Background Across England, around 60,000 people live in mental health supported accommodation: residential care, supported housing and floating outreach. Residential care and supported housing provide on-site support (residential care provides the highest level), whereas floating outreach staff visit people living in their own tenancies. Despite their abundance, little is known about the quality and outcomes of these services. Objectives The aim was to assess the quality, costs and effectiveness of mental health supported accommodation services in England. The objectives were (1) to adapt the Quality Indicator for Rehabilitative Care (QuIRC) and the Client Assessment of Treatment scale for use in mental health supported accommodation services; (2) to assess the quality and costs of these services in England and the proportion of people who ‘move on’ to less supported accommodation without placement breakdown (e.g. to move from residential care to supported housing or supported housing to floating outreach, or, for those receiving floating outreach, to manage with fewer hours of support); (3) to identify service and service user factors (including costs) associated with greater quality of life, autonomy and successful move-on; and (4) to carry out a feasibility trial to assess the required sample size and appropriate outcomes for a randomised evaluation of two existing models of supported accommodation. Design Objective 1 – focus groups with staff (n = 12) and service users (n = 16); psychometric testing in 52 services, repeated in 87 services (adapted QuIRC) and with 618 service users (adapted Client Assessment of Treatment scale). Objectives 2 and 3 – national survey and prospective cohort study involving 87 services (residential care, n = 22; supported housing, n = 35; floating outreach, n = 30) and 619 service users followed over 30 months; qualitative interviews with 30 staff and 30 service users. Objective 4 – individually randomised, parallel-group feasibility trial in three centres. Setting English mental health supported accommodation services. Participants Staff and users of mental health supported accomodation services. Interventions Feasibility trial involved two existing models of supported accommodation: supported housing and floating outreach. Main outcome measures Cohort study – proportion of participants who successfully moved to less supported accommodation at 30 months’ follow-up without placement breakdown. Feasibility trial – participant recruitment and withdrawal rates. Results The adapted QuIRC [QuIRC: Supported Accomodation (QuIRC-SA)] had excellent inter-rater reliability, and exploratory factor analysis confirmed its structural validity (all items loaded onto the relevant domain at the > ± 0.3 level). The adapted Client Assessment of Treatment for Supported Accommodation had good internal consistency (Cronbach’s alpha 0.89) and convergent validity (r _s = 0.369; p < 0.001). Supported housing services scored higher than residential care and floating outreach on six out of seven QuIRC-SA quality domains. Service users had a high prevalence of severe self-neglect (57%) and vulnerability to exploitation (37%). Those in supported housing (25%) and floating outreach (20%) experienced more crime than those in residential care (4%) but had greater autonomy. Residential care was the most expensive service (mean cost per resident per week was £581 for residential care, £261 for supported housing and £66 for floating outreach) but supported users with the greatest needs. After adjusting for clinical differences, quality of life was similar for users of supported housing and residential care (mean difference –0.138, 95% confidence interval –0.402 to 0.126; p = 0.306), whereas autonomy was greater for supported housing users (mean difference 0.145, 95% confidence interval 0.010 to 0.279; p = 0.035). Qualitative interviews showed that staff and service users shared an understanding of service goals and what constituted effective support. After adjusting for clinical differences, those in floating outreach were more likely to move on successfully at 30 months’ follow-up than those in residential care [odds ratio (OR) 7.96; p < 0.001] and supported housing (OR 2.74; p < 0.001), and this was more likely for users of supported housing than residential care (OR 2.90; p = 0.04). Successful move-on was positively associated with scores on two QuIRC-SA domains: the degree to which the service promoted ‘human rights’ (e.g. facilitating access to advocacy) and ‘recovery-based practice’ (e.g. holding therapeutic optimism and providing collaborative, individualised care planning). Service use costs for those who moved on were significantly lower than for those who did not. Recruitment in the feasibility trial was difficult: 1432 people were screened but only eight were randomised. Barriers included concerns about accommodation being decided at random and a perceived lack of equipoise among clinicians who felt that individuals needed to ‘step down’ from supported housing to floating outreach services. Conclusions We did not find clear evidence on the most effective model(s) of mental health supported accommodation. Indeed, our feasibility study suggests that trials comparing effectiveness cannot be conducted in this country. A range of options are required to provide appropriate support to individuals with differing needs. Future work Future research in this field requires alternatives to trials. Service planners should be guided by the mental health needs of the local population and the pros and cons of the different services that our study identified, rather than purely financial drivers. Trial registration Current Controlled Trials ISRCTN19689576. Funding This programme was funded by the National Institute for Heath Research (NIHR) Programme Grants for Applied Research programme and will be published in full in Programme Grants for Applied Research; Vol. 7, No. 7. See the NIHR Journals Library website for further project information. The fundholders are Camden and Islington NHS Foundation Trust and the research is a collaboration between University College London, Queen Mary University of London, King’s College London, the University of Ulster and Durham University.