6.2) and ‘information about others’ approval’ (BCT 6.3). Three per cent of detected BCTs were in this cluster and all three BCTs in this cluster were identified. Comparison of behaviour consists of experiencing demonstrations of the behaviour (BCT 6.1), usually as part of training sessions demonstrating successful de-escalation. ‘Social comparison’ (BCT 6.2) was found in everyday practice, through being compared with other wards or wards publicly acknowledging restraint rates, and receiving information about other people’s approval (BCT 6.3).

FIGURE 17.

Behaviour change technique cluster 6: comparison of behaviour.

Thirteen references,16,93,95,96,141,144,162,176,184,198,208,231,232 described de-escalation behaviour being demonstrated (BCT 6.1) as part of their training intervention, either by trainers or through role-play scenarios. ‘The Six Core Strategies intervention used Social comparison’ (BCT 6.2) at ward level, encouraging linked facilities to engage in healthy competition (e.g. by displaying data in open areas).

The Restraint Reduction Network promoted similar social comparison as ‘peer assessment’, in which one team or unit would be responsible for assessing the performance of another. This was felt to be effective in increasing motivation and engagement, and also in enhancing deeper understanding of assessment, and sharing good practice. 233

It became a feature at individual staff level within two interventions, one using Safewards. A well-publicised ‘star of the week’ initiative at Northumberland, Tyne and Wear NHS Foundation Trust was enthusiastically supported by staff and patients, and allowed a mild sense of competition between peers. 83

‘Information about others’ approval’ (BCT 6.3) was identified when there were descriptions of communication of support and approval from others. At ward level, one study described growing support from a consultant psychiatrist for intervention activities that had been instigated by nursing staff. It was also a feature of the social comparison activities when peers or service users nominated the best de-escalator or ‘star of the week’. Safety crosses were displayed publicly to show when the last incident of restraint had occurred:

Displaying safety crosses in the public area of the ward. This is a simple wall calendar that staff can mark in colour to show red days (when there was an incident of physical violence) or green days (incident-free). This evolved to include orange incidents, reflecting a near miss or build-up of hostility – which are not usually recorded in any form. This was an accessible way to share incident data and provided a focal point on the ward for staff, people using the service and visitors.

East London NHS Foundation Trust. 85 Contains public sector information licensed under the Open Government Licence v3.0

Natural consequences

As illustrated in Figure 18, in the BCT cluster ‘natural consequences’ a distinction is made between natural consequences and scheduled consequences such as a predetermined reward for performing a behaviour. A total of 3% of BCTs identified were in this cluster and included four BCTs: ‘information about health’ (BCT 5.1), ‘information about emotional consequences’ (BCT5.6), ‘salience of consequences (BCT5.2)’, ‘information about social and environmental consequences’ (BCT 5.3). ‘Monitoring of emotional consequences’ (BCT 5.4) and ‘anticipated regret’ (BCT 5.5) were not detected in any interventions.

FIGURE 18.

Behaviour change technique cluster 5: natural consequences.

The consequence of the desired behaviour (defined as successful de-escalation) was the elimination of the harmful effects of restrictive practices. Interventions described consequences of de-escalation failure predominantly on service users, although some interventions also referred to consequences for staff (e.g. Forster et al. ). 234 Some interventions merely talked about ‘negative’ consequences without specifying what these were, and so these were coded as health consequences. Health consequences (BCT 5.1) were specifically described in 23 interventions, primarily about physical risks to the service user. For example, Space Training235 included emphasis on understanding inherent risks of physical interventions, with a specific focus on ‘positional asphyxia’.

Some interventions (n = 7) emphasised physical and/or emotional health consequences, including interventions with a noticeable focus on the impact of restraint on the service user,123,143,195 and others that focused on health consequences for staff:70,141,236

. . . the goal of the program was both to reduce episodes of seclusion and restraint and reduce staff injuries. This intervention was designed to counteract the frequently encountered attitude that the real outcome of such programs is to place staff members at higher risk.

Forster et al. 234

. . . a series of mandated workshops on trauma-informed care, were created that included education on the neurobiological and psychosocial effects of trauma, the relationship of dissociative symptoms and self-harm to posttraumatic stress disorder (PTSD), and the retraumatization that occurs from being restrained or witnessing use of restraints and seclusion.

Chandler201

The broader social and environmental consequences of restrictive practices were referred to by only one study:

Has leadership evaluated the impact of reducing S/R [seclusion and restraint] on the whole environment? (This includes issues such as increased destruction of property; extended time involved in de-escalation attempt, additional admission assessment questions, debriefing activities and processes to document event, etc.)

Huckshorn133

‘Salience of consequences’ (BCT 5.2) was coded when interventions used particular methods to emphasise the consequences of not performing successful de-escalation and restrictive practices being used:

Each staff member experienced 5-point restraints first-hand, and many cited that experience as pivotal in their decision whether or not to restrain a patient in a state of agitation when queried 1 year after the course.

Forster et al. 234

The service users educated the project workers in consumer specialist meetings (1 hour per week) about their own experiences with violence and coercion, individual triggers of violence, and effective calming activities.

Putkonen et al17

This often included service user testimony about the consequences of their traumatic experiences of restrictive practices, for example Riley et al. 134

Repetition and substitution

As illustrated in Figure 19, 3% of the BCTs detected were in the cluster ‘repetition and substitution’ and consisted of three of the seven possible BCTs: ‘behavioural practice/rehearsal’ (BCT 8.1), ‘habit formation’ (BCT 8.3) and ‘reversal’ (BCT 8.4). ‘Behaviour substitution’ (BCT 8.2), ‘overcorrection’ (BCT 8.5), ‘generalisation of target behaviour’ (BCT 8.6) and ‘graded tasks’ (BCT 8.7) were not detected in any interventions.

FIGURE 19.

Behaviour change technique cluster 8: repetition and substitution.

