Impact of Supported Housing on Social Relationships Among Homeless Veterans

The availability of social support has a positive effect on overall health, functional capacity, and subjective well-being (1,2). Research indicates that supported housing, defined as access to rent subsidies combined with intensive case management (ICM) services, is not only associated with positive housing and clinical outcomes but also improves levels of social support among homeless adults with psychiatric or addictive disorders or both (3–8). Given the benefits associated with increased levels of social support, particularly familial and peer support, a more in-depth analysis of the types of support affected by supported housing programs is warranted.

This study examined multiple dimensions of social network structure and function in a sample of homeless veterans who participated in an experimental study of supported housing through the national Housing and Urban Development–Veterans Affairs Supported Housing (HUD-VASH) program in the 1990s. A previous analysis of data from this study demonstrated a significantly greater increase in social support among HUD-VASH participants but did not examine the kinds of relationships or types of support that were enhanced (5). In this study, we sought to extend those findings by examining changes in specific types of relationships and support over 18 months after study enrollment.

Methods

In this study, 460 homeless veterans with a diagnosis of a psychiatric or substance use disorder or both from four Veterans Affairs (VA) facilities were randomly assigned to HUD-VASH (housing vouchers plus ICM), ICM without designated vouchers, or standard VA treatment as usual. Baseline data were collected from January 1992 through March 1997. The analytic time frame for this analysis was limited to the first 18 months after baseline assessment. Participants provided written informed consent, and the protocol was approved by the Human Investigations Committees at each VA medical center. Additional details about conditions and methods have been reported elsewhere (3,5).

After a baseline assessment, veterans who were randomly assigned to HUD-VASH met with a case manager who assisted them in obtaining a Section 8 voucher and locating an apartment and who provided ICM support without a specified time limit. Two control groups received either ICM without a Section 8 voucher (provided by the same case managers serving the individuals randomly assigned to HUD-VASH) or VA standard care. Demographic, clinical, and community adjustment variables were collected at baseline and every three months thereafter. Social support was assessed via three objective measures (social network size, frequency of social contact, and availability of different types of assistance) and two subjective measures based on subscales from the Quality of Life Interview addressing satisfaction with support from family and from nonkin peers (9). [A description of all variables assessed is available in an online supplement to this report.]

Descriptive statistics were computed for baseline demographic, clinical, community adjustment, and social support variables. Between-group differences at baseline were assessed by using one-way analyses of variance and chi-square analyses. Longitudinal outcomes were assessed with linear mixed models (LMMs), which evaluated main effects for treatment condition, time effects across all follow-up periods (six, 12, and 18 months), and the interaction of treatment × time and controlled for baseline values of the dependent variable and any differences identified in baseline group comparisons. These LMMs were conducted both with the baseline observation included (to assess differential change from baseline over time) and without the baseline observation (to examine overall main effects of conditions).

Results

The sample was primarily male (N=438, 95%) with a mean±SD age of 42.5±7.8 years. Most of the sample was black, non-Hispanic (N=287, 63%), and most were either divorced (N=194, 43%) or never married (N=155, 34%). The only significant between-group difference found at baseline was for gender: the ICM group had a smaller proportion of males than the group receiving standard care. Veterans reported a mean network size of 10.3±9.0 members at the time of program entry. At baseline, 82% (N=375) of the sample indicated that they had at least one close family member, 67% (N=309) reported having at least one close peer, and 39% (N=180) reported that they had at least one close provider. Contact with family, providers, and peers occurred, on average, less than once per month, monthly, and more than once per month, respectively.

At baseline, veterans identified 1.4±1.6 persons available to provide tangible support, 2.1±1.7 persons available for instrumental support, and 3.3±2.3 persons available for emotional support. The number of sources across all types of support included 2.8±3.0 family members, 1.7±1.8 peers, 1.3±1.3 providers, and 1.0±1.0 people in other relationships. On average, veterans were mixed in their assessment of satisfaction with family relationships (4.0±1.6) and dissatisfied to mixed in their satisfaction with nonkin social relationships (3.6±1.3) (possible scores range from 1 to 7, with higher scores indicating more satisfaction). [Tables in the online supplement summarize these and other data for the sample.]

LMM analyses, including a random intercept, of the average score on social support measures across all follow-up observation periods (excluding the baseline observation but controlling for baseline values of the dependent variable and time) showed that participants in the HUD-VASH group reported significantly higher scores on almost all the composite social relationship variables at follow-up compared with the other two groups (Table 1). The largest effect sizes were observed for the provider support variables (Cohen’s d=.5 and .6 compared with ICM only and standard care, respectively). HUD-VASH participants reported an average gain across the follow-up periods of 1.2±2.7 close providers (compared with gains of .3±2.3 providers for ICM only and –.04±2.5 providers for standard care). Compared with veterans in the two control groups, HUD-VASH participants also reported that they could rely on providers for more types of support and had more frequent contact with providers.

