Are Violence Risk Assessment Tools Clinically Useful?

Abstract

Despite increasing concerns regarding the prevalence of violent behaviour in mainstream mental health settings, the impressive body of forensic research on violence risk assessment has thus far had only limited impact on front-line general mental health practice. The common objection raised by clinicians that risk assessment tools lack utility for clinical practice may contribute to this. The present paper argues that this objection, although understandable, is misplaced. Usage of appropriate, validated risk assessment tools can augment standard clinical approaches in a number of ways. Some of their advantages derive simply from having a well-structured approach, others from consideration of specific kinds of risk factors: ‘static’ and ‘dynamic’. The inappropriate use of tools without a firm evidence base, however, is unlikely to enhance clinical practice significantly.

Keywords

forensic risk assessment violence

More than two decades have passed since John Monahan's clarion call for improved research methodologies to investigate violence risk assessment in mentally disordered populations [1]. There has been no shortage of endeavour in the ensuing years: millions of research dollars have been spent [2]; risk assessment instruments of demonstrable predictive validity have been developed [3–6]; the key ethical [7–9], legal [10] and scientific [11, 12] debates arising from the findings have been well articulated. Most academics working in the field would now concur with the assertion that “unguided clinical judgement is unreliable … improved risk assessment depends on better assessment technology” [13].

Despite this laudable progress in the academic realm, the key lessons of research on violence risk assessment have not been systematically incorporated into the daily practice of most mental health professionals [14, 15]. Risk assessment technologies are generally used in a highly variable way, if at all. Locally developed, unstandardized, unvalidated schemes are often used in preference to standardized tools with demonstrated validity [16, 17]. Alternatively, full reliance may be placed on unstructured, unreliable and unfalsifiable clinical judgements even where there are clear indicators, such as past history of violence, of the need to optimize risk assessment [18].

Managerial enthusiasm for risk assessment tools, usually with scant consideration of the need for validation and standardization, or of the specific nature of the challenges faced by front-line clinical workers, has not helped matters. Naturally, if tools are not used in a way that demonstrably adds something to standard clinical processes, then the time expended on the paperwork will inevitably detract from the efficiency of service provision [19].

Descriptive research into how risk assessment is carried out in real-world settings is still in its infancy [20]. Recent Australian research, however, suggests that the belief that structured violence risk assessment tools lack utility – usefulness in enhancing day-to-day clinical mental health practice – may be a major contributor to this glaring theory-practice gap [21]. Two related concerns may also be relevant: the fear that attempting to formally predict adverse outcomes necessarily increases medicolegal liability for the prevention of such events (even where the capacity to prevent is limited); and the sense that risk assessment tools stigmatize some patients by labelling them as permanently and irretrievably ‘high risk’ or ‘dangerous’.

These beliefs have perhaps been especially pernicious and powerful in their effects because none is without some foundation in truth. This paper will explore their basis in more detail and suggest how they may be overcome in practice.

Academic/clinical divide in risk assessment

The academic/clinical divide is well recognized in mental health [22], and it may be that, given the more inherently values-focused nature of the field [23], the integration of empirical evidence into clinical practice is more challenging for psychiatry compared with other branches of medicine. In violence risk assessment, such integration has perhaps been further hindered by the promotion by some academics of their particular tools as the way of doing risk assessment, allied with scepticism as to the potential value of involving any clinical judgement in the process [3]. In addition, some violence-risk theoreticians minimize the importance of major mental illness and its constituent psychopathology, which have often been poorly integrated into risk technologies [3, 24], notwithstanding the enormous evidence base with respect to their importance [25]. It is scarcely surprising therefore that structured tools have at times been viewed as antithetical to real understanding of the kind achievable by experienced mental health clinicians [26].

Furthermore, the ultimate aim in clinical work is not risk assessment per se but risk management: prevention of adverse outcomes, by means of clinically legitimate and (hence) medicolegally defensible clinical decisions [27], rather than simply prediction. This contrasts with the aims of academic studies that seek to validate risk assessment tools by demonstrating the prediction of such adverse outcomes. It is unlikely therefore that simply developing new tools with even higher predictive accuracy than those currently available will overcome concerns regarding their real-world usefulness. Rather, to secure a lasting rapprochement between the academic and clinical spheres, new ways of embedding risk assessment technology into everyday practice are required. The challenge is to meet the needs and address the concerns of front-line clinicians, while still ensuring a scientifically valid basis for the decision-making processes that flow from such techniques.