Twenty-eight interventions described staff taking part in activities to practise and rehearse de-escalation skills. This was most typically in the form of role-play activities:

These team-building exercises, it was hoped, would highlight the different roles staff members play in the restraint process, crisis intervention, and de-escalation techniques. It was expected that these exercises would help to clarify the roles staff members play in a crisis and allow them to practice crisis management techniques.

Melin69

Detail was rarely provided about whether or not staff played the role of the service user. Some interventions did give details, with one using a fully immersive role-play scenario using actors in other roles.

There was little content in the interventions relating to habit, although one habit formation (BCT 8.3) technique was identified in Safewards, the requirement that staff:

. . . say something good about each patient at nursing shift handover . . .

Bowers et al. 16

There was one example of ‘habit reversal’ (BCT 8.4), used within four interventions,16,159,213,222 that encouraged staff to break the habit of saying ‘no’ to service user requests:

This has led to a culture of ‘say yes first’ at the trust, which helps patients to understand what needs to happen for a member of staff to say ‘yes’. The ‘reflect’ acronym helps staff to remember what they need to consider when answering a patient’s request: R – Reframe: What would it have taken to say yes? E – Easy: Was ‘no’ the easy option? F – Feeling: What would it have felt like? L – Listen: Did we listen? E – Explain: Did we explain? C – Creative: Were we creative enough? T – Time: Did we take the time?

East London NHS Foundation Trust, 2017. 85 Contains public sector information licensed under the Open Government Licence v3.0

First reflex should be saying ‘yes’, not ‘no’. Do you really need to say ‘no’? Can you justify saying ‘no’? Is this something that with a bit of effort or work or checking, you could say ‘yes’ or at least a partial ‘yes’ to?

Safewards213

Comparison of outcomes

As seen in Figure 20, cluster 9, ‘comparison of outcomes’, contained 2% of the BCTs with only one detected from the three within this cluster. ‘Pros and cons’ (BCT 9.2) and ‘comparative imagining of future outcomes’ (BCT 9.3) were not identified. The only BCT in this cluster was that of using a ‘credible source’ (BCT 9.1), usually as a way of imparting information. A variety of these credible sources were described within the interventions. Service users were used in 13 interventions:

When staff began to accept peers as co-workers and began to rely on them as a crucial part of the workforce, attitudes toward recovery changed significantly, and the tendency to use seclusion and restraint became more and more remote.

Ashcraft and Anthony193

FIGURE 20.

Behaviour change technique cluster 9: comparison of outcomes.

The psychological impact of restrictive practices was described by service users as part of training or other awareness-raising activities, this was also coded as ‘salience of consequences’ (BCT 5.2):

At the engagement sessions, delivered in partnership with service users, teams are introduced to No Force First and hear accounts of people’s experience of physical intervention.

Riley et al. 134

Clinical specialists were also described as credible sources in other interventions, either delivering training or describing the physical impact of restraint on the service user (e.g. a medical director in Madan et al. 108). One other type of credible source was academic researchers feeding back intervention data to ward staff (e.g. Mann-Poll et al. 212).

Reward and threat

As illustrated in Figure 21, cluster 10 (‘reward and threat’) contained 4 of the 11 possible BCTs. Future punishment, self-reward/incentive, non-specific reward/incentive and material reward/incentive (behaviour) were not detected. Two per cent of BCTs were in this cluster. Incentives and rewards could be ‘non-specific’ (BCT 10.6, BCT 10.3), ‘material’ (BCT 10.1, BCT 10.2) or ‘social’ (BCT 10.5, BCT 10.4). There could also be ‘self-incentive’ (BCT 10.7) and ‘self-reward’ (BCT 10.9), and could also be associated with ‘behaviour or outcome’ (BCT 10.8, BCT 10.10).

FIGURE 21.

Behaviour change technique cluster 10: reward and threat.

Most of the incentive and reward described within the interventions was of the social variety (BCT 10.4, BCT 10.5). In some interventions, wards or individuals were praised for reducing restrictive practices (e.g. Szypula and Martin89) through awarding certificates, celebrations of success, favourable publicity in service newsletters, and notice boards where positive messages can be left. For instance, a board in a staff office was used for staff to write supportive messages and compliments about the good work they were doing; this proved popular and eventually transferred into the patient area so that both patients and staff could write on it. 83

Blair and Moulton-Adelman222 reported the following:

Ongoing recognition of unit successes and individual staff initiatives related to improved patient care is encouraged. Such steps reinforce positive movement toward a therapeutic culture of care and improve overall morale and cohesiveness.

Blair and Moulton-Adelman222

Blair and Moulton-Adelman222 describe that success is acknowledged in many ways, including recognition of individuals in staff meetings, department-wide e-mails, and get-togethers organised across the unit for the purpose of celebrating the achievement of specific milestones. The celebrations help to reinforce awareness of the goals yet to be reached. 222

‘Material incentives and rewards’ (BCT 10.1, BCT 10.2) were few, although one intervention had a competition with a prize for the best performing ward, and cinema tickets were raffled among ward staff. 237 Continuing professional development (CPD) credits were used in one intervention as an incentive for attending training. 115

Regulation

Very few of the interventions used BCTs from the cluster ‘regulation’ (0.6%) and only one of the four possible BCTs was identified. As illustrated in Figure 22, ‘pharmacological support’ (BCT 11.1), ‘conserving mental resources’ (BCT 11.3) and ‘paradoxical instructions’ (BCT 11.4) were not identified. The only BCT detected in this cluster was ‘reduce negative emotions’ (BCT 11.2), which was targeted at reducing staff stress in order to promote the reduction of restrictive practices. Several interventions (n = 8) described a focus on reducing staff stress in general terms, although some addressed specific aspects of staff stress including anxiety,141 frustration236 or burnout:70

The strategy of allowing people to vent about possible negative outcomes was needed to create space for possibilities.

Ashcraft et al. 24

FIGURE 22.

Behaviour change technique cluster 11: regulation.

Other studies focused on reducing staff negative emotions during the post-incident debrief:

Opportunities are provided/scheduled to process the event with staff about their feelings, reactions, and safety.

