Report of the Committee on Serious Violent and Sexual Offenders

CHAPTER 2: RISK

Risk assessment and management

2.1 Our Committee's terms of reference require us to consider serious violent and sexual offenders who may present a continuing danger to the public. The central question, before considering any special sentencing powers or management approaches is - To what extent is it possible to know who presents a high risk of committing a serious offence? Our view is that the state of knowledge on risk assessment has improved in recent years and is likely to improve further as assessment tools are refined and validated for a Scottish population.

2.2 It is reasonable for decisions in the criminal justice and mental health systems to be informed by risk assessments, and this should be done to a greater extent than is currently the case. However, no current risk assessment procedure, nor any which is likely to be developed, can predict future human behaviour with anything approaching certainty. The risk assessor can determine whether an individual has a greater propensity to use violence than others, but cannot know when an individual's volition, or the circumstances in which he/she finds him/herself, will precipitate a violent act.

Definitions

2.3 Although words such as 'risk' are in everyday use, it is important to be clear as to what is meant. For our purposes, we are particularly concerned with risk assessment in relation to future serious violence and sexual violence. A definition which we have found helpful is that of Dr Stephen Hart, who defines violence risk assessment as:

'...the process of evaluating individuals to characterise the likelihood they will commit acts of violence and develop interventions to reduce that likelihood'⁸.

This definition is particularly helpful because it does not view risk assessment as an end in itself but links it to positive action to manage and reduce risk.

2.4 The term risk is preferred to 'dangerousness', because the term dangerousness implies a dispositional trait, inherent in an individual, that compels him/her to engage in a range of violent behaviour across a range of settings. That approach fails to take into account the complex interaction of psychological characteristics and situational factors in the production of violent acts. Violent individuals (including those who may have certain personality characteristics) are more likely to be violent in certain contexts. The response to risks presented by individuals should, therefore, not be restricted to an attempt to modify those characteristics in order to make the individuals less of a risk, but also seek to reduce the opportunities or triggers for violence.

2.5 We are concerned with risk of violence. Unfortunately, there is no universally agreed definition of violence or sexual violence in the risk assessment literature. This is important because it can be difficult to evaluate the effectiveness of different methods of risk assessment and risk management without a consistent definition of what is the particular undesired outcome.

2.6 In our view, the 'continuing danger to the public' referred to in our remit is the danger that the offender will commit further acts of serious violence, or sexual crimes. We have excluded from our consideration the risk that an offender may commit less serious crimes, such as minor breaches of the peace. The effect on the public and victims of less serious crimes is not to be minimised, but is a broader social problem than we are able to address.

2.7 Less serious offending behaviour may nevertheless be relevant to risk assessment, particularly if it suggests a pattern of behaviour that might lead to an act of serious violence.

2.8 The definition of risk assessment given above emphasises that it is a process, not a once-and-for-all event; the process should include assessment, review and re-assessment.

2.9 In short, risk assessment should be seen as an aid to making decisions as to what combination of controls and interventions should be applied to a person in order to manage the risk he or she presents.

Making better use of risk assessment

2.10 In our view, it is possible, within limits, to identify those offenders who pose a risk of future violence. No approach is without error. Furthermore, any assessment of risk is not necessarily wrong simply because an outcome occurred for which a low risk was predicted, whereas an outcome judged to be high risk did not. It is in the nature of any assessment of relative probabilities that such results will, from time to time, happen.

2.11 Nevertheless, the degree of accuracy with which we can predict risk of violence is comparable to, or greater, than that achieved in other important human decisions; for example, the likelihood of cardiac bypass surgery improving mortality rates, the effect of small class size on academic achievement, or the impact on mortality of chemotherapy for breast cancer.

2.12 Risk assessment is central to many decisions in the criminal justice process; overtly so in the case of decisions by the Parole Board and Scottish Ministers in relation to release of prisoners, but also in relation to sentencing, and the many day-to-day decisions that are made about offenders, such as whether to allow a prisoner to move to open conditions, or to allow home leave. Similarly, in mental health law, risk is a factor in the decision to detain or to discharge from detention (particularly for restricted patients following the Mental Health (Public Safety and Appeals) (Scotland) Act 1999), and it influences decisions concerning the level of security or freedom a patient may be accorded.

2.13 It is our view that the role of risk assessment in sentencing, management and release needs to be more clearly acknowledged, and the different types of risk assessment and management need to be better integrated.

2.14 Nevertheless, as a number of respondents to our consultation pointed out, any risk assessment process creates significant numbers of 'false positives' (people assessed as high risk who do not in fact offend) and 'false negatives' (people assessed as low risk who do offend).

