Background
In Western Europe and North America, the process of de-institutionalisation, defined as “…
the practice of caring for individuals in the community rather than in an institutional environment” (p.47), has led to a significant increase in community based care for people with severe mental illness (SMI) [
1]. Housing-based support, or supported accommodation, operates as a component of the broader mental health ‘care pathway’ and attempts to meet the needs of service users by providing focussed, flexible support. In this context, support aims to address functional impairment, develop practical living skills, improve social functioning and promote recovery and independence [
2]. Mental health supported accommodation services have been implemented widely; recent estimates indicate that, in the UK alone, over 60,000 individuals are currently receiving support in these settings [
3]. Due to high rates of service use, and expenditure related to staffing, support and infrastructure, this form of intervention is also extremely costly. However, despite the broad implementation of these services and the associated financial burden, little is known about their effectiveness.
Definitional issues are well documented in the literature, and present a significant obstacle to the assessment of the effectiveness of supported accommodation. Both within and between countries, supported accommodation services vary widely in terms of physical structure, staffing arrangements, levels of support, recovery focus, and discharge and move-on policies, contributing to confusion as to what exactly a supported accommodation service ‘looks’ like. Despite these issues being discussed in the literature for over 20 years, there have been few meaningful attempts to address them. As a result, the available literature is heterogeneous in nature and resistant to synthesis attempts, leaving researchers and policy makers with no clear summary of the bigger picture; that is, what works and for whom.
For these reasons, previous attempts to summarise the evidence base have been largely unsatisfactory. O’Malley and Croucher [
4] conducted a scoping study of supported accommodation services for people with mental health problems in the UK, aiming to explore evidence for models of good practice. After reviewing 131 studies from an original pool of 2506, the authors concluded that most services are based on the assumption that service users will progress from higher to lower levels of supported accommodation over time, however they could not identify any “
concrete evidence to support any particular model of housing support” (p.841). Due to the methodology used, the authors did not undertake an assessment of the quality of the publications, thus significantly limiting the validity of the findings. In addition, the study focussed solely on UK papers and is also now more than 10 years old. More recently, a Cochrane Review (initially conducted in 2002, and updated in 2006) [
5,
6] compared the efficacy of supported housing schemes, outreach support and standard care. The systematic review considered only randomised controlled trials (RCT) and quasi-randomised trials. A thorough search identified 139 potential studies for inclusion, but after review, none fulfilled the inclusion criteria. While the superiority of RCTs as a ‘gold standard’ for providing evidence for effectiveness is widely acknowledged, there is also a growing argument for considering other quantitative evidence beyond RCT studies. This is particularly salient in cases where RCTs are not possible due to ethical or pragmatic concerns, as is typically the case in in supported accommodation research. The Cochrane review provides a stark comment on the state of the literature in the field, however it does little to describe the existing evidence base.
In light of these observations, it is clear that there is an urgent need to summarise the current evidence as it relates to mental health supported accommodation services. We therefore undertook a comprehensive systematic review of data from quantitative studies in the field, incorporating evidence beyond that derived from RCTs alone. Our aim was to synthesise the current evidence on mental health and psychosocial outcomes for individuals residing in mental health supported accommodation, making comparisons between different models, where the quality of evidence allowed. Our objective was to report findings likely to be of interest to those providing and commissioning mental health supported accommodation, as well as policy makers, and to highlight areas for future research.
This review follows the PRISMA guidelines [
7]. The PRISMA checklist and review protocol are available and can be requested directly from the authors.
Discussion
This systematic review attempted to synthesise the literature on mental health and psychosocial outcomes associated with mental health supported accommodation services. Despite the initial aim of comparing and contrasting outcomes across supported accommodation models, the wide variation in accommodation services, in terms of structure, staffing and related variables, required us to group our findings by population sub-groups.
Homeless
The strongest evidence for supported accommodation comes from research with homeless mentally ill populations and the permanent supported housing model. Studies in this area demonstrate consistent evidence for improvements in housing retention and stability, and appropriate use of clinical services over time. There is also some indication that this form of support for this group is associated with improvements in symptoms, QoL and social functioning, but this evidence is inconsistent. The majority of studies reviewed found no change in substance use over time. As stated by Tsermberis [
97], “
Housing First and other supportive housing interventions may end homelessness but do not cure psychiatric disability, addiction, or poverty” (p.52). These findings are in line with the conclusions of a recent review of HF [
98].
