From flipping the coin to seeing both its sides

INTRODUCTION

Recently the editors of this journal sent me a disc which they had made from a videotaped presentation I gave on ‘dangerousness’ some fifteen years ago. The recording would win no prizes for quality. All the same, it is possible to discern from it the general direction of the talk. It is then possible to ask whether or not this very-dated lecture would pass muster in 2011. Although it is tempting to think that the content of the disc could be converted to a typescript (always assuming it would be possible to find someone patient enough to deal with the ‘ums’ and ‘ahs’ and the seemingly pointless digressions), there is no doubt that the same talk from such a typescript would today clear a hall fairly quickly. It is not so much that what was said was ‘wrong’, though there are instances of this (e.g. over-assertion of Monahan’s (1981) point that mental disorder is a non-correlate of violence), but that emphasis on some points was misplaced (e.g. the insinuation that obtainment of an actuarial violence prediction score necessarily outweighed other considerations). Despite the flattering claim by your editor (Dix, 2011, this issue) that this particular talk was influential to him and to others at the time, space precludes a detailed analysis of the video record. Nor is there any point in going over this bit of history too obsessively. It is more important to think about: (1) what has happened in the study of violence prediction, assessment and management over the past fifteen years; and (2) where should this field of study and practice be headed over the coming decade or so. This piece then, is a kind of ‘benchmark’ note. Only one additional point needs to be made about the conference in Gloucester: during a lunch break, the present author himself learned a very valuable lesson from one of the participants, a forensic nurse from Rampton (with whom, unfortunately, I lost touch the moment the conference ended). I refer below to this exact fluke later in this article.

VIOLENCE RISK ASSESSMENT AND MANAGEMENT: 1996–2011

By 1996 the idea that, when it comes to predicting violence, it might be best simply to ‘flip coins’ (Ennis & Litwack, 1974), was losing steam (though see Menzies et al. 1985; Stone, 1985). That year, 1996, was marked by Randy Borum’s paper in the American Psychologist. Borum (1996) reviewed the extant literature on violence assessment and, while emphasizing the value of actuarial predictions of violence (through reference to Harris et al. 1993; Menzies et al. 1985; and others), pointed to devices relying on structured professional judgment (SPJ). Particular stress was placed on Version 1 of the Historical/Clinical/Risk Management-20, HCR-20 (Webster et al. 1995). The first version of the HCR-20 was a rather rough and tumble affair. Yet it caught the attention of many who read Borum’s paper. Several colleagues from overseas were quick to express interest and to make translations (e.g. Sweden, Germany, the Netherlands). Probably, some of its appeal was due to the fact that it was based not only on the science of the time but also on the opinions held by individual practicing clinicians from the various mental health and correctional disciplines. However that may be, it was not long before the original version of the HCR-20 was in need of urgent revision. Version 2 was completed quickly (Webster et al. 1997). Colleagues then started publishing papers on the reliability and validity of the HCR-20 and its related schemes (e.g. SVR-20, SARA). From the beginning, these developments were captured by Kevin Douglas and associates in the form of an ‘Annotated Bibliography’ (Douglas et al. 2009).

Publication of the MacArthur study in 2001 was formative for a number of reasons (see Monahan et al. 2001). Through it, colleagues across the world came to realize the importance of studying risk assessment and management in civil populations. Heretofore the emphasis had been largely on forensic and correctional samples. The study yielded results that cut across different clinical sites and isolated predictors that were similar to, if not identical, with those obtained in law and mental health and in correctional and parole release programs. Of major interest in this study was the observation that various forms of mental disorder are only weakly positively correlated with subsequent violence (though as already noted, were not non-correlates as suggested earlier by John Monahan in 1981). But what it did suggest is that many such disorders, when associated with substance abuse, magnify violence risk very substantially. This enabled Monahan and colleagues to emphasize how limiting it can be to search for particular variables in isolation from one another, that it is absolutely necessary to develop schemes which will allow for the detailed study of interactions among variables both in the statistical mass but also in the individual case. The MacArthur study also confirmed that Hare-type psychopathy as indexed by the Screening Version (PCL: SV, Hart et al. 1995) gets hold of a substantial amount of variance as a predictor.

The year 2001 saw the publication of the HCR-20 Companion Guide (Douglas et al. 2001). Because the second version of the HCR-20 (Webster et al. 1997) was silent on the matter of risk management, it eventually occurred to the authors that colleagues needed a complementary source when considering how best to reduce or minimize violence risk to others. Accordingly, they edited a short manual with contributions from many colleagues. The idea was to show how insight could be improved, how clients could be helped to become amendable to treatments of various kinds, how support structures in the community can be built, and so on.

