Managing Clinical Risk

A Guide to Effective Practice

Caroline Logan, Lorraine Johnstone, Caroline Logan, Lorraine Johnstone

• Publisher: Willan
• Year: 2012
• Print ISBN: 9781843928546
• eBook ISBN: 9781136257827
• Edition: 1
• Topic: Social Sciences
• Subtopic: Criminology
• Index: Social Sciences

Violence directed towards others and violence directed towards oneself cause an immense amount of physical and psychological damage – to the harmed and the harmful person alike, to their families, and to the public at large. Managing clinical risk is an authoritative manual for practitioners working with harmful men, women, and young people, containing up-to-date information and guidance on what to do and how they can assess and manage clinical risk, communicate their concerns about risk, and account for their decisions about risk management to their clients and to the Courts.

This book provides an evidence-based understanding of risk in key areas of practice – violence, sexual violence, firesetting, suicide, and self-harm, working with individuals and organisations alike – and among special groups: women, young people, serving and former military personnel, clients with comorbid presentations, and clients with cognitive impairment. Further, it suggests and describes the skills practitioners need to understand and communicate their concerns to all who need to know about them through coverage of interviewing and risk formulation skills.

This is a guidebook to effective practice. All its contributors have a record of research, practice, and considered thinking in the area of clinical risk assessment and management. They all have a wide range of knowledge and experience about the notion of risk, conducting risk management in real world mental health, correctional, and community settings, and about working with clients with a label of high risk. Together, they combine theoretical and research knowledge with a wealth of practical skills in care and management, emphasising the collaborative and recovery-focused nature of modern risk management.

Part I
The need for change

1 Violence risk assessment

From prediction to understanding — or from what? To why?

David J. Cooke and Christine Michie

‘Everything in the future is uncertain, as is most of the past; even the present contains a lot of uncertainty…’

Dennis Lindley (2006, p. 7)

Risk management: the management of uncertainty

Perhaps Lindley's statement should be on the desk of all forensic practitioners. Risk by definition is about uncertainty (Gigerenzer, 2004). It follows, therefore, that risk management is about the management of uncertainty. Uncertainty about violence engenders anxiety in professionals and the public alike. Fortunately, the last two decades have seen dramatic strides in the technology available to assist uncertainty reduction — the technology of risk management. In retrospect, three eras of risk assessment can be discerned: the era of unstructured professional judgement, the actuarial era, and the era of risk management through structured professional judgement — this final era, we will argue, is transmuting into a fourth era, the era of risk formulation. It is important to recognize the achievements made so far (Otto and Douglas, 2010), but we must also be conscious of the limitations, and indeed the dangers, of some of the approaches that have gained acceptance within criminal justice systems. Lindley's quotation suggests that forensic practitioners should approach the task of risk assessment with transparency, circumspection and humbleness.

Violence risk assessment is at the centre of forensic practice and this is increasingly so as society becomes more risk averse. The focus of this chapter is the individual risky offender; the focus is on the provision of useful, valid, ethically sound information, which is probative and not prejudicial, information that can lead to principled and informed decisions. We emphasize this focus on the individual: the individual offender is the primary concern of the forensic practitioner when informing the decision-maker.

Nearly thirty years ago, John Monahan argued that violence risk assessment could only be improved if specific risk factors for violence could be identified. This stimulated a prodigious number of group studies and the field is much better informed about what risk factors need to be considered. We know ‘what?’ If we are to progress, it is now time to move to the question ‘why?’: why are people violent, what drives them to violence, what disinhibits or destabilizes them so the likelihood of violence is increased? This focus on ‘why?’ entails a shift of perspective, from a focus on groups to a greater focus on the individual — from a focus on statistics to a focus on psychological processes. In this chapter, we consider two broad issues which emphasize why such progress is now required.

