Deinstitutionalisation and the quest for alternatives to traditional psychiatric wards have been central projects in mental health policy-making, service development and services research in many high-income countries for the past 50 years. Throughout most of this period one of the proposed strategies for reducing dependence on the traditional ward has been development of residential alternatives to acute admission wards. Alternatives have taken the form both of community-based alternatives and of in-patient services that aim to adopt a markedly different model from traditional wards. ^{Reference Johnson, Gilburt, Lloyd-Evans, Osborn, Boardman and Leese1,Reference Johnson, Gilburt, Lloyd-Evans and Slade2} The oldest residential alternatives date back more than 40 years, including Loren Mosher's Soteria crisis house for people with schizophrenia of recent onset, ^{Reference Mosher3} and Paul Polak's network of family homes in which people in crisis could be supported. ^{Reference Polak, Kirby and Deitchman4}

Despite this long history, the idealism and energy invested in many model services and the enthusiasm of service users for residential alternatives, ⁵ they have not become a standard component in catchment area service systems in any country, nor have we had much robust evidence about their organisation, functioning and outcomes. A systematic review identified ten studies as relevant. ^{Reference Lloyd-Evans, Slade, Jagielska and Johnson6} Nine of these were from the USA, most samples were small and analyses were hampered by lack of power. The only conclusion from a meta-analysis was that there is some evidence that service user satisfaction may be greater with alternatives than with standard services. The papers in this supplement have been devoted to two UK studies that have aimed to substantially improve this evidence base. The Alternatives Study is a multiple methods investigation of six residential and in-patient alternatives to standard acute wards in different catchment areas across England. ^{Reference Lloyd-Evans, Slade, Jagielska and Johnson6–Reference Byford, Sharac, Lloyd-Evans, Gilburt, Osborn and Leese12} The Choices Study is a pilot patient-preference randomised controlled trial assessing the effectiveness and cost-effectiveness of two women's crisis houses. ^{Reference Howard, Flach, Leese, Byford, Killaspy and Cole13}

What do we know now about residential alternatives to acute admission?

Methodological challenges are an important barrier to empirical research in this area. Recruitment to research at the time of a mental health crisis is challenging, especially when the research design requires randomisation to take place before crisis intervention begins. ^{Reference Howard, Leese, Byford, Killaspy, Cole and Lawlor14,Reference Johnson, Nolan, Pilling, Sandor, Hoult and McKenzie15} Even if randomised controlled trials of reasonable quality can be conducted, they often leave many questions unanswered when complex mental health interventions are being evaluated. ^{Reference Slade and Priebe16} Our goal in the above-mentioned studies was to overcome some of the difficulties in obtaining clear evidence about complex service interventions by using a mixture of methods, including quantitative and qualitative, natural experimental and randomised, and by investigating eight alternatives in all. The matrix model of Tansella & Thornicroft advocates that we examine services in terms of inputs, processes and outcomes, taking account of national, local and individual levels. ^{Reference Tansella and Thornicroft17} A frequent criticism of mental health service evaluations is that they focus largely on outputs, without clear specification of inputs, such as the local social and service context, and processes, such as content of care. In the Alternatives and Choices studies this has been addressed by including an examination of the functioning of alternatives within their local service systems and by considering both the content and the outcomes of care.

Tansella summarises the main findings of each paper. ^{Reference Tansella18} Overall, what can we now conclude about residential alternatives to acute care within the UK National Health Service (NHS)? Perhaps the clearest conclusion we can draw from our studies echoes the only positive finding of the previous systematic review: ^{Reference Lloyd-Evans, Slade, Jagielska and Johnson6} a substantial quantitative investigation of service user satisfaction agrees with a qualitative investigation in indicating that service users prefer residential alternatives. ^{Reference Osborn, Lloyd-Evans, Johnson, Gilburt, Byford and Leese7,Reference Gilburt, Slade, Rose, Lloyd-Evans, Johnson and Osborn8} The quality of interpersonal relationships seems to be the aspect of services that most influences service users: they value residential alternatives because of better relationships, less coercion and a greater feeling of safety because of the absence of severely disturbed patients.

