Homelessness, defined as the absence of customary and regular access to a conventional dwelling or residence 1 , has been recognized as a growing social and public health problem in developed countries since the end of the 1980s. An estimated 100,000 people live on the streets in France^a. The number of homeless people is expected to rise in the current economic context of growing inequalities with potentially dramatic effects on health 2 3 . Indeed, homeless people in most Western countries have limited access to appropriate care for their complex health care needs 4 . Adherence to treatment and continuity of care are often poor 5 . They also have a greater risk of engaging in health-damaging behaviors, including tobacco, alcohol and drug dependence in particular 6 7 . In addition, it has been estimated that more than one in four homeless people suffer from serious mental illnesses, such as schizophrenia or bipolar disorder 8 9 10 11 . Providing care to those homeless people with severe mental illness is particularly challenging 12 : they are more likely to remain homeless for longer periods of time and to require more health support (that is, they tend to be in poorer physical health and to have more substance abuse co-morbidity) and more social support (that is, they have less social support, they encounter more barriers to employment and they have more contact with the legal system) than homeless people who do not suffer from mental illness 13 14 .

Over the last several decades, two main approaches have been investigated regarding treatment for homeless people with mental illness in the United States and Canada 15 16 . The Treatment First model, also called Continuum of Care, requires that homeless people first address mental health issues and/or drug misuse, moving progressively up a ‘staircase of transition’ before finally accessing independent housing 17 . In contrast to this model, which predominates in many European countries, including France, the Housing First model reverses this sequence by offering immediate access to stable housing 18 . Individuals can exercise some degree of choice regarding the location and type of housing they receive. No pre-conditions, such as being stabilized on medications or bringing substance abuse under control, are necessary. In addition, homeless people in the Housing First model access support from a multidisciplinary team, following a well-defined intensive case management (ICM) program for people with moderate needs (that is, participant/staff ratio = 20:1; availability: five days/week, eight hours/day) or assertive community treatment (ACT) for those with greater needs (that is, participant/staff ratio = 10:1; availability: seven days/week, twenty-four hours/day).

Recent reviews of the literature suggest that in North American contexts, the Housing First model is particularly appropriate for homeless people with mental illness 15 16 . Compared to Treatment First or Treatment-As-Usual programs, Housing First programs report greater residential stability and fewer arrests, greater control over drug and alcohol use, better health outcomes and well-being, and lower residential and health costs 18 19 20 21 22 23 24 . However, is the Housing First model applicable in European contexts 25 ? To our knowledge, no randomized studies have explored the impact of the Housing First model in Europe. Given the important differences in health and social care provision 26 27 , such models need to be tested and evaluated in these highly generous care contexts, such as in France, before implementation on a larger scale. In line with international recommendations on poverty policies and programs 28 29 and a recent French national report on health and homelessness 30 , the French Ministry of Health recommends applying rigorous techniques, such as randomized evaluations, to develop and test the Housing First model with real-world problems faced by the homeless in France. We thus designed a prospective randomized trial to assess the impact of a Housing First intervention in comparison with Treatment-As-Usual (that is,, control group receiving existing supports and services in each site) on health outcomes and costs for homeless people with severe mental illness. This project is supported both by private (less than 10%) funding sources and public funding sources, issued from different policymakers in charge of medical, social, housing and research services.

This study is the first large-scale attempt to assess the applicability of the Housing First model on homeless people with serious mental illness in France. Several studies have provided evidence regarding the effectiveness of the Housing First model in North American contexts 15 16 21 23 54 61 . However, the extent to which the Housing First model could be replicated in France needs to be rigorously investigated 30 . The management of health care and housing for homeless people with mental disorders in France is significantly different from health and social care services in the North American contexts where the aforementioned studies were carried out. Generalization from the findings and expected benefits of these studies of North American contexts to the French context would be hazardous. Indeed, the health care policy of universal coverage in France, as opposed to nonuniversal coverage, may have a significant impact on the management of care for homeless people with mental disorders 26 . A recent study has suggested that access to care for homeless people may be higher in France, with its near-universal coverage and free mental health system 5 . In addition, mobile mental health outreach teams including doctors, nurses and social workers have been created since 2005 to ensure the provision of health, mental health and social care for ‘hard to reach’ homeless individuals with and without severe mental disorders 35 . Finally, a range of social aid and housing services can be easily accessed by all citizens in France who meet appropriate criteria, including direct financial support, disability allowances and housing assistance. Of particular note is the 2007 legislation introducing a theoretically enforceable right to housing for all citizens. Thus, services provided in the current study for the Treatment-As-Usual group are far more extensive than in previous studies in other national contexts. The present study is therefore of great importance to confirm whether these previous findings can be generalized to the French-specific context.