Colton176

Scheduled consequences

Few aspects of cluster 14, ‘scheduled consequences’ (reward and punishment), were used in the interventions, with only 0.3% of BCTs being in this cluster and only one BCT identified: ‘remove punishment’ (BCT 14.10). ‘Behaviour cost’ (BCT 14.1), ‘punishment’ (BCT 14.2), ‘remove reward’ (BCT 14.3), ‘reward approximation’ (BCT 14.4), ‘rewarding completion’ (BCT 14.5), ‘situation-specific reward’ (BCT 14.6), ‘reward incompatible behaviour’ (BCT 14.7), ‘reward alternative behaviour’ (BCT 14.8) and ‘reduce reward frequency’ (BCT 14.9) were not identified (Figure 23).

FIGURE 23.

Behaviour change technique cluster 14: scheduled consequences.

The removal of punishment referred to the removal of an unpleasant consequence if the desired behaviour is performed. In five interventions,132,133,159,238,239 this was described as ensuring that debriefing sessions were supportive and not blaming to promote reflection on attempted de-escalation from the staff members involved, even if the episode had resulted in the use of a restrictive practice:

There was a focus on performance, but the approach was one of curiosity and help rather than summative judgement.

Lombardo et al. 159

Administrators have carefully fostered a welcoming, nonthreatening atmosphere that helps direct care staff overcome their initial trepidation about attending the meetings in the medical director’s office. Leaders have purposefully avoided any appearance of assigning blame or ‘second guessing’ decisions that have been made by direct care staff.

Allen et al. 238

Unused or little used behaviour change technique clusters in interventions

Two clusters of BCTs, ‘self-belief’ and ‘covert learning’, were not detected in any of the interventions (Box 1). The ‘self-belief’ category contains techniques that aim to promote self-efficacy, a determinant that has a key role in reducing lapses and coping with behavioural barriers. 240

Some clusters featured only by virtue of one BCT being coded (coded BCT in bold) (Box 2).

As illustrated in Box 2, there were two little-used BCT clusters: ‘regulation’ and ‘scheduled consequences’. Within these, ‘reduce negative emotions’ (highlighted in italics in Box 2) was coded in relation to stress reduction elements of an intervention, and ‘remove punishment’ (highlighted in italics in Box 2) referred to ensuring that, when de-escalation had failed and restrictive practices had been used, the staff member could debrief reflexively on what could have been done differently, rather than in anticipation of punishment.

It should be noted that, for a BCT to be identified in an intervention, there had to be evidence of its presence within the intervention materials. Therefore, there may be instances where a BCT remained unidentified owing to lack of evidence. See Appendix 14 for a full list of the BCTs that were not identified in the interventions.

Training

As illustrated in Figure 24, staff training was the procedure most frequently reported in interventions. The most frequent BCT detected was that of instruction, although that was coded for every training intervention even if no detail was provided. Following instruction, there were numerous other BCTs detected; Figure 24 shows those featuring in > 10% of interventions. Reframing the beliefs of participants often involved education about Trauma-Informed Care or, less frequently, the recovery model. ‘Problem-solving’ featured within training when participants were asked to consider their own settings and examine what antecedents led to incidents, and ‘action-planning’ featured where it was decided to make changes to avoid them. ‘Restructuring of the social environment’ via training might include aspects such as encouraging staff to try to have more contact with service users.

FIGURE 24.

Procedures: training: 135 interventions coded.

Training often involved staff practising or rehearsing de-escalation by taking part in role play, with each other, with the trainers or even with actors. Feedback on performance could involve social support from peers and trainers. Training was delivered by professional trainers from specific companies, sometimes using a ‘train the trainer’ model. Other interventions used senior clinical staff to deliver training and external trainers often came from clinical backgrounds. Service users were included in training, delivering information about the consequences of experiencing restrictive practices and increasing the salience of those consequences through telling of their personal experience.

Some interventions encouraged trainees to see themselves as a role model by taking their learning back to the workplace and sharing it with other staff, as well as leading by example.

Audit and feedback

Audit and feedback are when clinical performance-related data are collected and fed back to health-care staff, and are often available in electronic format. As illustrated in Figure 25, the audit and feedback procedures in the interventions typically consisted of data being collected from ward staff about any restrictive practices that occurred on the ward. These data might be the number of restraints or seclusion, the duration of seclusion or the use of PRN medication. Data are often entered by ward staff onto a centralised system. A number of different variables affected how the data were fed back to staff teams. The frequency of feedback was variable and could be daily, weekly, monthly or quarterly. The level of detail varied in terms of whether it was ward- or service-level data. The means of feedback could be in person or electronically, to an individual or a team. Data could be made public or retained within the staff team. No data were fed back at individual staff level. The resulting discussion about the data could involve problem-solving and action-planning, particularly if the rates of restrictive practice had not decreased in line with goals. Practical social support was evident here with staff making suggestions and supporting each other, and celebrating successes, a type of social reward. Some wards self-monitored their rates of restrictive practices, perhaps collecting more detail than would be collected by central incident reporting systems, for example about antecedents. Comparison to others might feature if wards were comparing their rates with other wards. Some interventions also monitored outcomes without feeding them back to staff.

FIGURE 25.

Procedures: audit and feedback: 53 interventions coded.

Nursing changes

As illustrated in Figure 26, the BCT detected most frequently in these interventions was ‘restructuring the social environment’. This often occurred through making staff more accessible and proactive to service user need as well as introducing new meetings to discuss ward safety (e.g. safety huddles), maybe with a regular prompt, often involving the ward staff monitoring their behaviour and providing each other with social support both practical and emotional. ‘Problem-solving and action-planning’ was carried out at individual and ward level with the overall milieu of the ward being the focus or individuals who might be experiencing distress. The underlying philosophy of care might be changed by adopting an approach such as Trauma-Informed Care or the recovery model. In these cases, ‘reframing’ would be used to change the beliefs of staff to view distress in a different way. Such approaches might encourage staff to see themselves as role models practising the new approach with service users and supporting other staff. Sometimes physical changes were made to service settings to make them more comfortable, or to include quiet areas: objects could be added such as exercise equipment or books and multimedia devices.

