Assessing violence risk in general adult psychiatry

Abstract

Aims and Method

We aimed to establish current practice in the risk assessment of harm to others within general adult psychiatry and review risk assessment documentation in use. Consultants working across 66 randomly selected trusts across England were surveyed. A qualitative analysis of risk assessment documentation was carried out.

Results

Data were obtained from 45 trusts (68%). Consultants reported that 30 (67%) of the trusts had standardised forms for risk assessment. Forty-one forms were subjected to content analysis. Wide variation was found in the methods used to identify risk factors and in approaches to quantifying risk.

Clinical Implications

Current risk assessment practice is highly variable, indicating a lack of consensus about suitable methods.

Government policy stipulates that service users’ risk of harm to others should be routinely assessed by specialist mental health services (Department of Health, 2000). Responsibility for this assessment and its documentation now affects the practice of all mental health professionals (Duggan, 1997). This increasing emphasis on risk has not been universally welcomed. It has been argued that it compounds the public perception of people with mental illness as dangerous (Petch, 2001), and consumes resources without a sound evidence base for doing so (Bindman et al, 2000). An analysis of inquiries into homicides by people with mental illness found the role of risk assessment to be limited and concluded that the most effective preventative strategy would be improving treatment for all patients, regardless of perceived risk (Munro & Rumgay, 2000).

Proponents of risk assessment argue that it simply requires basic clinical skills (Snowden, 1997), that the process itself can be valuable (Holloway, 1998), and that it is inseparable from risk management (Kennedy, 2001). Research has identified relevant dispositional, historical and situational risk factors for violence (Monahan & Steadman, 1994), and it has been suggested that actuarial methods might enhance predictive accuracy (Dolan & Doyle, 2000). Actuarial methods are, however, unlikely to be of use in populations with low base rates of violence, in which large numbers of false positives would be generated (Szmukler, 2001). This presents a difficulty for generalists who are required to assess violence risk routinely (Holloway, 1997), but without clear methods for doing so.

A previous study of the Supervision Register policy, which also required risk assessments within general psychiatric services, showed widespread variations between trusts in the criteria for identifying high-risk patients and suggested that several different methods were used (Bindman et al, 2000). We surveyed a representative sample of mental health trusts and aimed to first establish current violence risk assessment practice and second, describe and evaluate documentation produced at a local level with the intention of supporting violence risk assessment. This survey was a component of a wider project (the Clinical Assessment of Risk Decision Support, or ‘ CARDS’ study) to develop an evidence-based procedure for assessing violence risk in patients using adult mental health services (Watts et al, 2004).

Method

We designed a brief semi-structured questionnaire about the use of risk assessment documentation, training and guidelines in trusts (available from the authors). From a database of all mental health trusts in England developed for a previous study (Bindman et al, 1999), 66 were randomly selected. They were contacted and the names obtained of two general adult consultants for each trust. The consultants were sent the questionnaire by post and asked to send copies of their trust's risk assessment forms and guidelines for their use where these were available. A second questionnaire was sent to non-respondents after 6 weeks, followed by a telephone reminder. A content analysis of the risk assessment forms was then carried out. Principal themes and the different components used in assessments were identified independently by two of the authors (N.H. & J.B.), and a consensus reached.

Results

Survey response

At least one consultant responded from 45 of the 66 trusts (68% trust response rate). Of these 45 trusts, 30 (67%) were reported by the consultants to have standardised written forms for assessing the risk of violence and a further 2 (4%) were in the process of developing forms. Twenty-one trusts (47%) provided training for their use, mostly in the form of half-day multidisciplinary sessions incorporating wider risk issues. It was of note that where training was in place, many respondents commented that they had not attended it. Fifteen trusts (33%) provided written guidelines. A risk assessment form was usually completed at the time of referral in 20 trusts (44%); at hospital admission in 26 trusts (58%); before discharge in 25 trusts (56%); after expression of concern about safety to others in 20 trusts (44%); after a violent incident in 22 trusts (49%); or most often at the CPA review in 29 trusts (64%). Thirteen (29%) of the forms used included a proposed review date.

Structure and content of risk assessment forms

Forty-one forms were returned for analysis; several trusts had more than one form in use, depending on the hospital or community context. All violence risk assessments were subsumed within or appended to general risk assessment forms that included self-harm. In broad terms, three objectives from the forms emerged:

1. 1. identification of specific risk factors

2. 2. appraisal of overall risk

3. 3. risk management planning.

Risk factors were identified by tick box screens, narrative sections or by using a combination of both. Different components used with these approaches are shown in Box 1. Risk was summated in one of four ways, as shown in Box 2.

