Home > Methods and findings - Violence

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

National Collaborating Centre for Nursing and Supportive Care (UK). Violence: The Short-Term Management of Disturbed/Violent Behaviour in In-Patient Psychiatric Settings and Emergency Departments. London: Royal College of Nursing (UK); 2005 Feb. (NICE Clinical Guidelines, No. 25.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

7Methods and findings

7.1. Summary of development process

The methods used to develop this guideline are based on those outlined by Eccles and Mason (2001) and in the NICE technical manual (www.nice.org.uk).

The following sources of the evidence were used to inform the guideline:

  • Cochrane reviews: Salias and Fenton (2002), Carpenter and Berk (2002), Cure and Carpenter (2002), Aleman and Kahn (2001)
  • The Royal College of Psychiatrists' Clinical guidelines on the prevention and management of imminent violence (see agree tool in Appendix 11 and summary of St George's report in Appendix 12)
  • Other recent guidelines and reports (see Section 14)
  • Reviews of assessment processes, tools, tests and instruments for identifying those at risk (NCC-NSC)
  • Reviews of the interventions currently used in the UK for the short-term management of disturbed/violent behaviour in adult psychiatric in-patient settings (NCC-NSC)
  • Reviews of studies examining patients' views and experiences of the short-term management of disturbed/violent behaviour in adult psychiatric inpatient settings (NCC-NSC)
  • Reviews of the evidence on costs and economic evaluations (ScHARR)
  • Reviews on other topics and areas related to the short-term management of disturbed/violent behaviour in adult psychiatric in-patient settings in the UK (NCC-NSC)
  • The minority ethnic groups section was further informed by Breaking the circles of fear, Sainsbury's Centre for Mental Health (2002) – see Appendix 13 for appraisal with agree tool – and focus groups, run through black service user organisations and with relevant health care professionals, organised by the NCC-NSC (see Appendix 14).

The stages used to develop this guideline were as follows:

  • Develop scope of guideline
  • Convene multidisciplinary GDG
  • Review submission of evidence from stakeholders
  • Develop review questions
  • Identify sources of evidence
  • Retrieve potential evidence
  • Evaluate potential evidence
  • Undertake systematic reviews of the evidence
  • Extract relevant data from studies meeting methodological and clinical criteria
  • Interpret each paper, taking into account the results including, where reported, the beneficial and adverse effects of the interventions; costs and acceptability to service users; level of evidence; quality of studies; size; precision of effect and relevance; and generalisability of included studies to the scope of the guideline
  • Prepare evidence reviews and tables which summarise and grade the body of evidence
  • Draw up evidence statements
  • Formulate conclusions about the body of available evidence, based on the evidence reviews by taking into account the factors above
  • Trawl any recent and relevant guidance literature in areas where evidence is weak or lacking and present this to the GDG for comment
  • Send out evidence reviews for peer review
  • Agree final recommendations by formal consensus voting and apply recommendation gradings
  • Submit first drafts (short version and full versions) of guidelines for feedback from NICE registered stakeholders
  • Consideration by GDG of stakeholders comments following first stage consultation
  • Submit final drafts of all guideline versions (including Information for the public version and quick reference guide with algorithms) to NICE for second stage of consultation. The Guideline Review Panel (GRP) also comment at this stage
  • Consideration by GDG of stakeholders and GRP's comments
  • Final copy of the short form version is submitted to NICE for sign off, prior to publication of all edited versions.

7.2. Key clinical questions

The GDG identified the key clinical questions that were raised by the scope. Each of these questions related to an intervention or area addressed in the scope.

An algorithm detailing how the scope was translated into clinical questions can be found in Appendix 3.

Each of the clinical questions is outlined in the relevant methods section of this guideline.

7.3. Review methods

7.3.1. Search strategies

Search strategies were devised to identify the best available evidence for the interventions and related topics discussed in the guideline (see Appendix 4). It was recognised very early within the process that, in most instances, this evidence would not constitute meta-analyses, systematic reviews or randomised controlled trials (RCTs). Therefore searches were not limited to these study designs.

Where little evidence was available, studies were included in related areas, from which evidence could be extrapolated.

Searches were not limited to English language citations. Relevant European foreign language papers were translated. Unpublished and published papers were included.

The search strategies were structured as follows:

  • an overarching strategy for interventions – covering environment, prediction, de-escalation, observation, physical interventions, seclusion and rapid tranquillisation, along with service user and staff perspectives on these interventions – across a wide range of databases
  • a search of additional databases to identify guidance and reports not indexed in databases searched
  • a topic specific search strategy on major databases – see Appendix 4 for more details.

Hand searching was not undertaken, following NICE advice that exhaustive searching on every guideline review topic is not practical or efficient (Mason 2002).

Reference lists of relevant order papers were checked for articles of potential relevance.

Each evidence review was sent for peer review, prior to the first consultation phase, in an attempt to identify any further relevant papers. GDG members were invited to nominate any relevant research that may have been missed.

The databases searched, logs of results and all search strategies can be found in Appendix 4. Unless otherwise stated, all searches were run from 1985-2002/3. Searches began from this date as this guideline updates the RCPsych guideline, The management of imminent violence (1998), which was due for review. GDG members were asked throughout out the guideline development process whether any further relevant research had been identified, post search, that might impact on the recommendations.

For each intervention and related topic evidence of effectiveness, evidence of harm and cost effectiveness information was sought.

7.3.2. Sifting and reviewing the evidence

Once articles were retrieved, the following sifting process took place:

  • 1st sift: sift for material that potentially meets eligibility criteria on basis of title/abstract by two reviewers
  • 2nd sift: full papers ordered that appear relevant and eligible or where relevance/eligibility not clear from abstract
  • 3rd sift: full articles critically appraised and checked by one reviewer. More than 50 per cent of all articles in the guideline were then critically appraised by an independent reviewer as a quality check.

7.3.3. Data extraction

Study appraisal and methodological quality were assessed using checklists designed with assistance from the Centre for Statistics in Medicine at Oxford University. (Quality principles can be found in Appendix 10.) Data was abstracted by a single reviewer and evidence tables compiled. More than 50 per cent of all articles were then subject to a second quality assessment by a second reviewer. Any discrepancies between reviewers were resolved by discussion. Where needed, a third reviewer assisted with decisions on the inclusion or exclusion of a study.

The following were extracted where possible (the reporting of many studies sometimes lacked essential detail) and relevant:

Author, setting, number of participants at baseline and follow-up, methods and details of baseline and outcome measures, results including summary statistics and 95 per cent confidence intervals, and comments made on methodological quality.

Masked assessment, whereby data extractors are blind to the details of the journal, authors etc., was not undertaken because there is no evidence to support the claim that this minimises bias (Cullum et al. 2003).

7.3.4. Data synthesis

All studies were put into evidence tables and summarised using a qualitative narrative approach. No quantitative analysis was carried out for this review. Summary statistics of significance were reported in the evidence tables.

7.3.5. Appraisal of methodological quality

Very limited evidence for each of the review questions listed below was found. Therefore the resulting evidence reviews must be viewed as mapping exercises that aimed to highlight the range of research undertaken (which was often of mixed quality), in order to facilitate informed discussion by the GDG, to assist with deliberations around recommendation formulation and also to identify research gaps. Where a study was particularly weak it was excluded (see Appendix 6). It was considered particularly weak where the number of confounders and flaws were great enough to jeopardise the results. Concerns regarding the quality of individual studies are detailed in the relevant evidence table.

A large range of quality related concerns were commonly found across many of the studies included in these review. These included:

  • inappropriately small sample sizes
  • inter-rater reliability not always quantified where applicable
  • conclusions do not always appear to be supported by a study's results
  • methodologies are not always sound (that is, don't adhere to standard processes)
  • designs do not always appear appropriate - sometimes this is recognised by the authors
  • methods of analysis are not always clearly outlined
  • under-reporting
  • lack of detail about follow-up duration, losses to follow-up and drop-out rates
  • descriptions of interventions are not always adequate
  • description of how outcomes were measured are not always adequate or are sometimes lacking
  • poor reporting.

Where the studies in a review raise other, more specific, quality concerns, these are mentioned under the evidence summary for each review.

Authors were not contacted about any of the included studies, due to time constraints and the age of many of the studies.

In areas without sufficient evidence, previous guideline material was collated to help facilitate informed discussion by the GDG.

Clinicians and service users were also invited to give presentations on areas without sufficient evidence at guideline development group meetings to facilitate discussion. They acted as experts in these capacities. They sat within the group and entered fully into discussion. However, they were not GDG members and did not have voting rights, nor were they involved in drawing up the final wording of the recommendations.

The GDG then considered the evidence statements derived from the evidence reviews and used formal consensus methods (see Section 7.7) to derive recommendations and good practice points, particularly for those areas where research evidence was lacking or weak. They drew upon their own and others' clinical expertise and experience as necessary.

7.4. Evidence grading

Once individual papers had been assessed for methodological quality and relevance in terms of the clinical questions, they were graded according to the levels of evidence currently used by NICE.

Classification of evidence

Level of evidenceType of evidence
1++High quality meta-analyses, systematic review of RCTs, or RCTs with a very low risk of bias.
1 +Well conducted meta-analyses, systematic reviews of RCTs, or RCTs with a low risk of bias.
1-Meta-analyses, systematic review of RCTs, or RCTs with a high risk of bias*.
2++High quality systematic reviews of case-control or cohort studies. High quality case-control or cohort studies with a very low risk of confounding, bias or chance and a high probability that the relationship is causal.
2+Well-conducted case-control or cohort studies with a low risk of confounding, bias or chance and a moderate probability that the relationship is causal.
2-Case-control or cohort studies with a high risk of confounding, bias or chance and a significant risk that the relationship is not causal.*
3Non-analytic studies (for example, case reports, case series).
4Expert opinion, formal consensus.
*

Studies with a level of evidence ‘-’ should not be used as a basis for making a recommendation

The available evidence for each intervention and related topic was compiled into individual evidence reviews, including health economics information. A summary of all recent reports and guidelines on the topic was also compiled. All this information was then presented to the GDG. The methods and findings from each of these reviews are outlined in Section 7.8.

7.5. Grading recommendations

The grading of recommendations involves a process of assessment in which the available evidence is interpreted in relation to the clinical questions asked. Where evidence is lacking or is not directly related to every area covered by the clinical question, the recommendation will demand some degree of consensus. For example, it is possible to have sound methodological evidence in an area that is not particularly relevant to the target audience of the guideline. When applied to the target audience, this would therefore result in a lower grade of recommendation than the evidence might initially seem to suggest, since inferences would have to be made from the available evidence that are beyond the empirical data. This will be the case where the evidence only partially covers the clinical question that the guideline sets out to answer. Where no, or insufficient evidence is available, recommendations have to be arrived at using formal consensus methods alone.

In this guideline, D grade recommendations are differentiated from good practice points (GPP), which also have little or no evidence. Both carry a D grade status, but unlike D grade recommendations, GPPs are principles of practice.

The recommendations for this guideline were graded A to D, using the current NICE approach.

AAt least one meta-analysis, systematic review or RCT rated as 1++, and directly applicable to the target population, or
A systematic review of RCTs or a body of evidence consisting principally of studies rated as 1+, directly applicable to the target population and demonstrating overall consistency of results. Evidence drawn from a NICE technology appraisal.
BA body of evidence including studies rated as 2++, directly applicable to the target population and demonstrating overall consistency of results, or
Extrapolated evidence from studies rated as l++or 1+.
CA body of evidence including studies rated as 2+, directly applicable to the target population and demonstrating overall consistency of results, or
Extrapolated evidence from studies rated as 2++.
DEvidence level 3 or 4, or
Formal consensus.
D (GPP)A good practice point (GPP) is a recommendation for best practice based on the experience of the GDG.

The schizophrenia guideline used an older grading system that had only three grades A-C. Some recommendations that carry a C grade in the schizophrenia guideline carry a D grade in the current guideline. However, there is no difference in evidence level.

In the current guideline, good practice points, as well as D grade recommendations were arrived at using a formal consensus method.

7.6. Cost effectiveness review and analysis

7.6.1. Identification of papers

Searches were undertaken by ScHARR, alongside the clinical literature reviews, to identify relevant cost effectiveness, cost utility and cost-benefit analyses. Details of the databases searched and the search strategies can be found in Appendix 9. Titles and abstracts were sifted and relevant papers ordered by one reviewer.

7.6.2. Reviewing the evidence

Eligible papers were assessed by one reviewer using the Drummond checklist (Drummond et al. 1996). Evidence tables of the included studies were also produced by one reviewer.

7.6.3. Estimation of cost effectiveness

The scope of the guideline is broad and includes the assessment of risk, as well as the short-term management of disturbed/violent behaviour across the whole of range of adult in-patient settings. Little if any economic evidence was found for most areas of the guideline. Limited primary economic analysis was undertaken in relation to immediate life support (ILS) training. However, in many areas the evidence base was too weak to allow even limited primary economic analysis. See full details in Appendix 9.

7.7. Consensus process

Due to a dearth of good quality evidence, many of the recommendations in this guideline were arrived at solely, or in large part, by means of formal consensus methods. Three consensus meetings were held in March 2004.

A modified nominal group technique was used to finalise the recommendations and good practice points. An external facilitator was used to chair the meeting. The consensus process was facilitated by computerised voting consoles, which assured anonymity and allowed percentages to be quickly calculated. It also allowed the GDG to view the range of responses in the form of a graph immediately voting had occurred. Consensus was set at 80 per cent, unless a significant group within the GDG all voted against a recommendation. For example, if all the psychiatrists voted against a recommendation, even though 80 per cent agreement was achieved overall, it was considered that consensus had not been reached.

Prior to voting on each recommendation and good practice point, a discussion took place and modifications were made as necessary. The wording was re-typed if necessary and then displayed on a screen so that GDG members could see the recommendation or good practice point they were voting on. If consensus was achieved, the GDG moved on to discuss the next recommendation or good practice point. However, if consensus was not achieved, the recommendation or good practice points was discussed a second time, modifications made to reflect the concerns of the GDG and another vote took place. After debate on some areas, consensus was achieved for all recommendations.

7.8. Methods for individual evidence reviews

7.8.1. Introduction

This guideline is divided into a number of interventions and related topics. For most of these areas separate literature searches were undertaken (see Appendix 4). The number of papers found, included and excluded and the details of the resulting evidence base are discussed separately for each area. Section 7.3 details the reviewing process that was common to all areas of the guideline. Where there were deviations from this process, this is highlighted in the following relevant sections.

7.8.2. Prevention

7.8.2.1. Environment

7.8.2.1.1. Objectives

The original RCPsych evidence base on environment was examined. The following hypothesis was used to inform search strategies:

RCPsych hypothesis
  • Characteristics of the human and physical environment have powerful effects in mitigating and preventing, or exacerbating and precipitating the manifestation of violence.

After sifting and quality appraisal 17 papers were included by the RCPsych reviewer. However, the evidence base was too weak to offer support for this hypothesis.

Current guideline (update of RCPsych guideline)

Three review questions were identified by the GDG and used to inform all searches (see Appendix 4 for search strategies, databases searched and search logs). Unlike the RCPsych review, this review did not consider staff characteristics associated with an increased rate of disturbed/violent behaviour. Instead these were considered in the prediction review.

Review questions

The studies had to meet the following inclusion criteria:

7.8.2.1.2. Selection criteria
Types of studies

Systematic reviews through to before and after designs. Qualitative studies were also included (evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
  • Measurement of environmental factors that may impact on the short-term management of disturbed/violent behaviour.
  • Service users' and clinicians' views.
7.8.2.1.3. Clinical evidence

Seventy-five articles were identified, after combining the results of the main ‘intervention’ searches and the specific search for the ‘physical environment’. Thirty-two articles were ordered. Nine of the ordered articles were not directly related to the research questions and were therefore not included for the present review. Seven of the remaining 23 articles were included in the evidence review. In addition, seven papers from the RCPsych review were considered relevant and were included in this review. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Appraisal of methodological quality

In addition to the quality concerns mentioned in Section 7.3.5, the studies raised the following methodological concerns:

  • All of the studies are largely descriptive in content and none had a controlled comparison group.
  • The variation between studies in regard to the methods of data collection, outcome variables and statistical analysis make it impossible to aggregate the results.
  • Overall, the methodological quality, execution and reporting of the included studies are poor.
Included studies

Three of the studies are uncontrolled before-and-after designs with one an interrupted time-series design, one a correlation study and two qualitative studies.

Three of the studies (Rauter et al. 1997; Haller et al. 1996; Velasco et al. 1996) were concerned with the effects of a cigarette-smoking ban in in-patient psychiatric settings. Two of the studies (Haller et al. 1996; Velasco et al. 1996) concluded that smoking bans on locked in-patient wards do not have an effect on increased violence by service users. Haller et al. (1996) indicated that although staff were in favour of the intervention, service users held decidedly negative feelings towards the smoking ban. There was no increase in aggressive behaviour in the outcome measures used in the study. The prospective study by Velasco et al. (1996) found the number of verbal assaults increased after implementing the ban, as did the consumption of nicotine replacement products by the service users. The authors suggest that dangerous behaviour did not follow the implementation of the smoking ban. The methodological quality of the study by Rauter et al. (1997) is insufficient to draw any reliable conclusions regarding smoking bans in psychiatric inpatient facilities.

An interrupted time-series design was used to evaluate the effects of removing the Music Television (MTV) channel from the television of a maximum-security facility (Waite et al. 1992). The analysis indicated that there was a significant reduction in aggressive behaviour following the removal of MTV from the television. Although the study is well designed and executed, the relevance of these findings to the UK context is uncertain (for example, the number of psychiatric settings in the UK where a music television station is received is unknown).

Mistral et al. (2002) conducted a qualitative study in a intensive psychiatric care environment to evaluate the effect a £70,000 ward refit, training on risk assessment and control and restraint techniques, and clarity on rules and sanctions on staff illness, staff turnover, patient aggression and the rate of seclusion. Although results were not significant, there was a positive trend for all outcomes.

Johnson et al. (1997) conducted a qualitative study to explore service users and experiences prior to an aggressive incident. A phenomenological approach was used. Five key themes emerged: lack of space; relationships; restrictions on privileges; lack of power versus feelings of powerfulness during aggressive incidents; and ineffective self-empowerment strategies.

The study by Nijman (1999) is concerned with crowding in psychiatric in-patient units and aggressive behaviour. However, the study is poorly designed and the statistical methodology is flawed. The study is essentially a correlation study. The RCPsych review also included three studies that examined the issue of crowding (Palmstierna and Wistedt 1995; Lanza et al. 1994; Palmstierna et al. 1991). These studies suggested that crowding increased the rate of violent incidents.

A further study in the RCPsych review – Hunter and Love (1996) – used an uncontrolled before and after study to evaluate the effectiveness of procedural changes at mealtime on the number of violent incidents at mealtimes. A number of suggestions were implemented: plastic utensils were substituted for silverware; music selected by the hospital music therapists was played; the dinning room, gym and courtyard were left open after meals for service users with special privileges; and food service workers were trained in therapeutic communication. This study showed a significant (40 per cent) reduction in violent incidents.

Two further studies in the RCPsych review considered staff roles. A participant observer study by Katz and Kirkland (1990) suggested that good leadership, structured staff roles and predictable routines are associated with less violence on wards. While a retrospective cohort study (James et al. 1990) suggested that high staff turnover and extensive use of agency staff was associated with an increase in violent incidents.

Another descriptive study included in the RCPsych review found that video cameras detected more, but milder, episodes of violence than nurses (Crowner et al. 1994).

None of the above studies significantly changed the findings of the RCPsych review.

Evidence statements

Level of evidenceEvidence statement
Level 4The evidence suggests that environmental factors – such as crowding, banning smoking, high staff turnover and limit setting –affect on the incidence of disturbed/violent incidents. However, further research is needed to identify additional environmental factors.
Level 4The evidence suggests that both staff and service users believe that environmental factors – such as banning smoking, limit setting, medication, seclusion, physical interventions and communication – affect the incidence of disturbed/violent incidents. Further research is needed to identify additional factors.
7.8.2.1.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.2.2. Alarm systems

7.8.2.2.1. Objectives

No specific searches on alarm systems were undertaken in the RCPsych guideline. Any papers would have been included under the environment review.

In the current guideline, three review questions were identified by the GDG and used to inform all searches (see Appendix 4 for search strategy, databases searched and search log).

Review questions
7.8.2.2.2. Selection criteria
Types of studies

Systematic reviews through to before and after designs. Qualitative studies were also included (evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
  • Any measures of change to management of short-term disturbed/violent behaviour or rates of disturbed/violent episodes as a result of alarms.
  • Service users' and clinicians' views of alarms.
7.8.2.2.3. Clinical evidence

Eighty-one studies were identified in the initial sift. These were then subjected to two further sifts by two reviewers. After sifting for relevance and duplicates, 70 full papers were ordered. However, most were opinion pieces, anecdotal reports or fell outside the inclusion criteria for this review.

Seven papers were primary research studies. However, after critical appraisal and quality assessment only one of these contained information relevant to the research questions. References were checked for missing articles but no further studies were identified. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6.)

Included studies
  • The only included study did not address the question of effectiveness of alarms as measured by changes in incidence rates or impact on management and could have been excluded on this basis.
  • A summary of this study has been included to provide information on the existing research in this area.

A postal survey of 122 NHS and 19 private acute admission wards within the M25 area was conducted (Bowers et al. 2002), with the aim of assessing current safety and security measures. The questionnaire was divided into four sections: survey of banned items; searching policy; practice (for example, locking doors, counting cutlery); items present or absent (for example, alarms, intercom systems, CCTV). Response rate was 70 per cent, not including 17 discarded responses, because they were not from acute admissions wards. Results were analysed with descriptive statistics and Pearson correlation tests. Fifty-six percent of respondents had panic alarms that sounded in the whole unit. In 18 per cent they sounded in the ward only while 13 per cent of wards did not have any. Personal alarms were issued in 44 per cent of trusts whilst 45 per cent didn't use them. Forty-two percent of trusts had an emergency telephone extension and 45 per cent did not. Panic alarms were found in all rooms in 36 per cent of trusts, some rooms for 32 per cent and in only one room for 20 per cent. Whilst 3 per cent of trusts had panic alarms only in the office, 87 per cent did not. The authors note two types of unrelated security systems were identified by the report: type A (door security, restrictions and banned items) and type B (searches, guards and alarms).