Although the permanent supported housing approach has been shown to be effective in some domains, the intervention specifically targets mentally ill homeless populations, and the characteristics of the studied cohorts make generalising to other mental health populations troublesome. First, many of the samples used in these studies have relatively low rates of serious psychiatric illness. In the current review, we utilised a > 50% with psychiatric diagnosis cut-off point to ensure we included appropriate studies. However, even with this approach, it remains difficult to confidently apply the synthesised findings to other groups of people with mental health problems. Second, the presence of long term homelessness amongst the target population conflates the findings when considering the applicability to general SMI populations. It remains difficult to establish whether positive changes in psychosocial outcomes are attributable to intervention components that impact on homelessness (such as housing), or mental health (such as medication management) or both. Third, participants in the permanent supported housing studies typically present with higher rates of drug and alcohol use than comparable, non-homeless samples. There is danger that, due to the large and growing evidence base for these services, policy makers will attempt a wholesale import of the permanent supported housing model for use with psychiatric populations, without a reliable evidence base.
Deinstitutionalisation
Research on outcomes in supported accommodation for deinstitutionalised populations provided good evidence for improvement or non-deterioration in psychiatric symptoms, social functioning and rates of rehospitalisation. There was limited evidence for improvement in QoL and employment. Notably, a number of studies highlighted a consistent association between more restrictive settings and poorer outcomes, across psychiatric, social and QoL outcomes, for this group. Although, these findings are somewhat inconsistent, the threshold of ‘success’ for this population is radically different than for other groups. Due to the severity of clinical presentations and duration of institutionalised care, most researchers and clinicians consider the absence of deterioration as indicative of successful transition to community care. Indeed, one of the greatest challenges of the deinstitutionalisation process was to address the chronic psychiatric, social and behavioural difficulties of patients, while simultaneously maintaining their tenure in the community [
1]. Supported accommodation services appear to have contributed to the achievement of these goals; the reported findings, while not consistently demonstrating improvements across domains do, for the most part, highlight stability.
In reality, the deinstitutionalisation ‘story’ has already been told; in most European and north American countries, the deinstitutionalisation process commenced the late 1980s and early 1990s and, as such, the majority of the studies cited in this review are old, or report on longer-term follow-ups. It is generally accepted that community based settings are more humane and offer a better QoL than long term hospitalisation [
99]. The transition of people from long-stay wards to community-based care has been successful, and the evidence suggests that this group, for the most part, can be maintained in community settings without any significant deterioration [
100]. This is an important and well established finding but does little to guide us in the development and implementation of contemporary supported accommodation services.
General SMI
The reviewed papers in this group presented less clear evidence between supported accommodation and psychosocial outcomes for general SMI populations. While there was a trend toward reductions in rates of hospitalisation over time, the evidence was mixed with regards to symptoms, social functioning and QoL, with studies variously demonstrating improvement, no-change, or deterioration in these outcomes over time. These findings may reflect the heterogeneous nature of the literature.
This sub-group had the fewest number of studies overall, the fewest number of high-quality papers and the largest number of omitted low-quality studies, yet this population is growing rapidly, reflecting the broad adoption of the supported accommodation model, current approaches to community-based rehabilitation and the rejection of long-term hospitalisation as a form of treatment.. This observation highlights an urgent need for increased research in this area. As mentioned above, there is a genuine danger that due to the growing evidence base, the HF model is applied to this group despite the problems in generalising the research findings.
Strengths and limitations
The current review had a number of strengths. We applied a thorough search strategy, utilising a broad date range, and included a range of psychosocial outcomes and a variety of quantitative designs beyond RCTs. These methodological decisions enabled us to be comprehensive in our review and confident in capturing all key outcome studies.
