AIMS AND METHOD
We surveyed the usage and perceived utility of standardised risk measures in 29 forensic medium secure units (a 62% response rate).
The most common instruments were Historical Clinical Risk–20 (HCR–20) and Psychopathy Checklist – revised (PCL–R); both were rated highly for utility. The Risk Matrix 2000 (RM2000), Sex Offender Risk Appraisal Guide (SORAG) and Static-99 were the most common sex offender assessments, but the Sexual Violence Risks–20 (SVR–20) was rated more positively for its use of dynamic factors and relevance to treatment.
Most medium secure units use structured risk assessments and staff view them positively. As HCR–20 and PCL–R/PCL–SV (Psychopathy Checklist – Screening Version) are so widely used they should be the first choices considered by other services.
Violence risk assessment is central to the work of forensic mental health services. Standardised methods of assessment have become more common but there is great variation between services in the use of such instruments.
The alternatives to clinical assessment alone are actuarial methods (e.g. using the Violence Risk Appraisal Guide (VRAG); Quinsey et al, 1998) that prescribe the collection and interpretation of data relevant to risk (e.g. previous violence, substance misuse, psychopathy); and structured clinical methods (e.g. Historical Clinical Risk–20 (HCR–20); Webster et al, 1997) which require collection of similar data but also require the use of clinical discretion in using additional information and in how to interpret the data).
The current consensus is that structured clinical assessment is the best option for clinicians (Monahan et al, 2001). Recent guidelines from the Department of Health's Best Practice in Managing Risk (Department of Health, 2007) and the Royal College of Psychiatrists' Giving up the Culture of Blame: Risk Assessment and Risk Management in Psychiatric Practice (Morgan, 2007) have further supported the use of structured clinical methods for assessment of violence risk. However, the extent to which UK forensic mental health services have adopted these methods remains unknown.
The study aims were:
to measure how many medium secure services use structured violence risk and sexual offender risk assessment instruments;
to identify which instruments were used;
to measure their perceived utility.
A search was conducted for all medium secure forensic services on an internet database (www.theforensicdirectory.com), giving a sample of 47 adult medium secure forensic units (28 National Health Service (NHS) units, 19 independent units).
A questionnaire was designed for the study. It covered unit size, case mix and staffing. Regarding violence risk assessment methods (see online supplement) we asked specifically whether named instruments, i.e. Psychopathy Checklist – Revised (PCL–R; Hare, 1991), Psychopathy Checklist – Screening Version (PCL–SV; Hart et al, 1995), VRAG, Violence Risk Scale (VRS; www.psynergy.ca/pdf/vrssummary.pdf), Iterative Classification Tree (ICT; Monahan et al, 2000), Offender Group Reconviction Scale (OGRS; Taylor, 1999), Risk Assessment and Management Schedule (RAMAS; O'Rourke, 1995) and Historical Clinical Risk–20 (HCR–20), were used `frequently', `occasionally' or `never'. If used, respondents were asked to rate utility of each method in routine practice on a five-point Likert scale (ranging between `not useful' and `very useful'). If the unit accommodated people who were sex offenders, the process was repeated for sex offender risk assessments, namely the Rapid Risk Assessment for Sex Offence Recidivism (RRASOR; Hanson, 1997), Sex Offender Risk Appraisal Guide (SORAG; Quinsey et al, 1998), Static-99 (Hanson, 1997) and Risk Matrix 2000 (RM2000; Thornton, 2003). Respondents were asked to identify and rate any additional risk assessment instruments in use that were not listed on the questionnaire.
The questionnaire was sent in April 2007 to clinical directors with a covering letter (explaining the purpose of the study and maintenance of anonymity) and a stamped addressed envelope. Reminders were sent to non-respondents who were also followed-up by telephone. Respondents were telephoned for follow-up qualitative interviews, the notes of which were analysed by simple thematic analysis.
Responses were received from 29 (19 NHS, 10 independent) of 47 medium secure services surveyed, giving a 62% response rate. We attempted to obtain follow-up qualitative telephone interviews from all 29 responder units, out of which 11 (9 NHS, 2 independent units) were achieved. Clinicians at the other units were unavailable to give interviews.
Unit size ranged from 17 to 276 beds with a mean of 76.0 (median 59.5). Clinical teams included a mean of 5.5 consultant psychiatrist whole time equivalent posts (range 1–25) and 5.9 psychologist posts (range 1–19.5). Fifteen units (52%), all of which were NHS units, provided outreach or community forensic services, whereas 14 units (48%) did not. National Health Service and independent units differed significantly regarding the provision of outreach or community services, but not in any other way (regarding bed numbers, staff numbers and numbers of units using each risk assessment instrument).