In nine wards that had taken part in a previous study (Bowers et al. 2002), type A was associated positively with absconding rates and type B negatively with aggressive/angry behaviour. However, these results should be tested in a larger sample. The survey does not discuss the efficiency of alarms. The survey has been replicated in Northern Ireland. However, the results were only published in 2004. As such, they fall outside the cut-off point for searching on this topic, which was 2002. This study will be considered in the update.

Evidence statements

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine whether personal and institutional alarms and communication devices are an effective means of alerting staff to occurrences of disturbed/violent behaviour.
Level 4There is insufficient evidence to determine which principles of practice are necessary to ensure the effective use of personal and institutional alarms and communication devices.
Level 4There is insufficient evidence to ascertain staff and service users' views about the effectiveness of alarms in reducing disturbed/violent behaviour in psychiatric in-patient settings.
7.8.2.2.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.2.3. Prediction: antecedents, warning signs and risk assessment

7.8.2.3.1. Objectives

The original RCPsych guideline evidence base on prediction was examined. A list of excluded studies was available in the archived information received from the Royal College of Psychiatrists Research Institute. The following information was taken from the final report in the RCPsych guideline, which states that this hypothesis was used to inform the search strategies:

RCPsych hypothesis
  • It is possible, in acute clinical settings, to predict with reasonable accuracy which patients are more likely to become aggressive or violent in the near future.

After sifting and quality checks, 16 references on prediction were included in the RCPsych evidence review.

However, the included studies did not offer generalisable criteria in support of this hypothesis, so the RCPsych guideline concluded that:

The studies do not provide a clear consensus on items that would be clinically useful for short-term prediction across a variety of clinical settings. This does not mean that prediction (still less assessment) is impossible; only that no generalisation can be made from these results. (Royal College of Psychiatrists, 1998, p45)

Current guideline

The current guideline aims to assess whether research undertaken since 1995 now offers consensus on items and tools that are clinically useful in the short-term prediction of violent/aggressive behaviour. Four review questions were identified and used to inform the search strategy (see Appendix 4 for search strategy, databases searched and search logs).

Review questions

The studies had to meet the following inclusion criteria:

7.8.2.3.2. Selection criteria
Risk factors/antecedents/staff characteristics
Types of study

Prospective cohort studies (with or without controls) to before and after studies, and qualitative studies (level 2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
  • Measurement of risk factors/antecedents.
  • Staff characteristics associated with disturbed/violent behaviour.
Predictive instruments
Types of study

Prospective cohort studies (with or without controls) (level 2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of interventions

Actuarial checklists/tools and structured clinical judgement checklists/tools.

Types of outcome
Service user and staff perspectives
Types of study

Systematic reviews through to before and after designs. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome

Staff and service user views on risk factors, antecedents, predictive instruments and staff characteristics associated with disturbed/violent behaviour.

7.8.2.3.3. Clinical evidence

The same search strategy covered all the review questions. One thousand and twenty five studies were identified in the initial sift. After sifting for relevance and duplicates 290 full papers were ordered. However 40 further duplicates were later identified.

On scrutiny, 120 were opinion pieces, anecdotal reports, or fell outside the inclusion criteria for this review (see Appendix 7 for all primary research papers that fell outside the inclusion criteria). There were also 14 letters or editorials. A further 19 studies were on topics to be considered elsewhere in the guideline and were critically appraised in subsequent evidence reviews. Seventy-three primary research papers were identified; 61 met the inclusion criteria. No study offered evidence above level 2. References were checked but no further studies were identified. In addition, all papers (16) from the RCPsych prediction review, five papers from the RCPsych environment review and two papers from the RCPsych review of restraint and seclusion were considered relevant and were included in this review. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Included studies
I. Antecedents or warning signs

Eight studies were included that considered the antecedents or warning signs of short-term disturbed/violent behaviour. Four were UK studies, set in a variety of psychiatric in-patient environments. Three were US studies; two were set in veterans' medical centres; the other in a general psychiatric unit. The final study was undertaken in Norway, set in a secure unit. A range of study designs was used. One study was a prospective cohort, four were retrospective cohorts, one was a cross-sectional study within a prospective cohort, one was a survey and the other used semi-structured interviews. Although most of the studies associated violence with verbal abuse and aggressive/agitated behaviour, only the prospective cohort study is a sufficient design to allow predictors of violence to be discussed. The one prospective cohort study (Whittington and Patterson 1996) found no significant difference in behaviour between non-violent and violent cohorts in the 24 hours prior to an aggressive incident, although aggressive behaviour was the best predictor of short-term violence. However, they did note significant differences in behaviour three days prior to an aggressive incident. The violent cohort showed increased levels of anger and aggression (p=0.0001), verbal abuse, P<0.05), threatening gestures/stance (p<0.01) and abnormal activity level (p<0.05) compared to the nonviolent cohort. This study was set in a rural hospital in the UK. The predictive validity of this study and the generalisability of these findings need to be validated by further prospective cohort studies that examine antecedents of short-term violence across a variety of settings. Three further studies in the RCPsych review examined antecedents. Only one study (Sheridan et al. 1990) noted anything different, commenting that staff limit setting often preceded violent incidents.

None of the above studies significantly change the findings of the RCPsych review.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that the following may act as antecedents/warning signs for the occurrence of disturbed/violent incidents: verbal abuse, aggressive/agitated behaviour, threatening gestures and abnormal activity levels and staff-limit setting.
II. Clinical approaches to prediction (‘first generation’)

The literature searches yielded seven prospective cohort studies investigating clinical judgement as a means of violence prediction. One study (McNiel and Binder 1991), included in the RCPsych review, did not offer information on predictive validity. Two studies (McNiel and Binder 1995, included in the RCPsych review; Haim 2002) suggested low positive predictive values for clinical prediction. The first study was set in a locked facility in the US and the other in a forensic psychiatric hospital in Israel. Three studies (Nijman et al. 2002; Rabinowitz et al. 1999; McNiel et al. 1998) suggested that clinicians' judgement had a better predictive validity than the earlier analysis allowed and demonstrated positive predictive values of 41 per cent, 58 per cent and 75 per cent and negative predictive values of 76.9 per cent, 86 per cent and 98 per cent respectively. The three studies were conducted in a range of adult psychiatric in-patient settings, and were undertaken in different countries – the first in Israel, the second in Holland and the third in the US. In a final study (Hoptman et al 1999) the overall specificity arrived at by clinicians was 79 per cent, with a corresponding sensitivity of 54 per cent. This study was set in the US, in a forensic psychiatric hospital. It is a distinctive study in that 57.4 per cent of the participants were African-American and only 35 per cent Caucasian. In most of the other studies the majority of participants were of Caucasian origin.

Whilst four of the seven of the included studies suggest that clinicians are able to predict violence with a greater degree of accuracy than has previously been suggested, there are no identifiable features to explain the greater degree of clinician predictive accuracy found in these four studies. One study relates accuracy to confidence, but does not state if this is also related to experience. Another study admits that the inability of the study to provide an underlying reason for the high predictive value is a key weakness of the design. In order for clinicians' judgements alone to be recommended over and above other approaches to prediction, further studies that are tested across a variety of settings are required to validate these findings. Whilst the general trend of the studies stresses that clinicians may be able to predict violence with some degree of accuracy, there is a lack of consensus amongst the studies.

None of the studies significantly change the findings of the RCPsych review in which five further studies (Kirk 1989; Janofsky et al. 1988; McNiel et al. 1988; Apperson et al. 1983; Yesavage 1983) also indicated the low predictive validity of clinicians' judgement.

Evidence statement

Level of evidenceEvidence statement
Level 4The evidence suggests that clinicians' judgement has a relatively low predictive validity, only slightly better than chance.
III. Actuarial approaches (‘second generation’)
III.i. Risk factors

Whilst antecedents or warning signs are risk factors, in this review they are distinguished from more static variables which could be used to predict violence – such as diagnosis, demographic variables etc – which are referred to as risk factors. It is such risk factors that this next group of studies examines.

The literature search identified the following relevant studies: eight prospective cohort studies; six retrospective cohort studies; one cross-sectional study; two prospective case-controlled studies; and one retrospective/historic case-controlled study.

Seven prospective cohort studies identified the following significant risk factors:

  • community violence, male gender, young age, younger age at first hospitalisation, more frequent visitors – at least monthly, not having own clothing, low level of self-care functioning, number of admissions, duration of admission, coercive behaviour and lack of satisfaction of care, a diagnosis of organic psychotic condition, personality disorder, schizophrenia, and bipolar affective disorder.

Five further studies in the RCPsych reviews also examined risk factors. The studies noted the following risk factors:

  • history of violence, hostile suspiciousness, agitation-excitement, thinking disturbances, use of intoxicants, command hallucinations, impostor delusions and delusions about personal targets.

Only prior community violence/history of violence was mentioned in more than one study as a risk factor, and it was then only regarded as a risk factor within the first one to two days of admission and not afterwards.

The other prospective cohort study (Owen et al. 1998a) examined the risk factors amongst adult recidivists in psychiatric acute care settings, including geriatrics. It identified the following significant risk factors: being older, widowed, having personality disorder, or organic brain disorder, being detained under the Mental Health Act, and being highly sedated prior to the incident.

Within the studies there are no risk factors that consistently emerge, except for prior community violence. There is also no consensus within the studies as to how these various risk factors ought to be weighted. These studies only serve to illustrate that a huge range of variables are possible risk factors. However, most of the studies did suggest that demographic variables were largely irrelevant in risk prediction.

Most of the studies did not discuss the predictive validity of this approach. One study, which discusses community violence as a risk factor (Beck and Bonner 1988), notes a positive predictive validity for the first day of admission of 31 per cent – slightly worse than that averaged by clinicians' judgement in the studies discussed in Section II. One study (McNiel et al. 1998), notes that the actuarial approach was significantly better than clinician prediction. However, in this study clinician prediction was particularly low (true positives = 26.7 per cent, false positives = 73.3 per cent). Only one study (Rabinowitz and Garelik-Wyler 1999) shows a higher positive predictive value of 61.6 per cent and a negative predictive value of 69.3 per cent in predicting the violent group of service users. On the basis of these results it has not been possible to establish risk factors for the prediction of violence.

None of the studies significantly change the findings of the RCPsych review.

Evidence statements

Level of evidenceEvidence statement
Level 2 + to 2-The evidence suggests that the following may be risk factors for disturbed/violent behaviour: community violence, male gender, young age, younger age at first hospitalisation, not having own clothing, low level of self-care functioning, number of admissions, duration of admission, coercive behaviour + lack of satisfaction of care, a diagnosis of organic psychotic condition, personality disorder, schizophrenia, and bipolar affective disorder. However, further research is needed to reliably determine additional factors that may need to be considered across different populations.
Level 4There is insufficient evidence to determine how various risk factors associated with disturbed/violent behaviour in psychiatric in-patient settings ought to be weighted.
III.ii. Predictive actuarial tools

Eleven prospective cohort studies were identified through the literature searches. These studies assessed the predictive validity of a range of actuarial tools or checklists in adult in-patient psychiatric settings as a means of predicting violence (see Table 1, page 42). A wide range of risk factors was considered. Echoing the findings from the review of the actuarial approaches to determining risk factors considered in Section III.i above, there is no agreement amongst the tools examined here as to which risk factors are most important, or how the various risk factors ought to be weighted.

None of the studies took place in the UK. Five took place in the US; one in Australia; one in Sweden; one in Norway; one in Spain; one in Italy; and one in Taiwan. There is a need to test these actuarial tools in a European and a UK context. Research into prediction of violence in psychiatric in-patient settings involving the use of these tools is still at a preliminary stage. None of the studies considered whether prediction with a particular tool led to a decrease in disturbed/violent behaviour/incidents.

On the basis of the clinical evidence, no one tool emerged as the ‘gold standard’. Six studies (Chou et al. 2002; Ehmann et al. 2001; Arango et al. 1999; Krakowski et al. 1999; Almvik et al. 1998; Yesavage 1984) reported on the positive and negative predictive values established using this approach. With the exception of the Brøset violence checklist, which is still in a preliminary stage of development, the various actuarial tools showed consistently higher positive and negative predictive values than those established by clinical judgement alone in the aforementioned analysis (see Table 1, page 42). These results suggest that important developments have been made since the RCPsych guideline. Whilst more research is needed to validate the findings of these studies, and to test the instruments across a range of settings, the clinical evidence suggests that there is a trend towards greater predictive accuracy with actuarial tools than with clinical judgement alone.

Only three studies in the RCPsych review considered the use of actuarial tools. Since the completion of the RCPsych review, more studies as outlined here have examined the predictive accuracy of actuarial tools These studies suggest that actuarial tools offer greater predictive accuracy that clinical judgement alone.

Evidence statements

Level of evidenceEvidence statement
Level 4There is insufficient evidence on which to determine a ‘gold standard’ predictive actuarial tool.
Level 4The current evidence suggests a trend towards greater predictive accuracy with actuarial tools than with clinical judgement alone. However, further comparative research is needed.
IV. Structured clinical judgement instruments (‘third generation’)

Three studies examined the usefulness of instruments that measure structured clinical judgement. Two were European studies – one set in Sweden and the other in the UK – and the third was a US study. Two of these were prospective cohort studies and are described here.

The first prospective cohort study was undertaken in Sweden, (Belfage et al. 2000). It considered both the HCR-20 (historical/clinical/risk – 20-item scale, version 2) and the PCL:SV (psychopathy checklist: screening version) within a maximum security correctional setting. The study found that history of violence was not a good predictor of future violence. The authors suggest that this is unsurprising in a setting where all patients will score highly on the H-10 – the historical part of the tool. The R-5 showed the best predictive validity (p=0.004). Thirty of 41 participants were psychopaths for whom the R-5 was the only tool with any predictive validity (p = 0.002). The PCL:SV showed a higher score for violent recidivists and was significant when considered in conjunction to older age (p<0.1).

The second prospective cohort study (Hill et al. 1996) was conducted in the US in a state hospital to assess whether the PCL: SV was a good predictor of aggression amongst 55 male forensic psychiatric service users. The authors found that the PCL: SV total was a significant predictor of aggression. The PCL: SV was then reduced to presence or absence of psychopathy, where it again predicted aggression (multiple r = 0.69; R2=0.48; Beta = 0.69).

Three other studies that examine the effect size of the PCL and the PCL: SV are mentioned by Doyle and Dolan (2000). They also note a good predictive validity for the tool in a forensic setting. However, as of yet, insufficient research has been carried out to test the predictive validity of these instruments in UK adult psychiatric in-patient settings. More studies also need to be undertaken to validate the results of these studies. All studies are detailed in Table 2, page 45.

All these studies have taken place since the RCPsych review. They suggest that structured clinical judgement gives a greater predictive accuracy than clinicians' judgement alone, similar to that achieved by the use of actuarial tools.

Evidence statements

Level of evidenceEvidence statement
Level 4The evidence suggests that there is trend towards greater predictive accuracy with structured clinical judgement tools than with clinical judgement alone, similar to that achieved by the use of actuarial tools.
Level 4There is insufficient evidence on which to determine a ‘gold standard’ structured clinical judgement instrument.
V. Staff characteristics associated with the likelihood of aggressive incidents

Nine studies were identified which examined the relationship between staff characteristics and assault (Flannery et al. 1995; Flannery et al. 2001; Chou et al. 2001; Chou et al. 2002; Lanza et al. 1997; Owen et al. 1998b; Soares et al. 2000; Augestad and Vatten 1994; Ray 1988; Morrisson 1998). In addition a further five studies were identified in the RCPsych review (Whittington and Wykes 1996; Binder and McNeil 1994; Whittington 1994; Whittington and Wykes 1994; Carmel and Hunter 1991). All the studies, except one (Lanza, 1997), surveyed staff in a range of psychiatric settings. Lanza (1997) was set in a neuropsychiatric department in a veterans' hospital in the US. This study found no relationship between staff characteristics and assault. The other 12 studies identified the following characteristics as significant correlates of staff characteristics and the occurrence of aggressive incidents:

younger age, between youngest and oldest age, work experience, training in the management of violence, and grade, lack of training and limit setting/confrontation, authoritarianism, social restrictiveness, young age, limited supervision and gender. (It was also suggested that gender was non-significant (Binder and McNeil 1994). The study (Augestad and Vatten 1994) which emphasises gender as significant notes that whilst overall risk for men was significantly higher, the relative risk according to ward type was similar for men and women). Several other correlates were identified, but were not significant.

There was no consistency across the studies.

None of the studies significantly change the findings of the RCPsych review.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that the following staff characteristics may be associated with increased occurrence of incidents of disturbed/violent behaviour: younger age, level of experience, training and grade, gender, and involvement in limit-setting activities. However, further research is needed.
VI. Service user perspectives on reasons for assault

Seven studies were identified which examined service user perspectives on the causes of disturbed/violent behaviour.

One Canadian study surveyed a heterogeneous group of 12 thought-ordered individuals across two hospitals. Service users reported that factors that caused violence tended to be external rather than internal (that is caused by factors on the ward rather than by illness). The results of this study would need to be validated on a larger scale and replicated across other settings, before any evidence-based conclusions could be drawn about service user perspectives on the causes of aggressive incidents.

Another study (Whittington and Wykes 1996) conducted semi-structured interviews with staff in the UK to ascertain whether aversive stimulation (such as limit setting) had occurred prior to an assault. The results were verified by interviews with some service users and witnesses, indicating that some form of aversive stimulation often precedes an assault. However, the study is unclear about the number of service users who were interviewed or the extent to which they agreed with staff. The study reports that many staff believed that aversive stimulation trigger the majority of the aggressive incidents (see staff perspectives below).

One survey (Gillig et al. 1998) found that service users saw less of a causal connection between their own verbal abuse of staff and the physical abuse of staff than staff (p<0.05). However, they saw more of a causal connection between verbal abuse of service users by staff and physical violence against staff, than staff (p<0.05). This pattern and significance was echoed with regards to hostility and threats. Service users identified staff use of drugs and alcohol (p<0.05), the use of forced medication (p<0.05), restraints (p<0.05) and seclusion (p<0.05) as causes underlying violent incidents. They also stressed cross-cultural racism as a cause of violence.

Another study that used incident forms, a survey and interviews for data collection (Duxbury 2002) noted that service users believed that external and situational factors (such as interactions with staff and restrictive regimes) were largely to blame for violent incidents (p<0.001).

A further study using semi-structured interviews (Bensley et al. 1995), noted that service users, like staff, considered restrictions on service users' smoking, access to outdoors, defective staff clinical skills, service users not being treated with respect, as well as the use of seclusion and restraint, to contribute to violent incidents. Service users were also concerned that rules were not adequately explained.

Using video footage and interviews (Crowner et al. 1995) it was found that 12 per cent of service users argued that they had been playing with the victim, 12 per cent claimed that they had been subjected to verbal abuse, and 8 per cent claimed that they had been subjected to objectionable behaviour. The other 41 per cent gave a range of responses – from no response to anger at ward rules, and anger at unwanted sexual attention.

Ilkiwa-lavelle and Grenyer, (2003) found that service users believe improved handling of inter-personal conflicts would prevent violent incidents.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that service users regard external factors (such as limit setting, verbal abuse by staff and other service users, lack of respect by staff and harassment) as likely reasons for assault rather than internal factors (that is, caused by illness).

The RCPsych review did not specifically consider this issue.

VII. Staff perspectives on reasons for assault

Nine studies were identified which examine staff perspectives on reasons for assault.

Four studies (Gim et al. 1999; Duxbury 1999; Gillig et al. 1998; O'sullivan and Meagher 1998) carried out questionnaire surveys to assess health care professionals' views on the risk factors associated with psychiatric inpatient violence. The first study is set in Ireland; the second in Singapore; the third in the UK; and the fourth in the US. The following were proposed as risk factors by staff but not validated:

personality disorder, schizophrenia, substance abuse, intoxication, violent lifestyles, active hallucinations, paranoid ideas against others, noncompliance with treatment, internal factors and provocation.

Only personality disorder was mentioned in two of the studies, although all studies mention internal factors, including diagnosis. These suggest that staff emphasise diagnosis over other variables. However, more studies are needed before the generalisability of these findings can be assessed.

Two studies (Whittington 1996; Bensley 1995) conducted semi-structured interviews with staff. One study (Whittington 1996) attempted to assess whether they had caused aversive stimulation prior to an assault. After validating the results of the interviews by interviewing service users and witnesses, it was noted that 86 per cent of all assaults were preceded by some form of aversive stimulation (such as limit setting), although the authors suggest an interplay with diagnosis. These results require validation and replication across a variety of settings.

Using semi-structured interviews (Bensley 1995) found that staff believed inadequate staffing levels was the single factor that most contributed to assaults on staff. Like service users, staff were also concerned about service user restrictions on smoking, access to outdoors, staff clinical skills, service users being treated with respect, as well as the use of seclusion and restraint on the wards. Staff also mentioned a need for training in the management of violence, as well as concerns about the general physical environment. However, the study was of low quality.

Using incident forms, questionnaires and interviews (Duxbury 2002) noted that staff most commonly reported problematic interaction and restrictive environments as the causes of violence/aggression. However, staff did not consider their own personal interactions with service users to be problematic. Staff were unable to identify a cause for 26 per cent of all incidents. Staff attributed much more weight to internal factors (that is, illness) as underlying causes of violence/aggression than service users.

Using focus groups and surveys (Delaney et al. 2001), the following were noted as impacting on the possible risk of violence: service user history; service user status and mode of arrival; ongoing informal nurse assessment; individualised care; peer support and administrative responsiveness; nursing stress; and current policies/manuals. However, the study was of low quality.

Ilkiwa-lavelle (2003) found that staff believed the service user's illness to be a key causal factor, whereas service users believed inter-personal conflicts were relevant. Staff believed improved medical management would prevent violent incidents.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that staff users regard internal factors (that is, caused by illness) and the interplay between internal and external factors (such as staff limit setting) as contributing to disturbed/violent behaviour.