FIGURE 26.

Procedures: nursing changes: 48 interventions coded.

Service user involvement

As illustrated in Figure 27, the BCTs that were present with service user involvement were predominantly ‘restructuring the social environment’, ensuring that service users informed the development of interventions or parts of interventions. Service user committees were used for a variety of reasons including the ongoing monitoring of the behaviour of ward staff, reviewing the rates of restrictive practices and contributing towards problem-solving. Service users were part of delivering training, as a credible source testifying about the harms of restrictive practices, and were sometimes employed to provide staff with mentoring.

FIGURE 27.

Procedures: service user involvement: 37 interventions coded.

Care planning

As seen in Figure 28, care planning was one procedure used to reduce restrictive practices, primarily by attempting to problem-solve, by relating to service users’ needs while they were resident in a setting. The planning was frequently prompted by an admission, or might be carried out weekly with staff having a discussion with the service user about their current situation and/or previous experiences. It often included soliciting information about what ‘antecedents’ could cause distress and how they can be avoided. If a new method of care planning was introduced as part of the intervention, perhaps by placing greater value on service user perspectives, this might also be ‘restructuring the social environment’.

FIGURE 28.

Procedures: care planning: 37 interventions coded.

Communication strategies

As illustrated in Figure 29, procedures to improve communication focused on restructuring the social environment, often by introducing new fora for examining current behaviour and then problem-solving and action-planning with the provision of social support, sometimes involving service users.

FIGURE 29.

Procedures: communication: 32 interventions coded.

Procedures could add objects to support communication, for example a wall calendar allowing service users to book appointments with a doctor if they wished. These procedures were often informed by framing/reframing approaches such as Trauma-Informed Care and promoting service user involvement.

Debriefing

As illustrated in Figure 30, debriefing typically took place within a few hours of an incident of restraint. Debriefing after an incident when restrictive practices were used was an opportunity for both staff and service users to reflect on what happened (What antecedents had led up to the incident? What specific problems had arisen? Why had they not been solved? What could have been done differently? What would reduce the chances of the same thing happening again?) and its introduction was coded as a restructuring of the social environment. This involved problem-solving, action-planning and examining antecedents as well as receiving feedback on outcomes of the incident. The debrief was usually framed in positive terms so that staff did not feel that it was a form of punishment, rather an opportunity to reflect on what had happened and examine their behaviour. Debriefs involved a range of different people; they sometimes included only the staff member(s) involved but at other times included the service user, members of the broader staff team and senior managerial staff. One feature was the provision of social support, both practical and emotional.

FIGURE 30.

Procedures: debriefing: 30 interventions coded.

Sensory approaches

As illustrated in Figure 31, sensory procedures almost always included the addition of objects to the environment, usually sensory modulating equipment, available for service users to use on a ward. Some interventions restructured the physical environment by adding dedicated Sensory Rooms, quiet areas or even gardens. Assessments on admission for sensory preferences might become prompts to ensure that these facilities were tailored to the individual service user and involved problem-solving and action-planning.

FIGURE 31.

Procedures: sensory: 27 interventions coded.

Policy change

Procedures that involved a change of policy or rules restructured the social environment (Figure 32). Some also involved framing/reframing such as adopting a recovery approach, social support through restructuring meetings, prompts to carry out regular risk assessments, and problem-solving and goal-setting often through meetings. Objects added to the environment included posters displaying information about service rules.

FIGURE 32.

Procedures: policy: 26 interventions coded.

Risk assessment

As illustrated in Figure 33, the procedures for risk assessment were similar to care planning but focused on risk and safety. The most common BCT was the staff member identifying antecedents, problem-solving and action-planning with or about the service user. As with more general care planning, risk assessment was often prompted on admission to services or was carried out on a regular basis. This change in practice could sometimes be part of a broader change in the social environment.

FIGURE 33.

Procedures: risk assessment: 29 interventions coded.

Enhanced leadership

Procedures to enhance leadership were dominated by changes in the social environment (Figure 34), sometimes involving a wholesale adoption of an approach such as Trauma-Informed Care. These procedures included more contact with ward staff, the formation of project steering groups reviewing both behaviour and outcomes, and managers working on wards and being seen as role models in the attempts to reduce restrictive practices. Other aspects often informed through these mechanisms were problem-solving, action-planning and goal-setting for both behaviour and outcomes.

FIGURE 34.

Procedures: enhanced leadership: 29 interventions coded.

Role modelling

As illustrated in Figure 35, the development of staff as role models involved them self-identifying as a role model, operating as such within the staff team, predominantly by providing social support to their colleagues. The use of role models often involved restructuring the social environment in terms of introducing new staff who were the role models, or training existing staff to fulfil this position.

FIGURE 35.

Procedures: role models: 11 interventions coded.

Increased staffing

Increased staffing was rarely associated with any other BCT other than restructuring the social environment, coded in 91% (n = 11/12) of the interventions.

Activities for service users

As illustrated in Figure 36, few BCTs were detected in relation to the introduction of activities for service users. These changed the social environment to reduce the use of restrictive practices by improving the options open for staff when working with service users. They could include adding objects to the environment, such as sports equipment or games consoles.

FIGURE 36.

Procedures: activities: seven interventions coded.

Rapid response teams

Several interventions (n = 8) made use of rapid response teams although few BCTs were detected within these descriptions. The availability of teams restructured the social environment of the ward and included staff of varying professional backgrounds, including nursing and security staff who had received instruction in de-escalation. On arrival they provided practical, social support to ward staff; there were also elements of problem-solving and action-planning detected in two of them.