As the reviewed studies relate to supported accommodation only, it must be acknowledged that, by examining the outcomes in relation to subgroups, broader findings related to these groups may have been overlooked. For example, many of the Team for the Assessment of Psychiatric Services (TAPS) studies, and other programmes of research investigating outcomes for deinstitutionalised groups, were not included in this review as they did not explicitly consider supported housing, however data from these studies would have expanded and contextualised the reported findings as they relate to the deinstitutionalisation sub-group. The population-based conclusions, therefore, must be considered strictly in relation to supported accommodation.
As we were unable to examine differences in outcomes across models, the current review cannot comment on their relative merits in relation to outcomes. As described, there is a large variation in housing models; within each of the population sub-groups considered above, service models ranged from independent tenancies with outreach support to high-staffed, congregate residential settings. Inevitably, the characteristics of a service, such as the physical structure, staffing arrangements, levels of support, recovery focus, and discharge and move-on policies, will impact on service user outcomes, possibly beyond the influence of population characteristics. As a result, this review is limited in its ability to fully consider the effectiveness of mental health supported accommodation services.
By comparing services from different national contexts, we aimed to enhance our understanding of the critical components of these interventions and how contextual factors impact outcomes. However, due to the aforementioned variation in service models (evident even within countries), it was difficult to discern the impact of national level factors, such as legislation, funding barriers or statutory responsibilities. The international focus of this review makes it challenging to provide any specific recommendations for local policy makers and commissioners. A more targeted study, focusing on one country or region, would be better suited for this purpose. In line with the recovery approach, however, it is likely that any high-quality supported accommodation provision will comprise of a range of accommodation options, with the delivery of flexible, personalised support.
Finally, as we have considered evidence from non-RCT designs, the data presented herein, even from studies rated as ‘high quality’, should be interpreted with caution.
Conclusion
The mixed results of this study highlight the heterogeneity of the supported accommodation literature, in terms of research quality, experimental design, population, service types and outcomes assessed. There is some evidence that supported accommodation is effective across a range of psychosocial outcomes, with the most robust evidence showing the effectiveness of the HF model for homeless SMI and for other forms of supported accommodation for deinstitutionalised populations in reducing hospitalisation rates and improving appropriate service use. The evidence base for general SMI populations is less developed, and requires further research. Unfortunately, these broad observations reinforce the conclusions of Chilvers and colleagues [
5] in their recent Cochrane review: “
In the absence of evidence of their relative efficacy, decisions on the provision of alternative forms of accommodation and continued support for people with mental illness can only be based on a combination of professional judgement, patient preference and availability” (p.6).
The intention of the current review was to compare and contrast the effectiveness of various models of supported accommodation, across a range of psychosocial outcomes. However, as noted, this attempt was stymied by the large variation in service models, the lack of definitional consistency and, at times, poor reporting practices in the literature. In order to make assertions regarding the effectiveness of various models of supported accommodation, it is clear that a simple method of service categorisation, based on current reporting practices, is required. A taxonomy that can be applied retrospectively to existing research, and utilised in future studies, would allow effective synthesis of outcome data, facilitate an examination of efficacy and effectiveness, and strengthen follow-up/replication studies [
101]. While some attempts have been made to develop a supported accommodation taxonomy [
102,
103], these models are complex and have not been widely utilised. Recently, a new, simple classification system for supported accommodation services has been developed (The Simple Taxonomy for Supported Accommodation [STAX- SA] [
104]). Future research should consider utilising this tool to synthesise the available effectiveness evidence, comparing service user outcomes across service models.
Mental health supported accommodation services are widely implemented, however, currently we have no clear research base articulating what works and for whom. There is a clear need for high quality effectiveness research, improved reporting standards and consistent and meaningful descriptions of supported accommodation services in the literature. Researchers must prioritise experimental studies that compare outcomes across different service models. These developments should inform and improve mental health commissioning and service development decisions in the future.
Acknowledgements
We would like to acknowledge current and past members of the Quality and Effectiveness of Supported Tenancies (QuEST) research group at UCL/QMUL (Christian Dalton-Locke, Sarah Dowling, Isobel Harrison, Rose McGranahan, Sima Sandhu,) and the QuEST programme management group (Maurice Arbuthnott, Sarah Curtis, Sandra Eldridge, Michael King, Gerry Leavey, Gavin McCabe, Paul McCrone, Stefan Priebe and Geoff Shepherd) for their support with this project.