Objective: To assess the clinical effectiveness of an early discharge policy for patients with stroke by using a community based rehabilitation team. Design: Randomised controlled trial to compare conventional care with an early discharge policy. Setting: Two teaching hospitals in inner London. Subjects: 331 medically stable patients with stroke (mean age 71) who lived alone and were able to transfer independently or who lived with a resident carer and were able to transfer with help. Interventions: 167 patients received specialist community rehabilitation for up to 3 months after randomisation. 164 patients continued with conventional hospital and community care. Main outcome measures: Barthel score at 12 months. Secondary outcomes measured impairment with motoricity index, minimental state examination, and Frenchay aphasia screening test; disability with the Rivermead activity of daily living scales, hospital anxiety and depression scale, and 5 m walk; handicap with the Nottingham health profile; carer stress with caregiver strain index and patient and carer satisfaction. The main process measure was length of stay after randomisation. Results: One year after randomisation no significant differences in clinical outcomes were found apart from increased satisfaction with hospital care in the community therapy group. Length of stay after randomisation in the community therapy group was significantly reduced (12 v 18 days; P<0.0001). Patients with impairments were more likely to receive treatment in the community therapy group. Conclusions: Early discharge with specialist community rehabilitation after stroke is feasible, as clinically effective as conventional care, and acceptable to patients. Considerable reductions in use of hospital beds are achievable. Key messages Early discharge from hospital after stroke with specialist rehabilitation at home is feasible without an increase in readmission rates or stress to carers This randomised controlled trial shows this method to be as effective as conventional care when assessed with a range of measures of impairment, disability, handicap, carer stress, and patient and carer satisfaction at 1 year Significant reductions in bed usage can be achieved by the provision of a community rehabilitation team with no significant increase in rehabilitation services