2.15 More generally, risk assessment is an example of David Hume's famous dictum, that one cannot derive an 'ought' from an 'is'. Decisions about what level of risk justifies some form of special measure are, ultimately, matters of social policy, not scientific measurement. Where thresholds are set, those who are just above the threshold may not differ markedly from those just below the threshold. Therefore, a graded and flexible set of responses is needed, not simply an attempt to 'catch' the individuals presenting the highest risk.

2.16 Although, as we have said, risk assessment has developed markedly in recent years, it is likely to develop still further. We have sought, therefore, to recommend systems and procedures which will allow for the continued development and application of improved assessment methods.

2.17 Society, through its politicians, and the media, must learn to accept that, in dealing with human behaviour, risks can never be eliminated. Nothing is more likely to dissuade agencies, who may have something to contribute, from offering services than the certain knowledge that they will be pilloried if 'things go wrong'. If society is not prepared to accept any risk then it seems to us fundamentally wrong to expect any professional to carry what is an impossible burden. The issue is one upon which politicians must decide and then give a clear lead.

Approaches to risk assessment

2.18 There are three broad approaches which have been adopted to risk assessment in relation to future violent behaviour: clinical, actuarial, and structured clinical judgement. The distinction between the approaches lies in the process used rather than the variables considered; clinical variables (for example, a diagnosis of substance misuse) may be used in all three approaches. No approach has yet been fully validated for a Scottish offender population. In that connection, we will
later go on to make recommendations about current research priorities. However,
we believe that, on current evidence, the approach of structured clinical judgement is the most suitable one for the purposes of our criminal justice system.

2.19 The clinical approach to risk assessment is the approach that is currently most widespread in Scotland - both in clinical settings such as hospitals, and in non-clinical settings where decisions about future risk of violence are taken, e.g. in relation to parole and probation. Essentially, the clinical approach has been criticised as relying on 'an informal, "in the head" impressionistic, subjective conclusion, reached (somehow) by a human clinical judge⁹.' This characterisation sounds harsh, but the evidence is unequivocal that clinical judgements, even by experienced practitioners, are poor predictors of future violence. In his classic review of the field in 1981, John Monahan indicated that only one in three positive predictions of violence made by mental health professionals was accurate.¹⁰

2.20 A subjective, perhaps even instinctive, approach to risk assessment is not confined to the medical profession. It has historically been the case that sentencers, and those making decisions on release of offenders, have made such decisions with limited information and little guidance as to how to make use of what information they had. Decisions have, of necessity, been based as much on 'feel' as on solid evidence. We believe that, on the basis of developments in the evidence base for risk assessment, this approach should now change.

2.21 This is not to deny that a professional with experience and specialist skills can contribute greatly to risk assessment. It is simply to say that for any such assessment to have real predictive power it must be carried out in a structured way, having due regard to all relevant factors. Without this, the continued use of a clinical approach to risk assessment for violent and sexual offenders cannot be supported.

2.22 It must be stressed that we do not recommend that sentencing policy in general should be made more rigid. A sentencer must take account of many factors in determining the appropriate sentence for a particular offence. Most sentencing decisions are not based primarily on dealing with risk of future violence on the part of the offender, but on issues of appropriate punishment, deterrence and so forth. These are complex decisions, and it has long been the tradition in Scotland that it is for judges to have a wide discretion in making them.

2.23 However, there are some decisions, even with the system as it currently stands, which predominantly concern an appraisal of risk. It is our view that such decisions should be made on the best available evidence, and using the most accurate techniques. The unstructured clinical approach can amount to little more than an assertion by a clinician as to the level of risk, often set against a competing assertion by another clinician, with little basis for the court to decide between them.

2.24 The actuarial approach has been put forward as a more valid means of identifying risk of future violence. There is a variety of actuarial methods; the essence of them all is that they involve a formal, algorithmic, objective procedure to reach the decision as to risk. Annex 6 summarises the main actuarial instruments in current use.

2.25 We found the actuarial approach to be particularly favoured in the USA. There is strong evidence that actuarial approaches have greater predictive power than unstructured clinical approaches. However, they have a number of weaknesses, which lead us to the view that they should not be used, on their own, as the primary mechanism for undertaking risk assessment.

2.26 There is little evidence as to their accuracy for a Scottish population. This is not a trivial point. Many of the actuarial tools were based on analysis of re-offending amongst a specific, and often highly selected, population. The evidence we received in the USA and Canada was that the Violence Risk Assessment Guide (VRAG) was the most widely used and reliable predictive tool but there is as yet no evidence as to its applicability to Scotland.