An example of a structured narrative approach is shown in Box 3. This contrasts markedly with forms using tick boxes alone, which recorded answers to questions such as ‘Does the client have a history of violence?’ as simply ‘yes’ or ‘no’ without further elaboration. Although all forms shared the aims of identifying risk factors and appraising overall risk, it was noteworthy that only around half included a risk management plan or a section prompting further action from a positive screen, e.g. to collate further information or perform a more in-depth analysis.

Box 1. Methods for identifying risk indicators

Tick box sections

1. • Past history of violence

2. • Current factors increasing risk (e.g. substance misuse)

3. • Social/demographic factors

4. • Current symptoms (e.g. command hallucinations)

5. • Dispositional factors (e.g. impulsivity)

6. • Possession of weapons

7. • Threat posed to any identified individual

8. • Historical factors (e.g. childhood emotional deprivation)

Narrative sections

1. • Unstructured account of past violence

2. • Structured account of past violence, context and outcome

3. • Unstructured account of current risk indicators

4. • Structured account of current risk indicators

Box 2. Methods for summating risk indicators

1. 1. Dichotomous division: yes or no

2. 2. Grading: high/medium/low

3. 3. Scoring (weighted or otherwise)

4. 4. Narrative formulation

Box 3. Example of a structured narrative approach

1. 1. Is there evidence of violence particularly preceded by changes in mental state? If yes, describe. Give details of any plans or specific threats to harm a person. Do they have access to potential victims, particularly those incorporated into delusional systems? Are there specific trigger factors? Who needs to be aware of this risk?

2. 2. Is there evidence of poor compliance with treatment or disengaging from services? If yes, then how will this affect the management of any identified risk?

3. 3. Is there evidence of rootlessness or social restlessness (homelessness, social isolation, frequent changes of address or employment)? If yes, then how will this affect the management of any identified risks?

4. 4. Is there evidence of substance misuse? If yes, how does this affect the management of any identified risks?

5. 5. Are there aspects of their mental state that may constitute or exacerbate risk (these may include: persecutory delusions, morbid jealousy, passivity)? If yes, describe.

6. 6. History (chronology of events)

7. 7. Have any of the above factors changed recently? If yes, describe.

8. 8. Opinion

9. 9. What other information is required to complete the assessment? Include: seriousness, immediacy, volatility, specific interventions and treatment that will reduce risk, circumstances that may increase it. How long is this opinion current?

Discussion

Principal findings

The responses from the survey suggest that most trusts have standard risk forms incorporating the assessment of violence and around half provide training for their use, although many consultants do not attend. In comparing the forms themselves, there was evidence of striking variation. In identifying risk factors, forms varied considerably in their content and complexity. Unstructured narrative sections would appear to rely on the knowledge of the person completing the form as to what information is relevant; tick-box completion, by contrast, has the advantage of prompting the consideration of pertinent factors. However, simply ticking a box to indicate a risk factor as present arguably communicates little useful information. For example, how a patient understands and responds to passivity experiences, and what the person completing the form is describing when they say they are present, is crucial for putting a risk factor into the context of actual risk for that individual. Structured narrative sections appear to combine the best elements of both methods by directing the focus of inquiry while allowing risk factors to be contextualised. Some forms only identified discrete risk factors; here risk ‘ assessment’ is arguably a misnomer, as assessment implies some form of weighing up of available information.

The rationale behind using scoring or grading systems to summarise risk was not clear. Scores may be reproducible, and thus seem ‘scientific’, but their validity for use with the general population is questionable. There was often a lack of direction as to how a score or grade should be meaningfully interpreted. A score or grade as the conclusion of a risk assessment may be a false positive or negative, which may mislead management. A narrative summation has the advantage of collating what is actually known for that individual, and allows the balancing of risk and mitigating factors.

Around half of the forms did not include a plan for managing any identified risk. It could be argued that assessing risk should not be an end in itself. Clinical management plans are detailed elsewhere in care programme approach documentation, but these do not necessarily demand a specific focus on risk behaviour.

Strengths and weaknesses of the study

This study is the first to provide an overview of violence risk assessment practice in England. The trust response rate (68%) and the number of forms analysed (41) was sufficiently high to be representative. As this was the first study of its kind, we did not measure form quality in a reproducible way; there is not as yet any ‘gold standard’ for forms to be judged against. The main limitation of this study was the reliance on the consultants’ reports of trust policies and their best estimations of current practice in their area.

Implications

Trusts in England are in the main complying with the Department of Health's requirement to assess risk of violence. How this requirement is interpreted varies considerably between trusts. This is at odds with a culture of evidence-based practice.

Unanswered questions and future research

A consensus needs to be reached as to what risk assessment should entail in general psychiatry. Given the current state of knowledge, we would suggest that this should include semi-structured methods and not scoring or weighting systems, which in this context are somewhat specious.

Declaration of interest

This work was funded by a grant from the Policy Research Programme, Research and Development Division, Department of Health in England. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.