Standard clinical approaches

A traditional, clinical approach to violence risk assessment, at its best, involves the gathering of relevant data from multiple sources, generally using a recognized framework with various subheadings (e.g. ‘Personal history’, ‘Mental state’). It should include individualized, detailed, idiographic descriptions of past violent episodes (so-called anamnestic analyses), including a consideration of factors such as mental state, substance use, social context (such as relationship to, and behaviour of, the victim [28]) and engagement with treatment services at the relevant times. Previous episodes of violence will generally be due to some combination of personal and situational factors: a skilled clinician will weigh up how these factors interacted in the past, and the likelihood of similar such interactions in the future [11, 29]. Exactly how such data are translated into clinical decisions is little explored in the literature [20, 30], but clearly risk will be assessed as currently high to the extent that current mental state and circumstances match those of past episodes of violence. Appropriate management interventions are derived from this analysis, to both optimize overall psychosocial functioning and to reduce the likelihood of recreating past scenarios of risk.

Can structured tools add anything to this process? The method advocated by the majority of violence risk assessment experts in Australia (‘structured professional judgement’ [4]) supports and augments rather than replaces the need for such expert clinical processes. The term ‘decision support tool’ [31] explicitly acknowledges that structured risk assessment is an exercise in assisting with the making of appropriate decisions [32] rather than one of clairvoyance. The argument here is therefore not that formally structured assessment tools should replace clinical judgement, but that appropriate tools can augment traditional clinical methods, to a degree that makes their considerable implementation costs worthwhile. Some of the benefits of such augmentation are common to validated, structured tools in general, while others relate more specifically to the consideration of static and dynamic risk factors.

General benefits: value of a framework

Although an unthinking ‘tick-box’ approach to risk assessment is a parody of risk assessment and potentially worse than useless [33], the role of structured tools as aides-memoires should not be underestimated. They encourage all clinicians to gather and synthesize information in a methodical and consistent way, according to certain minimum standards. Ensuring that factors such as social context – often neglected in traditional clinical assessments [1] – are considered, is an obvious advantage.

The imposition of a framework also assists with countering common errors of judgement (so-called ‘heuristic biases’) to which all human decision-making is prone [34]. Such biases are more likely to influence decisions under conditions of cognitive overload, where many factors have to be considered simultaneously. Risk assessments generally involve just such a challenge. By assessing and summarizing risk according to a limited number of critical, evidence-based factors articulated in a structured tool, the clinician can more readily ‘see the wood for the trees’: information and opinions extraneous to violence risk are kept in the background and are less likely to influence decision-making.

Evidence-based frameworks for risk assessment also greatly facilitate communication of risk [35]. Such communication (and debate around risk estimates) can be problematic when full reliance is placed on opaque concepts such as ‘clinical experience’ and ‘intuition’. The importance of communicating risk estimates is increasingly recognized [36] as integral to optimal management. Structured tools enable the clear articulation of the basis for specific estimates of future risk, and clarify the sources of disagreement, where these occur. Ease of such communication between services will obviously increase further if and when all use widely recognized tools, rather than using locally developed procedures.

The full value of structured risk assessment tools becomes even clearer when the building blocks of such tools are considered: risk factors, that is, phenomena that have been empirically associated with future likelihood of violent behaviour. Some tools rely solely on historically derived, fixed, static risk factors (such as diagnosis, history of violence, history of childhood maladjustment etc.) [3]; some on dynamic risk factors, which are changeable [37]; others, including the Historical Clinical Risj-20 (HCR-20) [4], on a combination of both. The distinction is important because the role of each in assessment and management differs.

Utility of static risk factors

Analysis of static risk factors provides an estimate of long-term likelihood of violent behaviour: the patient's risk status [38]. More technically, it provides an estimate of the likelihood that the patient assessed is a member of a particular group with a specific, quantifiable likelihood of violence over a given future period of time [33].