A major strength of the current project is the proposition of a rigorous evaluation of two policy options (‘Housing First’ versus ‘Treatment-As-Usual’) with regard to answering the specific needs of the homeless in France. Although randomized experiments have become increasingly popular in recent years, medico-social experiments in real-life contexts using a randomized project design as in the current study are still relatively rare 29 . The advantage of this approach is that the results are more representative of everyday practices and provide significant evidence regarding the feasibility of using the Housing First intervention as part of ongoing routine care management in the French context. This thus increases the generalizability of the results and supports implementation 62 . Another advantage of randomized experiments is the interplay between theory and experimental research resulting from the close collaboration between researchers and implementers/policymakers 63 . In the present study, special attention has been paid to this sort of collaboration, involving a large number of professionals working closely together, including policymakers at national and local levels; researchers from multiple domains, including psychiatrists, public health professionals, sociologists and economists working within an independent national scientific committee (Institut de Recherche en Santé Publique); and health and social care professionals (doctors, nurses, social workers and peer workers). Finally, this study has been granted a significant budget, drawn from public funds with large teams of professionals, enabling the implementation of this complex intervention. Ambitious experiments as in the present study are all the more likely to be useful for influencing national policy in these areas 64 .

Another strength and originality of the present study is the proposition of a qualitative approach in addition to quantitative methods to investigate the complex health and social issues surrounding the current Housing First program. Qualitative research is particularly well suited to understanding complex phenomena, uncovering links among concept and behaviors, and generating and refining theory applicable to specific contexts 65 . In addition, mixed methods approaches, including qualitative and quantitative approaches, provide unique contributions to outcomes research 65 66 .

This study has several limitations. Even with the large overall sample size in this multicenter study, the sample is not representative of the homeless population as a whole. The inclusion criteria used in the present study only concern a particular section of the homeless population: those with high needs and severe mental illness. Generalization from the findings of this study to other homeless people would be hazardous. However, should Housing First prove to produce positive results in the French context, it is arguable that this may well remain true for less vulnerable individuals. Further work is needed to confirm this hypothesis, especially in populations with moderate needs and without mental disorders.

A further challenge for the present study will be to retain participants at the time of follow-up. Previous longitudinal studies on homeless individuals have managed to retain only 60 to 85% of participants over follow-up periods of 18 to 36 months 14 . As previously explained, attempts to improve follow-up rates in the current study will include using food coupons, obtaining contact details of significant others to help with locating participants and explaining to participants the importance of conducting follow-up.

Finally, we cannot exclude a series of limitations, such as potential Hawthorne effects (that is, individuals from the Housing First group may be conscious of being observed or indeed benefit from being observed, which may induce them to change their behavior for the duration of the experiment, or indeed for longer) or John Henry effects (individuals from the Treatment-As-Usual group may feel offended to be a comparison group and react by altering their behavior). Contamination bias between individuals in the two groups is also a possibility, although less likely because the number of patients included in the present study is relatively small compared to the total number of individuals seeking housing at each site. The qualitative analyses used in the present study should help us to analyze these different phenomena and take them into account when interpreting results.

The study started recruiting participants in August 2011, and the recruitment is ongoing.

^aSource: http://www.emmaus-france.org.uk, 2007.

^bAbsolutely homeless is defined as having had no fixed place to stay for at least the past seven nights with little likelihood of finding a place in the coming month.

^cPrecariously housed is defined as being housed in single room occupancy, rooming house, or hotel/motel as a primary residence AND in the past year having experienced two or more episodes of being absolutely homeless OR one episode of being absolutely homeless that lasted for more than four weeks in the preceding 12 months.

The authors declare that they have no competing interests.

AT, CF, VG and PA conceived and designed the study, including the statistical elements and power calculation. The study protocol and manuscript have been drafted by AT, CF, LB and PA. The study protocol and manuscript were critically reviewed by VG, CL, BV, PR, TG, BF, TA and CL. All authors read and approved the final manuscript.