The RCPsych review did not specifically consider this issue.

7.8.2.3.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.3. Training

7.8.3.1. Objectives

No specific searches on training systems were undertaken in the RCPsych guideline.

Current guideline

Two review questions were identified by the GDG and used to inform all searches:

  • What are the most effective and safe training programmes for the prevention of and the short-term management of disturbed/violent behaviour in adult psychiatric in-patient settings?
  • What are the views of staff and service users about the various training programmes in adult psychiatric inpatient settings and their content?

Included studies were subdivided into more specific review questions that related to each of the interventions and related topics covered in this guideline.

7.8.3.2. Selection criteria

Types of studies

Systematic reviews to controlled before-after studies. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcomes
  • Effectiveness of training packages in managing or reducing disturbed/violent behaviour.
  • Safety of training packages managing or reducing disturbed/violent behaviour.
  • View of staff and service users on the various training programmes for managing or reducing disturbed/violent behaviour.
  • Increased staff knowledge resulting from training.
  • Staff attitude change resulting from training.
  • Reduction in the number of disturbed/violent incidents.
  • Reduction in the number of staff days lost through illness.

7.8.3.3. Clinical evidence

Two hundred and thirty nine papers were identified in the initial sift. Eighty-four studies were ordered. After quality checking, 22 studies were included in the review. Eight studies were excluded. The remaining studies were overviews. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Appraisal of methodological quality

In addition to those mentioned in section 7.3.6, the following methodological concerns were raised by these studies:

  • All the studies included in this review have relatively small sample sizes.
  • For most of the studies the training package was not specified.
  • Where the training package was specified, insufficient details were provided to allow meaningful comparisons to be made with other training packages.
  • Long-term outcomes, such as improvement in service user care, were not measured.
Included studies

Fourteen before and after studies were included in this review. One of these before and after studies (Perkins and Leadbetter 2002) is conducted in an area outside of the scope of the guideline. However, this study is included here because the CALM technique that it evaluates is one of many training packages used in the UK for the short-term management of violence. Two pilot studies are also included. The first study (Frey and Weller 2000) is included, despite only being a pilot study, since it is the only study that considers the effectiveness of training service users, rather than staff, as a means of reducing inpatient violence in adult psychiatric settings. The second study (Bournemouth University, unpublished) is included since it is one of only two studies to evaluate the effectiveness of training in a clinical environment. Of the other studies included, one study is a retrospective cohort study, and seven are cross-sectional surveys.

As the studies often address multiple issues, the findings of the studies have been grouped together under topics. This means that some studies are referred to a number of times.

I.a. Review question: what are the most effective and safe training programmes for the prevention of and the short-term management of disturbed/violent behaviour in adult psychiatric in-patient settings?

The effectiveness of training staff in interventions for the short-term management of violence: general outcomes

Increased knowledge

In a before and after study (Calabro et al. 2002), significant increases were noted immediately after training in nonviolent crisis intervention (CPI) in post-test knowledge scores (p<0.001), positive change of attitude towards the techniques taught (p<0.001), self-efficacy (p<0.01) and positive change in behavioural intention (p<0.05). A further before and after study (Ilkiw-Lavalle et al. 2002), found that staff knowledge improved significantly immediately after training, with ancillary staff improving by the largest effect size (2.25). Staff with no prior training had the greatest improvements immediately after training (p<0.01) A before and after study (Paterson et al. 1992) noted a significant increase in knowledge immediately after training in post-test knowledge scores (p=<1%). Stress, as assessed by a general health questionnaire, was also significantly reduced (p=<1%). In a controlled before and after study (Rice 1985), significant improvements were found in all areas of knowledge (self-defence and patient restraint written test p<0.0001 for lesser security staff only) immediately after training.

None of these studies assess the effectiveness of training in a clinical environment.

Attitude changes

A before and after study (Collins 1994), found that staff were less likely to hold service users entirely responsible for their behaviour, and acknowledge facts such as service user fear as causative immediately after training and at six months post training. In a before and after study using the CALM techniques in a school for children with learning difficulties (Perkins 2002), no significant changes in staff attitude toward pupil aggression were noted.

In a before and after study (Collins 1994), staff confidence was found to increase immediately after training. Staff remained more confident six months post training. In a before and after study using the CALM techniques in a school for children with learning difficulties (Perkins and Leadbetter 2002), it was noted that 82 per cent of staff interviewed expressed increased confidence in their ability to deal with an aggressive incident six months post training.

None of these studies assess the effectiveness of training in a clinical environment

Number of disturbed/violent incidents

In a retrospective cohort study (Carmel and Hunter 1990), training in managing aggressive behaviour and CPR was not found to be significantly linked to the number of violent incidents on the wards. In a before and after study (Sjostrom et al. 2001), no significant reduction in the number of aggressive incidents using the social dysfunction aggression scale (SDAS-9) was noted six weeks after training.

In a controlled before and after study (Rice 1985), there was a significant reduction in violent incidents (p<0.05) for the 18 months after training. Taxis (2002) shows a dramatic decrease in the use of seclusion and restraint after a 42 month period of training.

In a before and after study (Whittington and Wykes 1996), it was noted that wards that sent the majority of their staff to a one-day training course that did not involve restraint training noticed a significant reduction in assaults (p<0.05) for the 28 days after training. Staff who took part in training had a 31 per cent lower rate of assault after training than those who did not take part. The decrease was unusual but did not reach significance.

Staff injuries and missed workdays

In a retrospective cohort study (Carmel and Hunter 1990), when wards highly compliant with training were compared to wards with low training compliance, a significant positive relationship was noted between those trained in managing aggressive behaviour and the number of staff injuries (p<0.005).

In a retrospective cohort study (Carmel and Hunter 1990), a significant relationship was noted between individual staff who were trained in either managing aggressive behaviour or CPR and a reduction in staff injuries based on monthly reports over the course of a year (p<0.001).

In a before and after study (Martin 1995), it was noted that two years after a training programme was initiated, although the number of aggressive incidents increased along with the level of aggression, the number of staff injuries fell and the number of missed work days fell resulting in a saving (in relation to missed workdays) of $173,960 (year 1); $2,478 (year 2); $2,414 (year 3). (NB: Not enough information reported to assess quality).

In a controlled before and after study (Rice 1985), there was a significant reduction in lost work days on wards that took part in the training relative to those that did not (p<0.001) for the 18 months post training.

In a before and after study, Sjöström et al. (2001) noted that six weeks after training, no significant reduction in the number of staff on sick leave was noted.

Prediction and risk assessment

In a cross-sectional survey of 193 UK trusts providing mental health services (Davis 2001), just over 50 of the 84 per cent replying provided training on risk assessment for harm to others.

A before and after study (Ilkiw-Lavalle et al. 2002), found that staff knowledge of prediction was significantly increased immediately after training (p<0.01).

None of these studies assess the effectiveness of training in a clinical environment.

De-escalation techniques

In a before and after study (Paterson et al. 1992),‘blinded’ raters judged that there was a significant increase in staff competence in de-escalation immediately after training.

In a before and after study (Philips and Rudestam 1995), judges rated that immediately after training, the experimental group of staff who received training placed significantly more value on non-aggressive responses to service user violence/aggression (p=0.05) than a control group who were not trained, or a control group who received only didactic training.

A controlled before and after study (Wondrak and Dolan 1992), trained student nurses to deal with verbal abuse. Using role-play, blind raters noted that there was significant improvement for those trained immediately after training, compared to those not trained in all areas except empathy, eye contact and anger levels. Post test, the attendees appeared more relaxed to blind raters (p=0.031), less upset (p=0.001) and had a more effective use of posture (p=0.005).On self-reported questionnaires, three areas achieved significance in those trained: feeling less angry (p=0.002); feeling less out of control (p=0.005); and feeling less threatened (p=0.035); in a similar situation.

In a before and after study using the CALM techniques in a school for children with learning difficulties (Perkins and Leadbetter 2002), semi-structured interviews suggested that verbal de-escalation appeared to have increased six months after training.

None of these studies assess the effectiveness of training in a clinical environment.

Restraint

In a before and after study (Paterson et al. 1992) ‘blinded’ raters judged that there was a significant increase in staff competence in control and restraint and disengagement immediately after training.

In a before and after study (Philips and Rudestam 1995), immediately after training, judges rated the experimental group of staff who received training as significantly more competent in physical skills as well, as displaying less fear and aggression (p=0.05), than a control group who were not trained, or a control group who received only didactic training.

In a controlled before and after study (Rice 1985), there were significant improvements in areas of skill (sensitive situations skill test p<0.001) and audiotaped simulations test (p<0.01) immediately after training. There was also a significant increase in the on-ward job reactions scale six weeks after training for maximum security workers, compared to controls (p<0.01). This scale measures how comfortable participants are in their interactions with service users.

In an unpublished pilot study conducted by Bournemouth University, the effectiveness of restraint and breakaway techniques were considered in a psychiatric intensive care unit (PICU). The study used a prospective cross-sectional approach over 32 months. During this period, 346 adverse incident forms were collected. They did not record the use of breakaway techniques, however, when 19/22 staff were interviewed retrospectively, three recalled using recognised breakaway techniques, one a restraint technique and one an unrecognised technique. Staff did not recall any techniques used being inappropriate or ineffective. They recalled problems with taking the client to the floor. All were satisfied with the training they had received, but wanted more frequent refresher courses and a greater emphasis on de-escalation.

Parkes (1996) conducted a before and after study in a 44-bed medium secure unit to assess the effectiveness of a four-day C&R training course. Interviews were conducted with all staff involved in a restraint incident for the 18 months prior to training and the 12 months after all staff had been trained. Data was collected on all 340 incidents involving physical restraint. One hundred and forty-nine incidents involving restraint occurred after training. For statistical purposes these were compared with 149 incidents immediately prior to training. Staff injuries during the restraint phase increased after training (p<0.05). Injuries to service users during restraint phase did not significantly alter post training. No other significant changes in injury rates were noted. Overall changes in injury rates were not significant. There were no significant changes in difficulty rating or risk rating after training. The modal number of staff restraining a person increased to three after training. The highest number of staff involved in a single restraint decreased from 10 to six after training.

Only two studies – Bournemouth University, (unpublished) and Parkes (1996) – assessed the effectiveness of training in a clinical environment. However, the first was a pilot.

Evidence statements

Level of evidenceEvidence statement
Level 4The limited evidence suggests that short-term improvements in knowledge, skills and reduction in stress occur after staff training in the management of disturbed/violent behaviour.
Level 4The lack of evaluations of the effectiveness of training in a clinical environment mean that a ‘gold standard’ training package for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings cannot be determined.
I.b. Review question: how effective was training service users in the management of their disturbed/violent behaviour?

One before and after pilot study (Frey and Weller 2000) examined the effect on incidences of aggressive and violent behaviour of training service users to respect themselves, peers and staff. Service users demonstrated a significant increase in knowledge, based on a questionnaire immediately after the training (p<0.05). A nurses' survey indicated a reduction in aggressive behaviour in the week after training (p<0.05). Authors claim that the training inadvertently led staff to alter their behaviour. It caused staff to become more aware of the causes of service user aggression, through examining the feedback. It is therefore unclear whether changes of staff behaviour or training service users had the most impact on reducing violence.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that training service users to respect themselves, peers and staff may reduce the occurrence of disturbed/violent incidents.
I.c. Review question: What are the views of staff about the various training programmes in adult psychiatric in-patient settings and their content?
General perspectives
Staff assessment of training needs

In a Canadian cross-sectional survey (Chaimowitz and Moscovitch 1991) of medical students to assess the adequacy of training to deal with violent incidents, 34.3 per cent thought that staff psychiatrists were adequately trained; 24.4 per cent thought that psychiatric residents were adequately trained; 50.4 per cent thought that nurses were adequately trained; and 79.5 per cent wanted improvements in education and training for staff. In a cross-sectional survey of medical trainees in New Zealand (Coverdale et al. 2001), only 30 per cent had training in the management of violence and only 36 per cent of these viewed it as adequate. Only 62 per cent of those who were psychiatry trainees had received training in the management of violence.

PICU ward managers at a PICU conference took part in a UK cross-sectional survey (Clinton et al. 2001). Eighty-one percent of respondents stated that they would attend a course on violence management, however only 17 per cent were aware of relevant courses in their locality.

In a before and after study (Ilkiw-Lavalle et al. 2002), it was noted that staff with previous training would have preferred to focus on special skills rather than repeat basic training.

In a cross-sectional study using semi-structured interviews (Southcott et al. 2000), it was noted that staff wanted more frequent refresher courses (three to six monthly). In an unpublished pilot study by Bournemouth University, it was noted that, of the staff interviewed, 19/22 wanted more frequent refresher courses and a greater emphasis on de-escalation.

Staff satisfaction with training

In a before and after study (Goodykoontz and Herrick 1990), it was found that staff felt more confident in their ability to handle violent situations after training. They stated, four months after training, that they were more likely to intervene than they had been, rather than waiting for hospital security. After training, they felt that they had a plan of how to proceed when faced with a violent incident.

In a controlled before and after study (Rice 1985), the training course was well received (mean = 5.5 on a 6-point scale where 6 is the best possible score). The results were little altered at six weeks and 15 months.

In a cross-sectional study using semi-structured interviews (Southcott et al. 2000), it was noted that staff were generally satisfied with the training that they had received and felt that the techniques that they had learnt were both effective and appropriate immediately after training. (All staff received training in control and restraint and breakaway techniques).

In an unpublished cross-sectional pilot study with retrospective interviews (Bournemouth University) it was noted that of all staff interviewed, 19/22, were satisfied with the training they had received.

In a before and after study using interviews with a standard form (Parkes, 1996), staff felt safer and more in control when relocating the service user after having received training. They felt that C&R techniques appeared more professional to observers than unauthorised holds. Staff felt that training made it easier to hold the service user for a protracted length of time.

Staff perspectives relating to specific interventions
Prediction and risk assessment

In a before and after study (Collins 1994), nurses believed that some prediction of violence was possible immediately after the training course and six months post training.

De-escalation techniques

In a before and after study (Beech 1999), students nurses were confident that they could manage verbal aggression immediately after training (p=0.0000).

In a cross-sectional study using semi-structured interviews four months after training (Southcott et al. 2002), it was noted that staff felt that de-escalation training should be provided before breakaway training.

Restraint

In a cross-sectional study using semi-structured interviews four months after training (Southcott 2002), it was noted that staff felt that the process of restraint was often messy and unco-ordinated and could be improved with better planning. (All staff received training in control and restraint and breakaway techniques).

Self-defence

In a before and after study (Beech 1999), immediately after training, student nurses believed that they would be able to protect themselves using reasonable force (p<0.0000).

Evidence statements

Level of evidenceEvidence statement
Level 4Staff perceive that training in the short-term management of disturbed/violent behaviour is beneficial and that it also increases confidence in dealing with disturbed/violent behaviour
Level 4The evidence suggests that staff often feel that their need for training is not met.
I.d. Review question: what are the views of service users about the various training programmes in adult psychiatric in-patient settings and their content?

In a controlled before and after study (Rice 1985), it was noted that after staff training, service users showed positive changes on a modified Coppersmith self-esteem inventory – a scale measuring depression and anxiety from an adjective checklist and on a modified feelings scale. The results were significant on maximum secure wards (p<0.05). The questions were given to each service user weekly from six weeks before training until six weeks after training. The researcher had wanted to assess service user responses to staff training through a ward atmosphere scale, but staff objected.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine service user perspectives on service user training to help service users manage their aggressive and/or disturbed/violent behaviour.
II. Current practices in training in the UK

Four cross-sectional studies were found that examined current training arrangements in the UK.

Wright et al. (2000) examined the policies for the management of violence in PICUs and RSUs. One hundred and twelve wards were surveyed, 33 policies were returned (representing a good geographical spread). Nine percent of policies were current and not awaiting update; 9 per cent were out of date; 27 per cent were undated. Less than two-thirds of the policies had a stated aim or a definition of violence. Three-quarters of the policies stressed the need to report incidents, have post-incident team support, review the incident, outlined expectations and responsibilities of staff, emphasised prevention and de-escalation and had a commitment to train all appropriate staff. Just over half also mentioned the need for refresher courses. However, where a commitment to training was mentioned, less than half stated who was responsible for ensuring training was provided. Ninety four percent of policies listed physical restraint as an acceptable method for managing violence. However, less than half the policies listed unacceptable methods of restraint.

Lee et al (2001) investigated the training that is undertaken in PICUs and RSUs. One hundred and twelve units were contacted (760 staff) and there was a 47 per cent response rate. It was noted that it was possible to identify a core curriculum of 12 techniques across a range of courses: taking the patient to the floor, three-person restraint team, sitting and standing the patient, negotiating stairways and doors, restraining hold, roles within team, turning the patient over, breakaways, entry into and exit from seclusion, blocking punches, blocking kicks, separating fighting patients. Eighty-two percent of staff were able to identify the organisation that provided their training. Most initial courses lasted for five days. Respondents in RSUs were significantly more likely to be taught breakaway techniques (p=0.03), entry and exit from vehicles (p=0.00017) and defence against weapons (p=0.02) than respondents in PICUs. The three techniques most commonly used in practice were verbal de-escalation, restraining holds and use of three-person team. Thirty one percent of respondents did not state that their courses contained ethical and safety issues or verbal de-escalation. While 39 per cent received training within three months of taking up post, 21 per cent did not receive training for a year or more and 8 per cent had received no training at the time of the survey. Ninety-eight percent stated that they expected to attend a refresher course. Confidence in the skills learnt was high (mean = 4.63 on a six-point scale).

Davis et al. (2001) approached clinical directors in 193 NHS trusts that provide mental health services to assess how much risk assessment training takes place. The survey had an 82 per cent response rate. Just over 50 per cent provided training on risk assessment for harm of others. Most trusts provided training on mental health legislation. Most trusts provided annual training courses, but these were not compulsory. Clinical directors noted that staff attendance was low, but that many staff received additional training as part of routine clinical work or courses such as MRCPsych. Around 50 per cent provided follow-up courses. The existence of written policies varied; most trusts had policies on observation.

Bleetman and Boatman (2001) conducted a cross-sectional questionnaire survey across 305 acute and community trusts, 30 ambulance trusts, 40 personal training organisations, and 63 corporate organisations. Mental health trusts were excluded. The aim was to provide an overview of control and restraint issues in the health services. The response rate was low. Acute and community trusts – 29 per cent; ambulance trusts – 30 per cent; training organisations – 45 per cent; corporate organisations – 13 per cent. The following results were noted: no significant difference in levels of confidence in the reporting process was found between those trusts using a specific aggressive/violent incident form and those using a general form. No significant results were found on the use of PPE or personal alarms. Training organisations reported the following results: 72 per cent stated staff were certificated to deliver training – but no standardisation; 56 per cent reported trainers were qualified first aiders; 78 per cent offered non-physical conflict management; and 67 per cent offered training in physical skills (types of skills outlined). Fifty percent knew skills taught were operationally effective. The evaluation of content of training packages appears subjective. This demonstrates the lack of standardisation in the UK. Authors note that it was not possible to reach any firm conclusions about the effectiveness of training techniques employed in the UK on the basis of this study.

Taken together, these four studies suggest that the following constitute the core curriculum of training courses in the UK:

  • taking the patient to the floor
  • three-person restraint team
  • sitting and standing the patient
  • negotiating stairways and doors
  • restraining hold
  • roles within team
  • turning the patient over
  • breakaways
  • entry into and exit from seclusion
  • blocking punches
  • blocking kicks
  • separating fighting patients.

However, the limited scope of these studies limits the generalisability of their findings.

Evidence statement

Level of evidenceEvidence statement
Level 4The evidence on current training practices in the UK indicates that there is a lack of standardisation in the way staff are targeted for courses, and in the range of interventions covered.
In addition, the effectiveness of training has not been adequately evaluated in a clinical environment.

7.8.3.4. Economic evidence

No studies containing relevant economic data were found. The following additional exploratory cost analysis was carried out (for full details see Appendix 9).

  • The cost effectiveness of life support training was considered. It was concluded that the cost per QALY of immediate life support (ILS) training with automated external defibrillators (AED) under scenario 1 is around £23,000. Sensitivity analysis shows that this may be cost effective (i.e. cost per QALY of £20K or less), if one or other of the factors are favourably different from scenario 1.
  • Scenario 5 suggests that advanced life support (ALS) training (where cost of training will be more than twice ILS) is highly unlikely to be cost effective.

7.8.4. Staff and service user perspectives

7.8.4.1. Staff and service user perspectives – general

7.8.4.1.1. Objectives

No specific searches on staff and service user perspectives were undertaken in the RCPsych guideline.

Current guideline

Three review questions were identified by the GDG to be addressed in this review:

Review questions
7.8.4.1.2. Selection criteria
Types of studies

Systematic reviews to before and after studies. Qualitative studies were also included (level 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome

General staff and service user perspectives on the short-term management of disturbed/violent behaviour.

7.8.4.1.3. Search strategy

No specific searches were undertaken to identify papers that discussed staff and/or service user perspectives on the short-term management of disturbed/violent behaviour in psychiatric in-patient settings, since all searches were broad enough to retrieve papers that examined staff and/or service user perspectives. The articles that form the basis of this review were identified by the various searches for each of the interventions covered in this guideline. However, rather than looking at a single intervention or area, we considered this topic in relation to the short-term management of disturbed/violent behaviour as a whole.

7.8.4.1.4. Clinical evidence

Eleven papers that examine general staff and service users' perspectives on the short-term management of disturbed/violent behaviour were identified by our searches. After critical appraisal, nine papers were included in this review. One was excluded. The other papers were overviews of a general nature, and were therefore ineligible. (Evidence tables of studies included only in this review can be found in Appendix 5. Evidence tables of studies excluded only from this review can be found in Appendix 6).

General: staff and service user attitudes towards disturbed/violent behaviour in psychiatric in-patient settings

Nine studies were identified which considered staff attitudes towards disturbed/violent behaviour. A range of study designs and perspectives were examined, making the results difficult to synthesise.