It is not difficult to see how such information may be used to stigmatize and reject patients, rather than to facilitate clinical care. Quantitative research methods, utilizing group data, can seem remote from clinical mental health practice, which emphasizes the understanding of the inner life of patients at an individual level. An unembellished statistical statement of the probability of behaving violently may appear to lack the practical value of a traditional clinically based consideration of past behaviours and psychopathology.

Furthermore, historically most of the impetus for the development of structured technologies to assist with violence risk assessment came not from health professionals, but from the correctional arena. The holy grail of a scientifically validated tool to reliably identify those people most likely to violently reoffend (with a view to consequent incapacitation) was first pursued with vigour in the US criminal justice system [39]. Similarly, even some of the technologies developed in the mental health field have been primarily aimed at assisting decision-making around predictive questions of long-term security, such as parole decisions or detention after the expiry of a prison sentence [3, 40, 41]. These technologies have tended to emphasize static (and hence unchangeable) variables. Data relating to such factors are generally fed into an algorithm that determines the probability of violence over a given time frame (generally of the order of several years), calculated by using an actuarial process that draws on reoffending data relating to other individuals who have similar scores.

With this background, it is scarcely surprising that many clinicians have assumed that the structure involved in formal risk assessment processes necessarily involves generating an ‘emotional distance’ [26] from patients. Clinicians are generally motivated by the desire to provide therapy and improve patients’ lives, rather than to incarcerate them with the aim of improving ‘public safety’ [42]. Many are justifiably wary of tools that emphasize the identification of high-risk individuals, labelled ‘as if they have an immutable quality of viciousness’ [43], notwithstanding the impressive predictive powers of the technologies involved.

The proper clinical use of static risk factors, and of tools based upon such phenomena, requires an appreciation of a wider debate about the underpinnings of human behaviour of all kinds, including violence. When considering the vagaries of human behaviour, the relative balance between the role of enduring personal factors (such as personality traits) on the one hand and transient situational factors (such as interpersonal provocations) on the other, has been a subject of longstanding debate in various disciplines. Social psychology has debated the importance of nomothetic traits versus idiographic situational determinants [44]; behavioural psychology has attempted to tease out the roles of environmental variables (stimuli and consequences) versus organismic variables (physiology and past learning history) [45]; criminology has historically swung between an emphasis on inherent biological deficits within individuals [46] and a converse emphasis on social/structural determinants of criminal behaviour [47]; criminological psychology [13, 48–51] has usefully examined the interplay between background or dispositional factors versus situational factors, which are more proximal to the behaviour under consideration. There is even a parallel dichotomy in accident and human error research, where a distinction is drawn between faulty, latent conditions in the system under consideration and acute, active failures proximate to the adverse outcome of interest [52].

These debates are of direct relevance to violence risk assessment. Humans tend to overestimate the degree to which behaviour (that of other people at least) is a result of stable personal factors and underestimate the degree to which it may be caused by fluctuating situational factors [34, 44]. This bias can become problematic in mental health, because to attribute others’ behaviour solely to fixed intrapersonal variables (static risk factors) may facilitate stigmatizing and coercive ways of working. Therapeutic nihilism and long-term detention may be understandable responses to recurring problematic behaviours if they stem solely from immutable internal factors.

Ethical risk assessment methods need to take these cognitive biases into account. However, the converse error is to totally ignore the role of the person and assume that situational circumstances at the relevant time were entirely responsible for past violent behaviours. Clearly, an understanding of both persons and the contexts in which their risk is likely to escalate is required for optimal risk assessment.

Rather than stigmatizing patients then, the proper use of risk assessment tools, in conjunction with anamnestic analyses of past episodes of violence, can aid understanding of the nexus between patient vulnerability to violent behaviour and contextual factors. A comprehensive analysis of static risk factors can help ensure that clinicians properly take account of historical material and consider pre-existing vulnerabilities for violence risk when planning long-term treatment strategies [8]. The UK psychiatric homicide inquiry literature suggests that such understanding is often deficient in practice [53]. Such structured understanding can help to avoid counter-transference errors, and ensure that the relevant data have been systematically sought, considered and processed in a transparent and logical way [54].