Beyond this prominent and important difference, we have found much that is similar in standard and alternative services, along with a few distinctive features of alternatives. The alternative services are highly integrated into local catchment area mental health service networks, accepting most of their referrals from other mental health services and working closely with other mental health professionals. They collaborate especially closely with the crisis resolution and home treatment teams now established nationwide in England. ^{Reference Glover and Johnson19} Working in synergy with these teams may have considerably enhanced the capacity of residential alternatives to manage relatively severe crises and to discharge patients quickly once initial difficulties improve. ^{Reference Lloyd-Evans, Gilburt and Johnson20} Service user populations overlap considerably between standard and alternative services, but the latter place restrictions on admissions of people with severe behavioural disturbance or who pose a significant risk to others, and most alternative services do not accept compulsory admissions. ^{Reference Johnson, Lloyd-Evans, Morant, Gilburt, Shepherd and Slade9} The most hospital-like of the community alternatives we investigated did, however, seem to be managing some detained patients successfully. ^{Reference Johnson, Lloyd-Evans, Morant, Gilburt, Shepherd and Slade9} Content of care was in most respects similar between standard and alternative services, apart from a trend towards fewer physical and more psychological interventions in the alternative services. ^{Reference Lloyd-Evans, Johnson, Morant, Gilburt, Osborn and Jagielska10} We found no evidence that staff spent more time with service users in the alternative services, even though staff both within these services and in other collaborating services believed this to be the case.

In terms of outcomes the total improvement during admission was greater in the standard services, but admissions to alternatives were shorter, suggesting that alternatives may discharge to other services at an earlier stage. ^{Reference Slade, Byford, Barrett, Lloyd-Evans, Gilburt and Osborn11} That this did not result in persisting differences in outcomes was suggested both by the similarity in outcomes at 3 months found in the Choices Study, ^{Reference Howard, Flach, Leese, Byford, Killaspy and Cole13} and by similar 1-year readmission rates in the Alternatives Study. ^{Reference Byford, Sharac, Lloyd-Evans, Gilburt, Osborn and Leese12} Over a year's follow-up there was evidence of a cost advantage for the alternatives, largely due to the lower costs of the initial admission.

An important limitation is that we were unable to draw clear conclusions about one of the models we attempted to evaluate. The service included as an example of a distinctive therapeutic model on a standard general acute ward showed little evidence of being different in any substantial way from the standard wards, and staff interviews suggested that the model had not been implemented as intended.

Where next for alternatives to admission?

So can we recommend to those who plan and commission services, in the NHS and elsewhere, that they include residential alternatives to acute admission as part of the spectrum of local acute services? Our findings suggest a cautious affirmative. If the driver for decision-making is cost-effectiveness, we showed that alternatives are associated with clinical improvement, but not to the same extent as standard services, that they cost less, and that post-discharge service use does not differ between people admitted to alternative and to standard services. If the experience of admission is a driver for decision-making, then the argument becomes more compelling, given the robust evidence from multiple sources that service users are more satisfied with the alternatives. Thus where alternatives are currently in operation and local stakeholders satisfied with them, there seems good reason to maintain them.

Our findings do not, however, amount to a persuasive case for seeing any type of alternative to acute admission as an essential component of local acute services. The care provided is not markedly different from that given on standard hospital wards, and evidence about cost-effectiveness is not definitive in view of the lesser improvement compared with a hospital stay in the alternatives.