I.a. Review question: do staff and service users perceive themselves to be safe in psychiatric inpatient settings?

In a survey (Baxter et al. 1992), it was noted that staff felt uncomfortable with the belief that they should be able to predict violent incidents and were concerned about the frequency with which violence occurred. They felt that there was a lack of support/protection from the hospital.

One cross-sectional study (Thomas et al. 1995) examined staff attitudes toward service user safety. Seventy five percent of nurses rated both themselves and service users as safe. However a smaller (unspecified) number of nurses stated that they believed that service users actually felt safe.

An overt researcher-as-participant study (Quirk et al. 2004) considered strategies used by service users to keep safe in adult psychiatric in-patient settings. The study was supplemented by interviews with staff, service users and advocacy work, as well as by results from a national audit. The following strategies for managing risk of violence were identified: avoiding risky situations, avoiding service users who explicitly warned others to keep away, finding a safe haven (like a bedroom), getting ‘specialled’ or not resisting it, using de-escalation techniques, allying oneself with someone high in the ‘pecking order’, making risk assessments of other service users (including proactive information gathering), warning staff about another service user, and getting discharged. The authors note that avoidance tactics were harder to employ in certain circumstances, like the canteen, where service users had to rely more heavily on staff. The authors concluded that service users take an active role in making a safe environment for themselves and are not passive recipients of safety interventions by staff. They suggest that, in part, this results from feeling unable to rely on staff to ensure their safety.

I.b. Review question: what impact does disturbed/violent behaviour in psychiatric inpatient settings have on staff and or service users?

Using mostly interviews and/or questionnaires one study (Wykes and Whittington 1998), used a case-control design; one study (Cheung et al. 1997), used a cross-sectional approach; eight studies were identified which sought to examined the impact of violence on staff. None were found that looked at the general impact of violence on service users.

Two studies looked specifically at the impact of physical assault. Poster and Ryan (1989) tracked staff responses to physical assault over the course of a year. Although the authors noted that 82 per cent of nurses had resolved the crisis by week six, they also noted that 21 per cent of staff met responder criteria six months after the event, and 16 per cent met responder criteria one year after the event. The authors argue that there is a need to support staff to help them cope with their responses to physical assault. Omérov (2002) used interviews to assess the impact of physical attack and found that 43 per cent of staff felt insulted by the attack and one-third of staff felt angry. Men were more likely to be frightened (p<0.05) and women were more likely to feel surprised (p<0.01) regardless of the outcome of the assault. Most staff felt very uncomfortable after the assault, brought the incident home, found it hard to relax, had frequent nightmares and found returning to work difficult. All but one staff member would have welcomed self-defence training and refresher courses. The majority of staff requested some kind of post-incident debriefing. Interviews were performed three days after an incident. There was no long-term follow-up.

Wykes and Whittington (1998) noted a significant difference between non-assaulted and assaulted nurses in terms of psychological distress in relation to the general health questionnaire anxiety scale. Participants were assessed twice – once within 10 days of the incident and once approximately four weeks later. Comparisons were made with baseline scores in a control group. There was a decrease in distress levels between the first time and the second. However, two new victims met the diagnosis criteria for post-traumatic stress disorder (PTSD) at the second time, according to the post-traumatic stress scale (PTSS).

In debriefing procedures, it was noted (Flannery et al. 1995) that staff who had been verbally assaulted had similar PTSD-like symptoms and disruption in mastery and meaning similar to those who had suffered physical or sexual assault. Time between debriefing and incident is not specified. There was no long-term follow-up.

Using SOAS, Cheung et al. (1997) noted that one-third of staff were emotionally shaken by the incidents they had been involved in, even though the rate of injuries requiring treatment was low.

Gillig et al. (1998) noted that 18 per cent of the staff they interviewed were considering changing their careers because of the emotional impact of violence/aggression.

In a questionnaire survey of Swedish and UK nurses (Nolan et al. 2001) it was found that less support was available for UK nurses following an incident (p<0.01), although they were significantly more likely to experience violence (p<001), sustain minor injuries and experience violence involving a weapon (p<0.05) than Swedish nurses. UK nurses reported lower self-esteem (p<0.05), and if they had experienced violence in the preceding 12 months were more likely than their Swedish counterparts to always find their jobs psychologically taxing (p<0.05). In the study overall, a significant positive correlation was found between self-esteem and feedback from line managers (p<0.05).

In a further questionnaire survey of Swedish nurses (Soares et al. 2000), it was noted that victims of violence were more likely to be less satisfied with their salary (p<0.05), complain of insufficient lighting and poor ventilation (p<0.001) complain about noise (p<0.001), find their psychological environment taxing (p<0.005), report that their work site was unpleasant (p<0.005), feel restless (p<0.05), feel less proud of their organisation (p<0.005), and state that they lacked resources (p<0.005).

I.c. Review question: what are staff and service users' attitudes towards the short-term management of disturbed/violent behaviour?

Eight studies attempted to elicit nurses' attitudes toward the management of disturbed/violent incidents.

Support/control emerged as a major theme in one survey (Lowe et al, 2002). Junior nurses were more likely to place an emphasis on limit setting and controlling strategies than senior nurses. Roper and Anderson (1991) conducted an ethnographic study on an in-patient emergency psychiatric unit to explore the variables underlying service user/staff interactions that might lead to violent incidents. Staff control emerged as a key theme, along with staff tension, helplessness/hopelessness, and counter-transference.

In a phenomological study, Cutcliffe (1999) noted a relationship between a nurse's ability to deal with an incident in a manner that promoted a therapeutic outcome and the nurse feeling supported in their work. Whilst there may not be a causal relationship between these two outcomes, this finding suggests the importance to nurses of being able to maintain a therapeutic relationship with service users.

Using semi-structured interviews, Spokes et al. (2002) found that nurses identified three key areas related to violence management: their clinical skills, personal characteristics (such as an ability to remain calm), and interpersonal skills.

Employing causal modelling, Morrison (1993) noted that psychiatric nurses disagree amongst themselves over how to define the seriousness of an incident.

Cutcliffe (1998) also noted that the decision to report an incident as violent depended on the therapeutic relationship between nurse and service user. Using unstructured interviews, Critchon (1997) noted that nurses' management strategies were dependent on the diagnosis and gender of the service users, as well as the seriousness of the aggressive action. For example, seclusion was felt more appropriate for male service users.

Again using unstructured interviews, this time with video vignettes, Critchon et al. (1998) noted that Canadian nurses tended to advocate more controlling measures – like PRN medication and seclusion – whilst UK nurses tended to opt for less controlling techniques – like de-escalation. UK nurses were also more likely to talk to the service user about what had happened.

Using a phenomenological approach, Carlsson et al. (2000) identified seven themes underlying nurses management strategies: respecting one's fear, respecting the client, touch, dialogue, situated knowledge, stability, mutual regard, and pliability.

Five studies examined service user perspectives of violence in psychiatric in-patient settings. Three studies used questionnaires; one study (Kumar and Ng 2001) conducted a focus group; and one study (Lancee 1995) used role-play scenarios.

One survey (Svensson and Hansson 1994) assessed the effect of personality traits, diagnosis and perceived coercion on service users' satisfaction with psychiatric in-patient settings. It was noted that service users with a higher level of ‘trait aggressive nonconformity’ were significantly less satisfied with the ward's physical and psychosocial environment (p<0.05), the treatment design (p<0.05) and the treatment programme (p<0.05). Service users with a higher level of trait sociability were more satisfied with the treatment programme (p<0.05). Service users with affective disorders had significantly better satisfaction than service users with schizophrenia concerning: information and influence (p=0.004), ward environment (p=0.005) and general satisfaction (p=0.003). Service users who were involuntarily admitted were less satisfied with care in the areas of staff-patient relationship, ward environment, treatment programme and general satisfaction (p<0.001). A significant two-way interaction was detected between perceived coercion and the personality trait, aggressive nonconformity (p=0.05). Service users who perceived improvement in their condition had higher satisfaction with ward environment (p<0.01), treatment design (p<0.01), treatment programme (p<0.001) and general satisfaction (p<0.01). The phenomenon of acquiescence was not related to reported levels of satisfaction. The authors comment that careful consideration needs to be given to how to collect satisfaction scores from service users with schizophrenia or who perceived coercion in connection with their treatment.

Another survey (Gillig et al. 1998) noted that service users reported more depression and worry (p<0.05) and a change in appetite (p<0.05) than staff as a result of violence in psychiatric in-patient settings.

A further survey (Thomas et al. 1995) investigated service user reactions to being assaulted. They noted that female service users were more likely than their male counterparts to feel happy with staff responses to an incident (39 per cent vs. 23 per cent). However, they noted that women were less likely than men to feel safe on the wards (57 per cent vs. 81 per cent).

Six service users took part in a focus group (Kumar 2001) to discuss the experiences of being either perpetrators, victims or witnesses of violence. Several members fell into several or all of these categories. Six overarching themes were identified: firstly, that an imbalance of power exists in the mental health system; secondly, that violence has psychological sequelae; thirdly, that the mental health service is not geared to help victims of ‘institutional violence’; fourthly, that the present mental health system fosters violence; fifthly, that a radical change is needed in the infrastructure of the mental health system; and sixthly, that reinforcement and reform may come from parallel efforts by staff and service users. Although acknowledging that the results may not be generalisable to a wider population, the authors argue that information saturation was achieved.

Role-play was used (Lancee 1995) to assess service user responses to different limit setting styles. Ninety-six service users participated, with limit setting styles ranging from belittlement to affective involvement with options. Service user anger at a particular limit setting style was the primary outcome variable. Three independent variables were considered: limit setting style, impulsivity and diagnosis. All proved significant (limit setting p<0.001, impulsivity p<0.001, diagnosis, p<0.05). The interaction between diagnosis and style had a greater significant (p<0.01). For all diagnostic groups, belittlement was most likely to cause anger. Impulsive service users were more likely to respond with anger to all limit setting styles than non-impulsive service users; non impulsive users had low anger for three limit setting styles – solution with options, affective involvement without options, and affective involvement with options. Service users with high impulsivity only responded with low anger to affective involvement with options. The same was also true of service users with schizophrenia. The sample size was too small to make other diagnosis specific observations. The authors argue that the study confirms that interpersonal factors play an important role in the management of anger in adult psychiatric inpatient settings.

Lanza et al. (1994) used interviews to compare staff and service user recollection of a violent incident. She found that with regard to ‘objective’ measures – such as limit setting and service users' actions during assault – there was general agreement between staff and service users. However, with regard to ‘subjective’ measures – such as the relationship between staff and service users, content of service users' speech, loudness of speech, number of staff and service users involved, and the cause of the incident – there was much less agreement.

Evidence statements

Level of evidenceEvidence statement
Level 4The limited evidence suggests that staff and service users believe that building therapeutic relationships, in which service users feel respected, leads to less disturbed/violent incidents. Key areas to be addressed in building such relationships include limit setting, and imbalance of power.
Level 4The limited evidence suggests that service users are adversely affected by in-patient disturbed/violent behaviour.
7.8.4.1.5. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.4.2. Minority ethnic groups

7.8.4.2.1. Objectives

No specific searches on minority ethnic groups were undertaken in the RCPsych guideline.

Current guideline

Two review questions were identified and used to inform all searches (see Appendix 4 for search strategies, databases searched and search logs).

  • Does race/ethnicity of a service user or staff member make a difference to how they are treated when they are involved in a disturbed/violent incident in adult inpatient settings?
  • Do staff and/or service users perceive that the race/ethnicity of a service user or staff member makes a difference to how they are treated when they are involved in a disturbed/violent incident in adult psychiatric in-patient settings?
7.8.4.2.2. Selection criteria
Types of studies

Systematic reviews to before and after studies. Qualitative studies were also included. (Evidence level 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
7.8.4.2.3. Clinical evidence

One hundred and sixty eight papers were identified by our searches. After sifting for duplicates and papers outside the scope, 41 were ordered. Only 23 of these papers were included. Ten were excluded. The rest were overviews or outside the scope of the review. There were 13 UK studies and 10 US studies. Some of the US studies are based in psychiatric services for veterans (ex-military), a specialised population. Study settings varied from general acute psychiatric to specialist services – forensic or psychiatric intensive care. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

To supplement the evidence base for this review we also conducted three focus groups, two with black service users and one with health care professionals with expertise in working with black service users (see Appendix 14).

Included papers covered three broad areas that fall within the scope of the guideline: prediction, interventions and admission. Special review questions were devised to focus the review in each of these areas.

I. Prediction

Seventeen studies were identified which addressed these questions. A range of study designs and perspectives were examined, making the results difficult to synthesis.

I.a. Review question: can disturbed/violent behaviour in psychiatric in-patient settings be linked to ethnicity?

The following studies suggested that black and minority ethnic service users exhibited higher levels of violence toward others than white service users: Dixon (2000); Commander et al. (1997a), (1997b); Sheehan et al. (1995); Lloyd and Moodley (1992); Chen et al. (1991); Chu, (1985); Lawson et al. (1984). Six of these were UK studies Dixon et al. (2000); Commander et al. (1997a); Commander et al. (1997b); Sheehan et al. 1995, Lloyd and Moodley 1992; Chen et al. (1991); and two were US studies (Chu, 1985; Lawson 1984). Four were prospective studies: Commander et al. (1997a); Commander et al. (1997b); Chu (1985); Lawson et al. (1984); two retrospective chart reviews (Dixon 2000; Sheehan et al. 1995); one cross sectional (Lloyd and Moodley 1992) and one case-control (Chen 1991).

The following studies found that levels of violence towards others were not related to ethnicity: Kho et al. (1998), a UK prospective study; and Feinstein and Holloway (2002), a UK cross sectional study. In addition, a qualitative UK study by Morley et al. (1991) found that 53 per cent of service users who were sectioned were not considered dangerous by their relatives.

The following studies suggested that other ethnic groups exhibited higher levels of violence toward others than black and minority ethnic service users: Kho et al. (1998) showed Asian patients to be more aggressive. Lawson et al. (1984) showed whites to be more violent, to make more threats and to commit more self-destructive acts.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence from these studies is conflicting; it is therefore not possible to ascertain if different cultural groups exhibit higher or lower levels of disturbed/violent behaviour than other groups.
I.b. Review question: are the tools used to predict disturbed/violent behaviour in psychiatric inpatient settings ethnically/racially biased?

A large number of tools were identified in the prediction evidence review. The majority of these make no mention of testing for racial bias. Therefore, it must be presumed that they have not been tested for racial bias. This is the case for the following tools which were found to indicate that black service users were more likely to be violent than white service users: Chu (1985) using the brief psychiatric rating scale and the Itil-Keskiner psychopathology rating scale.

Hutton et al. (1992) found that the overt hostility scale tended to suggest a greater propensity for aggressive or violent acts amongst black service users than occurred amongst white service users, and could lead to an erroneous interpretation as race was the only variable to emerge as a determinant of over hostility.

Choca et al. (1990) tested the cultural sensitivity of the Millon clinical multiaxial inventory to assess whether it was culturally fair. This personality instrument has weighted scores to provide different norms for black, white and Hispanic individuals to address potential bias. This study concluded that this test was a useful tool for prediction that takes account of racial bias, however some adjustment is needed to the item and scale levels.

Clinician prediction was also found to be at risk of racial bias: Hoptman et al. (1999) a US prospective study; McNiel and Binder (1995) a US retrospective chart review; and Strakowski et al. (1995) a US retrospective chart review. Minnis et al. (2001) surveyed British psychiatrists to test assessment bias in relation to violence. They suggested that racial stereotyping did not occur at first interview. Silver (2000) illustrates the effect of confounding according to the locality of the individual's residence and how this may effect reporting of results of violent incidents. Reubin et al. (1997) suggested that elevated levels of the enzyme creatine kinase can be used as a biological marker to predict aggression amongst African Americans. This finding could not verified from any other study.

Evidence statement

Level of evidenceEvidence statement
Level 4On the basis of the available evidence, it is not possible to determine a ‘gold standard’ tool for the prediction of disturbed/violent behaviour appropriate for use amongst different ethnic groups.
II. Interventions
II.a. Review question: is intervention choice for the short-term management of disturbed/violent behaviour ethnically/racially biased?

One study specifically addressed this question.

Chen et al. (1991) found a significantly higher number of African Caribbean service users were given high dose neuroleptic medication for disturbed/violent behaviour than service users from other ethnic backgrounds (p<0.03).

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence (one study) to assess whether African Caribbean service users are given rapid tranquillisation more often than service users from other ethnic backgrounds.
II.b. Review question: do staff and/or service users perceive that the race/ethnicity of a service user or staff member makes a difference to how they are treated when they are involved in a disturbed/violent incident in adult psychiatric in-patient settings?

Three studies examined attitudes of service users towards violence management in psychiatric in-patient settings in relation to ethnicity.

A qualitative UK study (Secker and Harding 2002), proposed key themes arising from interviews with African Caribbean service users relating to loss of control, experiences of racism and relationships with staff. Relationships with staff are very rarely experienced as positive.

A prospective UK study (Commander et al. 1997a), found that Asian and white service users are significantly more satisfied with in-patient treatment than black service users.

A UK descriptive survey (Wilson and Francis 1997), found that African Caribbean service users and African service users felt misunderstood as a consequence of being feared, ignored or stereotyped.

The two focus groups that the NCC-NSC commissioned from black service user organisations found that black service users perceived that they were given more restrictive interventions because of their race/ethnicity (see Appendix 14).

No studies were identified that examined staff perspectives on race/ethnicity in relation to the use of the interventions considered in this guideline for the short-term management of disturbed/violent behaviours in psychiatric in-patient settings.

The focus group which the NCC-NSC ran with nine health care professional who had experience of working with black and minority ethnic service users found that these health care professionals felt that the short-term management of disturbed/violent behaviour in the UK is racially/ethnically biased (see Appendix 14).

Evidence statements

Level of evidenceEvidence statement
Level 4The limited evidence base suggests that black/ethnic service users perceive that there is racial/ethnic bias in staff choice of intervention for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings. Staff-service user relationships, and feelings of being stereotyped, ignored and afraid, are key areas of concerns for this group.
Level 4Limited evidence from a focus group suggest that staff perceive that there is racial/ethnic bias in staff choice of intervention for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings.
III. Admission
III.a. Review question: are admission procedures ethnically/racially biased?

Commander et al. (1997a) mapped the pathways to admission for three ethnic groups (black, white and Asian). This study found that black service users were less likely to be receiving care from a health care professional prior to admission and that two-thirds of admissions involved the police.

Involvement of the police was examined in two studies, both from the US. Morley (1991) identified the role of police in admissions to hospital for African Caribbean service users experiencing psychotic symptoms. Commander et al. (1997a) noted that two-thirds of African Caribbean service user admissions involved the police and that the admission of Asian service users also had a higher level of police involvement than the admission of white service users. As both are US studies, it is difficult to generalise from them to the UK population. However, the two focus groups that the NCC-NSC ran with black service users also found that police involvement was often mentioned in connection with admission (see Appendix 14). Again, it is not possible to generalise on the basis of this small study to the UK population in general.

Evidence statement

Level of evidenceEvidence statement
Level 4Limited evidence suggests that black service users may be likely to have experienced police involvement during the admission process.
7.8.4.2.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.4.3. Gender

7.8.4.3.1. Objectives

No specific searches on gender were undertaken in the RCPsych guideline.

Current guideline

Two review questions were identified and used to inform all searches (see Appendix 4 for search strategies, databases searched and search logs).

7.8.4.3.2. Selection criteria
Types of studies

Systematic reviews to before and after studies. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcomes
7.8.4.3.3. Search strategy

Searches were run from 1998-2003/6, to capture current legislation, attitudes and organisation of care.

7.8.4.3.4. Clinical evidence

Three hundred and seventeen studies were identified in the initial sift. After sifting for relevance and duplicates, 20 full papers were ordered. Three met the inclusion criteria and 14 were excluded. All the other papers were opinion pieces, anecdotal reports, or fell outside the inclusion criteria for this review. References were checked but no further studies were identified. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Included studies
I. Review question: what impact does gender have on the short-term management of disturbed/violent behaviour in psychiatric inpatient settings?

One study was included that considered the gender differences among perpetrators of violent assaults resulting in injury to staff: a case-control study in the US (Lam et al. 2000). This study showed no difference in the proportion of male and female psychiatric in-patients perpetrating such violence (20 per cent of male patients vs. 18 per cent of female patients).

No studies were found that answered the question of whether male and female perpetrators of violence in the in-patient psychiatric setting were treated differently.

One small cross-sectional survey of 59 psychiatric inpatients (31 males, 28 females, representing only 39 per cent of eligible patients) was included that considered the different experiences of male and female patients who were potential or actual victims of other patients (Thomas et al. 1995). A similar proportion of the men and women reported harassment (physical or verbal or sexual) by other patients (68 per cent of males and 75 per cent of females) or having been hit (42 per cent of males and 36 per cent of females). More women were molested sexually (32 per cent of females and 7 per cent of males, p=0.01) and fewer females felt safe on the wards (57 per cent vs. 81 per cent of males, p=0.05). While many incidents were not reported to staff, more females were satisfied with the staff response when they did report an incident (25 per cent vs. 7 per cent of males, p=0.05). However, the small and possibly unrepresentative sample precludes generalisation from this data.

One case-control study from the US, involving more than 200 staff over a period of 2.5 years, examined whether the gender of staff was a factor in the risk of being assaulted by a psychiatric in-patient (Binder and McNiel 1994). In this study, staff gender was not associated with the risk of being assaulted for doctors, nurses or both disciplines together, but nurses were more likely to be assaulted than doctors.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that the gender of staff or service users does not impact on the incidence of disturbed/violent behaviour in psychiatric in-patient settings.
II. Review question: what are staff and service users' perspectives on whether gender has an impact on the short-term management of disturbed/violent in in-patient psychiatric settings?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence to determine staff and service user perspectives on the impact or influence of gender on the short-term management of disturbed/violent behaviour in psychiatric in-patient settings.
7.8.4.3.5. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.4.4. Other special concerns

7.8.4.4.1. Objectives

No specific searches on other special concerns were undertaken in the RCPsych guideline.