High-resolution anamnestic analyses of past episodes of violence are important but may, on their own, underplay the role of longstanding latent factors of concern within the perpetrator's make-up, and overemphasize the role of transient factors. The broader scanning approach to patient histories utilized by structured tools that draw upon static risk factors provides a necessary counterweight: a picture of the person and their risk status [38]. Such tools hence ensure that interindividual differences (such as personality traits), empirically related to risk, receive appropriate emphasis. As well as enhancing comprehensive understanding of the case, this may go some way to reducing the likelihood of inappropriate clinical complacency [18].

Allocating patients to different groups based on long-term future risk of violence, calculated from their static risk factor profile, may at first glance appear antithetical to individualized, patient-centred care. However, it is no different in principle to allocating to categories of other kinds, also derived from group data, such as ‘warranting trial of an antipsychotic’ or even ‘likely to benefit from a renal transplant’. The critical issue is not the utilization of group data per se, but rather whose interests are being served by the process. Where the interests of the patient are given primacy, there is little problem. However, violence risk assessment processes may of course be utilized with public protection as the primary goal. This is ethically troubling territory for the mental health professional, even more so where the processes rely entirely on static risk factors that are not open to therapeutic change.

None of this should distract clinicians from the potential clinical utility in having a robust sense of a patient's long-term risk status. Data that suggest that the individual is in a higher risk group can, for example, flag the need for increased monitoring and support, for forensic specialist evaluation, and a lower threshold for assertive intervention in the event of decompensation [18].

Utility of dynamic risk factors

Ideally, risk management processes mirror the quality assurance cycle: assess > plan > act > assess, explicitly linking risk assessment to congruent management interventions [55]. Consideration of idiosyncratic risk factors derived from a careful analysis of the patient's history (anamnestic analyses) are helpful but insufficient on their own to guide this process: a given individual may be violent in many different circumstances, and the past rarely repeats itself in every detail [56]. In order to link structured risk assessment most effectively with treatment, best practice involves regularly monitoring empirically validated dynamic risk factors. These are those factors (such as symptoms of mental illness, drug abuse, non-compliance, social stressors etc.) that change over time and which research has shown to correlate reliably with increased likelihood of violence. They provide an estimate of short-term likelihood of violent behaviour: the patient's risk state [38]. Several of the widely used structured violence risk assessment schemes incorporate such factors [4, 37]. Protective factors that diminish likelihood of violence are also included in some tools [37], further facilitating the construction of holistic care plans that challenge the dichotomy between care and containment [57, 58].

The benefits of explicitly joining the dots in this way – linking the structured assessment of dynamic risk factors to treatment planning, in an iterative process, are considerable [59, 60]. Levels of intervention can be appropriately matched to current risk level [24]; signs of escalation in risk state can be monitored, facilitating early intervention [53]; the efficacy or otherwise of given management plans can be regularly evaluated. Changes in risk state can be used as a measure of the efficacy of current management: clearly if dynamic risk factors are worsening, then this is an indication to reconsider current approaches, irrespective of whether there have actually been any violent incidents. This means that feedback on the effectiveness of treatment can be obtained without relying on the actual occurrence of adverse events. Structured risk assessments can therefore enhance treatment by providing a coherent, evidence-based framework for care plan revision [61].

Such processes can help assuage clinician anxiety about their capacities and responsibilities. Mainstream clinicians often view violence risk assessment as an arcane art, requiring highly trained specialists, rather than a readily learnable skill that largely draws upon pre-existing clinical knowledge. Furthermore, perhaps partly due to the fact that emphasis in most published studies has inevitably been on prediction rather than prevention, there is often anxiety about either “getting it wrong” by failing to predict an adverse outcome, or about being obligated to prevent an adverse outcome that a formalized risk assessment suggests is likely. An over-inflated concept of duty of care often results in the spectre of civil litigation influencing such thinking at both a clinical and a managerial level.

By explicitly linking management to assessment, a structured framework utilizing validated assessment tools is medicolegally protective, in the event of adverse outcomes. A documented demonstration that all appropriate and due care has been taken in order to maximize both patient and public safety is likely to rebut a claim of negligence, by demonstrating that the requisite standard of care was met [62].