Overview

The overall purpose of the study was to document the range of interventions developed to reduce restrictive practices, describe their characteristics and identify any overarching patterns. The search identified a disparate and complex collection of interventions in many sources and formats, steadily increasing over the 20-year search period since the introduction of the UK National Service Framework for Mental Health in 1999, and peaking in 2017. Most interventions can be found in research reports published in nursing journals. Various mental health-related and other non-governmental organisations provide resources for service providers in the form of links to training providers and examples of good practice in the reduction of restrictive practices.

Characteristics of interventions

The search found a total of 150 unique interventions to reduce restrictive practices. This included 29 intervention families (interventions with multiple records) and 121 standalone interventions (with only single records). Six Core Strategies was the largest intervention family and had been implemented in the broadest range of countries (n = 6) and clinical settings (n = 6). Safewards had also been implemented in multiple countries (n = 3) and clinical settings (n = 2). The highest numbers of interventions (and widest range) were found in the UK and the USA; others were identified in Canada, Australia, New Zealand, Singapore and several European countries. Most interventions were implemented in unspecified adult mental health inpatient settings, but, where more detail was provided, the most common setting in which to undertake an intervention to reduce restrictive practices was an acute ward.

Variation in intervention reporting

Unsurprisingly, given the broad inclusion criteria and the number of interventions subsequently identified, there was enormous variation in intervention reporting. The interventions were often intended to address multiple restrictive practices and related issues (e.g. service user behaviour, staff skills, quality, safety and adverse consequences) in varying combinations. The most common intervention target was seclusion or restraint; however, 11 interventions targeted PRN medication. Similarly, most interventions used multiple procedures, in some cases as many as 10, and, again, in varying combinations. The most common procedure was Training, although Changes to Nursing Approaches and Review of Incident Data were also often used. Despite the popularity of training as a key procedure, detailed descriptions of the training content and providers were often lacking. The least common procedures were Rapid Response Teams and Activities for Service Users. Forty-eight interventions reported involving service users in some way, but the type and extent of their involvement varied greatly. In some cases, service users were involved in multiple ways, whereas in others they had limited roles.

One hundred and three out of the 109 evaluations used a non-randomised design and there were just six RCTs; the remainder were qualitative or mixed-methods studies. Just two evaluations reported service user involvement in the evaluation. Seventy of the evaluations reported multiple outcome measures, and these most frequently focused on the incidence of seclusion or restraint; however, there was limited consistency as some noted the timing of incidents and others the severity. Only 18 made any reference to the cost of implementing the intervention or its financial impact.

Discussion of analysis

The study clearly shows that there has been an upturn in publications and research endeavour in this field in recent years. In the UK, at least, this is possibly a response to the publication of the Department of Health and Social Care’s response to events at Winterbourne View. Restrictive practices began to attract attention following deaths that occurred during their use, and this increased further when the abuse of patients at Winterbourne View Hospital was documented in 2011. 244

Nevertheless, regardless of an increase in the volume of the literature, the analysis highlighted gaps and huge inconsistency in study design and units of analysis across both sample size and settings. Reporting was generally poor and inconsistent, despite guidelines such as WIDER,61 which recognises the need to standardise reporting of complex social interventions aimed at changing behaviour and provides tools to address the issue. One explanation for this is that it is difficult to capture or measure who is exposed to the intervention, for what duration and at what intensity. To some extent, the level and clarity of detail provided were determined by the format of the retrieved record, and the nature of the intervention and its procedures. Although training might be conducive to measurement, many other procedures, including changes to policy, nursing approach or the physical environment, could be described as standalone events or as ongoing. Therefore, notwithstanding the advent of tools such as WIDER, the analysis showed that limitations remain when it comes to measuring fidelity, and dose, of interventions designed to reduce restrictive practices.

Discussion of quality assessment

The evidence base for interventions to reduce restrictive practices appears to be small; only six RCTs were identified, and variations in their targets and outcome measures precluded any meta-analysis. Most evaluations were published in nursing journals of low impact. The evaluation methods themselves were often poorly described and, in particular, lacked detail about the interventions or methodology that had been applied or the theoretical basis for the intervention. Forty-one potentially eligible evaluations of interventions were excluded from the analysis as they did not pass the MMAT screening questions.

Issues of quality and reporting are particularly problematic when trying to provide precise descriptions of interventions subject to evaluation because of the implications for replication and meta-analysis: ‘to facilitate replication, further development, and scale-up of the interventions’. 61 This clearly had an impact on the ability to apply the taxonomy, make meaningful comparisons of interventions or undertake a meta-analysis of the results. For those studies identified outside the academic literature, experience suggests that other types of reports/formats cannot be relied on to provide comprehensive details of an intervention.

On balance, the evidence suggests that interventions place greater emphasis on service users’ contribution to circumstances that lead to the use of restrictive practices than on the impact that restrictive practices (or their reduction) have on them. Many interventions reported that they sought to improve service user experience by, for example, promoting recovery or providing Trauma-Informed Care and included specific aims to improve quality of care, service user experience, a feeling of safety and recovery. Nevertheless, few reported examining service user outcomes such as injury or perception of safety. Although they measured the incidence, and sometimes duration, of restraint or seclusion, they did not report these incidents in terms of their impact on service users. Although service users were sometimes involved in intervention delivery, they were rarely involved in intervention design or evaluation. A good proportion of the interventions that were reported included procedures aimed at changing service user behaviour and engaging service users in identifying their own triggers, strategies and preferences. In contrast, interventions inviting managers and frontline professionals to reflect on their practice required a focus not on their triggers, for example, but instead on understanding service users’ histories.

Shaping knowledge

The majority of interventions used staff training as a mechanism to reduce restrictive practices. The training was usually a procedure of the intervention, for example to improve de-escalation skills, but could also form part of another procedure of the intervention such as training to use a risk assessment tool. Training was most likely to be part of a single-procedure intervention but, as the content was poorly described in many instances, it was not possible to determine the delivery, content, duration and frequency of the training. Training can contain a variety of different BCTs, for example social support, shaping knowledge, information about consequences, demonstration of the behaviour, rehearsal, identity, comparison of outcomes and self-belief. It may be that some of these BCTs are more effective in changing behaviour than others and some might have no discernible effect. Theorising how training might be expected to change behaviour would allow for the selection of some of these BCTs and their subsequent testing.