The alternatives we describe could not replace the requirement for acute in-patient services to care for the most behaviourally disturbed patients, who are often admitted compulsorily for longer periods. Indeed, the provision of alternative services might increase the density of behavioural disturbance on local in-patient units by removing those whose presentations are less challenging. This is not an argument against alternatives, but highlights the urgent need to design and test interventions to improve the experience and nature of standard acute in-patient psychiatric care. Regarding service users' experiences, our study results suggest that amount and quality of therapeutic contact are a key determinant of service users' experiences of acute admission: ^{Reference Gilburt, Slade, Rose, Lloyd-Evans, Johnson and Osborn8,Reference Lloyd-Evans, Johnson, Morant, Gilburt, Osborn and Jagielska10} rather than establishing alternatives, service planners and providers might legitimately decide that their primary priority should be to improve therapeutic contact for all who are admitted to acute facilities. The Tidal Model is one of a range of recent initiatives aimed at doing this. ^{Reference Barker and Buchanan-Barker21} Unfortunately, it did not seem to have been successfully implemented in the service included in our study. However, a variety of models and initiatives aimed at improving the quality of acute in-patient care warrant further investigation, ^{Reference Lelliott, Bennet, McGeorge and Turner22–25} including the simple but widely adopted ‘protected engagement time’ model recommended by the Chief Nursing Officer of the NHS as a means of increasing contact between nurses and patients. ²⁶ Nonetheless, our findings suggest that the availability of residential alternatives to admission benefits at least some service users. The question thus arises of how such services might be improved, in terms of both the range of service users to whom they are available and their outcomes. We suggest that two (in some ways contrasting) directions for service development might be fruitful. Both our initial survey and our subsequent study indicated that the more hospital-like, clinically oriented alternatives served groups who were closer in clinical profile and severity to those on acute wards. ^{5,Reference Johnson, Lloyd-Evans, Morant, Gilburt, Shepherd and Slade9} Thus, if choice is to be provided for people with severe psychotic and bipolar disorders, including some of those compulsorily admitted, services that aim (as in hospital) to provide a full range of interventions of proven effectiveness for these groups are desirable. These interventions do not all need to be directly provided by the staff of the residential alternative, but close links with community professionals in services such as crisis teams will be needed if they are to be readily available. Thus services in the clinical crisis house and crisis team beds groups in our typology of alternatives may thrive by aiming to provide as full as possible a range of effective acute interventions, but in a milieu that is markedly different and more acceptable to service users than hospital.

Our qualitative studies, however, made it apparent that alternative services were often valued when they took approaches that were different from those available in hospital. This seemed to apply especially to some groups not seen as well served by acute wards, such as people with depression or personality disorders. Both in the first phase of the Alternatives Study and in the later studies, we were struck that few community alternatives explicitly aimed to implement models that were radically different from hospital care. Indeed, it is rare for such alternative services to be based on a predefined theory or model, and unconventional forms of service organisation, such as user-led services, are also uncommon in the NHS. Thus the implementation and evaluation of more distinctive models that provide truly alternative forms of care would be of interest in terms both of service development and research. Such models are found outside the NHS; for example, user-led in-patient services based on a recovery model exist in several countries. ^{Reference Slade27} A randomised controlled trial in the USA compared admission to a consumer-run crisis residential programme (n = 196) with standard admission (n = 197), and showed that the alternative was associated with greater symptomatic improvement and strongly associated with greater satisfaction. ^{Reference Greenfield, Stoneking, Humphreys, Sundby and Bond28} Other parts of the world also offer residential alternatives with approaches that appear markedly different from standard in-patient services. These include replications of the Soteria model in Switzerland and Germany, ^{Reference Ciompi and Hoffmann29} Italian services that closely integrate community mental health centres with acute residential care, ^{Reference Mezzina and Johnson30} and services in various parts of the USA that are based on placement with families. ^{Reference Bennett31} The research evidence regarding many of these models remains limited, so that multiple methods investigations of the kind that we have conducted would be of considerable interest.

In research terms, valuable work remains to be done in investigation of the models that we have described. Longer-term investigation of a broader range of outcomes, and replication of our work at other services and in other countries, would be valuable. Further research comparing alternative services with crisis team and acute day services would also be of value in helping to identify which acute service models work best for which situations and service users. Future research will be more informative if it focuses on clearly defined models with specified aims, theoretical underpinnings and content, with research aims including identification of critical ingredients and mechanisms in successful services and assessment of model fidelity. The typology of alternatives that we have developed offers a starting point for this work. In terms of critical ingredients, our qualitative research suggests that the therapeutic relationship is central to patient experiences: the contribution of good therapeutic relationships to good outcomes and the factors that promote or impede such relationships could fruitfully be investigated in future studies. The new Medical Research Council Framework for Evaluation of Complex Interventions provides relevant guidance, ^{Reference Craig, Macintyre, Macintyre, Michie, Nazareth and Pettigrew32} advocating the use of experimental methods where this is feasible, but with considerable flexibility of approach often required to make evaluations feasible and valid. Our hope is that the studies discussed here will be among the first substantial steps in developing a robust evidence base regarding residential alternatives to hospital, so that a consensus can at last be achieved regarding their value, their role within local acute care systems, and how they can most effectively contribute to service users' recovery following acute crises.