ABSTRACTWith research showing a high prevalence of psychosis in prisons, its effective management is essential for clinical and criminal outcomes. In a matched sample of released prisoners and discharged patients with psychosis (124 pairs) we investigated whether group participation (prison vs. hospital) affected the likelihood of reoffending as well as time to reoffending. Statistical analysis was completed using multilevel logistic regression and multilevel survival analysis. We found that prison cases were more likely to commit any offence within a given period of time (1, 2 or 3 years). Moreover, at any particular time 3 times as many patients with psychosis released from prisons reoffended proportionally to their matched controls discharged from hospitals (HR = 2.92, 95% CI = 1.99, 4.29, P < 0.001). In conclusion, release from prison carries higher risk of a future offending and reduced time to reoffending among offenders with psychosis. Notwithstanding limitations inherent in observational study designs, we advocate that in addition to clinical need and the humanitarian argument, offenders with psychosis should be treated in secure hospitals to reduce future recidivism.