Services for homeless adults have undergone several historical paradigm shifts since the 1980s, when homelessness reemerged as a major national problem (1,2). In the current paradigm, permanent supported housing is the preferred intervention for homeless adults who have serious mental illness and, as is often the case, comorbid addiction problems (3–5).
Supported housing, broadly defined as subsidized housing with case management support, often uses an intensive, community-based model, such as assertive community treatment (ACT) (6–8). However, intensive case management involves frequent individual meetings and face-to-face contact (up to several times per week) between clients and clinicians, often in the client’s home or community. Such meetings can be time- and resource-intensive for staff and socially isolating (9,10), or even stigmatizing (11), for clients. For these reasons, ACT-like models may not be feasible for use as a population-based, public health approach to the problem of homelessness.
Intensive case management was initially developed for the most frequently hospitalized clients with the most severe mental illnesses (12–14). However, it may not be the most clinically effective or efficient or the most desirable recovery-oriented mode of treatment for many supported-housing clients, such as those with primary addiction problems or less disabling mental illnesses. The U.S. Department of Housing and Urban Development–Veterans Affairs Supportive Housing (HUD-VASH) program provides supported housing for a diagnostically diverse population of homeless veterans, many of whom are without a severe mental illness or a comorbid substance use disorder (15). A previous demonstration by the U.S. Department of Veterans Affairs (VA) showed the feasibility of assisting homeless veterans to search for housing through supportive groups rather than through intensive community-based case management (16). Several studies have shown that clients who receive intensive case management can be successfully transferred to less intensive mainstream services (17,18).
This study evaluated a group intensive peer-support (GIPS) model of case management that was developed at one site in the HUD-VASH program (19). The GIPS model may be timely, given that HUD-VASH is expected to serve over 60,000 clients in the next few years (20). Scaling up the supported-housing approach as a way to comprehensively address homelessness is likely to require increased client-to-staff ratios, making intensive case management models less practical. The GIPS model offers an alternative approach that may use case manager time more efficiently, builds on the strength of mutual peer support, and more directly promotes continuity of care, social integration, and client recovery.
This study used administrative evaluation data to examine the impact of implementing the new GIPS model within the HUD-VASH program at one demonstration site in New England. We compared the outcomes and processes of treatment and housing acquisition at the demonstration site before and after implementation of GIPS with outcomes from other HUD-VASH sites across the country during the same time frame. Adjustment was made for differences in client characteristics at the time of program entry. We hypothesized that implementation of GIPS would lead to increased social integration and case manager contacts and to no significant decline in housing acquisition times.
In the GIPS model, case management support is provided during weekly, hour-long group meetings led by case managers in which clients serve as active peers. The groups provide support and facilitate peer exchanges and mutual learning as clients undertake the steps required to obtain a housing voucher, find an apartment, transition to living in the community, and develop fuller social lives. Groups are based on rolling enrollment, so clients at different stages of housing acquisition and recovery can advise, inform, and provide emotional support to each other. In instances of staff turnover, the groups provide a foundation for continuity of care.
Details of the GIPS model have been described elsewhere (19), but the basic rationale is that psychoeducation, assistance with practical needs, peer support, and participation in community activities may be best provided in a group format rather than through individual, community-based intensive case management. Clients are assigned individual case managers, and intensive individual case management is provided as needed. However, case manager–led group meetings of active peers are the default mode of case management support.
The potential effects of GIPS are theorized to be related to peer support and active client participation. Treatment delivered in groups has been found to reduce social isolation, provide positive modeling behaviors, encourage candid feedback, and provide emotional support (21–23). Moreover, by requiring clients to attend group meetings rather than delivering case management services to them, the program encourages clients to become more active in their recovery and provides greater opportunities for social integration. Such opportunities allow clients to succeed in domains of life other than symptom reduction (24). For practical reasons, GIPS may result in more case manager–client interactions because case managers are able to see groups of clients more often than each client individually.
