Multicomponent organisational programmes reduce seclusion and restraint in acute psychiatric inpatient care, but the supporting designs are weak

Introduction

Coercive containment remains a routine feature of acute psychiatric inpatient care. A meta-analysis cited within a Canadian pilot study estimated that, on average, 14.4% of inpatients in mental health settings experience physical or mechanical restraint, 15.8% experience seclusion and 25.6% experience chemical restraint.¹ Harms are well documented rather than speculative: a systematic review of coercive interventions in adult psychiatry, summarised in a review of child and adolescent practice, estimated the incidence of post-traumatic stress disorder after seclusion or restraint at between 25% and 47%.² Cochrane reviews cited within a realist review of youth settings concluded that no controlled studies evaluate the value of seclusion or restraint in serious mental illness, which leaves the practices without a demonstrated therapeutic rationale.⁶

Regulatory frameworks have narrowed the permissible grounds for these practices. The Project BETA consensus statement summarised the Centers for Medicare and Medicaid Services conditions of participation, under which seclusion is defined as involuntary confinement in a room a patient is physically prevented from leaving, and a drug counts as a restraint when it is used to manage behaviour rather than as standard treatment for the patient's condition.³ Australian national safety priorities have gone further, framing reduction and, where possible, elimination as an explicit policy goal, although a nursing commentary from that period noted that many clinicians regarded elimination in acute care as unrealistic.⁴

This review considers what interventions have been evaluated against measured rates of seclusion and restraint in adult acute inpatient settings, what the underlying designs permit, and which contextual factors appear to govern transfer between wards. The direct adult evidence is dominated by uncontrolled before-and-after evaluations, with one appraisal of comparative studies providing the most rigorous synthesis.⁵ Several of the most quantitatively detailed reduction results come from child, adolescent and residential settings; these are treated here as extrapolation and labelled as such, since no controlled adult replication of those specific programmes was identified among the sources reviewed.^2,6

Multicomponent organisational programmes

The strongest synthesis of adult acute settings sets a cautious ceiling on what can be claimed. Gaynes and colleagues identified 17 comparative studies of violence prevention and de-escalation strategies, and rated only two approaches — structured risk assessment and multimodal interventions consistent with Six Core Strategies principles — as supported at low strength of evidence for reducing seclusion and restraint.⁵ Every other intervention examined, whether preventive or de-escalatory, fell into the insufficient category, and all but one study carried a medium or high risk of bias.⁵

Larger observational reductions have nonetheless been reported over long implementation periods. The Project BETA consensus statement summarised an 11-year programme across nine Pennsylvania state hospitals in which the seclusion rate fell from 4.2 to 0.3 episodes per 1,000 patient-days and mean seclusion duration from 10.8 to 1.3 hours, while the restraint rate fell from 3.5 to 1.2 episodes per 1,000 patient-days.³ The same source reported that data on staff injury indicated no increased risk of assault, and described a separate five-year hospital initiative in which regression analysis attributed most of a 75% reduction in incidents to changes in how critical cases were identified and reviewed.³ Results across these series are not uniform: the statement also described a retrospective analysis at a large inner-city hospital where restraint use fell significantly but assaults on patients and staff rose, which the authors attributed in part to staff having received no specific training in managing violent patients.³

Whether reduction raises risk to staff has not been settled by controlled work. An Australian nursing commentary observed that no controlled studies had examined the effect of reducing seclusion and restraint on occupational injury, while anecdotal accounts from the United States indicated no increase in assaults and possibly fewer.⁴ The same commentary noted that Pennsylvania state hospitals had virtually eliminated both practices and that systematic use of the six core interventions had produced marked reductions across varied settings, though those claims rested on programme reports rather than trials.⁴

Evidence from younger populations is more granular but does not transfer directly to adult acute wards. In a review of 18 studies of child and adolescent inpatient units, a Six Core Strategies implementation across three United States children's facilities reported statistically significant falls in combined seclusion and restraint of 62%, 86% and 69% per 1,000 client days, and a decade-long quality improvement project at a paediatric psychiatric hospital reported mechanical restraints falling from 485 to zero and physical restraints from 3,033 to 379.² A feasibility study in an Australian acute adolescent ward, using archival mixed-methods data, reported elimination of seclusion together with significant reductions in restraint use and nursing sick leave in the 12 months after implementation of the same framework.⁷ Both remain uncontrolled pre-post designs, and the adolescent findings are extrapolation when applied to adults.