The HCR-20 Companion Guide (Douglas et al. 2001) includes a ‘risk tracking device’. This is a simple form which lists the 5 Clinical (C) variables and 5 Risk Management (R) variables. It is to be completed for each client individually at suitable intervals (e.g. daily, weekly, monthly). The idea is to be able to index change over time, to find out whether risk is declining (or increasing). It was this particular notion of risk tracking that was introduced to me by a nurse from Rampton at the Gloucester conference in 1999. He showed me week-by-week data from his own ten or so clients. And it was this that led us to include the risk tracking chart in the HCR-20 CG.

Fortunately, the literature up to the end of the first decade of the twenty-first century has recently been summarized. The edited book by Otto & Douglas (2010) is the new standard on the scientific side (see also the perceptive review by Hanson, 2009) and the UK Department of Health (2007) complements this on the policy and practice side (see also Risk Management Authority, 2005). Yet there has been one development, not noted hereto this point that I would wish to emphasize. This has been the evolution of the Short Term Assessment of Risk and Treatability (START Version 1.0, Webster et al. 2004; Version 1.1, Webster et al. 2009).

It was in about 2000 that, as a research consultant to the Minimum Secure Forensic Unit of St. Joseph’s Healthcare, Hamilton, Ontario, I was called upon to guide an implementation of the HCR-20. Early in the process I encountered some ‘resistance’ from the staff. This is not unusual and can often be overcome given some patience and willingness to compromise (see Wright & Webster, in press). Yet this particular ‘contrary opinion’, voiced by one of the senior nurses, Connie Middleton, caught my attention. She did not think that the HCR-20 quite ‘did it’ for one of her clients. So then began a long process aimed at coming up with something which would suffice for Connie (and her client). Along the way, she prevailed upon Mary-Lou Martin, a highly experienced Clinical Nurse Specialist, to join our meetings. Mary-Lou’s condition for attendance was that any eventual scheme must give weight to client strengths. In the end we came up with 20 dynamic variables which were to be scored for both vulnerability (on a scale of 0, 1 or 2) and strength (on its own scale of 0, 1 or 2). Moreover, it was decided necessary to look beyond violence risk to others and to bring into consideration related risks such as self-harm, suicide, taking unauthorized leave, substance abuse, self-neglect, and so on. Although, it may be that the START scheme is complex to score, at least until the user has become thoroughly familiar with its workings, we would argue that assessing violence and related risks is no simple matter, that these risks are almost always inter-related and must be taken into account as risk management plans are evolved. We are now wondering if there might be value, at least as a thought exercise, to see if the items in the Hare PCL-R could be pushed to the other (strength) side and so create a new complimentary which might help clinicians frame, with their clients, plans which are more constructive and more likely to succeed than otherwise would have been the case.

Connie Middleton has now retired. The main emphasis of the project has shifted to the Forensic Psychiatric Services Commission in Port Coquitlam, British Columbia, where it is led by Johann Brink and Tonia Nicholls. As is clear from the 2009 publication of Version 1.1 of START, colleagues from several parts of the world are struggling hard to make START ‘work’ (see for example Braithwaite et al. 2010; Wilson et al. 2010; Nonstad et al. 2010). Aside from the continuing necessity of establishing reliability, validity, and other such characteristics, attention is now in being devoted to the exacting problems entailed in implementing and monitoring the use of this scheme across different settings (i.e. mental health, forensic, and correctional).

TOWARD A ‘NEW LECTURE’

So what might a modern lecture look like? It should still start out talking about the 2x2 table, false negatives, false positives, and so on (see Undrill, 2011, this issue). The point still has to be made that even with the most vigorous application of any or all of the many new assessment schemes, prediction errors will still occur (Otto & Douglas, 2010; Hanson, 2009). They cannot be prevented (Mossman, 2000). No ‘fail safe’ method exists. This is something understood by all capable clinicians and researchers. The only ambition, and it is an obviously important one, must be to minimize false positive and false negative errors. This inherent limitation is not something comprehended by many senior administrators and bureaucrats. They do not want too many false positives (as they are chronically short of beds and do not wish to harbour people who could be helped safely in the community). And, for sure, they want no false negatives (with all of the attendant inquiries, commissions, investigations, inquests, and so on (see Maden, 2007)).