The GIPS model was developed through interaction between researchers at the VA New England Mental Illness Research, Education, and Clinical Center (MIRECC), HUD-VASH staff at the VA Connecticut Healthcare System, and colleagues at the National Center on Homelessness Among Veterans. Ongoing oversight and support for program implementation were provided through weekly joint meetings between MIRECC and HUD-VASH staff. The study was approved by the institutional review boards at the VA Connecticut Healthcare System and at Yale University.
+
Source of data and study design
Data for clients were obtained from VA administrative data collected by HUD-VASH case managers at all sites nationally from April 2009 to April 2011, which included the year before (April 2009–March 2010) and the year after (April 2010–March 2011) implementation of GIPS. Formal admission forms were completed by referring clinicians, including VA homeless outreach staff. HUD-VASH case managers completed structured progress reports on the process of housing acquisition for each client every three months after program entry.
GIPS was implemented at one demonstration site, the Errera Community Care Center of the VA Connecticut Healthcare System, in April 2010. The research design was a pre-post nonequivalent cohort study that compared outcomes, service delivery, and timing of housing acquisition for clients of the demonstration site before (N=102) and after (N=167) GIPS was implemented with those of clients of other sites across the country before (N=9,659) and after (N=21,318) implementation of GIPS. There was no overlap of clients during the pre-GIPS and post-GIPS periods.
+
Background characteristics and mental health diagnoses.
Information on age, gender, race, marital status, military service war era, lifetime episodes of homelessness, and lifetime years incarcerated was based on client self-report and confirmed by case managers.
Clients were asked to indicate the number of nights in the past 30 days spent in the following locations: the client’s own place (apartment, room, or house), someone else's place (apartment, room, or house of a family member or friend), temporary housing (transitional housing, transient hotel, or boarding home), residential treatment (VA or non-VA residential treatment program), institution (hospital, nursing home, prison, or jail), and homeless (shelter, outdoors, or in vehicles).
Clients were asked about the number of days they worked in the past month and about the amount of income they received in the past month from employment (including compensated work therapy and supported employment), disability compensation or public assistance (including service-connected VA disability and federal welfare payments), and all other sources.
Mental health status was assessed with both self-report and clinician-rated measures. Self-report measures included a mental health symptoms score, which is the sum of scores on eight items from the psychiatric subscale of the Addiction Severity Index (25), and a social quality of life score, which is the mean of scores on three items from the Heinrichs Carpenter Quality of Life Scale (26). Clinician-rated measures included an observed-psychosis rating scale (27), the Global Assessment of Functioning (GAF) (28), and the clinician Alcohol Use Scale and Drug Use Scale (29).
+
Case manager contacts and services.
Every three months, case managers reported the number of activities among a list of 29 items that they had performed for clients during the prior three-month period. The items include six in the area of general support, eight treatment items, eight housing items, and seven items related to employment and income. The items chosen were summed for a total score of case manager activities.
Case managers rated their working alliance with clients on five items containing statements such as “we are working on mutually agreed upon goals” on a 7-point Likert scale, with 0 indicating never and 6 indicating always (30).
Case managers reported the frequency of face-to-face meetings with clients on a 5-point Likert scale, with 0 indicating no contact and 4 indicating contact at least twice a week. They also rated the time spent with clients per week, with 0 indicating no face-to-face visits and 5 indicating more than 120 minutes. These data were used to calculate a product score representing the amount of total time spent by case managers in face-to-face contacts. Case managers also reported the frequency of contact with clients’ families or caregivers.
Case managers also reported the proportion of face-to-face contacts with clients spent in the client’s apartment and elsewhere in the community on a 5-point scale, with 0 indicating none and 5 indicating 81% or more.
Clients must complete a series of steps to obtain HUD-VASH housing (19). Once clients are referred to the program, they work with the HUD-VASH team to obtain a housing voucher, a process that includes documenting income requirements, attending an interview with the HUD office, and completing necessary paperwork. After being issued a voucher, clients conduct a housing search to find an apartment that will accept vouchers and meets their living preferences. Once a lease is signed, clients set up a payment plan for rent and make arrangements to move in. Case managers recorded the number of days needed to complete each stage of the housing acquisition process (admission into the HUD-VASH program, receipt of a housing voucher, signing of an apartment lease, and moving in) and the total number of days between admission and moving in.