First, we provide a detailed account of the misuse of so-called actuarial models. In our view, the field has become overconfident about the ability of actuarial procedures to make reliable predictions about whether individuals will be violent (cf. Craig and Beech, 2009; Hanson and Howard, 2010). This misplaced confidence is underpinned by the failure to appreciate the problems of making predictions — with any certainty — for individuals based on aggregate statistics, and importantly, a misunderstanding of the logic and purpose of true actuarial approaches.

Second, we argue that the field of risk assessment can only progress by shifting away from the actuarial paradigm. In our view, for the field to progress, it is necessary to generate both a taxonomy of risk processes and systematic guidance for risk formulation. The development of criteria to validate individual risk formulations is an important — indeed pressing — challenge for the field.

The promise and the peril of the actuarial approach

‘The Guide is definite. Reality is frequently inaccurate.’

Douglas Adams

The unstructured clinical approaches of former years have attracted great opprobrium because of their lack of an evidence base, their idiosyncratic nature, and their lack of transparency, replicability and utility (Quinsey, Harris, Rice and Cormier, 1998). As a reaction, attempts were made to impose structure on the decision-making of clinicians through the use of so-called actuarial risk assessment instruments (ARAI) (e.g., Hanson and Thornton, 1999; Quinsey, et al., 1998). These procedures have proved popular amongst practitioners and administators alike. The adoption of ARAI techniques has become widespread: ‘In North America and the United Kingdom, actuarial risk assessment has permeated the entire criminal justice system’ (Craig and Beech, 2009; p. 197). Khiroya, Weaver and Maden (2009) demonstrated that the Risk Matrix 2000 (RM2000) and Static-99 were the most commonly used sex-offender assessments in English medium secure forensic units.

While some commentators have contended that the superiority of actuarial scales is a given (Craig and Beech, 2009), other commentators have been less sanguine (Hart, et al., 2003). Empirical evidence suggests that actuarial approaches perform as well as structured professional judgement approaches (SPJ) in terms of predictive validity (e.g., Hanson and Morton-Bouron, 2009; Singh, Grann and Fazel, 2011; Yang, Wong and Coid, 2010a). However, as we demonstrate below, where the actuarial approaches merely claim to estimate the likelihood of an offence, the SPJ approach provides so much more in terms of formulation. Importantly, SPJ approaches can lead to interventions that are both effective and proportionate.

It is perhaps telling that the proponents of actuarial approaches have derided clinicians' understanding of research methods and statistics (Harris, 2003). But, as we will see, the statistical underpinning of the actuarial approach is, in reality, very shaky. Despite this, proponents have argued for the wholesale adoption of actuarial methods:

What we are advising is not the addition of actuarial methods to existing practice, but rather the complete replacement of existing practice with actuarial methods … actuarial methods are too good and clinical judgment too poor to risk contaminating the former with the latter.

(Quinsey, et al., 1998, p. 171)

In our view, this is a dangerous and scientifically untenable position (Allport, 1940; Harcourt, 2006). We will now explain why this is so.

The basis of the actuarial approach to violence risk assessment

The actuarial paradigm appears simple. A group of offenders, usually prisoners, is assessed — generally characteristics that are easy to measure (and not necessarily clearly criminogenic) are recorded, for example, age, marital status, history of offending, type of victims, and so on. Sometimes the cohort of prisoners is followed up and new criminal convictions are identified from criminal records. More generally, the cohort is followed back, that is, the files of prisoners who have been released and whose convictions status has been monitored, are reviewed. Statistical methods may be used to estimate the characteristics associated with the observed likelihood of reconviction for the group of offenders. Typically, the development sample (the group on which the statistical model was developed) will be divided into risk groups — high, medium or low — then the proportion of each group who reoffended is provided. This information about a group is then used to make a prediction about a new individual, the person who is the focus of the risk assessment decision. Generally, the decision-maker is provided with information about this new individual's likelihood of reoffending using a process of analogy; that is, ‘This man resembles offenders who are likely to recidivate, therefore, he is likely to recidivate’ (Hart, 2003, p. 385). Craig and Beech (2009) recommended the following method of communication:

Actuarial risk assessment of Mr X using Risk Matrix/Sexual indicates that his score falls within the ‘medium’ risk category, such a score is associated with a 13% likelihood over five years, 16% over 10 years, and 19% over 15 years, of being reconvicted for a sexual offence (for known and convicted sexual offenders) in a group of sexual offenders with the same score.