Given that people with a mental disorder can at times behave violently for the same reasons as the rest of the population, it is appropriate that structured risk assessment tools go beyond a consideration of psychiatric (in the narrow sense of that term) risk factors. Risk levels may of course be elevated even when symptomatic control of a person's mental illness has been achieved. This can lead to much angst on the part of services, who may feel pressured to take responsibility for such risk. Despite calls from some in the forensic mental health field to embrace violent behaviour (whatever the aetiology) as a viable treatment target [63], overstretched public mental health systems understandably tend to retreat into a narrower focus on eradication of acute psychiatric symptoms. The precise role that mental health services should play with respect to other risk factors, such as persistent substance use, antisocial attitudes or social disharmony and disadvantage, will continue to be an area of active debate as services are remoulded to meet 21st century societal needs. At both a political and a clinical level, however, this debate is surely best served by explicit acknowledgement (and measurement) of all those factors that relate to risk in any given case. When violence risk assessments reveal dynamic factors that do not appear to be easily modifiable in practice (whether due to resource limitations, patient factors, service philosophy, or other reasons), rather than this being a trigger for angst, it can promote open discussion around the appropriate roles of mental health services, training needs, resource deficits, potential avenues for research and the like. Thus, the process can yield data to assist in clinical governance processes [64].

Conclusion

The challenges involved in the assessment and management of violence risk are considerable but will be greatly assisted by the integration of empirical research findings into clinical practice. The barriers to this are significant but not insurmountable. The fears that tools add nothing of utility, can stigmatize patients or increase medicolegal liability are no longer warranted, provided the technology is applied in an appropriate way.

Although the same practice principles apply irrespective of the challenge, precisely which tools are appropriate will vary according to the clinical task. The assessment and management of acute violence in an inpatient unit, for example, will be assisted by quite different tools [65] than those relevant to long-term violence risk in the community. Similarly, risk of violence to self (suicide and other deliberate self-harm) now has a considerable literature supportive of specific structured tools in augmenting routine clinical practice [66]. Locally developed tools with no empirical foundation, many of which attempt in a single document to assess clinical risk in all its many forms, throughout all the various arms of a clinical service, are unlikely to provide any improvement over standard clinical care.

Future integration of theory with practice, to develop violence risk assessment that is both empirically based and clinically useful will not be best served by a ‘top down’ approach of academics telling clinicians what they should do, but rather by an active dialogue, whereby “researchers can be informed of constraints of clinical practice which perhaps impede implementation of research products” [20]. The sooner such products are widely implemented, the sooner such a healthy dialogue can commence.

Footnotes

Acknowledgements

Thanks to Dr Adam Brett of Western Australia Forensic Services for helpful comments on an earlier draft of this manuscript.

References

1. Monahan

. The prediction of violent behavior: toward a second generation of theory and policy. Am J Psychiatry 1984; 141: 10–15.

2. Monahan

Steadman

Silver

. Rethinking risk assessment: the Macarthur study of mental disorder and violence. Oxford University Press, New York 2001.

3. Quinsey

Harris

Rice

Cormier

. Violent offenders: appraising and managing risk. American Psychological Association, Washington, DC 1998.

4. Webster

Douglas

Eaves

Hart

. The HCR-20: Assessing Risk for Violence Version Two. Simon Fraser University, British Columbia 1997.

5. Hare

. Psychopathy Checklist Revised, 2nd edn. Toronto: Multi-Health Systems, 2003.

6. Steadman

Silver

Monahan

. A classification tree approach to the development of actuarial violence risk assessment tools. Law Hum Behav 2000; 24: 83–1000.

7. Grisso

Appelbaum

. Is it unethical to offer predictions of future violence?. Law Hum Behav 1992; 16: 621–633.

8. Mullen

. Dangerousness, risk, and the prediction of probability. New Oxford textbook of psychiatry, Gelder

Lopez-Ibor

Andreasen

. Oxford University Press, Oxford 2000; 2066–2078.

9. Litwack

. Some questions for the field of violence risk assessment and forensic mental health: or, “back to basics” revisited. Int J Forensic Ment Health 2002; 1: 171–178.

10.

10. Rogers

. The uncritical acceptance of risk assessment in forensic practice. Law Hum Behav 2000; 24: 595–605.

11.