One-way analyses of variance (ANOVAs) and chi square tests compared differences in baseline characteristics of HUD-VASH enrollees at the demonstration site and at the other sites during the years before and after April 2010. Post hoc group comparisons were made with Tukey's honestly significant difference (HSD) test and pairwise chi square tests.
Mixed linear models were used to examine the association of site (demonstration site versus other sites), cohort (before and after April 2010), and the interaction of site and cohort on clients’ outcomes, case manager contacts, and service delivery over six months. The site × cohort interaction effect was the principal independent variable of interest. Adjustment was made for sociodemographic characteristics and mental health diagnoses that were significantly different across groups at baseline by including them as covariates. Baseline values of dependent variables were also included as covariates. An autoregressive covariance structure was specified, and least-squares means were calculated for site and cohort.
Time elapsed between steps in the process of housing acquisition by group was studied by two-way ANOVAs that examined site × cohort interaction effects. Significance for all omnibus statistical tests was set at p=.01.
Table 1 shows the baseline characteristics of clients of the demonstration site and at all other sites before and after GIPS implementation. Differences were found in age, race, marital status, and war era by treatment site. In addition, clients of the demonstration site after GIPS implementation had more homeless episodes in the past three years than clients of the other sites before and after GIPS implementation. Clients of other sites were more likely than clients of the demonstration site to have a history of incarceration. The proportions of clients with clinical diagnoses of alcohol abuse or dependence, mood disorder, personality disorder, and posttraumatic stress disorder were higher at the demonstration site than at other sites.
At baseline, clients of the demonstration site had spent more nights in residential treatment in the past month than clients of other sites. Clients of the demonstration site after implementation of GIPS also worked the fewest days in the past month and earned the least employment income compared with clients of other sites before and after implementation of GIPS. Clients of the demonstration site had higher observed-psychosis ratings and lower GAF scores than clients of other sites before and after implementation of GIPS. At baseline, social quality of life scores of clients of the demonstration site before GIPS implementation were lower than the scores of clients of other sites before and after implementation of GIPS.
A significant time × cohort interaction effect revealed that over the course of six months, clients enrolled in GIPS showed a greater increase in social quality-of-life scores than clients of other sites (Table 2). There were no other significant differences between sites or any significant site × cohort interactions related to housing, employment and income, or clinical outcomes, suggesting that GIPS implementation had no adverse effects on these variables.
Case managers at the demonstration site reported a higher proportion of face-to-face contacts with clients in the community, and case managers at the other sites reported a higher proportion of face-to-face contacts in clients’ apartments. However, there were no significant site × cohort interactions for location of face-to-face contact (Table 3). After GIPS implementation, general support activities, housing assistance activities, and sum of case manager activities were significantly greater at the demonstration site than at other sites, and the site × cohort interaction effect was significant.
To examine suppression effects of covariates, analyses were repeated without covariates except baseline values of dependent variables. There was no change in the main results except for a greater reduction in nights in residential treatment at the demonstration site than at other sites after GIPS implementation, and the site × cohort interaction was significant (F=7.66, df=1 and 13,620, p=.006). As found previously, there was greater increase in social integration (F=15.60, df=1 and 15,142, p<.001) and case manager activities (F=11.57, df=1 and 15,101, p<.001) at the demonstration site after GIPS implementation, and significant site × cohort interaction effects were sustained.
Finally, examination of housing acquisition showed that after GIPS implementation, clients of the demonstration site took significantly more time than clients of other sites to be admitted to the program after referral. However, they obtained a housing voucher significantly faster than clients of other sites (Table 4).
This study is the first comparative evaluation of the GIPS model within the VA’s HUD-VASH program or, to our knowledge, of any group-oriented case management model of supported-housing assistance for people with mental illness who are homeless. The results suggest that GIPS may be a viable alternative form of case management for supported housing with the advantage of improving social integration without resulting in any loss of effectiveness on other measures.