Implementation conditions and fidelity

Programme effects appear to depend heavily on how implementation is conducted. An integrative review of 19 articles on Safewards in adult inpatient and forensic units, appraised with the Mixed Methods Appraisal Tool and a narrative instrument for discussion papers, found that the model can reduce containment and conflict, but that this outcome varied across the literature and that fidelity measures were themselves limited.⁸ Safewards comprises six originating domains of conflict — staff team, patient community, patient characteristics, life outside hospital, regulatory framework and physical environment — and ten discrete staff interventions, and the original cluster randomised trial indicated that the model could reduce restrictive practices, which prompted wide international uptake.⁸ The review also identified staff involvement in implementation as a recurring determinant of success, and noted that only two of the included papers examined the model from the consumer point of view.⁸

Trauma-informed care shows a comparable dependence on organisational preconditions. Bryson and colleagues screened 693 abstracts published between 2000 and 2015 and retained 13 articles describing whole-system implementations in youth psychiatric or residential settings, covering the Attachment, Self-Regulation and Competency framework, Six Core Strategies, Collaborative Problem Solving, the Sanctuary Model, Risking Connection and the Fairy Tale Model.⁶ Five factors were associated with successful implementation: senior leadership commitment, sufficient staff support, amplifying the voices of patients and families, aligning policy and programming with trauma-informed principles, and using data to motivate change.⁶ The authors concluded that coercive measures may be reduced either by targeting them directly or by implementing broader therapeutic models, and that the relative efficacy of the two routes remains untested; the population is youth, so inference to adult acute wards is extrapolation.⁶

Where those preconditions are absent, reduction appears to stall. Nineteen studies from 11 Eastern Mediterranean countries, synthesised narratively under PRISMA 2020 guidance, reported nursing knowledge deficits across multiple settings, predominantly negative attitudes towards restraint, and recurrent problems of inadequate training, chronic understaffing and limited access to alternatives.⁹ Individuals with mental illness in those studies consistently described restraint as humiliating, punitive and physically distressing.⁹ The included studies were predominantly cross-sectional with convenience samples, so the review supports description of the barriers rather than estimation of their effect.⁹

Training, dedicated roles and post-incident practice

De-escalation training is mandated widely but has been evaluated poorly. The EDITION programme co-designed an enhanced de-escalation intervention and tested it in an uncontrolled feasibility trial across ten inpatient wards in two United Kingdom mental health trusts, spanning acute, psychiatric intensive care and low, medium and high secure forensic settings.¹⁰ Conflict and containment were measured with the Patient-Staff Conflict Checklist over 24 weeks divided into pre-intervention, embedding and post-intervention phases; regression analyses indicated that reductions in both conflict and containment were predicted by study phase, and no adverse events related to the intervention occurred.¹⁰ The uncontrolled design and self-selecting sample mean this constitutes a feasibility signal rather than an effect estimate.

One of the few controlled evaluations tested a dedicated staff role rather than a training package. A bi-phasic time series study compared a psychiatric intensive care unit that created a reducing restrictive interventions advocate — a post that met patients and carers and advised, trained, supervised and debriefed the multidisciplinary team — with a control unit, across 19 months before and 19 months after implementation.¹¹ The intervention unit showed significant reductions in seclusion, full restraint and standing holds, and interviews with the postholder, ward manager and organisational lead described safety culture changes considered unlikely without the role.¹¹

Post-incident debriefing, a component of most multicomponent programmes, has almost no standalone evidence. A Cochrane review located a single randomised controlled trial of debriefing after coercive measures in adults with schizophrenia or schizophrenia-type psychosis, in which 422 people were randomised and 109 participated.¹² The trial found a point estimate favouring more rather than fewer repeat seclusions, with a wide confidence interval spanning no effect, and no evidence of benefit for peritraumatic distress or satisfaction with care; all outcomes were graded very low certainty.¹² This apparent divergence from the packaged programmes is best read as a difference in what was tested — one isolated component against treatment as usual, rather than an organisational change strategy.

Patient participation and perceived coercion

Interventions that formalise patient preferences have been evaluated mainly on process and climate outcomes. A pilot mixed-methods study of the Joint Crisis Plan, conducted across a psychiatric emergency department, a psychiatric intensive care unit, an acute care unit and a forensic unit in Quebec, combined 16 interviews with a pre-post comparison involving 57 healthcare providers and 53 inpatients.¹ Implementation difficulties complicated evaluation of effects on seclusion and restraint use, but a moderate change in violence prevention climate, of effect size 0.40, was detected from the provider perspective.¹ Qualitatively, inpatients valued the reflective process of completing the plan while providers focused on its technical aspects.¹