Current guideline

Two review questions were identified and used to inform all searches (see Appendix 4 for search strategies, databases searched and search logs).

  • What special considerations are needed in the short-term management of disturbed/violent behaviour where the service user has physical disabilities?
  • What are the staff and service users' perspectives of the considerations needed for the short-term management of disturbed/violent behaviour where the service user has physical disabilities?
7.8.4.4.2. Selection criteria
Types of studies

Systematic reviews to before and after studies. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
7.8.4.4.3. Search strategies

Searches were run from 1998-2003/6, to capture current legislation, attitudes and organisation of care.

7.8.4.4.4. Clinical evidence

Nine papers were found in our searches. However all were excluded as none of them addressed the review questions.

Evidence statements

Level of evidenceEvidence statement
Level 4There is no evidence that identifies the special considerations that are needed in relation to the short-term management of disturbed/violent behaviour where the service user has physical disabilities.
Level 4There is no evidence to determine staff and service users' perspectives on what special considerations are required in relation to the short-term management of disturbed/violent behaviour where the service user has physical disabilities.
7.8.4.4.5. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.5. Psychosocial interventions

Original RCPsych guideline

The original RCPsych evidence review covering all psychological interventions, including both de-escalation techniques and observation, was examined. Their searches were undertaken based on the following review questions and hypotheses:

RCPsych review questions

  • Can psychological interventions have the effects of reducing aggressive behaviour?
  • Are particular psychological interventions more effective in reducing aggressive behaviour?

RCPsych hypotheses

  • That psychological interventions can have the effect of reducing levels of disturbed/violent behaviour.

RCPsych sub-hypotheses

  • That psychological interventions have no effect in reducing levels of aggressive behaviour.
  • That particular psychological interventions are more effective in reducing levels of aggressive behaviour.
  • That psychological interventions are similar in terms of reducing aggressive behaviour.

After sifting and quality checks, only eight papers relating to psychological interventions were included by the RCPsych reviewer.

The reviewer indicated that no evidence had been found on which evidence-based recommendations could be made:

In conclusion, I found it impossible to answer our original hypotheses. We had no good evidence to support any of our original hypotheses. […] We are unable to comment on whether any intervention is more effective that any other in reducing levels of aggression (RCPsych unpublished evidence review).

7.8.5.1. De-escalation techniques

7.8.5.1.1. Objectives
Current guideline

The current guideline focuses more specifically on particular psychological interventions – that is, de-escalation and observation. Two review questions were identified and used to inform the search strategy (see Appendix 4 for search strategy, databases searched and search logs).

Review questions
7.8.5.1.2. Selection criteria
Types of studies

Systematic reviews through to before and after designs. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
7.8.5.1.3. Clinical evidence

One hundred and ten studies were identified in the initial sift. After sifting for relevance and duplicates, 10 full papers were ordered. Seven were opinion pieces, anecdotal reports, or fell outside the inclusion criteria for this review. References were checked and four further studies were identified and ordered. However, only four studies were primary research papers, three of which proved relevant to the research question. No study offered evidence above level III. No additional studies from the RCPsych review were included. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Included studies

A prospective observation study (Jambunathan and Bellaire 1996) attempted to evaluate the effectiveness of crisis prevention (CPI) techniques in preventing the need for mechanical restraint and seclusion. Techniques were assigned levels and these were linked with a stage of escalation (see table below). Ten registered nurses prospectively collected data in four-hour shifts. All were trained in CPI on 12-hour initial training and four-hour refresher course. The study evaluated aggressive incidents including a wide cross-section of psychiatric patients in a state-run in-patient psychiatric facility. The study suggests that CPI techniques allow most conflict situations (84.2 per cent) to be resolved without the need for mechanical restraint or seclusion. Most service users' behaviour cues (76.6 per cent) and most staff interventions (69.4 per cent) occurred at level 2. However, there appeared to be a lack of intervention at level 1. It was also noted that more than 50 per cent of the incidents occurred on admission units. More medication was administered where staff did not have an in-depth knowledge of the service users.

Table 3Outline of CPI techniques

Escalation level and associated behaviourLEVEL 1: Anxiety
(Change or increase in behaviours such as crying, pacing, rocking, wringing hands, raising voice)
LEVEL 2: Defensive
(Begins to loose rationality, becomes verbal, yelling, belligerent, sarcastic, intimidates, uses verbal threats, shakes fists)
LEVEL 3: Acting out
(Loses control, physical episode)
LEVEL 4: Tension reduction
(Regains control)
CPI techniques usedSupportive, emphatic, active listening, asking questions, discussing thoughts/feelings, reducing stimuli, refocusing tasks, offering medication as needed.Limit setting, allow verbal release, isolate situation, assemble a team, planning for de-escalation or physical control.Physically holding patient, escorting patient to safe area, mechanical restraints or seclusion if CPI unsuccessful.Attempt to regain therapeutic rapport, coping mechanisms, contracting.

(N.B. Most CPI techniques were verbal de-escalation techniques, however physical restraint is included in the third stage as a CPI technique.)

While the study is reasonably designed, it is a non-experimental pilot study, observers were not blinded to either staff or service users, and staff were informed that observations would be carried out prior to the study. The results also do not allow the different service user groups to be analysed independently. No supporting evidence is offered for selection of antecedents of violence, or for their division into four levels of escalation. Supporting evidence is also lacking with regard to the relationship between levels of escalation and CPI techniques.

A before and after study at a veterans' medical centre, (Richmond et al. 1996) also measured whether the implementation of de-escalation techniques reduced the use of restraint and seclusion – unlike Jambunathan and Bellaire (1996), they did not treat physical restraint as a de-escalation technique. They suggested that training in verbal de-escalation, ‘time out’, relaxation techniques, medication, diversional activities and decreased stimulation led to a 47 per cent decrease in restraint use and a 31 per cent decrease in the use of seclusion. Whilst these results seem promising, the study design is non-experimental and confounders are not explored.

Staff perspectives

A qualitative study (Johnson and Hauser 2001) used unstructured interviews to elicit nurses' views on how to de-escalate the escalating service user. The author reported that expert nurses were able to develop an awareness of where service users are on the continuum of escalation, noticing early behavioural and verbal signs, which allowed them to successfully implement de-escalation techniques. However, the sample size was very small, and the method non-experimental.

Service user perspectives

A triangulation study using incident forms, questionnaires and interviews, (Duxbury 2002), noted that service users were not aware of staff using de-escalation techniques (p<0.000).

The RCPsych review did not include any studies that evaluated de-escalation techniques. Therefore the findings of this review alter those of the RCPsych psychosocial interventions review, although the evidence presented above is limited.

Evidence statement

Level of evidenceEvidence statement
Level 4The limited evidence suggests that de-escalation techniques decrease rates of disturbed/violent behaviour.
7.8.5.1.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.5.2. Observation

7.8.5.2.1. Objectives
Current guideline

Two review questions were identified and used to inform all searches (see Appendix 4 for search strategy, databases searched and search logs).

Review questions
7.8.5.2.2. Selection criteria
Types of studies

Systematic reviews through to before and after designs. Qualitative studies were also included. (Evidence Levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
  • The effectiveness of observation techniques at decreasing the number of disturbed/violent incidents and potentially violent incidents, without the use of other interventions.
  • Staff and service user perspectives on observation techniques.
Types of outcome excluded
7.8.5.2.3. Clinical evidence

Seventy-five studies were identified in the initial sift. After sifting for relevance and duplicates 22 full papers were ordered. However, 12 were opinion pieces, anecdotal reports, or fell outside the inclusion criteria for this review.

Fourteen studies were primary research, however, only nine studies proved relevant to the research question. No study offers evidence above level III. References were checked for missing articles but no further relevant primary studies were identified. No additional studies from the RCPsych review were included. (Evidence tables of included studies are found in Appendix 5. Evidence tables of excluded tables are found in Appendix 6).

Appraisal of methodological quality

In addition to the quality concerns mentioned above, these studies raised the following methodological concerns:

These studies were included to provide a systematic review of the research that has been conducted on observation in psychiatric in-patient settings. Gaps in the research can be readily identified and low graded evidence statements were presented to assist the GDG in their deliberations.

Included studies

Several statewide surveys to establish the reason why constant observation (CO) was used have been conducted in the US. Torkelson and Dobal (1999) carried out a six-month statewide survey focusing primarily on surgical and medical units and their use of CO. Stratified randomisation was used to select hospitals. Authors found that the decision to initiate and discontinue CO could be made by a wide spectrum of people (clinician, nurse, family member). The most common reason for CO was either danger to self or others. Although 84/89 hospitals agreed to participate, very little information on cost was provided. The results are difficult to analyse from the perspective of violence to others, since this is not differentiated from violence to self. The analysis is also flawed. No firm conclusions of the effectiveness of CO are offered.

Moore et al. (1995) also undertook a statewide survey. Hospitals were selected by stratified randomisation; 19/26 agreed to participate, however only 15 made use of constant observation (CO). Of these, only six were psychiatric hospitals. Again a wide spectrum of people made the decision to initiate CO. Those observing requested more training and information. There was a lack of information on costs. No attempt was made to differentiate between different interpretations of CO. The hospitals that used CO to combat violence are not specified.

Bowers et al. (2000) carried out a random stratified sample survey of constant observation (CO) policies in England and Wales. There was no consistency amongst trusts (see Evidence table for full details). Of 26 policies supplied, only two used the same terminology – constant observation had different meanings in different locations; level 1 meant either high or low levels of observation. Differences also existed between official policy and questionnaire responses. It was noted that this is particularly worrying, as agency staff are often used to carry out CO. The report does not discuss the effectiveness of CO.

Shugar and Rehluk (1990) conducted a retrospective controlled cohort study to consider the effectiveness of close observation (CO). Carried out in a psychiatric teaching unit, this study examined the use of CO in both civil and forensic patients. Various predictors signalling the need for CO were identified. However, most patients had CO supplemented by medication, so that it is difficult to assess the efficacy of CO. The authors acknowledge this and therefore only offer tentative conclusions. The authors suggest that CO should only be used as a short-term measure but offer no evidence. While an interesting study, the design is weak and the conclusions may therefore be limited.

Philips et al. (1977b) used a retrospective/historic two-year cross-sectional survey to assess whether a correlation existed between involvement in constant observation (CO) and absenteeism. A statistically positive correlation (p<0.05) was noted. Discrepancies were explained in terms of reduced staffing levels forcing a reduction in CO, and high demand for CO obligating nurses not to take sick leave. There is a lack of essential information in this study. The conclusions drawn should be interpreted with caution.

Philips et al. (1977a) identified the type of service users who usually receive continuous observation (CO). Using a retrospective 10-year cohort study, they found that service users receiving CO were most likely to be female and suffering from either schizophrenia or depression. (CO was used for suicide risk for both types of service users, and for behavioural reasons with service users suffering from schizophrenia). The age range for service users with depression was between 30-50, while for schizophrenia, it was between 15-29 years and 35-40 years. Staff concerns about CO related to the length of time an individual nurse was engaged in CO (entire shift) and the effect of CO on other service users and staff within the ward. Seventy-five percent were in favour of a special unit for CO, with 45 per cent suggesting they would be prepared to work there full-time, 34 per cent sometimes and 21 per cent never. The author stresses that more research is necessary to elicit the therapeutic value of CO – a procedure that is identified as cost effective but time consuming.

Bowles and Dodds (2001) report the effect of dismantling the formal observation policy in a 21-bedded acute ward in Bradford. They argue that formal observation became redundant and, after 18 months, one-to-one observation was not used at all, with five to 10 minute checks used only rarely. The number of suicides did not increase, but the levels of absconding were almost halved, with self-harm falling by two-thirds and violence and aggression by a third. Staff sickness was also reduced by two-thirds. Removal of the policy has also meant a saving of £45,000 over 12 months. They state that service users are now more involved in their care and in ward decisions. The authors argue that the removal of this policy – which they describe as an ‘outmoded ritual of mental health nursing’ – has freed up nurses' time, allowing activities to be set up and time to be ‘gifted’ to service users as required. Ninety-five percent of service users now receive daily one-to-one time with a nurse, which the authors argue is the most valuable intervention. While the authors acknowledge that the study is too small for the results to be generalised, they insist that it should bring the practice of formal observation into question. This study does not provide enough information about their previous formal observation policy and so is open to a number of interpretations. The one-to-one interventions implemented, once formal observation was dismantled, could be viewed as a more therapeutic and appropriate form of formal observation.

A three-and-a-half month prospective audit was conducted in a psychiatric intensive care unit (Lehane and Rees 1996). It examined responses to incidents that would have formerly led to seclusion. The author notes that one-to-one nursing was used in 86 per cent of cases but does not offer any information on its effectiveness. The sample size is relatively small.

Service user perspectives

One study examined service user perspectives on observation. Jones et al. (2000) conducted a three-month survey in one mental health trust to assess service users' feelings about and preferences within constant and close observation – the highest level out of four levels employed within the trust. The study revealed that mental health service users – including those who exhibited aggressive behaviour, but particularly those with suicidal tendencies – preferred to be observed and felt safest when observed by either nurses they knew or nurses who talked to them.

Staff perspectives

One study considered staff perspectives on observation. Neilson and Brennan (2001) carried out a retrospective/historic audit to determine staff knowledge of and attitudes toward a new hospital special observation (SO) policy and differences between wards with respect to these two variables. This was assessed by a knowledge questionnaire, semi-structured interviews and a score-schedule of a randomised sample of 144 special observation record sheets (SORS). The hospital policy had four levels of SO – red, amber, blue and green – in order of decreasing urgency. Nurses were purposely selected to ensure a broad mix of trained and untrained staff. All staff demonstrated good knowledge of the policy. However, although 35.29 per cent stated that communication and documentation had improved since its implementation, authorising signatures and reasons for SO were often not stated. Staff also felt that decisions about SO were too medically driven (94.2 per cent) and that there was poor medical review of SO (32.36 per cent). In addition, 82.4 per cent felt that blue level was used too frequently, without clinical assessment of need. They also stated that it was impossible for staffing levels to meet current demands of SO (73.6 per cent); 29.41 per cent felt that red level could provoke disturbed patients and 23.6 per cent felt that gender needed greater consideration when allocating staff to SO.

Yonge and Stewin (1992) conducted qualitative research using ‘ethnography’ – a programme for textual analysis. Findings suggested that close observation (CO) is a procedure that nurses find stressful. Nurses felt that they were also on CO and had to find ways of dealing with emotions caused by this encounter. Meal times and bathroom visits were flagged up as particularly stressful for nurses. None of the nurses interviewed accompanied the patient into the bathroom, even where this was in breach of hospital policy. Nurses also supported one another in handover, attempting to limit the repetitive questions for the patient. At the same time, some saw CO as an opportunity to develop a quality relationship with the patient. Nurses expressed different preferences for certain types of CO patients – psychotic, depressed.

The RCPsych review did not include any studies that evaluated observation. Therefore the findings of this review alter the findings of the RCPsych psychosocial interventions review, although the evidence presented above is limited.

Evidence statements

Level of evidenceEvidence statement
Level 4It is not possible to ascertain the effectiveness of observation on the basis of the available evidence.
Level 4The limited evidence suggests that service users prefer to be observed by a nurse that they know and that most staff find observation a stressful procedure.
7.8.5.2.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.6. Other interventions

7.8.6.1. Physical interventions and seclusion

7.8.6.1.1. Objectives

The original RCPsych guideline evidence base on restraint and seclusion was examined. A list of excluded studies was available in the archived information received from the Royal College of Psychiatrists' Research Institute. The following information was taken from the final report in the RCPsych guideline, which states that these hypotheses were used to inform their search strategies:

RCPsych hypothesis
  • Restraint when skilfully applied by trained and supervised staff, according to monitored protocols and the context of other methods, is an effective and safe means of coping with overtly violent behaviour.
  • When properly used and explained, restraint can be acceptable both to users of services and to staff.
  • Seclusion is unnecessary if restraint is properly applied in association with other methods of good practice.

After sifting and quality checks, 16 references on restraint and seclusion were included in the RCPsych evidence review.

However, the included studies did not offer generalisable criteria in support of these hypothesis so the RCPsych review concluded that:

No strongly evidence-based conclusions can be drawn from the quantitative evidence.

Current guideline

Three review questions were identified and used to inform all searches (see Appendix 4 for search strategies, databases searched and search logs). Physical intervention includes the use of pain compliance.

Review questions
7.8.6.1.2. Selection criteria
Types of study

Systematic reviews through to before and after designs. Qualitative studies were also included. (Evidence levels 1-2).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
7.8.6.1.3. Clinical evidence

One hundred and thirty three studies were identified in the initial sift. After sifting for relevance and duplicates 84 full papers were ordered. After quality appraisal, 21 papers were included, seven papers were excluded and 60 were opinion pieces, anecdotal reports, or fell outside the inclusion criteria for this review. References were checked and 14 further studies were identified and ordered. None met the inclusion criteria. In addition, 10 studies from the RCPsych review were included. (Evidence tables of all included studies can be found in Appendix 5. Evidence tables of all excluded tables can be found in Appendix 6).

Quantitative evidence

A Cochrane review undertaken by Salias and Fenton (2001) focused on the effectiveness of restraint or seclusion or strategies designed to reduce the need for restraint or seclusion in the treatment of mental illness. It found no trials that met the minimum criteria. It concluded:

In the absence of any controlled trials in those with serious mental illness, no recommendation can be made about the effectiveness, benefit or harmfulness of seclusion or restraint. In view of data from non-randomised studies, use should be minimised for ethical reasons.

This Cochrane systematic review is currently being updated. Contact with the author suggests that the conclusions are unlikely to change.

Four studies in the RCPsych review considered the role of the seclusion room in a psychiatric in-patient setting (Brooks et al. 1994; Craig et al. 1989; Hafner et al. 1989; Kingdon and Bakewell 1988). Two of these studies suggest that use of seclusion rooms reduce violent incidents. However, one study (Kingdon and Bakewell 1988) suggests that violent incidents are better reduced by improved staffing patterns, education and management participation. The fourth study (Brooks et al. 1994) suggests that both levels of restraint and seclusion are increased by overcrowding.

No studies examined the use of pain compliance in physical interventions.

Evidence statements

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine the effectiveness and safety of either physical interventions or seclusion for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings.
Level 4There is insufficient evidence to determine the effectiveness or acceptability of pain compliance as a technique for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings.
Sudden death

Evidence from published case series that link physical interventions to adverse reactions was collated (see Evidence tables of included studies, Appendix 5). It was not possible to determine whether there was a relationship between physical interventions and an increased likelihood of sudden death.

Three studies were conducted in attempt to show the relationship between restraint and positional asphyxia. Parkes (2000), Schmidt and Snowden (1999) and Chan et al. (1997) conducted experimental studies on healthy subjects. All suggest that restraint in the prone position does not result in effects likely to cause death and that other factors need to be in situ.

This issue was not specifically addressed in the RCPsych review.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine the relationship between physical interventions and sudden death.
Staff and service user perspectives

Six studies examined staff perspectives on physical intervention and seclusion:

One study focused on staff decision-making processes. A sample of 64 nurses – a response rate of 77 per cent – were asked about knowledge and experience gained, as well as values and concerns with regard to the practice of seclusion. Key themes emerged around safety and the abuse of seclusion. However, in a further study, nurses stated that seclusion had a place in real world practice (Alty 1997).

Muir-Cochrane (1996) found that, in a sample of seven nurses, the core underlying themes relating to the use of seclusion were control, and that staff saw themselves as gatekeepers who maintain control.

Mason (1997) took a random sample of 25 nurses in a forensic hospital and found decision-making around the use of physical interventions and seclusion to be based upon the need to stick to an original decision. They found that a feeling of being under the gaze of others in authority led to the need to balance the responsibility of an untoward incident, against care of the individual. Furthermore, they found that nurses felt a need to justify their actions to those in authority and therefore tended to adopt positions of safety.

Marangos-Frost and Wells (2000), in an ethnographic study, interviewed six nurses on an unlocked psychiatric ward. They found themes consistent with other studies presented here, in particular concerning the decision-making dilemma of choosing between risking harm to the patient and others, or restraining – both equally unwelcome options. However, the findings were based on those that self-harmed, as well as those that harmed others, and the nurses were very experienced, therefore not representative of the usual ward staff team.

Holzworth and Wills (1999) found a sample of nine nurses in a short-term psychiatric hospital had a preference for seclusion over restraint. However, they noted that overall, there was inconsistency between nursing staff in selecting seclusion, restraint or observation. Lemonidou (2002) conducted a descriptive survey of nurses' attitudes and choice of restrictive intervention. The study involved 190 nurses in adult psychiatric inpatients in five hospitals in Greece. The findings suggest that nurses prefer seclusion to restraint. It was also noted that nearly half of the service users continue to be aggressive after restraints were removed. However, this study focused on the use of mechanical restraints.

The RCPsych review also contained two further studies (Tooke and Brown 1992; Soliday 1985), which compared staff and service user perceptions of seclusion. Service users who had been secluded had less favourable attitudes towards seclusion than either staff or service users who had not been secluded. Some service users saw seclusion as punitive. However, other service users saw a need for seclusion. In one of the studies, staff stated that they could not see how they could cope without access to a seclusion room.

Four studies examined service user perspectives on physical intervention and seclusion:

A relatively recent UK-based study carried out by Sequeria and Halstead (2002) with a sample of 14 in-patients who had been interviewed 12 hours after being restrained, found emergent themes of anger with a sense of injustice, and that service users felt that the intervention was unwarranted. The researchers also noted that anxiety continued long after the incident, along with mental upset. Contrary to this, the study also found that female patients restrained by female staff welcomed the safe feeling of containment, which even led to them seeking restraint.

A study by Gallop et al. (1999) highlighted concerns raised about the effects of restraint on those who may have previously suffered sexual abuse. They found that service users reported negative experiences of being rendered powerless and being degraded. In this study, six out of the 10 participants were restrained for self-harm. Therefore, care is required whilst extrapolating to this review. Nevertheless the study highlights the need to be aware of previous history when considering this intervention.

