OBJECTIVES: To test the hypotheses that older people and their informal carers are not disadvantaged by home-based rehabilitation (HBR) relative to day hospital rehabilitation (DHR) and that HBR is less costly. DESIGN: Two-arm randomised controlled trial. SETTING: Four trusts in England providing both HBR and DHR. PARTICIPANTS: Clinical staff reviewed consecutive referrals to identify subjects who were potentially suitable for randomisation according to the defined inclusion criteria. INTERVENTIONS: Patients were randomised to receive either HBR or DHR. MAIN OUTCOME MEASURES: The primary outcome measure was the Nottingham Extended Activities of Daily Living (NEADL) scale. Secondary outcome measures included the EuroQol 5 dimensions (EQ-5D), Hospital Anxiety and Depression Scale (HADS), Therapy Outcome Measures (TOMs), hospital admissions and the General Health Questionnaire (GHQ-30) for carers. RESULTS: Overall, 89 subjects were randomised and 42 received rehabilitation in each arm of the trial. At the primary end point of 6 months there were 32 and 33 patients in the HBR and DHR arms respectively. Estimated mean scores on the NEADL scale at 6 months, after adjustment for baseline, were not significantly in favour of either HBR or DHR [DHR 30.78 (SD 15.01), HBR 32.11 (SD 16.89), p = 0.37; mean difference -2.139 (95% CI -6.870 to 2.592)]. Analysis of the non-inferiority of HBR over DHR using a 'non-inferiority' limit (10%) applied to the confidence interval estimates for the different outcome measures at 6 months' follow-up demonstrated non-inferiority for the NEADL scale, EQ-5D and HADS anxiety scale and some advantage for HBR on the HADS depression scale, of borderline statistical significance. Similar results were seen at 3 and 12 months' follow-up, with a statistically significant difference in the mean EQ-5D(index) score in favour of DHR at 3 months (p = 0.047). At the end of rehabilitation, a greater proportion of the DHR group showed a positive direction of change from their initial assessment with respect to therapist-rated clinical outcomes; however, a lower proportion of HBR patients showed a negative direction of change and, overall, median scores on the TOMs scales did not differ between the two groups. Fewer patients in the HBR group were admitted to hospital on any occasion over the 12-month observation period [18 (43%) versus 22 (52%)]; however, this difference was not statistically significant. The psychological well-being of patients' carers, measured at 3, 6 and 12 months, was unaffected by whether rehabilitation took place at day hospital or at home. As the primary outcome measure and EQ-5D(index) scores at 6 months showed no significant differences between the two arms of the trial, a cost-minimisation analysis was undertaken. Neither the public costs nor the total costs at the 6-month follow-up point (an average of 213 days' total follow-up) or the 12-month follow-up point (an average of 395 days' total follow-up) were significantly different between the groups. CONCLUSIONS: Compared with DHR, providing rehabilitation in patients' own homes confers no particular disadvantage for patients and carers. The cost of providing HBR does not appear to be significantly different from that of providing DHR. Rehabilitation providers and purchasers need to consider the place of care in the light of local needs, to provide the benefits of both kinds of services. Caution is required when interpreting the results of the RCT because a large proportion of potentially eligible subjects were not recruited to the trial, the required sample size was not achieved and there was a relatively large loss to follow-up. TRIAL REGISTRATION: Current Controlled Trials ISRCTN71801032.