The second largest set of effect sizes favoring HUD-VASH was found on the total support index (Cohen’s d=.6 and .4 compared with ICM only and standard care, respectively), the emotional support index (Cohen’s d=.7 and .3 compared with ICM only and standard care, respectively), total network size (Cohen’s d=.6 and .4 compared with ICM only and standard care, respectively), and satisfaction with nonkin social relationships (Cohen’s d=.6 compared with ICM only). HUD-VASH participants gained an average of 3.5±8.9 network members over the course of the 18-month follow-up (mostly providers and other veterans), compared with an average gain of .6±7.7 members for ICM only and 1.3±8.6 members for standard care. Unexpectedly, the standard care group showed a greater increase than the ICM-only group in receipt of social support and in subjective quality of nonkin social relationships.

A second set of LMMs, which included the baseline observation as a response variable along with all follow-up observations, was conducted to assess the main effects of time and differential rates of change in social support between groups (group × time interactions). In these analyses, significant main effects for time were found for all social network variables, showing that on average, participants demonstrated significant gains from baseline on all structural and functional network characteristics. Only two significant group × time interactions were found, which showed that compared with ICM only and standard care clients, HUD-VASH clients had significantly greater gains over time on the number of close providers and frequency of contact with providers. [A table in the online supplement provides additional details.]

Discussion

This study examined the longitudinal impact of a supported housing intervention for homeless veterans on social network structure and support during an 18-month randomized clinical trial. Veterans who were offered ICM along with a housing voucher gained an average of 3.3 network members (primarily peers and providers) over the course of 18 months. Compared with veterans who received ICM without vouchers and veterans in standard care, those who received ICM and housing vouchers had more frequent contact with their network members; identified more sources of tangible, instrumental, and emotional support; and were more satisfied with their nonkin social relationships. These data suggest that access to housing itself may be a facilitator of positive changes in other areas of life—in this case, overall levels of social support (10,11). An examination of posthousing activities associated with HUD-VASH may shed more light on the ways in which supports are being strengthened through participation in the program (12).

These analyses also revealed that the increases in network size and contact were primarily attributable to enhancements in relationships with VA providers and veteran peers, with no differential effect on family relationships or nonveteran peers. Thus the social support impact appears to be concentrated on people who are likely to be directly or proximately involved in the HUD-VASH program itself. There may be a need for greater emphasis on helping individuals establish or reestablish natural, nonprogrammatic supports among networks of family and friends, relationships that have been associated with greater satisfaction with support networks (family) and several key processes of recovery from mental illness (2). Perhaps an active family intervention, which has not yet been emphasized as a feature of supported housing, could fruitfully improve these relationships. With evidence suggesting that social support, particularly natural supports, may facilitate successful housing (13–15) and provide a buffer against future stresses, this is an area worthy of further exploration.

An unexpected finding was that the group that received the ICM support without the housing voucher reported significantly less availability of emotional support and less satisfaction with nonkin relationships than the group that received standard VA treatment. Given that both the HUD-VASH and the ICM-only groups were offered similar models of clinical case management support and by the same case managers, this finding may reflect an unexamined and unintended impact of the randomization process. For some participants, working on housing issues in the absence of a rent subsidy while peers received the most effective pathway out of homelessness may have been more stressful and discouraging than continuing with “business as usual,” as with the standard care group.

Finally, this study highlights some important characteristics of homeless veterans’ social support on entry into the program. On average, veterans were able to identify about ten close individuals, the majority of whom were family members, followed by veterans and nonveteran friends. Eighty-two percent of our sample identified at least one close family member, and 67% identified at least one close peer as part of their network of support. Furthermore, nearly 100% of veterans in all three conditions reported having at least one source of instrumental support and one source of emotional support.

Additional research is needed to explore the longer-term effects of having strengthened social networks through housing. Do these relationships provide protective or buffering effects for challenging times in the future? Are they sustained over time, or are patterns of relationships likely to revert to prehousing levels? Are there particular supports that can help individuals and their supportive networks adapt to periods when challenges are likely to reemerge? These questions, among others, are germane to today’s supported housing programs and are currently unexplored.

Several limitations require comment. First, the experimental trial was conducted in the 1990s. Models of case management and supported housing, as well as network characteristics, may look different today. For example, given the current emphasis on peer-run programs and supports throughout the VA (12), a parallel social network analysis of the impact of today’s supported housing programs might reveal larger effects on peer and familial relationships than found in the analysis reported here. This question warrants further empirical exploration.

Second, many aspects of the case management services were not examined or were imprecisely measured in the original study, making it difficult to determine specific programmatic elements that may have contributed to the expanded social support networks among HUD-VASH participants. Furthermore, without a “housing only” comparison group, we could not determine whether the positive effects on social support observed for HUD-VASH participants were due to the housing itself or to the combination of housing plus support.

Finally, most veterans in the sample were male and were among those who had already been referred to a supported housing program. The social networks of this sample may have been larger and perhaps more diverse than those of homeless veterans who were not connected with such programming, female homeless veterans, and nonveteran homeless individuals.

Conclusions

Supported housing appears to play a positive role in facilitating social connections among homeless male veterans, especially in establishing relationships with other veterans and providers. Further study of the process of expanding social support and of interventions that may enhance family and peer support specifically is warranted.