Bonner et al. (2002) conducted a pilot study in the UK, where he interviewed two members of staff and a service user involved for each incident that resulted in physical restraint. Initial findings suggested that failed communication is an antecedent of restraint. The study further suggested that restraint was used as a last resort to contain and support the patient. The study also suggests that both patient and staff can suffer trauma and distress after the incident and that support, post incident, is important to both groups.

Using a questionnaire survey (Mann et al. 1993), suggested a range of attitudes towards the seclusion room, with only a minority of service users suggesting that there should be no such room. The authors noted that many service users reported that the room was helpful. They noted that these tended to be service users who had no history of substance abuse (p<0.05). Conversely, service users with no history of substance abuse more often reported that the room was like a padded cell (p<0.05). Service users who used the room for the first time were more likely to report that it was stuffy (p<0.05) and to describe it as torture (p<0.05). Non-compliers were less likely to label the room safe and secure (p<0.05). Service users with disorders other than depressive disorders were more likely to report that once in the seclusion room, it is difficult to get out (p<0.05).

Three further studies in the RCPsych review (Wise et al. 1988; Hammill et al. 1986; Binder and McCoy 1983) surveyed service users who had experienced seclusion. Another study (Eriksson and Westrin 1995) surveyed service users about seclusion, rapid tranquillisation and restraint. These studies found a mixed response to these measures. Whilst service users sometimes saw a need for them, on many occasions they felt that they had been used unnecessarily.

None of these studies significantly change the findings of the RCPsych review.

Evidence statements

Level of evidenceEvidence statement
Level 4The available evidence suggests that staff may find using seclusion and physical interventions traumatic, but also consider that these interventions serve a necessary function.
Level 4The limited evidence suggests that service users may find seclusion and restraint degrading, although some service users believe that measures, such as seclusion and physical intervention, are sometimes justified.
7.8.6.1.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

7.8.6.2. Rapid tranquillisation

7.8.6.2.1. Objectives

The aim of this review is to examine the evidence on the efficacy and safety of medications currently used for rapid tranquillisation (i/v, i/m, oral (any form) in psychiatric inpatient settings). It builds on the original RCPsych guideline, the NICE schizophrenia guideline (2002) and a recent New Zealand health technology appraisal (2001) all of which also examined this intervention.

I. Original RCPsych guideline

The original RCPsych evidence review on rapid tranquillisation was examined. Their searches were undertaken based on the following review questions and hypotheses:

RCPsych review aim
  • To produce a systematic review of the use of medication in managing violent incidents in clinical settings in which mental health care is provided.
RCPsych hypotheses and sub-hypotheses
  • Is medication effective and safe in preventing and managing violent incidents?
  • What are the contra-indications to using medication to manage disturbed/violent behaviour?
  • Is the effectiveness of medication related to general sedation or to specific therapeutic effect on an underlying disorder?
  • Is medication effective in the management of disturbed/violent behaviour irrespective of the aetiology of disturbed/violent behaviour?
The RCPsych review inclusion criteria

Any controlled or non-controlled (including qualitative) research studies.

The RCPsych outcome measures:

  • The reduction of disturbed/violent behaviour
  • An increase of safety to patients and staff
  • Measure of therapeutic effectiveness
  • Measure of harmful effects of medication
  • Measure of the threshold of the use of medication.

Searches were made from 1986-1996 on Medline, Cinahl, Embase and Psychlit.

After sifting and quality checks, 15 references relating to rapid tranquillisation were included by the RCPsych reviewer. Of these, only six were controlled studies that were judged to be fair to good.

The reviewer indicated that no evidence had been found on which evidence-based recommendations could be made.

The review of the literature has shown that using strictly evidence-based criteria, no individual or combined psycho-pharmacological agent (s) is the definitive intervention during or just prior to an act of violence […] It is clear that more research needs to be carried out if the questions posed in our hypotheses are to be satisfactorily answered.

II. Schizophrenia guideline

The systematic review for the NICE schizophrenia guideline (2002) drew on Broadstock (2001) and the Cochrane review of zuclopenthixol acetate (Fenton M et al. 2001). They identified six phase III randomised controlled trials that addressed the issue of rapid tranquillisation in relation to schizophrenia. These have been added to the evidence review for this guideline.

III. New Zealand health technology appraisal (HTA) (M Broadstock 2001)

This systematic review identified 12 phase III randomised controlled trials that addressed the issue of rapid tranquillisation. These have been added to the evidence review for this guideline. One of these studies (Thomas et al. 1992) was excluded from the current review, as most of the participants had a primary diagnosis of intoxication and therefore falls outside the population considered in this guideline. Another study (Salzman et al. 1991) was excluded, as it did not appear to be a randomised controlled trial.

Current guideline

The current guideline focuses on the efficacy of rapid tranquillisation as an intervention for the short-term management of disturbed/violent behaviour. However, unlike the RCPsych review, dual diagnosis is not included in the scope. This has meant that some papers included by the RCPsych and the New Zealand HTA are excluded from this review. The inclusion criteria are also more stringent, as only systematic reviews to phase III randomised controlled trials are included, as noted below, which has led to further exclusions from the RCPsych review.

The following questions were identified and used to inform the search strategy of the current review (see Appendix 4 for search strategy, databases searched and search logs).

Review questions
7.8.6.2.2. Selection criteria
Types of study

Systematic reviews through to phase III randomised controlled trials (evidence level 1). Qualitative studies and surveys were also included to obtain information on staff and service user views to answer review question 3.

Interventions

Rapid tranquillisation i/v, i/m, and oral (any form).

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All adult in-patient mental health settings, excluding learning disability.

Types of outcome
  • Decrease in hostility/aggression
  • Tranquillisation
  • Sedation/somnolence
  • Side effects and adverse reactions
  • Satisfaction with care
  • Economic outcomes (considered in separate economic review – see Appendix 9).
7.8.6.2.3. Search strategy

In addition to the searches covering 1985-2003, searches were also conducted covering 1969-1985 on all databases to ensure that no research papers were missed in the RCPsych review.

7.8.6.2.4. Clinical evidence

Eighty-three studies were identified in the initial sift. After sifting for relevance and duplicates 71 full papers were ordered. A further paper was identified through the stakeholder consultation process.

Medication

Twenty-three papers were opinion pieces, anecdotal reports, letters, or fell outside the inclusion criteria for this review. References were checked and one further study was identified.

Eight papers were systematic reviews, seven of which proved relevant to the research question. However, one systematic review was a duplicate, so only six systematic reviews were included. Eleven randomised controlled trials were retrieved; one was excluded because it was already included in an included systematic review. Ten further randomised controlled trials that were not mentioned in any of these systematic reviews were located, all of which proved relevant to the research question. All trials were phase III as advised by NICE. All evidence is level 1.

The rapid tranquillisation search strategy was broad enough to have found any articles on PRN that related to the short-term management of disturbed/violent behaviour. One systematic review (Whicher et al.2002) was found and five further studies were identified.

Service user and staff perceptions

Seven surveys and qualitative studies examined staff and service user perceptions and preferences in relation to rapid tranquillisation and PRN medication. These studies were included, as they address the acceptability of this intervention to staff and service users, and doctors' preferences concerning rapid tranquillisation medications, and therefore supplement the evidence base from randomised controlled trials. All evidence is level 3.

Sudden death

Although sudden death was not searched for specifically, one case study on sudden death was retrieved by the searches. Three further studies were located in references to studies ordered for this review. No study presented evidence above level 2.

Additional search

In addition to the search from 1985-2003, a further search was conducted from 1969-1985 to ensure that no research papers were missed by the RCPsych review. Seventy papers were identified in this search. After sifting for relevance and duplicates, 30 full papers were ordered. A further seven papers were included, 19 papers were excluded and 13 papers were opinion pieces, anecdotal reports, letters, or fell outside the inclusion criteria for this review. References were checked, but no further studies were identified.

Of the seven included papers, three were randomised controlled studies that examined drugs used for rapid tranquillisation. Three were cross-sectional surveys, which examined the use of PRN medication in various relevant settings, and one was a cross-sectional study of service users' attitudes to an incidence of forcible medication.

(Evidence tables of all included studies can be found in Appendix 5. Evidence tables of all excluded tables can be found in Appendix 6).

Appraisal of methodological quality

Common methodological shortcomings were:

  • inappropriately small sample sizes (number needed to treat (NNT) not always stated or sufficient)
  • participants not always sufficiently agitated to require rapid tranquillisation
  • outcome measures not always sufficiently defined
  • intention to treat analysis not always clearly described
  • statistical measures (OR, RR, CI) not clearly reported.
Included studies
I. Systematic reviews

Six systematic reviews were used to inform this review.

Efficacy of medication

Four of these reviews looked at the efficacy of a particular medication. Three reviews were all of very high quality (Cure and Carpenter 2001; Fenton et al. 2001; Carpenter and Berk 2000); a further review was of a lesser quality, but the quality was not overly compromised (Aleman and Kahn 2001). These reviews were used as part of the evidence base for assessing the efficacy of the medications droperidol, zuclopenthixol acetate, clotiapine and risperidone. Three of these reviews examined the efficacy of these medications specifically with reference to rapid tranquillisation (Carpenter and Berk 2002; Fenton et al. 2002; Cure and Carpenter 2001). These three reviews looked at the efficacy of droperidol, clotiapine and zuclopenthixol acetate respectively. All three reviews concluded that the use of these medications in emergency psychiatry was currently only justified in terms of clinical (that is, expert opinion), rather than research evidence. Clotiapine is off patent and currently unavailable in the UK since the manufacturers, Novartis, found the production off patent costs prohibitive. Droperidol was voluntarily withdrawn by the manufacturer, Janssen-Cilag Ltd, from the end of March 2001, amid concerns over the medication's safety as an oral treatment for chronic conditions. Cost effectiveness of production resulted in other forms of the medication also being withdrawn.

The reviewers suggest that further research is needed into the efficacy of droperidol for the purposes of rapid tranquillisation, for which it seemed to be safe, although the evidence underlying the medication's safety is also clinical rather than research-based. Likewise authors of the review on clotiapine also stressed that whilst they did not want to discourage the use of this medication for rapid tranquillisation, more research is needed to establish its efficacy and safety in relation to other medications used for this intervention. Similar conclusions were reached about zuclopenthixol acetate, with the reviewers stressing that there was no evidence to suggest that this medication was either safer or more effective than other medications currently used for rapid tranquillisation. Furthermore zuclopenthixol acetate is slow acting and therefore is normally no longer recommended for rapid tranquillisation. This evidence is graded at level 1.

The evidence from these reviews indicates that none of these medications emerges as a gold standard medication for use in rapid tranquillisation. Zuclopethixol acetate is slow acting and not normally used for rapid tranquillisation and both droperidol and clotiapine are unavailable in the UK for rapid tranquillisation. The evidence tables and meta-analyses relating to these medications are given in Appendix 8.

Evidence statement

Level of evidenceEvidence statement
Level 1+The available evidence suggests that zuclopenthixol acetate should not normally be used for rapid tranquillisation.
Risperidone

The other systematic review which was included as part of the evidence base for this guideline examined risperidone (Aleman and Kahn 2001), an atypical antipsychotic. This study did not focus on the issue of rapid tranquillisation. However, it was included because it informed the review about the action of this agent and its various side effects. (All meta-analyses from this systematic review are included in Appendix 8).

The review of risperidone (Aleman and Kahn 2001) considered this medication's function for the management of aggression, but excluded a study because it looked specifically at violence, although there were additional quality issues underlying this exclusion. Some attempt was made to counter the heterogeneity of the studies, by carrying out analyses of only double-blinded randomised studies and those with similar doses, in order to assess the significance of various methodological differences between the studies. The reviewers argue that there was a clear superiority of risperidone over conventional antipsychotics (mostly haloperidol). However, the authors acknowledge that the service users in the included studies did not have chronic aggressive behaviour, which limits the generalisability of the result in relation to rapid tranquillisation. The reviewers also note that risperidone is not available as an intramuscular preparation, which further limits its suitability for an emergency situation. The authors' conclusions on the efficacy and appropriateness of risperidone appear to be overly optimistic in relation to the evidence base and should, therefore, be interpreted with caution.

All the papers included in the above reviews are not included separately in the evidence tables. The literature was trawled for further studies that would add to the evidence base of these reviews, but no additional studies were located.

Systematic reviews assessing the safety and efficacy of rapid tranquillisation

Two further systematic reviews were included in this evidence review, but were considered principally in terms of background information. Both reviews had a similar aim to the current review – that is to assess the efficacy and safety of rapid tranquillisation as an intervention for the short-term management of violence (NICE schizophrenia guideline 2002; Broadstock 2001). Both of these reviews stressed the dearth of the evidence base. Neither suggested that one medication emerged as the gold standard medication for use in rapid tranquillisation. The schizophrenia guideline undertook limited metaanalysis while Broadstock did not. Both studies identified medications that they believed were safe and efficacious for rapid tranquillisation on the basis of very limited research evidence. The NICE schizophrenia guideline closely followed the Broadstock review, which is being used as a base for this current review. In addition, many of the recommendations from the schizophrenia guideline were drawn verbatim from the RCPsych review. In addition to recommending the use of intramuscular haloperidol (a conventional antipsychotic) and intramuscular lorazepam (a benzodiazepine), they also suggested that intramuscular olanzapine (an atypical antipsychotic) should be considered for use in rapid tranquillisation for service users with schizophrenia. They suggested that a combination of haloperidol and lorazepam should be used only in exceptional circumstances and also that the BNF limits should not be exceeded, except within exceptional circumstances. They recommended the use of intravenous medications, as the original RCPsych guideline had done, for exceptional circumstances. They also recommended that i/m diazepam and i/m chlorpromazine should not be used for rapid tranquillisation.

The New Zealand health technology appraisal, undertaken by Marita Broadstock, (2001) suggested that benzodiazepines and antipsychotics seemed to be reasonably safe and effective for rapid tranquillisation and that no significant differences in terms of effectiveness were noted between them or between single and combination regimes. The review argued that there was some evidence to suggest that droperidol may be faster acting that other antipsychotics, but equally safe and effective. They found no studies that appraised the effectiveness of valproate or atypicals and therefore did not comment on their efficacy or safety for use in rapid tranquillisation. They also noted that there was some evidence to suggest that there were less extrapyramidal symptoms (EPS) associated with benzodiazepines (lorazepam), or benzodiazepines and antipsychotics (lorazepam and haloperidol) in combination, than when antipsychotics were used alone (haloperidol, clotiapine). Broadstock (2001) notes that the conclusions arrived at are broadly consistent with those found in the RCPsych guideline.

Conclusion

The current guideline is not considering studies where the primary diagnosis is alcoholism or substance abuse. Therefore studies included in these reviews were included or excluded from this guideline review, on the basis of the inclusion criteria outlined above. One study (Thomas et al. 1992), included in both Broadstock (2001) and the RCPsych guideline was excluded in this guideline's review because the study population did not necessary or primarily have a psychiatric illness, and because most of the participants were intoxicated. This trial was also excluded from the Cure and Carpenter (2001) systematic review on these grounds. All other randomised controlled trials included in these reviews were included in the current review and are assessed below, in conjunction with several additional randomised controlled trials identified in the searches.

II. Randomised controlled trials

Nineteen randomised controlled trials, not included in the systematic reviews by Cure and Carpenter 2001; Fenton M et al. 2001; Carpenter and Berk 2000 are included in this review. One of these studies reported on two different trials (Garza-Trevino et al. 1989). Eleven of these studies were conducted in the USA. One took place in Israel, one took place in Brazil, one took place in India, one took place in Denmark and the other three were multi-country studies. Unless otherwise stated, all studies compared intramuscular loading (i/m) of various medications. Some studies switched to oral loading after the first 24hours. Where this occurs, it is indicated in the evidence tables.

Haloperidol vs. lorazepam (vs. haloperidol and lorazepam)

Four studies compared the benzodiazepine (i/m) lorazepam and the conventional anti-psychotic (i/m) haloperidol. (Foster et al. 1999; Bieniek et al. 1998; Battaglia et al. 1997; Garza-Trevino et al. – study I, 1989). Two of the studies (Foster et al. 1999; Battaglia et al. 1997) evaluated the efficacy and safety of these two medications against each other. Foster et al. (1999) noted no significant difference between the agitation scores for lorazepam and haloperidol at one hour on the BPRS (p=0.39, WMD=3.26 [-4.16, 10.68] 95% CI), but did note a significant difference in favour of lorazepam at one hour on the CGI (p=0.002, WMD=0.67 [0.25, 1.09] 95% CI). Battaglia et al. (1997) noted no significant difference between haloperidol and lorazepam at one hour, based on mean ABS score (p=0.27, WMD=-1.92 [-5.31, 1.47] 95% CI).Two of the studies (Battaglia et al; 1997; Garza-Trevino et al. – study I, 1989) considered the efficacy and safety of these two medications against a combination of haloperidol and lorazepam. Battaglia et al. (1997) noted a significant difference between haloperidol and combination at one hour, based on ABS score (p=0.03, WMD=3.85 [0.46, 7.24] 95% CI) and between lorazepam and the combination at one hour based on ABS score (p=0.005, WMD=-5.77 [-9.76, -1.78] 95% CI). However, it is unclear if the combination would have been superior if the dose of the single agents had been equivalent to that of the combination. Garza-Trevino et al., study I, 1989 found that the combination was more likely to lead to tranquillisation in 30 minutes. These findings were replicated in ANOVAS. However, the authors suggest that it is unclear whether the combination would have been superior if the dose of the single agents had been equivalent to that of the combination.

One study (Bieniek et al. 1998) only considered the efficacy and safety of haloperidol against that of a combination of haloperidol and lorazepam. No differences were noted between the two groups with ANOVAS, but non-parametric tests indicated that a greater percentage of participants improved after 60 minutes in the combined group.

Whilst two of the studies compared the same single doses, and two of the studies compared the same combined doses, there were many methodological problems with these studies. These problems included: relatively small sample sizes, short follow-up periods, side effects not being recorded, many comparisons performed with no adjustment for p-value, baseline and information not recorded. The studies were also heterogeneous. One study (Garza-Trevino et al. 1989) was not double-blinded, one study (Battaglia et al. 1997) considered sleep as a desirable endpoint, (the other studies did not), and combination doses were not equivalent to single medication doses. With such heterogeneity, meta-analysis was not appropriate.

In terms of efficacy, no study found the antipsychotic to differ from the benzodiazepine. However, given the side effects caused by haloperidol (for example, dystonia), all authors suggested that lorazepam may be the preferred course of treatment. All three studies (Foster et al. 1999; Bienien et al. 1997; Garza-Trevino et al. – study I, 1989) that compared combination against a single medication or medications suggested the superiority of the combination in terms of efficacy. However, two studies note that, since the single dose was not equivalent to the combined dose, it remains unclear whether the combined doses were more effective simply because of the strength of dose. The study (Battaglia et al. 1997) that did not comment on this, regarded sleep as a desirable endpoint and therefore viewed effectiveness in terms of a different outcome. One study (Foster et al. 1999) noted that there is need for more dose response studies.

Conclusion

The studies suggest that the medications (i/m) haloperidol and (i/m) lorazepam are equally effective. It is also suggested that the combination of haloperidol (i/m) with lorazepam (i/m) is also effective. Although there are many methodological problems with the studies, the body of evidence suggests that, either as single agents or in combination, these medications are efficacious. However the dose response comparisons between the combination or the medications as single agents is unclear on the evidence of these trials. Whilst more extrapyramidal side effects (EPS) are recorded with haloperidol, the chosen medication should be dictated by individual service user histories.

Evidence statement

Level of evidenceEvidence statement
Level 1+ to 1-The available evidence suggests that (i/m) haloperidol and (i/m) lorazepam are equally effective. However, there is no firm evidence for either the efficacy or safety of haloperidol or lorazepam as single agents, or in combination. Haloperidol has been found to result in an increased likelihood of more extrapyramidal effects.
Olanzapine vs. haloperidol vs. placebo

Two trials (Brier et al. 2002; Wright et al. 2001) evaluated (i/m) olanzapine against (i/m) haloperidol and against (i/m) placebo. Both studies were large multi-site, multi-country studies. However, they included participants with schizophrenia (571 participants in total). Both authors were involved in both studies and both work for Eli Lilly who sponsored both studies.

It is unclear whether the participants actually required rapid tranquillisation, since all gave consent before being included in the study. Objective measures of behaviour were used in both studies at baseline (PANSS – positive and negative syndrome scale). There was no long-term follow up with either study. Both olanzapine and haloperidol were significantly more effective than placebo in reducing agitation at two and 24 hours in both studies. At 30 minutes, a dose of 5.0mg, 7.5mg or 10mg were significantly more effective than the placebo. Olanzapine was significantly more effective than haloperidol in reducing agitation at 15, 30 and 45 minutes (Wright et al. 2001). In Brier et al. (2002), group sizes did not allow comparison with placebo. Acute dystonia was not associated with olanzapine, but was found in 7 per cent of the haloperidol group. (Wright et al. 2001) Brier et al. (2002) also found that olanzapine was not associated with dystonia. There were no differences between olanzapine, haloperidol and placebo in terms of hypotension and clinically relevant changes in the QTc interval (Brier et al. 2002). On this basis, Brier et al. (2002) suggests that olanzapine has a safer profile than haloperidol. The schizophrenia guideline examined both of these studies and undertook a meta-analysis that slightly favoured olanzapine (see Appendix 8 for results).

Conclusion

Olanzapine (i/m) would appear to be both effective and safe for use in rapid tranquillisation for service users with schizophrenia. It would also appear to have fewer side effects than (i/m) haloperidol and more rapid onset of action. However, in these two trials the populations were only suffering from moderate disturbance where rapid tranquillisation was not necessarily required. Therefore further RCTs with appropriate populations are needed to verify the findings from these trials for use of olanzapine in those who are disturbed/violent. (Level 1 and level 2.)

It has subsequently been noted that a warning has been issued by the manufacturers advising against the use of olanzapine outside the SPC recommended dose, as adverse effects have been recorded.

