Violent figures, risky stories

Doctor (2004, this issue) raises the important question of how to assess and manage violent behaviour. He claims that actuarial models of risk assessment based on epidemiology have failed, and that indeed the most reliable risk assessment is one based on clinical methods. He proposes that psychodynamic psychotherapy be used for this purpose, to uncover the meaning of the violent behaviour. Such therapy will, Doctor argues, help the violent person be aware of and understand the function of their behaviour. This understanding in turn will reduce the need in the patient to act out violently.

But with this argument Doctor only addresses the second part of his topic: how to manage violent behaviour. Individual clinical assessment may be suitable for understanding the individual patient, but it may be doubted whether psychodynamic psychotherapy will help us assess the risk of future violent behaviour. For such risk assessments we do seem to need figures.

We cannot ignore the many studies that have shown that clinical methods are outperformed by actuarial methods in predicting violence. To come to a valid prediction, only a small number of cues need be taken into account, and the single most predictive cue is past violence. Although clinicians often make moderately valid short- and long-term predictions of violence on the basis of interviews and demographic data, significantly more accurate results have been obtained with statistical prediction rules. Criminal history variables are the best predictors, and clinical variables show the smallest effect sizes (e.g. see Mossman, 1994; Gardner et al, 1996; Bonta et al, 1998).

Indeed, ‘very crude terms’, as Doctor chooses to call them, are not insufficient for prediction at all, and the assessor does not need all that much information to make a valid prediction. So who needs clinical methods? Doctor advocates their use. Maybe not because he underestimates the power of statistics, but because he is simply more concerned with understanding and managing (violent) patients than with predictions of their violent behaviour.

Of course statistics do not help us to understand our patients. Also, stories are much more compelling than statistics. Newman (2003) makes a convincing case for the power of stories over statistics. He juxtaposes the eye-witness account of a flight attendant describing the distress of a mother who lost her baby in a crash after he had advised her to place the infant on the floor of the aeroplane with calculations regarding the evidence that providing child restraints in aeroplanes would save hardly any lives and cost millions. Personal stories have much more impact on decision-makers than calculations of costs and benefits. We can identify with the mother, but not with an amount of money. People do not die or commit violent acts statistically; they really die and act violently.

Clinical v. actuarial: need we choose?

Mental health professionals should use what works best. Research in evidence-based medicine tells us that we do well to use actuarial methods, since using the evidence there is improves patient outcomes. But to understand the patient, such external evidence should be complemented with individual clinical experience and judgement, and the patient’s unique story is quite important (Greenhalgh, 2002). In Greenhalgh’s view, no one ever needs to choose between evidence-based practice and clinical expertise. Clinical expertise generates the hypotheses that may then be tested scientifically against the available evidence; and the evidence figures in the hypotheses.

Introducing narratives in the clinical encounter has clear advantages. The clinician could use the available evidence, in the form of an illness-script or a DSM classification or a nosological or other model, as a skeleton explanation. They could then flesh out this skeleton with individual patient data, thereby creating a well-founded yet personalised narrative or story. The patient could also present their narrative, and the clinician would match this story to actuarial evidence about the hypothesised illness. The result is a healthy mix of statistical and clinical input: a story that facilitates communication between clinician and patient, and that at the same time incorporates the available evidence. Like Greenhalgh, I see no need to choose between the clinical and the statistical. On the contrary: why not have the best of both? Incorporating statistics would keep clinicians from being drawn into the patient’s narrative, which is the most available and vivid explanation of their behaviour, but possibly not the best one. Adding clinical insights to statistical explanations would give meaning to the figures. It improves our understanding of why this individual would, for example, perform violent acts and how to manage it.

Understanding violent patients through a narrative-based approach is just as insufficient for predicting violence as understanding the meaning of violent behaviour through a psychodynamic approach. If narrative-based medicine, or the psychodynamic approach, is to make a difference, common key elements in patients’ narratives should be taken up in the cues that are used in prediction. It remains to be tested whether the predictive value of actuarial methods using these cues is then really improved, over just using the single cue of past violence. Until that time, it seems irresponsible to trade in actuarial methods for clinical methods in the prediction of the risk of violence.

Conclusions

Mental health professionals can use whatever suits their professional expertise in trying to understand their violent patients: psychodynamic psychotherapy, the patient’s narrative, or their own stories based on their experience and training. But their methods ought not to be used in predictions without validation. The effectiveness of the different approaches in predicting violence needs to be established through well-designed comparative studies. That is, the evidence-base of the chosen approach should be uncovered. Meanwhile, for the sake of the safety of the general public and the patients themselves, actuarial methods cannot be discarded. Indeed, clinicians should ‘[retreat] emotionally into … a scientific attitude’ (Doctor, 2004, this issue) not to blind them to what happens, but to add to the scientific value of their predictions. Clinical psychology and psychiatry are sciences, not arts. The bottom line is, as Holloway (2004, this issue) puts it, that ‘we all need to learn how to combine clinical wisdom with reliable evidence’.