The new lecture, like the old one, will still have to emphasize that, properly applied, SPJ schemes like the HCR-20 and START are supported by scientific evidence, that they help get beyond ‘flipping coins’, that they can improve clinical care. The new lecture, though, will have to point to the importance of isolating client strengths and working in therapeutic alliances. It could be that the new lecture should continue the coin metaphor but change it to emphasize the importance of ‘seeing both sides’ (i.e. strengths and risks). The new lecture will have to place weight on how institutions and programs can sometimes have iatrogenic effects. Rather than decrease violence they can and do, in some unfortunate instances, accentuate it (see Maden, 2007).

This new lecture will have to stress that, generally helpful though some of the risk assessment schemes developed over the past decade or two have been, they cannot be allowed to degenerate into becoming ‘tools’ applied mindlessly to individuals within correctional facilities, forensic services, and mental health services. That all or most of these various devices have been shown in recent meta-analytic and other studies to ‘predict’ violence against others, at least to some moderate degree (e.g. Yang et al. 2010; Singh et al. 2010), does not give staff members the right to try to ‘force’ treatment. Properly used, schemes should take into account changeable ‘dynamic’ variables as well as relatively fixed ‘static’ ones which ought to bring the client fully into the planning picture. The emphasis on preventing risk of violence against others has got to make room for thoughtful consideration of a range of inter-related risks. These are crucial in the therapeutic process. Potential for self-harm and suicide are obvious risks that require to be explored (e.g. Bouch & Marshall, 2003). Our experience with START so far, is that colleagues and clients alike appreciate being called upon to devote attention to the considerations of factors that are health promotive and health protective. We are also beginning to realize that schemes like START can begin to predict, at least to a degree, not only risk of violence against others but, importantly, success in adjusting to institutional routines and to life in the community.

Since in 1996, the time of Borum’s article, there were only a few largely unvalidated SPJ schemes available, there was no mention of ‘implementation issues’. Yet this now needs to be a topic of research and discussion in its own right. It has taken work to produce clinical guidelines (e.g. Department of Health, 2007, National Institute for Clinical Excellence, 2005). Now the question is: How do these principles become actualized? How are clinicians in the various mental health, forensic and correctional disciplines to be encouraged to use these schemes fairly (in concert with clients’ wishes). How are minimal standards of practice to be attained? What might an ideal staff training and support program look like?

Related to this, and in need of mention in the new lecture, is the present unsatisfactory state of knowledge with respect to the internal functioning of the interdisciplinary mental health team. To be sure, there has been some mention of this topic in recent years (see Webster & Hucker, 2007). Yet, when sitting in as a guest at many a team meeting, it soon becomes evident that one does not have to be a sociologist or social psychologist to witness alarming power imbalances and even actual conduct that gets close to being ‘unprofessional’ (e.g. outright rudeness). When it comes to decision making, better ways have to be found to make all disciplines fully included. This applies especially perhaps to nurses, those who are employed in both the hospital and community contexts (Privitera, 2011). These colleagues, who work so closely with clients and, increasingly, in varied capacities, possess expertise which is easily lost or underplayed.

At a more general level, the new lecture will have to drive home the point that proper professional application of SPJ schemes requires not only clinical sophistication, which it undoubtedly does, but also a thoughtful, balanced and respectful approach both to the clients we serve (and protect) and to the society we protect (and serve). As well as this, clinicians, researchers and administrators must make sure that, if some particular type of risk assessment device is being ‘imported’, this scheme will be tested in situ to determine the extent to which it has demonstrable applicability to the specific sub-populations under study. Some researchers have been very punctilious about conducting these kinds of necessary ‘confirmatory’ studies (e.g. Gray et al. 2007). Such studies do not necessarily show that a much-researched assessment device will actually do the job when put to the test in the local situation (Doyle et al. 2010). The ‘new lecture’ will have to get across the idea that, whereas in the mid 1990s the risk assessment task seemed mainly one of isolating the risk factors pertinent in the particular case, the new task is that of figuring out how the factors, once isolated, interacted with one another according to circumstances as they prevailed at the time of the index event, the extent to which these same factors operate in the present, and the likelihood that they will be in play in the future. We shall need to develop suitable technologies, ones applicable to the individual case, to help create interventions which will work in the specific instance. It is clear that total scores on the HCR-20, START, and related schemes, should be viewed as having even less importance than was proposed in the original 1999 lecture, that we now need to find ways of helping clinicians use information derived from item-by-item consideration to create explanatory models to fit the particular case. Certainly, the new focus needs to be on how the best possible decisions can be made. Some of the scientific emphasis previously directly at clients needs to be redirected toward ourselves as clinicians and as researchers in important positions of trust.