Evidence statement

Level of evidenceEvidence statement
Level 1+ to 2The evidence suggests that olanzapine (i/m) is safe and effective for rapid tranquillisation for service users with schizophrenia. However the study population did not necessarily require rapid tranquillisation. The evidence suggests olanzapine (i/m) has fewer side effects and is more rapid in onset than (i/m) haloperidol.
Ziprasidone (vs. haloperidol)

Three studies considered the efficacy and safety of (i/m) ziprasidone (Daniel et al.2001; Lesem et al. 2001; Brook et al. 2000). Two considered this medication only (Lesem 2001; Daniel et al. 2001), measuring the effectiveness and safety of different doses (2mg vs. 10mg and 2mg vs. 20mg). The other study (Brook et al. 2000) compared (i/m) ziprasidone with (i/m) haloperidol.

Brook et al. (2000) found significant reductions in agitation from baseline at day three that favoured ziprasidone. (BPRS total WMD=-3.06 [-5.68, -0.44]95% CI; BPRS agitation score WMD=-1.13 [-2.23, 0.03] 95% CI;CGI-SWMD=-0.34 [-0.55, -0.13]. However, the paper contained insufficient data to calculate the relative reductions in agitation at earlier time periods.

The other two studies, Daniel et al. (2001) and Lesem et al. (2001), employed the same methodology; both were sponsored by Pfizer. It was not clear whether either of the studies dealt with truly agitated participants, since all gave consent. Both 10mg and 20mg were noted to be significantly more effective than 2mgs. A reduction of two or more on BARS scale at two hours after initial injection showed the following significant result for 10mgs (p=0.003, OR=0.32 [0.15,068] 95% CI), and the following for 20mgs (p<0.00001, OR0.04 [0.01, 0.16] 95% CI). Lesem et al. (2001) argues that 10mg is a therapeutic dose but probably at the lower end of the spectrum, especially given the agitation levels of the participants in the study. There was no significant difference in side effects with any of the doses. Most side effects were moderate, suggesting the reasonable safety of this atypical antipsychotic. The study Brook et al. (2000), comparing haloperidol with ziprasidone, did not consider rapid tranquillisation, but the management of acute psychosis. In this context, compared to haloperidol, ziprasidone was significantly more effective in managing aggression by day seven. However, no firm conclusions about the relative effectiveness of ziprasidone compared to haloperidol as an agent for rapid tranquillisation can be arrived at on the basis of this study.

Conclusion

It would seem that (i/m) ziprasidone 20mg is relative safe for use in rapid tranquillisation. Its effectiveness needs further testing with more highly agitated participants. Whilst meta-analysis of these two studies would have been possible, the value of doing so appears limited, given that the participants did not necessarily require rapid tranquillisation and no comparison with other medications was made (level 1). The relative effectiveness and safety of (i/m) ziprasidone compared to (i/m) haloperidol as an agent for rapid tranquillisation cannot be established on the basis of this evidence (level 2). Ziprasidone received a black box warning in 2002 in relation to its QTc prolonging potential, which may be increased in situations of high arousal.

Evidence statements

Level of evidenceEvidence statement
Level 1-The evidence suggests ziprasidone (i/m) is safe for use in rapid tranquillisation, however its effectiveness with highly agitated service users is not known.
Level 2The relative effectiveness and safety of (i/m) ziprasidone compared to (i/m) haloperidol as an agent for rapid tranquillisation cannot be established on the basis of this evidence.
Loxapine vs. haloperidol (vs. thiothixene)

Four studies examined the use of (i/m) loxapine. Three compared it to haloperidol (IM) (Tuason 1986; Fruensgaard et al. 1977; Paprocki and Versiani 1977) and the other to (i/m) thiothixene (Dubin and Weiss 1986). Neither Tuason 1986 nor Dubin and Weiss 1986 were double-blinded and in one of these studies (Dubin and Weiss 1986), it is unclear whether participants required rapid tranquillisation. However, the other two studies (Fruensgaard et al. 1977; Paprocki and Versiani 1977) were double blinded. All medications achieved significant improvement from baseline and there were no significant difference in numbers of adverse reactions between groups, except in one study (Fruensgaard et al. 1977) where (i/m) loxapine 50mgs produced more pronounced sedation (p<0.025). Haloperidol 5mg did not differ in median time to rapid tranquillisation from loxapine 25mg (Tuason 1986) (at two hours OR=0.32 [0.09,1.22] 95% CI based on CGI ratings for global improvement. The same is also true of the mean BPRS scores at two hours WMD =1.10 [0.78, 1,42]). Nor did haloperidol 5mg differ in median time to rapid tranquillisation from loxapine 50mgs (Fruensgaard et. al. 1977; Paprocki and Versiani, 1977). There is insufficient data to calculate 95 per cent CI for these two studies. However, thiothixene 10mgs (Dubin and Weiss 1986) was significantly less tranquillising in the initial phase of treatment than loxapine 25mgs (60min vs. 95min) p=0.0008 OR=9.00 [2.49,32.57] 95% CI.

Conclusion

The studies had various limitations which make it difficult to formulate firm conclusions about the relative effectiveness and safety of (i/m) loxapine compared to either (i/m) haloperidol or (i/m) thiothixene for use in rapid tranquillisation. However, at the doses prescribed, loxapine would appear to provide a more rapid tranquillising effect than thiothixene (level 1).

Evidence statement

Level of evidenceEvidence statement
Level 1-The evidence is not conclusive for the effectiveness or safety of (i/m) loxapine compared with either (i/m) haloperidol or (i/m) thiothixine.
Thiothixene and lorazepam vs. haloperidol and phenobarbital

A further study also considered the use of (i/m) thiothixene in combination with (i/m) lorazepam against (i/m) haloperidol in combination with (i/m) phenobarbital (Garza-Trevino et al. 1989 – study II). This study was not double-blinded, there was a very short follow-up period (24 hours) and the side effects were not described, although the authors claim that there were few indications of over-sedation or dystonic reactions. There appeared to be no difference in effectiveness between the two groups. The authors argue therefore that a combination of antipsychotic and a hypnosedative is a useful intervention for the management of agitated behaviour.

Conclusions

It is difficult to generalise concerning the effectiveness and safety of these medication combinations on the basis of only one study, given the various limitations noted above.

Evidence statement

Level of evidenceEvidence statement
Level 1-The evidence for thiothixene and lorazepam vs. haloperidol and phenobarbital combinations is inconclusive.
Haloperidol (vs. flunitrazepam, vs. molidone, vs. midazolam and sodium amytal vs. midazolam, vs. chlorpromazine)

Four trials (Binder and McNiel 1999; Dorevitch et al. 1999; Wyant et al. 1990; Reschke 1974) evaluated the efficacy and safety of five further medications against (i/m) haloperidol; (i/m) flunitrazepam, (i/m) molindone, and (i/m) midazolam, (i/m) sodium amytal, and (i/m) chlorpromazine. Neither of the first two studies showed a significant difference between haloperidol and the other medication in terms of effectiveness. Flunitrazepam showed a slightly quicker reduction in aggression at 30 minutes, but this did not reach significance at 90 minutes on the OAS scale (p=0.37 OR=3.00 [0.27, 33.08]). Molidone showed slightly less reduction in symptoms at three hours. Erythema at injection site was slightly more common for molidone than haloperidol. This side effect is not discussed in relation to (i/m) flunitrazepam. Both studies had small sample sizes and neither used objective measures to evaluate behaviour at baseline. In the study of molidone there was no adjustment to the p value to account for the many comparisons and outcomes (outcomes were not restricted to rapid tranquillisation). It was also difficult to assess whether side effects resulted from the oral phase of the intervention.

The study of (i/m) haloperidol vs. (i/m) midazolam or (i/m) sodium amytal (Wyant et al. 1990) randomly assigned participants to either (i/m) haloperidol 10mg, (i/m) midazolam 5mg or (i/m) sodium amytal 250mg. Over two hours, (i/m) sodium amytal and (i/m) midazolam proved significantly more effective than haloperidol in terms of mean global ratings for motor agitation (p< or =0.05), but there was no significant difference in hostility rating (p<0.10). This study has several limitations, not least being a very small sample size. It is also only single blinded. In addition, there is insufficient data in the paper to calculate 95 per cent CI. Side effects are not mentioned. On both these counts it could have been excluded, but is reported here to illustrate the available research on these medication combinations. The authors recognise the need for a large-scale future study comparing midazolam with lorazepam.

In the study of (i/m) haloperidol vs. (i/m) chlorpromazine aggression was significantly more effectively controlled with (i/m) haloperidol 5mgs and 2mgs (p<0.05) compared to (i/m) haloperidol 1mg, (i/m) chlorpromazine 25mgs or (i/m) placebo. More adverse reactions were noted with haloperidol (transient hypertension, drowsiness (awake), dry mouth and mild EPS) than chlorpromazine, although there was greater somnolence with chloropromazine. The study had a very small sample size. Most participants in this study were women.

Conclusions

Given the limitations of the studies, no firm conclusion can be reached about the relative superiority of these medications compared to haloperidol, although both flunitrazepam and molindone appear to be reasonably safe and effective within these trials. The study of midazolam and sodium amytal did not mention side effects. These trials' limitations mean that no firm conclusions about these medications can be drawn. The study of midazolam used sleep as a desirable endpoint, making comparisons with other studies difficult. Chlorpromazine (i/m) was slower acting than haloperidol, although it had fewer side effects. Chlorpromazine (i/m) is no longer considered a suitable medication for rapid tranquillisation, since it is a local irritant if given intramuscularly; carries a risk of cardiovascular complications; causes hypotension due to adrenergic receptor blocking effects, especially in the doses required for rapid tranquillisation; and is erratically absorbed. Its effect on QTc intervals also suggests that it is unsuitable for use in rapid tranquillisation.

Evidence statement

Level of evidenceEvidence statement
Level 2-The evidence for haloperidol (vs. flunitrazepam, vs. molidone, vs. midazolam and sodium amytal vs. midazolam, vs. chlorpromazine) is inconclusive.
Haloperidol plus promethazine (vs. lorazepam vs. midazolam)

One study (Trec 2003) compared (i/m) halperidol-promethazine with (i/m) midazolam. Clinicians decided doses within a range of 7.5-15mgs of (i/m) midazolam and 5mgs of (i/m) halperidol plus 25-50mgs (i/m) promethazine. More somnolence was noted in the midazolam group. One man suffered respiratory depression with (i/m) midazolam 15mgs and recovered after being given (i/v) flumazenil 0.25mgs. One woman with epilepsy suffered a grande mal seizure with (i/m) haloperidol 5mgs and (i/m) promethazine 50mgs. When ratios of those either asleep or tranquil at one hour were considered, the study favours midazolam (OR=0.49 [0.22, 1.09] 95% CI, NNT=12.5 [6.4, 77.7] 95% CI).However, a larger percentage of these patients were asleep in the midazolam group than in the haloperidol + promethazine group (93 per cent compared to 87 per cent). If only those patients who were tranquil at one hour are considered, the treatment favours haloperidol + promethazine (OR=2.91 [1.64, 5.18] 95% CI). No definitions are provided for tranquil or asleep. The preferred outcome in the UK context is considered to be tranquil (calm) and conscious.

The other study (Alexander 2004), compared (i/m) haloperidol and promethazine combined with (i/m) lorazepam. Doses were (i/m) haloperidol 10mgs plus (i/m) promethazine 25-50mgs or (i/m) lorazepam 4mgs. (i/m) Haloperidol plus (i/m) promethazine was significantly more likely to induce sleep for all time periods (p=0.00). (i/m) Haloperidol plus (i/m) promethazine also resulted in quicker onset of tranquillisation (p=0.0001)/sleep (p=0.0000). Four people in the lorazepam group were never tranquil; one person in the haloperidol plus promethazine group was never tranquil. No adverse reactions were noted with haloperidol plus promethazine. One person in the lorazepam group with history of bronchial asthma complained of moderate worsening of respiratory difficulty; one person reported nausea and dizziness. There was no dystonia. Sleep was considered the desirable endpoint. When ratios of those either asleep or tranquil at one hour are considered, the study favours haloperidol (OR=5.44 [1.16, 25.52] 95% CI). However, a larger percentage of these patients were asleep in the haloperidol + promethazine group than in the lorazepam group (98 per cent compared to 90 per cent). If only those patients who were tranquil at one hour are considered, the treatment favours lorazepam (OR=0.33 [1.18, 0.58] 95% CI).

Conclusions

Unlike most of the other studies in this review, both were large studies of a high methodological quality. Despite this, after consultation with two independent methodological advisers, it was decided that meta-analysis would not be appropriate for the following reasons. Firstly, it is not clear that the two benzodiazepines (midazolam and lorazepam) are sufficiently similar clinically to be treated as a single class, nor is it clear that the two benzodiazepine doses are equivalent, which could make the effect size vary. Secondly, the primary outcome was four hours in TREC (2003) but 15 minutes in Alexander (2004). These were rated by blinded assessors (Alexander 2004). While TREC (2003) did take measurements at 20 minutes, these were not made by blinded raters. There is also a danger of masking differences in effect when combining different time points.

One of these trials (Alexander 2004) considered sleep the primary desirable outcome. However, the study did detail numbers asleep and numbers tranquil at each endpoint (as did Huf 2003). Alexander (2004) argues that sleep is a safer option for staff however, no significant difference in injury rates were noted with lorazepam, which was less sleep inducing. Neither study mentioned whether monitoring procedures – for example, observation, ECG, etc. – were put in place once participants were classified as asleep. There is disagreement between the studies as to whether haloperidol plus promethazine is actually more likely to induce sleep than a benzodiazepine. As sleep is not normally considered a desirable endpoint for rapid tranquillisation, the studies suggest that haloperidol plus promethazine may be effective in rapid tranquillisation when sleep is a desirable outcome, which has been suggested as being a safer option for staff. However, if tranquil (calm) is the desirable endpoint, lorazepam alone is favoured.

Few patients treated with (i/m) haloperidol plus (i/m) promethazine suffered dystonic reactions, since promethazine has anti-cholingeric properties.

Evidence statement

Level of evidenceEvidence statement
Level 1+The available evidence suggests that haloperidol with promethazine i/m is effective in rapid tranquillisation by inducing sleep. The evidence suggests lorazepam i/m is effective in rapid tranquillisation by calming the service user.
Heterogeneity

The included studies had many heterogeneous aspects/deficiencies in reporting, such as differing settings, and insufficient time to allow NNT to be calculated. Furthermore, some studies did not contain sufficient data to allow the OR to be calculated. The studies also had different comparator medications, doses and outcomes (for example, sleep as the endpoint and sleep as an adverse effect). In addition, the term sleep is often loosely defined in these studies, which further complicated any comparison. Follow-up periods also differed across studies. After methodological advice, it was considered that in the face of such heterogeneity, meta-analysis would be inappropriate.

Appraisal of methodological quality

There were also many methodological quality issues – for instance, most studies did not report their method of randomisation nor how they ensured blinding/lack of bias. It was felt that meta-analysing studies of low quality might therefore be misleading. Furthermore the outcome measurements were not comparable or always clearly defined. Actual dosages given in the studies also make comparisons within and between studies difficult.

Evidence statements

Level of evidenceEvidence statement
Level 1+ to 1-A gold standard medication for rapid tranquillisation has not yet been established
Level 1+ to 1-There appear to be no conclusive benefits in terms of effectiveness of one antipsychotic over another; of antipsychotics over benzodiazepines; or of combination medications over single medication regimes for rapid tranquillisation.
Level 1+ to 1-The body of evidence suggests rapid tranquillisation as an intervention for the short-term management of disturbed/violent behaviour is both reasonably effective and reasonably safe. This evidence suggests that both benzodiazepines and antipsychotics appear to be effective and reasonably safe for use in rapid tranquillisation.
Level 4It is not possible to determine the safety or effectiveness of medications other than antipsychotics and benzodiazepines for rapid tranquillisation.
III. Studies other than randomised controlled trials

Most of the non-randomised studies which examined rapid tranquillisation that were identified investigated medications which were considered in the randomised controlled trials or systematic reviews discussed above. One study discussed a medication not considered elsewhere. This study (Lee et al. 1992) was a prospective cohort study (n = 10) that considered the use of lithium citrate. There were considerable limitations to this study noted by the authors, such as a lack of a wash out period and the lack of a double-blind control. They suggest that this medication may be an alternative treatment to neuroleptics or benzodiazepines, particularly for service users who demonstrate persistently agitated behaviour, despite treatment with neuroleptics, benzodiazepines, barbiturates and/or antihistamines. However, they recognised the need for a randomised controlled trial.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence of the safety or effectiveness of lithium citrate for use in rapid tranquillisation for those who demonstrate persistently agitated behaviour.
IV. Sudden death and rapid tranquillisation

No randomised controlled trials had death as an adverse event. However, four case studies were identified in which death was linked to medication.

One case study reported in a letter, noted that a young adult who was given (i/v) diazepam 20mg and (i/v) haloperidol 20mgs by a GP and psychiatrist at home suffered a fatal cardiac arrest (Quesnstedt et al. 1992). One other case of sudden death (Dolan et al. 1995) occurred after rapid tranquillisation. In this case, post-mortem examination revealed some congestion of the pulmonary parenchyma. The author expressed concern that the autopsy had partially attributed the cause of death to medication, when sudden death in similar circumstances, without the presence of medication, is known to occur.

In another study (Lynch and Kotsos 2001), a white female was found at home with a fatal dose of benzatropine. It is unclear whether a suicide attempt was made.

In the final study (Kumar 1997) an aggressive service user was restrained and stopped breathing before rapid tranquillisation took place, but a toxic level of chlorpromazine was found in his blood after treatment with chlorpromazine, zuclopenthixol acetate and zuclopenthixol decanoate in the weeks preceding his death. One study did not specify the ethnicity of the deceased; the other three studies all noted that the deceased were Caucasian; three service users were male, the other was female.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine whether there is an association between sudden death and the pharmacological interventions used for the short-term management of disturbed/violent behaviour.
V. Qualitative studies and surveys

Our searches identified seven qualitative studies which were deemed to be of reasonable quality to merit inclusion in this review. Three were conducted in the UK, one in the USA, one in Israel and one in Sweden.

Two of studies set in the UK did not specifically ask staff and service users about their feelings and beliefs about rapid tranquillisation.

The first study (Hyde et al. 1998) compared user dissatisfaction scores with incidents of rapid tranquillisation and found that there was no correlation between a service user satisfaction score and their experience of this intervention. However, as the questionnaire did not ask service users to score their satisfaction/dissatisfaction with this intervention, there are many other confounders that could obscure this correlation. It is unclear what the aims of the questionnaire were. The other UK study (Duxbury 2002) asked staff, service users and medical staff for their views on the approaches used to manage violence. Again, there was no specific reference to rapid tranquillisation, but a disparity was noted between actual violence reported in incident forms and the use of rapid tranquillisation. The authors suggest that this is a worrying trend. They note that both staff and service users agreed that there was a need for greater alternatives to restraint, seclusion and rapid tranquillisation.

The third study (Burgess 1997) reviewed doctors' medication preferences for rapid tranquillisation. The study took the form of two cross-sectional surveys conducted in Oxfordshire – the first in 1990 and the second in 1994-5 – to assess the effect of guidance issues by the Royal College of Psychiatrists. The following outcome measures were considered: drug of choice; route of choice; other drugs used if situation not rapidly resolved by drug of choice; mean dose (second study only); desired endpoint (second study only); time to desired endpoint (second study only). Doctors were asked to rate the same scenario of a psychotic patient. Study 1 showed the following results: 56 per cent response rate (sent to consultants only). The drug of choice was chlorpromazine. The route of choice was intramuscular (93 per cent) against intravenous (7 per cent). Forty-two other drugs were used if the situation was not rapidly resolved by drug of choice, suggesting no consensus. These included neuroleptics, benzodiazepines, anticonvulsants and paraldehyde.

Study 2 showed the following results: 77 per cent response rate from registrars, junior doctors and consultants – 69 per cent response rate from consultants. The drug of choice was chlorpromazine (25 per cent), with haloperidol + lorazepam (22 per cent) and haloperidol alone (16 per cent). Junior doctors were more likely than consultants to use a short acting antipsychotic and a benzodiazepine as first line treatment (p<0.05). The route of choice was intramuscular/oral (93 per cent) against intravenous (7 per cent). Mean dose: chlorpromazine 103.4mg, haloperidol alone 9.6 mg, haloperidol + lorazepam 9.2mgs of haloperidol. Other drugs used if the situation was not rapidly resolved by drug of choice included increasing the dose of neuroleptic medication, adding benzodiazepine and adding clopixol acuphase. 14 per cent would use clopixol acuphase as first line treatment either alone or with other drugs; 62 per cent would consider clopixol acuphase at some point in the first 24 hours. The desired endpoint was: patient non-sedated but calm, 59 per cent; patient sedated but mobile, 31 per cent; patient asleep, 10 per cent. Time to desired endpoint was: within one hour, 21 per cent; within 6 hours, 43 per cent; within 12 hours, 16%; within 24 hours, 10%; more than 24 hours, 6 per cent.

The authors noted that by 1990, only 7 per cent of doctors used a short acting antipsychotic + a benzodiazepine, with 35 per cent in 1994 (p<0.001). However only consultants were surveyed in 1990 and the study showed that significantly more junior doctors than consultants would choose this drug combination in 1994 (p<0.05). The authors also noted that: a) haloperidol continued to be prescribed at a higher dosage equivalent than chlorpromazine, even when used in combination with a benzodiazepine; b) i/v administration is unpopular despite reports that diazepam plus haloperidol is the most rapid effect method of rapid tranquillisation; c) that while the preferred endpoint is non-sedated but calm, most advocated highly sedative drugs, i.e. benzodiazepines. The authors posited that the introduction of local protocols, in line with RCPsych guidance, many have further altered doctors' choices and suggested that an additional survey is required to monitor this.

The other four studies surveyed service users, with the aim of discovering their views about forcible medication. One study (Haglund et al. 2003), also interviewed staff. In the study, service user and nurse perceptions of forced medication differed. Nurses focused more on positive effects of medications, while service users stressed the negatives. Less service users retrospectively approved of forced medication than anticipated by nurses. Nurses mentioned no alternatives to rapid tranquillisation.