Online Table DS1 shows the frequency of use and perceived utility of violence risk assessments (see online data supplement). The PCL–R and HCR–20 were used by most units, often being used together. Clinicians were trained in the use of the PCL–R and described it as familiar, simple and well validated. Many clinicians described the HCR–20 as an `aide-memoire' in clinical practice, and used it to guide clinical management. Clinicians also liked the fact that staff of all disciplines can use it, so it is easily incorporated into team working.
The START (Short-Term Assessment of Risk and Treatability; Webster et al, 2004), was rarely used but received the highest utility rating. Two forensic units used self-generated scales (mean utility rating 3.50); a clinician at one of these units expressed concern about the scale's lack of external validation and unfamiliarity to clinicians in other services.
Online Table DS1 also summarises the findings regarding the frequency of use and perceived utility of sex offender risk assessments. Sex offender risk assessments were being used in 20 (69%) of the responder medium secure units. The RM2000, Static-99 and SORAG were used by the most units. The Sexual Violence Risks–20 (SVR-20; Boer et al, 1997) scored highest for utility (rated 5 out of 5 by all six units that used it). Important factors in determining the utility of a sex offender risk assessment instrument were familiarity, training, validation and clinical usefulness of risk assessment scores.
Qualitative interviews found the HCR–20 was favoured because it was accessible to all disciplines; it provided comprehensive information about violence risk; it helped with risk management; it was tailored to the individual because it included specific risk scenarios; its dynamic content allowed monitoring of change; and it was widely understood by other clinicians.
The PCL–R was thought to be useful in cases of suspected psychopathy. Clinicians liked the sophisticated psychological training, and the scale was widely used, thus facilitating communication.
Actuarial assessments were used frequently, but clinicians thought it best to use more than one and incorporate scores into a wider risk assessment involving clinical judgement. Thus, they were used as part of structured clinical assessment rather than as stand alone measures. Reasons for adopting a specific method included encouragement or insistence by the local healthcare trusts, and research evidence. Clinicians were unsure of best practices for sex offender risk assessment, as there was a bewildering array of tools, many developed from US prison populations with limited validation in UK populations. Personal preferences influenced choice of tools. The SVR–20 was highly rated because of the specialist training received before use, and clinical utility of the scores.
Forensic units used the results of risk assessments to predict risk scenarios; to reduce risk of absconding; to inform decision-making in CPA (Care Programme Approach) meetings and ward rounds; and to guide treatment, management and rehabilitation of individuals.
This study is the first to examine the use of structured violence and sex offender risk assessments in UK medium secure forensic units and shows that these methods have been widely adopted in a relatively short time.
The study had limitations. First, the reliability of the new internet database (www.theforensicdirectory.com, used to identify the sample) is unknown and some forensic units may have been omitted from the sample. Second, although the response rate is respectable for surveys of this kind, given the small target population we would have preferred a higher response rate in both the postal survey and telephone interviews. There may be bias in that units that have not adopted structured methods may be less likely to respond.
Recent research showed actuarial risk assessment instruments, namely the VRAG (for violence risk) and Static-99 (for sex offender risk), which have high `margins of error' at the group level, but so high at the individual level as to render risk estimates virtually meaningless (Hart et al, 2007).
Initially, it may appear worrying that actuarial methods of violence risk assessment (PCL–R) and sex offender risk assessment (RM2000, Static–99 and SORAG) were popular in our study. However, they were generally used appropriately as a supplement to clinical assessment. Clinicians acknowledged their limitations and used them responsibly, tailoring them to fit the individual patient. Both the VRAG and Static–99 were rated relatively low for clinical utility (3.60 and 3.33 respectively), but were nevertheless considered useful additions to comprehensive assessment.
Structured clinical instruments, particularly HCR–20 and SVR–20, scored higher for utility and were used by most units. Our interviews suggested that clinicians were persuaded of the value of these instruments in summarising risk factors and assisting the development of management plans (Doyle & Dolan, 2006).
The benefits of structured clinical risk assessment operate along two dimensions that are somewhat independent. First, they may increase accuracy of risk assessments, although it is debatable whether any particular instrument is superior to another. Specialist services are taking a sensible approach in applying more than one measure. The second major benefit is in providing transparency, plus a shared language for describing and communicating about risk. These are desirable goals in their own right, being fully consistent with broader aims of greater accountability and tighter clinical governance. Success in these aims depends less on actual instruments used than on consistency between services. Communication is best, and scrutiny easiest, when services use the same measures. Our study suggests the HCR–20 and PCL–R (or PCL–SV) are becoming the de facto standard within medium security, which should make them the first choices for other services.
Thanks to all the forensic units and clinicians that participated in the study.
- © 2009 Royal College of Psychiatrists