(Craig and Beech, 2009, p. 205)

This is a form of inductive logic where it is argued that because Mr X belongs to a group, then the best point estimate of his risk of reoffending is the average for the group. While this may be technically correct, it is fundamentally misleading because of the huge degree of uncertainty associated with this estimate. And of course, argument by analogue falls down further when the offender being assessed is different from the standardization sample; for example, when first offenders are compared with prisoners, or when internet offenders are compared with contact offenders.

While we understand this desire for certainty, we agree with Gigerenzer that certainty in prognostications is an illusion (Gigerenzer, 2002). Unfortunately, this illusion of certainty is bolstered because the actuarial methods wear the clothes of science — samples are collected, data are analysed, arcane statistics are generated, but the product is inherently misleading. Let us consider that contention in more detail.

A signal case

‘There is no such uncertainty as a sure thing.’

Robert Burns

Our interest in the uncertainty associated with actuarial tests was triggered when the first author was contacted by a Scottish sheriff (a judge) who was concerned about conflicting expert opinions about an offender. One opinion was based on a frequently used actuarial tool — the RM2000 (Thornton, 2003) — and it was deemed that the offender was at ‘high’ risk of reoffending. The second opinion, based on a structured clinical judgement approach — the Sexual Violence Risk-20 (SVR-20, Boer, Wilson, Gauthier and Hart, 1997) — concluded the offender was ‘low’ risk. The sheriff was concerned; he had to pass a sentence. The first author's initial reaction was one of disquiet when he realized that the actuarial judgment of ‘high’ risk was founded on only three pieces of information: the offender was 18 years of age, he had not lived in an intimate relationship with someone for two years or more, and he had not met his victim face-to-face before (he and the victim had been in regular contact by phone for five weeks prior to the offence). The offender was convicted of a statutory offence of having sexual intercourse with a minor (see Cooke, 2010a for a fuller account). This case triggered a process of analysis that led to the publication of Hart, Michie and Cooke (2007), a paper which some have regarded as ‘controversial’ (e.g., Craig and Beech, 2009, p. 203). Below we consider and clarify some of the issues raised in response to previous papers on this issue (Cooke and Michie, 2010; Hart, Michie and Cooke, 2007). We believe that the quantification of uncertainty is key, and that uncertainty is ignored at our peril.

Quantifying uncertainty

‘Statistics means never having to say you're certain.’

Anon

Authors and advocates of ARAIs explicitly state that their purpose is prediction; that is, the statement that a specified event will occur in the future. For example, ‘RM2000/S is a prediction tool for sexual violence’ (RM2000 Scoring Guide February 2007, p. 3), and ‘Static-2002 predicts sexual, violent and any recidivism as well as other actuarial risk tools …’ (Static-2002, Unpublished Manual, p. 1). How certain or uncertain are these predictions? Statistical methods allow the quantification of uncertainty; indeed, a core function of statistical methods is the estimation and quantification of uncertainty. By definition, all estimates are subject to error: the precision of a parameter estimate (e.g., mean rate of sexual recidivism of a group) is specified by a confidence interval (CI), that is, the range of values within which an unknown population parameter is likely to fall. The interval's width provides a measure of the precision — or certainty — associated with the estimate of the population parameter and is determined, in part, by the sample size used to estimate the population parameter.

When we are considering predictions for an indiv...

Table of contents

• Cover Page
• Half Title page
• Issues in Forensic Psychology
• Title Page
• Copyright Page
• Contents
• List of contributors
• Foreword
• Preface
• Abbreviations
• Part I The need for change
• Part II Key areas of practice
• Part III Key client groups
• Part IV Key practice skills
• Postscript
• Index