Goals and planning

The use of goals and planning to reduce restrictive practices included strategic goals and planning at several levels (i.e. individual planning with service users, ward team goals, and at the wider organisational or health system level). It is of interest whether or not health system level goals and planning have as much impact as those generated at ward level and whether or not goals and planning with individual service users is more effective at reducing restrictive practices. Some interventions claimed that those initiatives generated at ward level had greater ‘buy-in’ from staff, whereas others claimed similar importance of top-down initiation and support. These often occurred simultaneously in the same intervention so were difficult to distinguish. Coding of BCTs found at individual, team and system levels would go some way to address this difficulty, although that was beyond the scope of this study as it sought to describe content overall and via individual procedures. Theorising why using ‘goals and planning’ employed at these different levels might effect change would allow these effects to be tested in isolation.

Antecedents

The social environment of mental health services was thought to contribute to the use of restrictive practices, and many interventions sought to improve it through the promotion of interaction between service users and staff. This became a broad category, somewhat reducing its sensitivity. The physical environment was frequently involved, although generally limited to changes made to wards to make them more therapeutic, something recently highlighted by the CQC in the UK. There was less attention paid to the overall architecture and design of wards, which was unsurprising when few services have resources for a full redesign. There may be aspects of overall structure, such as visibility of staff and service users, natural light, opportunities for private space or fresh air, that reduce restrictive practices. Again, these modifications were often made in conjunction with other intervention procedures, meaning that the effect on behaviour of physical restructuring alone is unknown.

Feedback and monitoring

This was a frequently identified component of many of the interventions and has a range of BCTs associated with it. As described above, it is difficult to identify feedback on behaviour within these interventions as successful de-escalation may not be recorded, whereas outcomes of the (failed) behaviour are routinely recorded. As well as the feedback and monitoring BCTs, other BCTs relating to social support, information about antecedents, social comparison and reward within these procedures were also detected; it is of interest what role these play in these procedures. There was little description of how the data were fed back to staff, whether delivered verbally, in writing, electronically or during a meeting. It was often unclear what level of detail was provided in the data, whether it was at ward level or service level, over what time period and how identifiable incidents would be to those involved. These might be important factors to test to better determine whether or not feedback and monitoring changes behaviour, and again the use of theory would allow these factors to be tested.

Targeting behaviour of staff versus behaviour of service users

One of the challenges of this approach was associated with the target that the intervention was focused on, that is staff or service user behaviour, and subsequently how to code this. Many instances were detected in interventions when a BCT was being used that was targeting the service user but that could also change staff behaviour, for example the introduction of a sensory room aiming to change the behaviour of service users and facilitate self-care to reduce distress. In that respect it was targeting service user behaviour; however, it also changed staff behaviour because the introduction of a sensory room provides an additional resource that they can use to support people experiencing distress. This has come about through either restructuring the physical environment or adding objects to the environment to either reduce antecedents to conflict or promote de-escalatory behaviour if conflict arises. As such, this was included and coded as a BCT. The interpersonal relationships between service users and health-care staff are extremely important and it cannot be assumed that the escalation of conflict behaviour is always generated by the service user: the staff member may contribute equally, whether intentionally or not. An intervention such as a sensory room can therefore facilitate de-escalatory behaviour from the staff member as it might promote calmer behaviour from the service user.

A further example was apparent in ‘action-planning’, and this often included care planning with an individual service user. It could be seen that this was an attempt to change the service user’s behaviour in the event of them becoming distressed and, on one level, this may prove the case; however, the activity of action-planning with the service user may also cause the staff to action-plan. They may discover that when this service user becomes distressed, they are made more distressed by a noisy environment. They can then plan to help the service user to find a quieter space in the event of distress, which in turn successfully avoids the use of restrictive practices. As interventions can affect the behaviour of staff, service users and visitors in the therapeutic milieu to different degrees, it is important that this is acknowledged while retaining the focus on staff behaviour and how it might be changed.

The one BCT that lacked sensitivity in this study was also one of the most frequently used, namely ‘restructuring the social environment’. This was used to capture a broad range of changes implemented by interventions from a strategic management level through to everyday interactions between service users. The decision to create subcategories (e.g. access to staff, management support, promote socialisation) for this BCT to retain this detail addresses this difficulty to a certain degree but it might be considered a limitation when viewing the overall summaries of the BCTs coded in this study.

Future development of BCT coding could consider the use of subcategories. Some of this information could be seen as contextual rather than relating to a specific BCT. For example, it is unclear whether a new type of staff meeting is a BCT in itself or the BCT is only what happens during the meeting, for example problem-solving and practical social support. This may be a grey area that was specific to these data and not easily addressed through the use of subcategories. Presseau et al. 47 also found that this code required more specificity when coding system-level interventions, and recommended the addition of what is ‘restructured’, preferably within an explicit programme theory.

‘Behaviour’ versus ‘outcomes of behaviour’

For this study, the target ‘outcome of the behaviour’ was an absence of something happening, such as reduced seclusion episodes or fewer restraints. The target ‘behaviour’ was successful de-escalation, which is itself not an absence of carrying out restrictive practice but a proactive attempt to avoid it by using other strategies. Although it was clear for this study that the desired ‘behaviour’ was that which sought to reduce the use of restrictive practice, often described as de-escalatory behaviour, this was almost never measured or even accounted for among the interventions. The focus was always on the ‘outcomes’ of the behaviour, such as the number of restraints or seclusions. This meant that feedback about the ‘behaviour’ could rarely be identified as the focus was on feedback on ‘outcomes of behaviour’. This somewhat fails to acknowledge successful de-escalation.

Behaviour change technique dosage

The frequency of use of BCTs whether they were present or not within an intervention was summarised, not how many times evidence of their use was detected. Therefore, this does not reflect the ‘dose’ of each BCT. It is possible to report this cumulatively using NVivo software, but is not meaningful across a large number of interventions.

