Housing First and the Risk of Failure: A Comment on Westermeyer and Lee (2013).

Over the last 5 years, community policies in response to homelessness have shifted toward offering permanent housing accompanied by treatment supports, without requiring treatment success as a precondition. The US Department of Veterans Affairs (VA) has embraced this "Housing First" approach. A 2013 report sounds a contrarian note. In a 16-person quasi-experimental study, 8 veterans who entered VA's permanent supportive housing did poorly, whereas 8 veterans who remained in more traditional treatment did well. In this commentary, we suggest that the report was problematic in the conceptualization of the matters it sought to address and in its science. Nonetheless, it highlights challenges that must not be ignored. From this report and other research, we now know that even more attention is required to support clinical recovery for Housing First clients. Successful implementation of Housing First requires guidance from agency leaders, and their support for clinical staff when individual clients fare poorly.

In the past year, there has been considerable publicity surrounding an impressive reduction in the number of veterans homeless each night, down to approximately 50,000 in 2014 (from 75,609 at a single point in time in 2009) (Phillips and Kesling, 2014; United States Department of Housing and Urban Development & United States Department of Veterans Affairs, 2009). This dramatic decline is due in part to the provision of housing to over 74,000 veterans over the last 4 years (Phillips and Kesling, 2014) through a collaboration between the Department of Housing and Urban Development (HUD) and the Department of Veterans Affairs (VA). In this program, known as HUD-VASH, HUD provides vouchers for private market rental apartments, whereas VA identifies and assists veterans in locating housing and in their recovery (Department of Veterans Affairs, 2011). Recently, VA has adopted an evidence-based approach to housing and recovery known as “Housing First,” which entails a) removing traditional preconditions to housing such as completing treatment or proving continuous sobriety, b) providing extensive support for recovery, and c) delivering recovery-oriented services according to the consumer’s choice (Austin et al., 2014a; Tsemberis, 2010; Tsemberis et al., 2004). Like many communities throughout the world, VA has made Housing First central to its efforts to end homelessness (Schoenhard, 2012; Hwang et al., 2013).

In 2013, the Journal of Nervous and Mental Disease published a study from Minneapolis that offers a skeptical view of both Housing First as a recovery approach and HUD-VASH as a program (Westermeyer and Lee, 2013). Westermeyer and Lee compare 1-year outcomes for the first 8 HUD-VASH clients in Minneapolis to 1-year outcomes of 8 matched clients who had been placed in recovery programs according to criteria of the American Society of Addiction Medicine (ASAM). When compared to these ASAM-placed clients, the HUD-VASH clients demonstrated very poor outcomes. Seven of 8 clients returned to heavy substance or alcohol use within 1 month of placement. At the 1-year mark, 8 were classified as “readdicted” (although this term was not defined), 7 lost housing, and 4 had new addiction-related medical diagnoses, compared to none in the group placed using ASAM criteria. The authors propose that the HUD-VASH program was based on the belief that clients were “beyond recovery” and that these outcomes data support “close monitoring and relevant contingencies” in the treatment of substance use disorders.

In this commentary we suggest that the Minneapolis report, as published, is problematic both in its conceptualization of the matters it sought to address and in its science. Despite those limitations, it highlights pressing challenges in the adoption of Housing First, challenges that are evident in the literature and in our team’s research.

The question of whether Housing First embodies an unwise confidence in the potential to house persons with active addictions is one that has been raised previously (Kertesz et al., 2009). However, analyses of housing outcomes have been mostly reassuring in regard to retention of clients with addiction at the time of housing placement, (Edens et al., 2011, 2014) even as analyses concerning other outcomes (health, addiction recovery, health costs) have produced mixed findings (Basu et al., 2012; Mares and Rosenheck, 2011; Tsai et al., 2014). Does the Minneapolis study convey a useful warning for communities or agencies that embrace Housing First? This commentary will suggest that useful lessons can be gleaned from the Minneapolis report, provided its significant scientific limitations are acknowledged at the outset.

LIMITATIONS

Conceptual

The introduction itself promises a comparison of “permanent HUD-VASH and Housing First programs” against “patients placed using ASAM criteria.” This purpose, as stated, conflates placement triage (ASAM placement criteria) (Mee-Lee, 2001), therapeutic principles (Housing First, abstinence contingencies, or other treatment modalities), and material arrangements for housing. This conflation produces confusion.

A method for placing clients (such as those advanced by ASAM) is not a method of treatment. In principle, superb triage can be undermined if available treatment options are poorly executed. We can speculate that the 8 non–Housing First clients in Minneapolis were directed to a typical combination of VA-contracted shelters and other programs. However, the published information is insufficient to explain why the outcomes described for these 8 clients placed according to ASAM criteria (who experienced no relapse, no work loss, no homelessness) are so radically different from results obtained in nearly all of the homeless addiction treatment literature to date, including programs that applied intensive drug testing and strict abstinence contingencies (Kertesz et al., 2009; Kertesz et al., 2007; Milby et al., 2008; Orwin et al., 2005; Schumacher et al., 2007).