After implementation of GIPS, clients of the demonstration site reported a greater increase in social quality of life than clients of other sites during the same time, and case managers reported providing a greater increase in clinical services than case managers at other sites. These findings partially support the hypothesis that GIPS implementation would be associated with greater client social integration. Although case managers did not report spending more time face to face with clients, they did report conducting more activities, especially to provide general support and housing assistance.
Although clients of the demonstration site appeared to have more extensive histories of homelessness and psychopathology than clients of other sites, after adjusting for these differences, there were no differences in the clients’ housing, clinical, or psychosocial outcomes after GIPS implementation. Initiation of the potentially less individually intensive GIPS program at the demonstration site resulted in no worsening of outcomes compared with other sites and was associated with faster acquisition of vouchers after program admission. Although the time from referral to program admission was greater at the demonstration site after GIPS implementation compared with other sites, HUD-VASH staff did not have a controlling role in the clinical process between referral and admission. As a result, this difference was not likely a result of GIPS implementation and may have been due to procedural difficulties in making referrals experienced by some VA clinicians at the demonstration site.
The GIPS model may be particularly useful in supported-housing programs that emphasize social integration, that serve an increasing number of homeless clients with less severe mental illness and more substance abuse problems, and that face growing client-staff ratios because of staff turnover, delays in hiring, or other reasons. Because GIPS is provided mostly in groups and relies heavily on peer support, it may offer better continuity of care in the face of staff turnover. The popular Housing First model has been shown to be effective for homeless adults with severe mental illness (31), but alternative supported-housing models, such as GIPS, should be considered for other homeless populations.
In addition, any effort to comprehensively address the problem of homelessness will require large-scale implementation of a public health approach and cannot practically accommodate individual-based community case management. Furthermore, the current paradigm of mental health services is focused on empowering clients and being recovery oriented (32). Supported-housing programs that strive to move in this direction may benefit from GIPS or a similar model that incorporates peer support, solicits participation from clients, and promotes client choice, learning, and independence.
This study had several limitations that require comment. Because there was no random assignment to treatment conditions, we cannot conclude that our results reflect the causal effects of GIPS. Also, even though regression was used to address significant differences between groups and cohorts, we could not control for unmeasured effects, such as differences in staff turnover or delays in hiring. Although we found improvements in social integration, we do not know whether case manager ratings of activities reflect the program’s emphasis on group meetings. Case managers reported increased social activities by clients, but there was no information about whether the activities occurred within the program, outside the program, or both. Furthermore, VA administrative data rely on the diligence and accuracy of VA clinicians in reporting on their work and client outcomes. Although large samples were used for the bulk of the analyses, the use of smaller samples in the analyses of process times may have undermined their power. In addition, we did not have data to evaluate the extent or impact of incomplete reporting. Strengths of the study included the use of multiple control sites and analysis of outcomes along several dimensions.
These data suggest that the GIPS approach is a viable form of case management for supported housing and may have an advantage in fostering social integration and recovery. It represents a recovery-oriented approach that can be scaled up to address homelessness.
This work was funded, in part, by a Career Development Award that was provided to Dr.Tsai by the U.S. Department of Veterans Affairs (VA) New England Healthcare System and supported by the VA and the Veterans Health Administration Office of Research and Development. The authors thank Laurie Harkness, Ph.D., for the support of the Errera Community Care Center and the U.S. Department of Housing and Urban Development–Veterans Affairs Supportive Housing (HUD-VASH) team at the VA Connecticut Healthcare System, led by John Sullivan, L.C.S.W., and all the case managers involved. Vincent Kane, L.C.S.W., director of the National Center on Homelessness Among Veterans and of the HUD-VASH program, provided invaluable support. Elina Stefanovics, Ph.D., and Chris Cryan, B.A., provided data management. The views presented here are those of the authors and do not represent the position of any federal agency or of the United States government.
The authors report no competing interests.