11. Litwack

. Actuarial versus clinical assessments of dangerousness. Psychol Public Policy Law 2001; 7: 409–443.

12.

12. Mossman

. Understanding prediction instruments. Textbook of forensic psychiatry, Simon

Gold

. American Psychiatric Publishing, Arlington, VA 2004; 501–523.

13.

Blackburn

Risk Assessment and prediction. In: McGuire

, Mason

, O'Kane

, eds. Behaviour, crime and legal processes. New York: Wiley, 2000:177–204.

14.

14. Webster

Cox

. Integration of nomothetic and idiographic positions in risk assessment: implications for practice and the education of psychologists and other mental health professionals. Am Psychol 1997; 52: 1245–1247.

15.

15. Webster

Muller-Iberner

Fransson

. Violence risk assessment: using structured clinical guides professionally. Int J Forensic Ment Health 2002; 1: 185–193.

16.

16. Higgins

Watts

Bindman

Slade

Thornicroft

. Assessing violence risk in general adult psychiatry. Psychiatr Bull 2005; 29: 131–133.

17.

17. McCall

. A review of approaches to forensic risk assessment in Australia and New Zealand. Psychiatry Psychol Law 2003; 10: 221–226.

18.

Maden

Review of homicides by patients with severe mental illness. London: National Institute for Mental Health 2006.

19.

19. Maden

. Standardised risk assessment: why all the fuss?. Psychiatr Bull 2003; 27: 201–204.

20.

20. Elbogen

. The process of violence risk assessment: A review of the descriptive research. Aggression Violent Behav 2002; 7: 591–604.

21.

21. Byrne

, Gagliardi

. Mental health staff beliefs, knowledge and self-reported behaviour about risk assessment. In: Human rights in a world of fear and diminishing resources. Melbourne: Australian and New Zealand Association of Psychiatry, Psychology and Law (Victoria), 2006:13–14.

22.

22. Douglas

Cox

. Violence risk assessment: science and practice. Leg Criminol Psychol 1999; 4: 149–184.

23.

23. Fulford

KWM

. Analytic philosophy, brain science, and the concept of mental illness. In: Bloch

, Chodoff

, Green

, eds. Psychiatric ethics, 3rd edn. Oxford: Oxford University Press, 1999:161–192.

24.

24. Andrews

, Bonta

The psychology of criminal conduct, 3rd edn. Cincinnati, OH: Anderson Publishing, 2003.

25.

25. Mullen

. Schizophrenia and violence: from correlations to preventative strategies. Adv Psychiatr Treat 2006; 12: 239–248.

26.

26. Doctor

. Psychodynamic lessons in risk assessment and management. Adv Psychiatr Treat 2004; 10: 267–276.

27.

27. Heilbrun

. Prediction versus management models relevant to risk assessment: the importance of legal decision-making context. Law Hum Behav 1997; 21: 347–359.

28.

28. Scott

. Assessing dangerousness in criminals. Br J Psychiatry 1977; 131: 127–142.

29.

29. Meloy

. Violent and homicidal behavior in primitive mental states. J Am Acad Psychoanal 1988; 16: 381–394.

30.

30. McNiel

Binder

. Correlates of accuracy in the assessment of psychiatric inpatients' risk of violence. Am J Psychiatry 1995; 152: 901–906.

31.

31. McNiel

Gregory

Lam

Binder

Sullivan

. Utility of decision support tools for assessing acute risk of violence. J Consult Clin Psychol 2003; 71: 945–953.

32.

32. Vinestock

. Risk assessment. “A word to the wise”?. Adv Psychiatr Treat 1996; 2: 3–10.

33.

33. Reid

. Risk assessment, prediction, and forseeability. J Psychiatr Pract 2003; 9: 82–86.

34.

34. Plous

. The psychology of judgment and decision making. McGraw-Hill, New York 1993.

35.

35. Kumar

Simpson

AIF

. Application of risk assessment for violence methods to general adult psychiatry: a selective literature review. Aust N Z J Psychiatry 2005; 39: 328–335.

36.

36. Arya

Nicholls

. Identifying and communicating clinical risk. Australas Psychiatry 2005; 13: 366–370.

37.