Appetite for the tool preceded any demonstration of its performance. In a pre-implementation survey at the same Quebec institution, 84.8% of individuals with severe mental illness and healthcare providers judged that a joint crisis plan would meet a critical need in the province, while identifying lack of organisational support, insufficient training and perceived complexity as likely barriers.¹ Training in the pilot combined a 90-minute online module on the concept of crisis and shared decision-making with a 90-minute in-person session using clinical scenarios, followed by a one-month trial period during which the research team attended the wards several times weekly.¹

Clinician beliefs remain a substantial constraint on such approaches. In an online survey of 512 nurses across Australian psychiatric inpatient and emergency settings, respondents generally regarded seclusion and restraint as necessary last-resort methods and tended to disagree that containment could be eliminated, while accepting that its use damages relationships with consumers.¹³ Seclusion was viewed significantly more favourably than mechanical restraint, and empathy, rapport and trauma-informed principles were the factors respondents most often identified as reducing the likelihood of containment.¹³

Doedens and colleagues, reviewing the nursing literature systematically, traced a shift over two decades from a therapeutic paradigm, in which coercive measures were held to benefit patients, to a safety paradigm in which they are viewed as undesirable but necessary for ward safety.¹⁴ That review found the literature relating staff characteristics to actual use of coercion to be inconclusive, reported that nurses' perception of how intrusive a measure is appears influenced by how often they use it, and identified nurses' own perceived safety as the more promising target for further study.¹⁴

Variation in use, ward context and cost

Where restraint occurs, and whether it escalates, appear to be partly separable problems. A retrospective observational study of 41,685 incidents recorded across a large English National Health Service trust between June 2024 and May 2025 found restraint prevalence highest in older adult wards (12.93%) and low secure units (12.86%), and higher on mixed (11.55%) and female (9.47%) wards than male wards (6.73%).¹⁵ Escalation from restraint to seclusion was highest in medium secure units (6.22%) and acute wards (4.72%) and minimal in high secure and older adult wards, and multilevel modelling suggested that ward-level context contributed to variation in seclusion outcomes.¹⁵ Counterfactual simulation in the same study suggested that, in high-risk settings, interrupting escalation may lower seclusion rates more than an equivalent reduction in restraint incidents.

Staffing recurs as the resource on which relational alternatives depend. Thematic analysis of semi-structured interviews with 130 staff and 32 service users during a restraint minimisation initiative generated five themes, including insufficient staff to do the job, detriment to staff and service users, the burden of non-clinical paperwork, false economies, and the observation that alternative interventions such as de-escalation could not be delivered at prevailing staffing levels.¹⁶ The finding is qualitative and drawn from a single initiative, so it identifies a mechanism of failure rather than quantifying it.

Economic evidence is thin relative to the sums involved. The EDITION report noted that physical restraint costs United Kingdom mental health services £6.1 million and enhanced observations £88 million per annum.¹⁰ Luigi and colleagues appraised 20 economic studies of non-pharmacological interventions targeting aggression and restrictive practices, identified through searches covering 2000 to 2025 and mostly conducted on acute wards: eleven reported intervention costs only, twelve reported savings, of which eight permitted cost analyses, and just two interventions yielded both clinical benefits and net savings.¹⁷ Reporting quality was weak, with few sensitivity analyses to model uncertainty, heterogeneity or distributional effects.¹⁷

Conclusions and outlook

What is reasonably established is narrow. Multicomponent organisational programmes that combine leadership commitment, routine use of incident data, workforce development, prevention tools, patient and family involvement and structured debriefing are consistently accompanied by lower rates of seclusion and restraint across long observational series, and structured risk assessment carries comparable support; both, however, rest on low strength of evidence in the only systematic appraisal of adult acute settings.^3,5 Reduction has not been shown to increase assaults on staff in the series that reported injury data.³

What remains open is whether these associations are causal, and which components carry the effect. Isolated components perform poorly when tested alone, as the single randomised trial of post-incident debriefing illustrates.¹² Controlled evaluation is scarce: the strongest adult design among the sources reviewed compared one intervention unit against one control unit over matched 19-month periods,¹¹ while the largest United Kingdom training evaluation was explicitly a feasibility trial without a control arm.¹⁰ Ward-level variation in escalation and staffing may account for a meaningful share of the differences between wards that are otherwise attributed to programmes.^15,16

The evidence that would settle these questions is identifiable. Definitive multi-site randomised or cluster-randomised trials with prospectively defined containment outcomes and fidelity measurement, of the kind the EDITION feasibility work was designed to enable, would separate programme effect from secular trend.¹⁰ Future evaluations would benefit from reporting cost alongside clinical outcomes, given how few existing economic studies permit any cost-benefit inference,¹⁷ and from incorporating consumer-reported outcomes, which remain the largest gap in the Safewards literature.⁸