Who will behave violently in the next two years?

Psychologists and psychiatrists often get the blame on the rare occasions that a former mental health patient goes on to commit a violent act (see note below). They are expected to be able to predict which patients are a risk. But a new American study has found a sample of 67 psychologists, psychiatrists, nurses and social workers were unable to use archived hospital admission evaluations and clinical notes to predict which of 52 patients went on to behave violently in the next two years. The finding comes from an investigation by Michael Odeh and colleagues into the cues used by mental health professionals to predict whether a patient is likely to be violent.

The 67 clinicians were asked to use the patient reports to predict how likely it was that each patient went on to be violent in the two years following their hospital admission. Afterwards the clinicians were asked which information they had used to come to that judgment.

The thirteen most commonly used cues for predicting dangerousness were: past assaults, non-compliance with medication, history of substance abuse, presence of psychosis, violent thoughts, previous admission to a psychiatric hospital, paranoid delusions, a diagnosis of mental illness, uncooperativeness, a history of poor impulse control, prior use of a weapon, hostility, and family problems. However, these cues did not accurately predict which patients went on to behave violently in the next two years.

The research also revealed professional differences. For example, nurses and social workers cited ‘hostility’ three times as frequently as psychologists and psychiatrists. They were also twice as likely to cite delusions, medication compliance and family problems as relevant cues predicting dangerousness – perhaps, the authors suggested, because of their more direct involvement in patient care.

While noting that the clinicians in this study never actually had the opportunity to meet the patients they were assessing, the researchers concluded that “the findings in this study may be further justification for a more structured approach to help clinicians evaluate risk factors when making clinical predictions”.

An earlier study by the same researchers suggested that although individual clinicians’ predictions of violence were inaccurate, accuracy was achieved by aggregating the judgments of multiple clinicians.

Odeh, M.S., Zeiss, R.A. & Huss, M.T. (2006). Cues they use: Clinicians’ endorsement of risk cues in predictions of dangerousness. Behavioural Sciences and The Law, 24, 147-156.

Post written by Christian Jarrett (@psych_writer) for the BPS Research Digest.

Note, 95 per cent of murders are not committed by psychiatric patients and most psychiatric patients are not dangerous.