More troubling questions concern the actual treatment availability for the HUD-VASH clients in this case series. The Housing First approach requires consumer-centered recovery services to engage clients and reduce substance use–related harms (Gilmer et al., 2013; Ridgway and Zipple, 1990). Although addiction treatment is not compulsory, regular and intensive engagement with clients to promote recovery is required. In short, Housing First should never be construed as “housing only,” but rather as a comprehensive recovery approach that is executed well when clients are constructively engaged and offered treatment (Gilmer et al., 2014; Nelson et al., 2014). Unfortunately, there are no data to suggest this intensive approach was available to the first 8 HUD-VASH clients in the Minneapolis report. In fact, it is strongly implied that all treatment terminated at the point that clients were housed. In essence, the observed poor outcomes early in this program’s execution could potentially represent shortcomings in fidelity to the expectations of Housing First, rather than a mistake in the choice to house.

Scientific Design

A central scientific issue with the results reported in Minneapolis study concern the potential for bias. The HUD-VASH clients in this study were the first 8 placed in apartments; they are in effect a pseudo-random sample. Were the 8 ASAM-placed comparators similarly sequential? The availability of complete 1-year follow-up data for all 8 comparators raises the question of whether these 8 represent “retained clients” rather than a pseudo-random sample of entrants to more traditional programs. In the literature, retention of homeless clients in treatment programs is low (in a review of 15 different interventions, loss of 2/3 or more was typical) (Orwin et al., 1999), and in the most rigorous scientific trials, addiction treatment outcomes are modest (Milby et al., 2009; Orwin et al., 2005; Sosin et al., 1995). The near-perfect results among the 8 Minneapolis comparators might be due to analysis focus on persons “retained,” rather than a random sample of program entrants.

Contextualization Within the Literature

The poor outcomes for the initial 8 HUD-VASH clients in the Minneapolis program are certainly troubling. Job loss (5/8), loss of partner (5/8), “readdiction” (8/8), and loss of housing by 7 of 8 HUD-VASH clients all stand out. Poor outcomes are perhaps more common than advocates would like, but most Housing First research suggests that they are not quite so prevalent as reported among these 8 clients. An analysis of 29,143 HUD-VASH clients (based on data from 2008 to 2011, technically before formal adoption of Housing First) found no effect of substance use disorder on housing success (Tsai et al., 2014) In fact, 65% to 69% met criterion of >60 days housed in the past 90 days at the 6-month mark (communication from Dr. Jack Tsai, September 6, 2014).

Similarly, the 11-site Collaborative Initiative on Chronic Homelessness (n = 756) reported that clients typically attained 80 days housed in the preceding 90, which should also be reassuring (Edens et al., 2011). Follow-on analysis reported that heavy users of stimulant drugs were somewhat less successful in attaining housing stability, but the difference was not particularly marked (Edens et al., 2014). Among supportive housing interventions in Chicago, 117 of 178 (65%) clients attained stable housing at 18 months (Sadowski et al., 2009). Given this literature, 2 prudent inferences are justified. First, results as poor as those from Minneapolis are a striking exception. Second, permanent supportive housing programs, including those that use a Housing First approach, cannot guarantee success for all who enter, though their successes are certainly more common than the Minneapolis report suggests.

POSITIVE LESSONS

The Minneapolis report sounds an important cautionary note meriting serious consideration, apart from any concerns regarding the scientific design of the study. This report should spur useful (and perhaps humble) reflections on how efforts to embrace Housing First can be improved, provided the right management and resources are brought to bear. The lessons below are derived in part from our own qualitative work (Austin et al., 2014b; Kertesz et al., 2014), but also from the literature (Tsemberis, 2010). In essence, the key ingredients for success include a number of elements that require action either by national leaders or local institutional leadership. Among these is the necessity for strong resource supports for clinical care in combination with leadership actions necessary to foster institutional change.

Assurance of Clinical Resources

Westermeyer and Lee’s assertion that the Housing First approach treats clients as “beyond recovery” is not the view of Housing First originators (Tsemberis et al., 2004). Treatment and therapeutic supports are required for Housing First to deliver a client-centered recovery. When Housing First is not implemented with an assertive community treatment team, the case manager is expected to broker and coordinate a range of services, from family/community integration to medical, mental health, and addiction treatment, education and vocational assistance, with all supports directed by each client’s strengths and interests (Tsemberis, 2010). But is that support sufficiently robust in VA at this time?

The Canadian experimental analysis of Housing First (“Chez Soi”) carried out across 5 cities targeted a client to case manager ratio of 20 to 1 (Hwang et al., 2012). This was reduced to 16 to 1 when the needs of the clients proved too intense for the original staffing ratio. Across the 8 VA HUD-VASH programs our research team has examined, case management ratios were typically at 25 to 1 or even 35 to 1. At some medical centers, staff had no regular workspace. Training in core competencies such as cognitive behavioral therapy, motivational interviewing, and harm reduction was variable. In short, we observed that in at least at some VA medical centers, staff were not yet fully prepared, trained, or sufficiently numerous to maintain the intensive clinical supports which define Housing First.