2.27 Even their predictive capacity may be compromised if there are factors particular to the individual which affect risk levels, but are not contained in the instrument, e.g. the fact that the individual has acquired a physical disability, or made overt, specific threats.

2.28 Actuarial methods are limited to prediction: they do not generate any strategies for managing risk. Furthermore, many of the risk factors are historical, e.g. history of childhood disturbance, and cannot be modified, so they may be of little assistance in considering when a person should progress through a progamme of care with changing levels of security.

2.29 An actuarial approach is also difficult for a court to apply confidently, since what it gives is an overall 'score'. It is often difficult to dispute the score, and so the court may have little choice other than to accept the actuarial finding, or enter into debate about its scientific merit, rather than the more relevant issue of the risk presented by the individual before it.

2.30 In recent years a new approach to the assessment of risk has been developed, that of structured clinical judgement. It should be stressed that this is an approach, not a particular instrument. The approach seeks to combine the systematic and evidence-based elements of the actuarial approach with the sensitivity to individual risk factors of the clinical approach. It both assesses risk and aids discussion of how best the risk should be managed.

2.31 Structured clinical judgement requires consideration of risk factors that have received empirical support in the literature. Decision making is assisted by guidelines that have been developed to reflect the current state of knowledge. Guidelines of this sort are increasingly common in general medicine, and are becoming more prevalent with the drive towards evidence-based medicine, although they are less frequently used, to date, in psychology or psychiatry.

2.32 Our terms of reference ask that we inform the development of a medical protocol to respond to the needs of personality disordered offenders. As we go on to explain [see paragraph 11.19], we are not convinced that a medical protocol is the best mechanism for dealing with services for, and treatment of, offenders with personality disorders. We do believe, however, that protocols need to be developed for the multi-professional assessment of risk and risk management, and a structured clinical approach is the most suitable foundation currently available for such protocols.

2.33 Our Committee received an extremely helpful presentation from one of the authors of the HCR-20, one of the best known and best researched of the risk assessment tools which are based on the structured clinical approach. Annex 6 describes the main features of this tool. Early results suggest that the HCR-20 shows considerable promise for the prediction of future violence and the management of those who pose a risk of violence, although, again, research is needed into its suitability for the Scottish population.

2.34 The structured clinical approach is based on assessment by trained people with appropriate expertise. It supports a multi-disciplinary approach, rather than the more traditional model of investing a particular professional with an assumed unique insight into the danger presented by an individual. At its basic level, it requires due consideration to be given in risk assessment to a wide range of factors which have been shown empirically to have a bearing on risk. These include both historical factors (such as a history of previous violence) and those which may be subject to change (such as active symptoms of mental illness). It also has regard to questions relevant in risk management, such as the extent to which the individual would be exposed to destabilising factors.

2.35 The outcome can be presented in various ways, according to the purpose of the assessment. In the context of assessment for a court, the outcome may be represented as indicating high, medium or low risk. It can also be used to prepare a full description of the level and type of risk presented by an individual.

2.36 In our Committee's view, it is necessary to have a risk assessment process that is not only defensible in terms of its scientific basis, but also one which can be applied for practical purposes. In particular, it must meet the needs of sentencers and others who take decisions on release and discharge of offenders. A court considering a special disposal based on risk should require evidence that is soundly based, and that can be tested: in cross-examination and against competing evidence. In the Scottish tradition it is also important that the expert witness does not take over the decision-making authority from the sentencer. It is the sentencer who decides what weight to attach to the expert evidence.

2.37 The structured clinical approach, in our view, meets these requirements. It was, after all, developed with these requirements very much in mind. The basis of a finding by an assessor that an individual presents, for example, a high risk of future violence, can be set out and the individual components of that assessment can be challenged.

2.38 While knowledge is improving, there is a great need for more research. In particular, we recommend:

Developing a co-ordinated risk assessment and management process

2.39 It is our view that a properly co-ordinated risk management process should be developed for serious violent and sexual offenders. We believe that this should have the following key features, namely it should:

be based on best practice:

• be empirically based
• take a multi-disciplinary and multi-modal approach to information
• be based on access to all relevant information
• allow sufficient time for assessment - where appropriate on a residential basis
• use all appropriate technology

be dynamic:

• be sensitive to changing conditions
• continue over time

produce practical outcomes:

• identify factors which aggravate or mitigate risk
• lead to management strategies
• involve systematic monitoring of risk outcomes

be open and regularly monitored:

• be transparent and open to challenge
• contain quality control systems

Best practice

2.40 Risk assessment should be empirically based, i.e. supported by acceptable scientific evidence. For many years, risk assessment was not based on any evidence. Such evidence as was available suggested that attempts to gauge risk were little or no more successful than chance. In such circumstances, it is understandable that people relied on intuitive approaches. Today, the developing body of evidence in relation to risk assessment and risk management has rendered the intuitive approach unsustainable.

