People with severe mental illness and substance use disorders (dual disorder) often have considerable contact with the criminal justice system.

To test the effects of client characteristics on six criminal justice outcomes among homeless (at intake) people with mental illness and substance misuse disorders.

The sample was of participants in a randomized controlled trial comparing standard treatment, assertive community treatment (ACT) and integrated treatment (IT). Data were analysed using hierarchical logistic regression.

Half the sample was arrested and a quarter incarcerated during the two‐year follow‐up period. The regression models explained between 22% and 35% of the variance of the following criminal justice measures: (1) major offences, (2) minor offences, (3) substance‐use‐related offences, (4) incarcerations, (5) arrests, and (6) summons. Prior criminal behaviour was the strongest predictor of all of the dependent variables; in general, demographic and diagnostic variables were not. Similarly, neither the type nor the amount of mental health treatment received predicted subsequent criminal behaviour.

Elsewhere the authors have shown that ACT and IT had advantages for health and stability of accommodation but these analyses suggest that more specialized interventions are needed to reduce criminal behaviour in dual disorder individuals. Copyright � 2005 John Wiley & Sons, Ltd.