37. Webster

Martin

Brink

Nicholls

Middleton

. START: short term assessment of risk and treatment. Forensic Psychiatric Services Commission, British Columbia 2004.

38.

38. Douglas

Skeem

. Violence risk assessment: getting specific about being dynamic. Psychol Public Policy Law 2005; 11: 347–383.

39.

39. Greenwood

. Selective incapacitation. Rand, Santa Monica, CA 1982.

40.

40. Appelbaum

. Law and psychiatry: dangerous severe personality disorders: England's experiment in using psychiatry for public protection. Psychiatr Services 2005; 56: 397–399.

41.

41. DSPD Programme. Dangerous and severe personality disorder (DSPD) high secure services for men: planning and delivery guide. Department of Health, Home Office, HM Prison Service (England and Wales), 2005.

42.

42. Szmukler

. Risk assessment: ‘numbers’ and ‘values’. Psychiatr Bull 2003; 27: 205–207.

43.

43. Lindqvist

, Skipworth

Evidence-based rehabilitation in forensic psychiatry. Br J Psychiatry 2000; 176.

44.

44. Bem

Allen

. On predicting some of the people some of the time: the search for cross-situational consistencies in behavior. Psychol Rev 1974; 81: 506–520.

45.

45. Nelson

Hayes

. Conceptual foundations of behavioural assessment. Guilford Press, Guilford 1986.

46.

46. Lombroso

. L'uomo delinquente. Horpli, Milan 1876.

47.

47. Merton

. Social structure and anomie. Am Sociol Rev 1939; 3: 672–682.

48.

48. McGuire

. Psychological processes in crime. Understanding psychology and crime, McGuire

. Open University Press, Maidenhead 2004; 55–74.

49.

49. Farrington

. The causes and prevention of violence. Violence in healthcare, Shepherd

. Oxford University Press, Oxford 2001; 1–27.

50.

50. Pfohl

. Deciding on dangerousness: prediction of violence as social control. Punishment and penal discipline, Platt

Takagi

. Crime and Social Justice Associates, Berkeley 1980; 129–141.

51.

51. Steadman

. A situational approach to violence. Int J Law Psychiatry 1982; 5: 171–182.

52.

52. Reason

. Human error: models and management. BMJ 2000; 320: 768–770.

53.

53. Prins

. Will they do it again?. Routledge, London 1999.

54.

54. Otto

. Assessing and managing violence risk in outpatient settings. J Clin Psychol 2000; 56: 1239–1262.

55.

55. Wack

. The ongoing risk assessment in the treatment of forensic patients on conditional release status. Psychiatr Q 1993; 64: 275–293.

56.

56. Douglas

Kropp

. A prevention-based paradigm for violence risk assessment. Crim Justice Behav 2002; 29: 617–658.

57.

57. Mullen

. Care and containment in forensic psychiatry. Crim Behav Mental Health 1993; 3: 212–225.

58.

58. Ward

. The management of risk and the design of Good Lives. Aust Psychologist 2002; 37: 172–179.

59.

59. Glancy

. The clinical use of risk assessment. Can J Psychiatry 2005; 50: 12–17.

60.

60. Dvoskin

Heilbrun

. Risk assessment and release decision-making: toward resolving the great debate. J Am Acad Psychiatry Law 2001; 29: 6–10.

61.

61. Douglas

Webster

Hart

Eaves

Ogloff

JRP

. HCR-20 companion guide. Mental Health Law and Policy Institute, Simon Fraser University, Burnaby BC 2001.

62.

62. Council of the Shire of Wyong v. Shirt 146 CLR 40. 1980.

63.

63. Hodgins

. The major mental disorders and crime: stop debating and start treating and preventing. Int J Law Psychiatry 2001; 24: 427–446.

64.

64. Donaldson

Halligan

. Implementing clinical governance: turning vision into reality. BMJ 2001; 322: 1413–1417.

65.

65. Ogloff

Daffern

. The dynamic appraisal of situational aggression: an instrument to assess risk for imminent aggression in psychiatric inpatients. Behav Sci Law 2006; 24: 799–813.

66.

66. Simon

. Suicide risk assessment: is clinical experience enough?. J Am Acad Psychiatry Law 2006; 34: 276–278.