The use of theory was not explicit in many of the interventions, which is not unusual in health services research. 66 This is a significant weakness of this evidence as the explicit use of theory would allow better comparison between interventions in terms of what assumptions the intervention is based on, how the techniques that have been employed are expected to have an effect, and how effective this ultimately was. It is recommended that developers of future interventions make more explicit use of theory.

Comparison with previous research

There is little previous research with which to compare the results of the current study. The only other study to apply the BCT taxonomy to interventions seeking to change professional behaviour, Presseau et al. ,47 reported some similar findings, albeit in a different area of health care (diabetes care).

Pressau et al. 47 identified 21 BCTs in 11 clusters, compared with 43 in 14 clusters identified in the current study. The BCTs identified most frequently in Pressau et al. 47 differed from those reported here, which is not unexpected as the two studies concerned interventions targeting different issues. Pressau et al. 47 did not find covert learning, self-belief (similar to the current study findings) or use of regulation and scheduled consequences (in contrast with the current study findings) evident in intervention materials. The current study detected small numbers of one of the BCTs in each of these categories. In addition, Pressau et al. 47 detected no BCTs targeting identity, whereas role modelling and ‘reframing’ occurred within many of the interventions reported in the current study. This latter point may be related to the reality that moving to a Trauma-Informed Care system involves reframing perspectives on restrictive practices as a key element.

Therefore, the current study contributes further insights into the potential of the BCT taxonomy to understand mechanisms for change in the behaviour of staff in health settings. Potentially, further research into the use of BCTs may facilitate the emergence of a recognisable set of BCTs likely to be effective across diverse health settings, and others which are particularly helpful in specific contexts.

Strengths

As demonstrated in Chapters 4–6, the study was successful in meeting objectives 1 and 2. The current report provides a high level of detail concerning the range, components and application of interventions using BCTs.

A rigorous approach to coding documents was employed, supported by qualitative analysis software (NVivo). All intervention information was stored in one project and the initial coding frame took the form of the BCT taxonomy. Adopting a deductive qualitative analysis for the BCT coding enabled within-BCT-category coding, which allowed the identification of subthemes within the more populated BCTs. It also supported the use of memos to allow communication between the two coders about areas of divergence. This approach was, on reflection, considered to have been a success and will be repeated with a similar study.

Limitations

The broad search strategy that included a wide range of grey literature in diverse formats, together with no criteria for exclusion on the basis of quality, was developed in response to the research team’s familiarity with the research field and prior awareness that, although the research evidence was likely to be sparse, there were a large number of interventions being used in practice, with potential to be mapped onto the BCT taxonomy. Our novel approach to searching was comprehensive, and effective in its results, yet records of interventions were disparate in terms of focus and quality. The strategy was thus useful for identifying the number and range of available interventions; however, it does raise questions about how such diverse evidence can be assimilated and appraised in a meaningful way.

Quality of description

With some exceptions, the quality of the description of the content of interventions was very poor. This meant that it is not possible to detect precisely what is happening without information from individual staff. Given that the reporting of the content of interventions is poor, this is a weakness of all reviews of this sort. Contact details for obtaining further information for 140 records were extracted but there were only 14 replies to 100 requests for further information. Four interventions devised by commercial training companies charged a fee for access to their materials. Those resources that had been developed often lacked sufficient detail for coding or had been poorly stored, for example citing outdated URL links. Standards for reporting the development of health research [e.g. via the Enhancing the QUAlity and Transparency Of health Research (EQUATOR) network]245 do not often appear to have been applied. The use of appropriate guidelines would greatly enhance the quality of research in this area. The reporting of outcomes is a particular difficulty owing to the diverse nature of outcomes used and metrics applied.

Our approach to BCT coding, for example the overcoding of certain BCTs, predominantly ‘instruction on how to perform the behaviour’ (BCT 4.1), may be considered a risk because this may have enabled the recognition of specifics of training that may not have occurred. Given that this BCT has been used in nearly all of the interventions, it was unfortunate that it was frequently very poorly described and, as aspects of training can vary dramatically, an understanding as to the exact content of much of the training was not gained. However, if this code had not been used whenever training was involved then it would have been under-reported. On balance, it was concluded to be preferable to over-report, with an accompanying call for clarity, rather than disregard. ‘Restructuring the social environment’ (BCT 12.2) was also used in a wide variety of circumstances and consequently there is some overlap with contextual factors such as service user involvement. It was thought better to have some duplication in this important area than not to code it. The impact of joint working between service users and health-care staff might deter the use of restrictive practices, potentially caused by staff knowing the service user through this different relationship.

Setting

As this was an exploratory piece of work it was limited to adult mental health services. Although a range of adult inpatient settings were included (including acute, forensic and PICU services) this review excluded a range of services and settings where restrictive practices are commonplace. This included Child and Adolescent Mental Health Services, and interventions targeting people with learning disabilities and organic conditions. It may be worthwhile conducting similar research to explore the techniques used to reduce restrictive practices across other settings.

English language as an inclusion criterion

It was beyond the study remit to include material that was not reported in English. It is likely that, by limiting the review to records in English, it was predisposed to identifying interventions from the USA, Canada, Europe and Australia. It is difficult to assess how many additional records might have been included in this review had this limit not been imposed. There are international efforts to reduce restrictive practices across adult psychiatric institutions. For example, there are widespread efforts across Europe that are published in non-English-language journals. Further afield, the use of Pasung (physical restraint and confinement) in Indonesia has been under considerable scrutiny. It is likely that some interventions in some countries that aim to reduce restrictive practices may not have been included.

