Psychological Medicine

Original Article

Actuarial prediction of violent recidivism in mentally disordered offenders

ROBERT J. SNOWDENa1 c1, NICOLA S. GRAYa1a2, JOHN TAYLORa3 and MALCOLM J. MacCULLOCHa1a2a4

a1 School of Psychology, Cardiff University, Cardiff, UK

a2 South Wales Forensic Psychiatric Service, Caswell Clinic, Glanrhyd Hospital, Mid Glamorgan, UK

a3 Kneesworth House Hospital, Bassingbourne-cum-kneesworth, Royston, Hertfordshire, UK

a4 Department of Psychological Medicine, University of Wales College of Medicine, Cardiff, UK

ABSTRACT

Background Actuarial instruments may be useful in predicting long-term violence in mentally disordered patients. We compared two instruments that differ in terms of what they are designed to predict (general versus violent recividism) and the inclusion of stable mental health variables.

Method A large sample of mentally disordered patients were scored on two risk assessment instruments, the Violence Risk Appraisal Guide (VRAG) and the Offender Group Reconviction Scale (OGRS), based on information at the point of discharge. Their criminal histories for at least 2 years following discharge were obtained from official records.

Results Both instruments were good predictors of both violent and general offending. Over shorter periods (<1 year) the VRAG had very good predictive validities for both types of offences [areas under the receiver operating characteristic curves (AUCs) >0·85], which were significantly better than the OGRS. For longer follow-up periods the instruments had approximately equal prediction accuracy. However, both instruments predicted far more offences than were in fact recorded.

Conclusions The VRAG is a very good predictor of future violence in the UK sample. The OGRS may also be of value as it can be completed quickly and without the need for mental health variables. Caution is needed, however, as both instruments appeared to over-predict the levels of reconvictions in this sample.

(Online publication May 31 2007)

Correspondence

c1 Address for correspondence: Professor R. J. Snowden, School of Psychology, Cardiff University, Cardiff CF10 3AT, Wales, UK. (Email: snowden@cardiff.ac.uk)

Metrics