Sensitivity to Clinical Risks at the Time of Transition

Moving patients directly out of residential treatment programs into supportive housing represents a delicate transition; doing so without careful negotiations with current treatment providers is likely to be risky. In a study of HUD-VASH that predates the massive expansion of the last few years, O'Connell et al. reported that substance use >15 days per month was associated with modestly worse housing outcomes; on average, clients who came from residential treatment programs were at additional risk (O'Connell et al., 2013). Nonetheless, clients with active substance use appear to obtain housing results that are as good or nearly as good as those obtained among those without such disorders (Edens et al., 2011, 2014; Tsai et al., 2014). In the non-VA Collaborative Initiative on Chronic Homelessness, clients attained >70 days housed in the preceding 90, regardless of whether they had prior alcohol or drug use at time of placement (Edens et al., 2011). In a randomized trial from Vancouver, scattered-site Housing First interventions were associated with reductions in criminal behavior, compared to both treatment as usual and congregate living situations. Statistical analyses focused on averages mask the reality that for certain individuals, the departure from treatment is frankly adverse, and that for others, living alone in a scattered-site apartment is a burden that entails loneliness and risk (Schutt and Goldfinger, 2011). However, the available research data demonstrate clearly that these risks can be substantively remediated for most, provided communities invest the clinical resources to support client recovery and listen carefully to the clients and the people who care for them.

Leadership

A great deal of scholarship has focused on the challenges of leading a major organizational change in health care settings (Damschroder et al., 2009; Van Deusen Lukas et al., 2007). VA’s ambitious goal of ending veteran homelessness over 5 years, a major expansion of housing resources and staff, and the embrace of a new interventional philosophy (Housing First) represents a transformative effort, and organizational stress is nearly unavoidable. Those stresses often find expression among the clinical teams charged with providing direct care. If not anticipated, these stresses may prove toxic. The Minneapolis report certainly implies a measure of stress between the HUD-VASH and addiction treatment teams (“Five placements occurred without the knowledge of the multidisciplinary treatment team”). Conflict is a byproduct of change. Organizational leaders and mid-level management ideally should mitigate such tensions or put them to productive use, but no productive outcome is possible if leadership is not publicly engaged and directly asking for information, even information that could be embarrassing or challenging at first blush.

Elsewhere we have reported how VA medical centers differentiated themselves from less successful comparators in the degree to which leadership was able to engage and organize change (Kertesz et al., 2014). In our view, the ingredients for effective community leadership focused on ending homelessness are not unique to VA and should inform the work of mayors, county commissions, or nonprofit philanthropic agencies (e.g., Community Foundations or the United Way).

We have reported (Kertesz et al., 2014) that effective leaders encouraged top-to-bottom alignment of housing endeavors by taking specific steps. They included homeless program directors in senior leadership meetings. They designated key aides to maintain a continuous flow of communication to the entities charged with implementing Housing First. Effective leaders inquired about and personally resolved challenges related to workspace, hiring processes, vehicles, computers, and collaboration with municipal, federal, and philanthropic leaders. They joined staff on the front lines and in outreach activities, making clear that they wished to understand the unique challenges of finding and housing highly vulnerable clients. Similarly, effective leaders assembled cross-departmental workgroups (or cross-agency workgroups) that included both senior and mid-level managers; the effect of such mechanisms was to encourage communication and bypass the resistance of strictly “up the chain” reporting structures. By contrast, less effective leadership was evident when organizational planning focused purely on the number of clients housed, and when the groups charged with housing highly vulnerable clients did so more or less under the radar. In such situations, resource requirements—such as space, vehicles, or workgroup alignment—came to leadership’s attention later than would be ideal.

Explicit Support for Staff When Failures Come to Light

The Minneapolis report evokes what our interviewing team has come to term as the “nightmare scenario” in which clients are evicted from housing after behavioral deterioration, or even die. That fear is often mentioned by social workers, many of whom had already helped to house dozens or hundreds of clients successfully. As one told us, “we are just one newspaper headline away from disaster.” It’s a valid fear. It has already been noted, by VA authors, that the medical and mental disorders so strongly associated with homelessness don’t disappear after housing is offered (O'Toole et al., 2013).

A constructive implementation of Housing First has to seek generously funded clinical supports and embrace a client-centered philosophy. Effective implementation requires concrete leadership action, as we have noted. But there may be also be a separate and salutary effect of describing the “nightmare scenario,” as the Minneapolis report helps to do.

Serious, committed work requires frank acceptance of the risk of failure. Clinicians on the front lines bear an intimate burden, both because they care for their clients and because they are at risk of institutional blame when adverse outcomes find their way into the public eye. If nothing else, the Minneapolis report should remind us that another important task for institutional leaders will be to offer an embrace that is both sincere and public when failures, inevitable as they are, come to light.