2.41 We received evidence that the practice of risk assessment could be greatly improved. In particular a great many respondents (including the majority of local authority respondents and bodies such as the Mental Welfare Commission, the Royal College of Nursing, and the Association of Directors of Social Work) called for a multi-disciplinary approach to the assessment of offenders with personality disorders or mental disorders, particularly prior to sentencing.

2.42 A number of local authorities specifically called for a national framework for such multi-disciplinary assessment. The Scottish Executive has produced a framework for risk assessment in local authorities¹¹, and this could form the basis of a broader approach. We address this in our recommendations regarding the Risk Management Authority (See Chapter 3).

2.43 Several respondents took the view that risk assessment should be uniform across different agencies. Consistency is to be welcomed, but uniformity may not be possible nor desirable since the different agencies are assessing risk for different purposes. We believe that there should be compatibility between different approaches, and that quality should be standardised.

A dynamic approach

2.44 The dynamic approach, continuing over time, assumes that a decision about risk should not be seen as a 'once and for all' event. Risk requires to be assessed at many stages. Many factors will be common to each risk assessment, and so it is in the interests of accuracy and efficiency if risk assessments are carried out using a common language, and sharing common information. It is also important to be able to assess what may have changed in the presenting risk, which can only be done if assessments are carried out in a consistent way.

2.45 It is also important that risk assessment is consistent with, and leads to, practical outcomes. The goal is not simply to measure risk, but to work out how best to reduce risk.

Access to information

2.46 An area where considerable improvement is possible is that of access to relevant information¹². In our recommendations on sentencing, we emphasise that any risk assessment is ultimately dependent on the quality of available information. Many people have pointed out to us the difficulty faced by a person seeking to assess the risk presented by an offender when the principal source of information is the offender him/herself. Among the factors that are of great significance in assessing risk is knowledge of all the circumstances of the current crime, and those of previous crimes. It is often of great importance to be aware of matters such as episodes of antisocial behaviour in childhood and adolescence.

2.47 Lady Cosgrove's Expert Panel has undertaken a considerable amount of work on this issue, and we are grateful to them for sharing it with us. In overcoming the barriers to the proper use of information it is important to identify the nature of the various difficulties. Some of these are structural (for example, issues about where information is stored and in what form). Some are attitudinal, arising at worst from mistrust of other professionals, and reluctance to give up control of knowledge. Related to these are issues of professional duties of confidentiality. It has been argued that the problems presented by confidentiality are often overstated, and can be alleviated by a clearer understanding of the limits of client confidentiality. Nevertheless, the problem does exist. Finally, there are serious issues over the extent to which information that may be contested, or unproven, can be taken into account in a sentencing process (discussed at paragraphs 5.11-5.26).

New procedures

2.48 In North America, considerable reliance is placed on technological devices for risk assessment which are little used here, notably the penile plethysmograph. In our view the penile plethysmograph, when operated by appropriately trained staff, may add to the evidence available in a risk assessment for sexual violence and avoid the reliance on self reporting by the offender.

2.49 The Home Office, as part of its initiative on Dangerous Severely Personality Disordered People, is carrying out work on techniques for improving risk assessment. Although our recommendations do not accord in every respect with the current English proposals, many issues of risk assessment are common to both, and we would encourage the Scottish Executive (in due course, the Risk Management Authority) to make arrangements to share information with the Home Office on the development of risk assessment procedures.

Training issues

2.50 A more systematic use of risk assessment, and a more consistent understanding of its application, require a considerable investment in training. This is increasingly recognised in some professions, such as social work, but needs to be more widely adopted for all who participate in risk assessment, or have to make decisions based on risk. We include in this the Judiciary and the Parole Board, Scottish Executive officials (insofar as they provide advice to Ministers on risk assessment and management), social workers, prison officers, prosecutors, psychiatrists and psychologists.

2.51 As we will say in Chapter 3, the proposed Risk Management Authority would play a role in setting standards, but it will be for the individual agencies and professional bodies to ensure that the necessary training is delivered. We would encourage the development of joint training across professional boundaries, and also evaluation of the effectiveness of training in risk assessment.