Multiple procedures

The multiple components and multiple aims of the interventions causes further obscurity. Although a single target might on the face of it be ‘reducing restrictive practices’, this is a target made up of a range of different practices. Many of the interventions had multiple components; however, without individual testing of the components of interventions, resources might be wasted on elements that are simply not effective, or are perhaps effective only when coterminous with another intervention. This study shows that complex interventions have been developed and implemented, potentially ahead of the rigorous testing required to develop a robust evidence base. For example, the current review did not find evidence to answer questions such as Is repeated staff training necessary? How important is action-planning? Is overall care planning better than specific safety planning or risk assessment?

Outcomes

Restrictive practices include a variety of outcomes and they are often reported in diverse ways using a range of units of analysis, for example restraints per service user, or restraints per day. Studying the reduction in the use of PRN medication might require a quite different approach to outcomes evaluation, compared with reducing restraint. For example, where one restrictive practice is reduced, another might increase.

The potential for so much variety here makes it difficult to compare studies and meta-analyse outcome data. There is little consideration of service user-reported outcome measures, and the development of such measures would be a helpful addition to the increasing body of research in this area.

Behaviour change techniques showing the most promise

When interventions were examined by target, type of study or reported findings, it was apparent that there was a small group of BCTs that were most frequently found across all the interventions, such as ‘instruction on how to perform the behaviour’, ‘restructuring the social environment’, ‘problem-solving’ and ‘action-planning’.

A complication of this finding was that the BCT ‘restructuring the social environment’ was a broad category that, for the purposes of this study, encompassed several subcategories. There were some variations that might tentatively indicate that some BCTs were more effective in reducing restrictive practices than others. The RCTs appeared to make more use of ‘prompts or cues’ and ‘information about antecedents’ than those studies that reported no positive findings. Caution in interpreting this finding is required owing to the small number of RCTs and, indeed, the small number of studies reporting no effects.

Because of the generally weak evidence base around interventions evaluation, including the small number of RCTs, this review cannot confidently recommend the testing of specific BCTs that show promise. Instead, there appears to be a strong argument for intervention developers to consider the specific components of interventions, and why and how these individual components might be tested.

Undetectable behaviour change techniques

Fifty BCTs were either not found at all, or found only rarely. Of these, some will not be applicable, some might have been presented in practice but were undetectable from available reports, and others might never have been considered and yet could be effective. Videos of training sessions and observational work would help elaborate on this and could be considered in future studies.

As BCTs using reward and threat were rarely detected, it may be valuable to explore the potential of this area to have an impact on the reduction of restrictive practices.

Pro re nata medication

That so few studies have focused on chemical restraint/rapid tranquillisation is concerning. There is widespread and frequent use of chemical restraint (rapid tranquillisation) in acute mental health wards. Given that most medicines given at this time are either benzodiazepines or antipsychotics, further research in this area is warranted. It is also clear that the process of offering and/or refusing PRN medication can lead to an escalation in the use of restrictive practices, and additional understanding of this process may lead to a subsequent reduction in restrictive practices.

Databases searched for unpublished studies

• ClinicalTrials.gov (US National Institutes of Health).

• Dissertations & Theses A&I (ProQuest): 1743 – present.

• Evidence Search (NICE).

• International Clinical Trials Registry Platform (World Health Organization).

• OpenGrey (INIST-CNRS).

Mental health organisations

1. European Violence in Psychiatry Group (EViPRG) (http://eviprg.eu/).

2. Mental Welfare Commission for Scotland (www.mwcscot.org.uk/).

3. National Mental Health Consumer and Carer Forum (https://nmhccf.org.au/).

4. CPI (www.crisisprevention.com/).

5. Safe Crisis management (https://safecrisismanagement.com).

6. Healthcare Improvement Scotland (http://ihub.scot/).

Health-care organisations

Societies/colleges

Charities

1. Rethink (www.rethink.org/).

2. MIND (www.mind.org.uk/).

3. Mental Health Foundation (www.mentalhealth.org.uk/statistics).

4. National Alliance on Mental Illness (www.nami.org/).

5. Mental Health Australia (https://mhaustralia.org).

National government health departments

1. Department of Health and Social Care (UK) (www.gov.uk).

2. Australian Government. Department of Health (https://agedcare.health.gov.au).

3. US Department of Health and Human Services (HHS) (www.hhs.gov/).

4. Ministry of Health. New Zealand (www.health.govt.nz/).

Health-care quality agencies

1. NICE Quality Standards (www.nice.org.uk/standards-and-indicators).

2. CQC (UK) (www.cqc.org.uk/).

3. NHS Improvement (https://improvement.nhs.uk/).

4. AHRQ (www.ahrq.gov/).

5. Australian Commission on Safety and Quality in Health Care (www.safetyandquality.gov.au/).

6. Health Quality Ontario (www.hqontario.ca/).

7. Health Quality & Safety Commission New Zealand (www.hqsc.govt.nz/).

Training providers

1. Skills for Health (UK) (www.skillsforhealth.org.uk/).

2. CPI (www.crisisprevention.com/).

3. Restraint Reduction Network (http://restraintreductionnetwork.org/).

4. Dynamic intervention (https://dynamicinterventions.org).

5. Challenging Behaviour Foundation (UK) (www.challengingbehaviour.org.uk/).

6. Royal College of Nursing (www.rcn.org.uk/).

7. CALM (www.calmtraining.co.uk/).

8. BILD (www.bild.org.uk/).

9. Caring Solutions (www.caringsolutionsukltd.com/).

10. Centre for Mental Health (www.centreformentalhealth.org.uk/).

11. Scottish Association for Mental Health (SAMH) (www.samh.org.uk/).

12. Support in Mind Scotland (www.supportinmindscotland.org.uk/).

Assumptions

• We are applying the BCT taxonomy to interventions that aim to change staff behaviour in terms of how staff interact with patients.

• The intervention target is interaction between staff and patients: ‘The final target of most interventions was to organically change interactions between patients and between staff and patients’. 16

• Objects added to make environment more comfortable, etc., are for patients but aimed at increasing quality of interaction and reducing conflict and aggression (which leads to restrictive practices).

• ‘Self’ can include individual and groups of staff as opposed to management.

• Knowing that data are being collected is not the same as receiving feedback.