Objectives: To define the characteristics and determine the effectiveness of organised inpatient (stroke unit) care compared with conventional care in reducing death, dependency, and the requirement for long term institutional care after stroke. Design: Systematic review of all randomised trials which compared organised inpatient stroke care with the contemporary conventional care. Specialist stroke unit interventions were defined as either a ward or team exclusively managing stroke (dedicated stroke unit) or a ward or team specialising in the management of disabling illnesses, which include stroke (mixed assessment/rehabilitation unit). Conventional care was usually provided in a general medical ward. Setting: 19 trials (of which three had two treatment arms). 12 trials randomised a total of 2060 patients to a dedicated stroke unit or a general medical ward, six trials (647 patients) compared a mixed assessment/rehabilitation unit with a general medical ward, and four trials (542 patients) compared a dedicated stroke unit with a mixed assessment/rehabilitation unit. Main outcome measures: Death, institutionalisation, and dependency. Results: Organised inpatient (stroke unit) care, when compared with conventional care, was best characterised by coordinated multidisciplinary rehabilitation, programmes of education and training in stroke, and specialisation of medical and nursing staff. The stroke unit care was usually housed in a geographically discrete ward. Stroke unit care was associated with a long term (median one year follow up) reduction of death (odds ratio 0.83, 95% confidence interval 0.69 to 0.98; PConclusions: Organised stroke unit care resulted in long term reductions in death, dependency, and the need for institutional care. The observed benefits were not restricted to any particular subgroup of patients or model of stroke unit care. No systematic increase in the use of resources (in terms of length of stay) was apparent.