However, all service users mentioned at least one alternative (dialogue, more explanation of ill-health, coaxing, waiting, no medication, no injection). Nurses perceived these measures necessary to improve health. The authors noted that service users were more likely to accept forced medication from a nurse they knew.

Another study (Schmeid and Ernst 1983) asked service users to rate the retrospective acceptability of both seclusion and rapid tranquillisation. Service users were asked immediately after the intervention and also once psychosis passed. Service users found rapid tranquillisation more unacceptable than seclusion or restraint (p=0.01; for men only p=0.001). However, many service users were unclear on their feelings about these interventions and did not know which staff member had instigated the intervention. The authors note that this was particularly true for male service users, many of whom had alcohol related problems. This is now an old study and although rapid tranquillisation was the least acceptable of these interventions, it is not possible to generalise whether service users in the UK also prefer seclusion to rapid tranquillisation. The author suggests that it is beneficial for service users to discuss their feelings after the use of one of these interventions.

Greenberg et al. (1996) conducted structured telephone interviews with service users two weeks after discharge. There was a 50 per cent loss to follow-up. The authors noted that 60 per cent of service users interviewed retrospectively agreed that rapid tranquillisation had been beneficial, while 43 per cent felt that they should be coerced in similar situations in the future. However, the authors acknowledge that the responses of those lost to follow-up might have been more negative.

In another study, where semi-structured interviews were conducted on the ward, a far more negative response to rapid tranquillisation was noted. Only 4/11 service users retrospectively approved of rapid tranquillisation and one of these did so only in the most vague of terms. The rest all disapproved. However staff believed that 7/11 service users had retrospectively approved of the use of this intervention.

Schwartz et al. (1988), in a small study in one hospital setting, assessed service users' mental state both before and after forcible medication (service users with organic brain disorder were excluded). They argued that those who retrospectively disagreed with the treatment did not recognise that they had an illness that had required hospitalisation; did not agree that the hospitalisation was necessary; or that it had been helpful (p<0.01). They also noted significant differences between those who did and those who did not retrospectively agree with forcible medication on the BPRS scales for thought disturbances and hostile suspiciousness. At discharge, the following eight items were significant for those who did not retrospectively agree with the forcible medication: conceptual disorganisation (p<0.05); mannerisms and posturing (p<0.01); grandiosity; hostility; suspiciousness; unco-operativeness; unusual thought content (p<0.001). In the light of these findings, the authors argue that judicial review of forcible medication is seldom required.

Evidence statement

Level of evidenceEvidence statement
Level 4The evidence suggests that service users' retrospective responses to rapid tranquillisation are variable.
Both staff and service users agree that, where possible and appropriate, alternatives to restraint, seclusion and rapid tranquillisation are preferable.
VI. PRN medication

Although only rapid tranquillisation is mentioned directly in the scope, pro re nata (PRN) – as needed – medication is also sometimes used in a similar way to rapid tranquillisation in psychiatric in-patient settings. For this reason, a review of the use of PRN medication in this context was undertaken.

Effectiveness

One systematic review (Whicher et al. 2002) was found that examined the efficacy of PRN for the short-term management of aggressive behaviour. No randomised trials were found which met the inclusion criteria for this review.

Whicher et al. (2002) conclude that:

This common current practice has no support from randomised trials. Current practice is based on clinical experience and habit rather than high quality evidence. Current practice, therefore, outside of a well designed, conducted and reported randomised trial, is therefore difficult to justify.

No further studies were identified which examined the effectiveness of PRN medication as a pharmacological intervention for the short-term management of disturbed/violent behaviour.

Staff perspectives

A study by Geffin et al. (2002a) examined the beliefs of doctors and nurses in in-patient psychiatric units about PRN medication for psychiatric disorders. They concluded that nurses and doctors have different views about the effectiveness of antipsychotics and benzodiazepines. They noted that some of these views are at odds with the known properties of these medications. They argue therefore that doctors should always specify the usage when writing PRN prescriptions. They argue that further education is needed to achieve best practice in PRN medication.

A second survey by Geffin et al. (2002b) examined the uses of PRN medication in two large psychiatric units. The authors noted that while a maximum daily dose was normally specified (87 per cent), indications for use were only specified in 6 per cent of cases. Staff noted medication-related morbidity in 37 per cent of service users taking PRN medication, compared to 3 per cent on only regular scheduled medication. Forty-nine percent of PRN medication was given for agitation. However, administration records frequently failed to specify a reason for use in 48 per cent of cases. Nearly two-thirds of administrations (64 per cent) had no recorded outcome. Of the remaining 26 per cent, 76 per cent were reported as being partially or completely effective, with the remainder recorded as ineffective. Higher daily doses of PRN medication were given to manic patients, males, younger patients and those with substance abuse disorders. Co-prescription of typical antipsychotics PRN with atypical antipsychotics was common (64 per cent).

Three further studies were found which also examined the use of PRN medication. One was set in Canada (Craven et al. 1987); two were set in the USA (Walker 1991; Craig and Bracken 1995). A finding of two of these studies (Craven et al. 1987; Walker 1991) was that indications were not always included in the prescriptions. Where an indication was given, only one was specified, although the prescriptions was often used for a number of indications. Minimum intervals between doses were not always stated and maximum daily doses were also not always specified. In addition very few of the prescriptions specified an end-date.

The other study (Craig and Bracken 1995), noted a difference between the use of PRN medication with service users who had a discrete period of disruptive behaviour and those who had intermittent periods throughout their stay. The following results were significant: half of the intermittent vs. more than 90 per cent of those with discrete episodes had an increase in their medication or had their medication changed during the study month (p=0.03). The authors noted that 12/27 service users who had antipsychotic medication serum levels drawn either before or after the study were found to be in a sub-therapeutic range, with two-thirds below detection. They suggest the importance of checking serum levels to ensure that these are adequate, in order to reduce the need for PRN medication.

Evidence statement

Level of evidenceEvidence statement
Level 4The evidence suggests that the use of PRN medication for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings is inconsistent and may not be appropriately administered or monitored.
Health care professionals require education on the appropriate use of PRN medication
7.8.6.2.5. Economic evidence

There are two papers that are economic analyses of rapid tranquillisation. One is a UK-based cost minimisation study (Hyde et al. 1998) and the other is a cost consequence analysis, based in Canada (Laurier et al. 1997). Both compare zuclopenthixol acetate with haloperidol. The two papers are in disagreement regarding which of the two medications is more cost effective and thus the literature is inconclusive. Zuclopenthixol acetate is no longer recommended for RT due to a long onset of action. See discussion above. (For details of the health economics review, see Appendix 9).

7.8.7. Emergency departments

7.8.7.1. Objectives

No specific searches into the short-term management of disturbed/violent behaviour in emergency departments were undertaken in the RCPsych guideline.

Two review questions were identified and used to inform all searches (See Appendix 4 for search strategies, databases searched and search logs)

Review questions
  • How is disturbed/violent behaviour by psychiatric patients best managed in the short-term in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?
  • What are the views of staff and service users about the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

These questions were addressed in relation to the various interventions and related topics covered in this guideline and specific review questions were devised (see below).

7.8.7.2. Selection criteria

Types of studies

Systematic reviews to before and after studies (levels 1-2). Qualitative studies were also included.

Types of participants

Adult psychiatric service users <16 years, excluding people with a primary diagnosis of substance abuse, older persons with an organic mental disorder (for example, any form of dementia) or a progressive neurological disease (for example, Parkinson's disease).

Types of setting

All emergency departments.

Types of outcome

7.8.7.3. Clinical evidence

Eighty-one papers were identified in the initial sift. After sifting for relevance and duplicates, 50 full papers were ordered; 18 met the inclusion criteria; 12 were excluded. All the other papers were opinion pieces, anecdotal reports or fell outside the inclusion criteria for this review. References were checked but no further studies were identified. (Evidence tables of included studies can be found in Appendix 5. Evidence tables of excluded studies can be found in Appendix 6).

Appraisal of methodological quality

In addition to the quality concerns raised above, these studies had the following methodological problems:

  • confounders not considered or taken into account
  • inappropriate sample size
  • much anecdotal ‘evidence’ based on author's experience and reflection rather than on primary research.
I. Prevention

Seventeen studies addressed issues around prevention of disturbed/violent behaviour in emergency departments.

I.i. Environment
I.i.a. Review question: how does the environment and organisation impact on disturbed/violent behaviour by psychiatric patients in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?

No studies addressed this review question.

I.i.b. Review question: what are the views of staff and service users about how the environment and organisation impacts on the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

Lillywhite et al. (1995) audited the interview rooms and surveyed the staff of a mental health service to assess the safety of interview rooms in a general hospital outpatients' department, general hospital emergency department and a psychiatric hospital, according to 10 safety criteria, and to assess medical staff's ratings of the relative importance of these 10 criteria. Emergency department rooms scored least well in terms of suitability for interviewing potentially aggressive patients, scoring poorly on every criteria other than ‘alarm bell.’ This was due to: a) the isolated position from other staff, especially at night when most psychiatric assessments take place b) cubicles used are cramped, with inadequate seating and a lack of privacy c) the emergency department is where junior doctors are most likely to assess disturbed and potentially violent patients, and unlikely to have the support of psychiatric trained nurses or access to their notes prior to assessment. Features felt to be most important with agitated/potentially violent patients were space, access, layout, weapons, alarm and ease of exit. The study indicated a large disparity between the features of the ideal interviewing situation and those actually available. The authors recommend safety features should be incorporated, and that violent incidents should be monitored and logged, reviewed and acted upon. The numbers of staff surveyed were low (22) and of those only three were emergency department staff which, divided between three sites, provides a weak basis for generalisation.

Burns and Harm (1993) conducted a questionnaire and interview study of 682 emergency nurses. Interviewees reported feeling that there was a lack of concern for their personal safety. Emergency nurses found debriefings helpful. It was suggested that critical incident stress debriefing teams should formulate strategies for involving nurses in the debriefing process and that nurse peers should play a significant role in the debriefings.

Cembrowicz and Shepherd (1992) conducted a survey of staff in a UK emergency department, which concluded that fitted furniture and padded seating should be installed, and that potential weapons should be stored out of sight. They suggest that CCTV may deter casual hooliganism, but will be ignored by the highly intoxicated. In their view, security officers tend to be under-trained. They suggest that the use of uniforms may aggravate a violent situation.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence to determine how the environment and organisation impacts on the incidence of disturbed/violent behaviour by psychiatric patients in an emergency department.
The limited evidence suggests that staff feel that the environment in emergency departments often puts them at risk of disturbed/violent behaviour.
I.ii. Prediction: antecedents, warning signs and risk factors
I.ii.a. Review question: how is disturbed/violent behaviour by psychiatric patients best predicted in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting, and what are the key risk factors?

In a survey of staff and records in the violent incident book of the A&E department of a UK general hospital, Cembrowicz and Shepherd (1992) reported that in four of the last 20 incidents recorded, the violent patient was known to have a psychiatric illness. They found nursing staff and male doctors were most frequently assaulted; receptionists least frequently. The recording of information by staff was haphazard, sometimes due to time pressures, because they hadn't been injured themselves or because they wanted to avoid being blamed if they made frequent reports. If an assailant used a weapon, it was most likely to be any implement that happened to be ready-to-hand. The study concluded that staff need to be aware of body language which signals an angry outburst (flared nostrils, staring eyes, aggressive stance, pointing and pacing) and the risks of violent behaviour associated with intoxication.

Beck et al. (1991) conducted a case-control study using record review of 99 patients identified over six months as violent (evidence of assault or battery) or potentially violent (verbal threat or staff impression of poor control and anger or agitation); 95 control patients, judged not to be violent or potentially violent; and staff interviews. They found that women were more often violent, and men were more often potentially violent, made threats or were a source of concern to staff (p<0.03). Study patients were, on average, four years younger than control patients (p<0.005), more often brought in by police, and subsequently hospitalised (62 per cent vs. 29 per cent).

Cooper (1988) undertook a retrospective survey of patients referred by a general emergency department to a psychiatric unit. He analysed the antecedents and mapped the course of violent behaviour. Thirty percent had physically attacked another person immediately prior to presentation in the emergency department. Most of this violence was perpetrated by non-psychotic individuals in the throws of an interpersonal crisis. Twenty-five percent were found to be acutely intoxicated with alcohol, but intoxication may have gone undetected in many more. The majority of patients referred from an emergency department to a psychiatric ward were judged to be non-psychotic, presenting with situational crises and personality disorders rather than a major mental illness.

In a cross-sectional survey of all 130 qualified staff in the emergency departments of two hospitals, Lee et al. (2001) found that greater self-efficacy (judgements of what one can accomplish, rather than skills one possesses) was observed in higher grade staff who had experienced higher levels of aggression. The author notes that the nature of the association between self-efficacy and levels of violent behaviour encountered is not illuminated by this study. This study did not differentiate between violence committed by people with psychiatric illness and people without psychiatric illness.

Evidence statement

Level of evidenceEvidence statement
Level 4Limited evidence suggests that heightened arousal, depressive symptoms and alcohol intoxication are antecedents of disturbed/violent behaviour in emergency departments.
I.ii.b. Review question: what are the views of staff and service users about prediction of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence on risk factors for disturbed/violent behaviour in emergency departments from the perspective of staff and service users.
II. Training
II.a. Review question: how effective is training in the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments immediately prior to admission to an adult psychiatric in-patient setting?

No studies addressed this review question.

II.b. Review question: What are the views of staff and service users about training in the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence to determine the effectiveness of training in the short-term management of disturbed/violent behaviour in emergency departments, nor staff views on such training.
III. Minority ethnic groups, gender and other special concerns
III.i.a. Review question: how do ethnicity, gender or other special concerns impact on the short-term management of disturbed/violent behaviour by psychiatric patients in A&E settings, immediately prior to admission to an adult psychiatric in-patient setting?

No studies addressed this review question.

III.i.b. Review question: what are the views of staff and service users about the impact of ethnicity, gender or other special concerns on the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence to ascertain whether:
ethnicity, gender or special concerns (for example, disability) impact on or influence the approach to short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments.
There is no evidence to determine staff or service user perspectives on whether ethnicity, gender or other special concerns influence or have an impact on the short-term management of disturbed/violent behaviour in emergency departments.
IV. Psychosocial techniques
IV.i. De-escalation techniques
IV.i.a. Review question: are de-escalation techniques an effective tool for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?

Lee et al. (2001) found that nurses' aggression management training did not appear to equip them with the skills required to manage violent behaviour in emergency departments. He recommends that aggression management training should encourage nurses to examine their beliefs about violence and should focus on diffusion and de-escalation of violence, rather than control and restraint techniques.

Lane (1986) reports case studies of three patients admitted to the emergency department. He describes how techniques employing empathy (a combination of the suspension of judgement and sympathetic and creative imagination) were used to manage violence where more severe measures (restraint and medical management) might otherwise have been used. Generalisability is very limited.

IV.i.b. Review question: what are the views of staff and service users about the use of de-escalation techniques for the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence to suggest that de-escalation techniques are an effective tool for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments.
IV.ii. Observation
IV.ii.a. Review question: is observation an effective tool for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?

No studies addressed this review question.

IV.ii.b. Review question: what are the views of staff and service users about the use of observation for the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence on whether observation is an effective tool for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments.
V. Other Interventions
V.i. Pharmacological interventions: rapid tranquillisation and PRN medication
V.i.a. Review question: are pharmacological interventions effective and safe for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?

Roberts and Geeting (2001) describe the use of ketamine to tranquillise a dangerous and violent patient on admission to the emergency department. Within two to three minutes of intramuscular administration of 480mg ketamine (5mg/kg), violent activity had completely ceased. Mild sinus tachycardia and transient hypertension were observed 10 minutes after initial sedation, but all vital signs were normal 50 minutes after ketamine administration. Ketamine effects dissipated within two hours, however additional aliquots of lorazepam were required to control agitation over the next 12 hours. No immediate complications from ketamine or emergence phenomena were observed.

The authors argue that given ketamine's wide safety profile, potent anaesthetic effects, rapid onset, ease of intramuscular administration, absence of respiratory depression and short duration of action, it is considered useful for immediate tranquillisation of selected, undifferentiated, uncontrollable adults, who are in a life-threatening situation that requires immediate medical intervention. Concomitant use of benzodiazepines and selected use of atropine are suggested to ameliorate emergence phenomena, and to dry excessive oral secretions. The authors stress that, after use, close monitoring of cardiovascular parameters is essential. This is a case study of a single patient, so it is not possible to generalise.

V.i.b. Review question: what are the views of staff and service users about the use of pharmacological interventions for the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

Binder and McNiel (1999) conducted a cross-sectional survey of 20 medical directors of psychiatric emergency departments. The aim was to assess how acutely violent patients were managed in psychiatric emergency rooms, to examine medical directors' practices and to investigate emergency department characteristics. Fourteen reported that acutely violent patients are usually put in restraints and medicated intramuscularly or intravenously and given a medical assessment only after they were less agitated. Thirteen used the same acute medication regimen for all violent patients, regardless of eventual diagnosis. Eleven used haloperidol plus lorazepam, with or without benzatropine. Five used droperidol, in one case alongside lorazepam and diphenhydramine. Fifteen stated that the intramuscular route was the most common, two preferred the intravenous route whenever possible. All 20 felt that their preferred medication regimen was effective for calming the violent patient, usually after one dose and always after one to two repeated doses. Only three stated that agitated patients will usually take medications orally and that mechanical restraints are rarely used. Factors cited as allowing them to use less coercive techniques included: a system where most contacts were with people who were known to the system or who had case management protocols; a computerised system where information on patients is available within 30 to 60 seconds; a less violent patient population; and the availability of nurse clinicians who know the population. Results suggest that the strategies most frequently advocated in recent review articles for the assessment and management of violent patients are not generally applied by those responsible for the emergency management of acutely violent patients. Clinicians appeared to place the highest priority on prevention of patient and staff injuries, by rapidly reducing violent behaviour through restraints and intramuscular medications – typically a combination of neuroleptics and benzodiazepines – irrespective of diagnosis. The authors suggest that one could argue that these practices involve risks of excessive coercion, overmedicating patients, and exacerbating underlying medical conditions. On the other hand, the clinical experience of practitioners suggests that these strategies rapidly ameliorate acute violence and thereby reduce the risk of injury.

No studies were found on the use of PRN in emergency departments.

Evidence statement

Level of evidenceEvidence statement
Level 4There is no evidence to suggest that different medication regimes are either more effective or safe for rapid tranquillisation in emergency departments than medication regimes commonly used for the short-term management of disturbed/violent behaviour in psychiatric in-patient settings.
The limited evidence suggests that staff in emergency departments may use inappropriate strategies for dealing with the short-term management of disturbed/violent behaviour in service users.
V.ii. Physical interventions
V.ii.a. Review question: are physical interventions effective and safe for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments, immediately prior to admission to an adult psychiatric in-patient setting?

In Beck et al. (1991), a study in a psychiatric emergency service, psychotic patients who were restrained were 6.36 times more likely to be hospitalised than were psychotic patients who were not restrained (p<0.05). Non-psychotic patients who were restrained were 5.36 times more likely to be hospitalised than non-psychotic patients who were not restrained (p<0.03). Patients brought in by police were more likely to be put into restraints than patients brought in by others, and more likely than patients who came in unaccompanied.

V.ii.b. Review question: what are the views of staff and service users about the use of physical interventions for the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

Foust and Rhee (1993) conducted a prospective descriptive questionnaire study of all staff in an emergency department over nine months, to determine the incidence of battery (wilful and unlawful use of force or violence on the person of another) against emergency department medical staff by patients or visitors. Over the course of the study period, 19 instances of battery occurred. In eight cases, battery occurred when the patient was restrained, and four when the patient was restrained, but restraint was being modified.

The department's unusual restraint policies were described, whereby all patients placed on psychiatric or substance abuse ‘holds’ (requiring that they are a danger to themselves, to others or gravely disabled) be restrained with a loosely applied cloth belt that encircles the abdomen. Four of the incidents occurred when the patient was in abdominal restraint only, and eight other incidents when in abdominal and extremity restraint. The restraint procedures were not described. It was noted that 79 per cent of battery was carried out by patients with psychiatric problems or who were intoxicated. Consistent with other studies, incidents were significantly under-reported. In only four cases were hospital incident forms filled in. Authors recommend that strategies to prevent or control violence should be concentrated on evening and nightshifts.

Evidence statement

Level of evidenceEvidence statement
Level 4There is insufficient evidence to assess whether physical interventions are either effective or safe for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments, on the basis of the available literature.
V.iii. Seclusion
V.iii.a. Review question: is seclusion effective and safe for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency, immediately prior to admission to an adult psychiatric in-patient setting?

No studies addressed this review question.

V.iii.b. Review question: what are the views of staff and service users about the use of seclusion for the short-term management of disturbed/violent behaviour by psychiatric patients in an emergency department?

No studies addressed this review question.

Evidence statements

Level of evidenceEvidence statement
Level 4There is an absence of evidence to determine whether seclusion is either effective or safe for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments.
There is an absence of evidence to determine staff or service user views on the use of seclusion for the short-term management of disturbed/violent behaviour by psychiatric patients in emergency departments.

7.8.7.4. Economic evidence

No studies containing relevant economic data were found (see Appendix 9).

Copyright © 2006, Royal College of Nursing.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means electronic,mechanical, photocopying, recording or otherwise,without prior permission of the Publishers or a licence permitting restricted copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. This publication may not be lent, resold, hired out or otherwise disposed of by ways of trade in any form of binding or cover other than that in which it is published,without the prior consent of the Publishers.

Cover of Violence
Violence: The Short-Term Management of Disturbed/Violent Behaviour in In-Patient Psychiatric Settings and Emergency Departments.
NICE Clinical Guidelines, No. 25.
National Collaborating Centre for Nursing and Supportive Care (UK).

Download

NICE (National Institute for Health and Care Excellence)

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...