Social Relationships of Dually Diagnosed Homeless Adults Following Enrollment in Housing First or Traditional Treatment Services

Abstract

Objective

Strong and effective social support is a critical element of mental health recovery, yet social support is often lacking for adults experiencing homelessness. This study examines differences in the social networks of participants newly enrolled in programs that use either a Housing First (HF) approach (i.e., provides immediate access to permanent housing with ongoing consumer-driven support services) or a treatment first (TF) approach (i.e., traditional clinician-driven staircse model that requires temporary or transitional housing and treatment placements before accessing permanent housing).

Method

We use a mixed-methods social network analysis approach to assess group differences of 75 individuals based on program type (HF or TF) and program retention.

Results

Quantitative results show that compared with TF, HF participants have a greater proportion of staff members in their network. TF participants are more likely than HF participants to maintain mixed-quality relationships (i.e., relationships with elements of support and conflict). As compared with participants who remain in a program, those who disengage from programs have a greater proportion of mixed relationships and relationships that grow distant. Qualitative analyses suggest that HF participants regard housing as providing a stable foundation from which to reconnect or restore broken relationships. However, HF participants are guarded about close relationships for fear of being exploited due to their newly acquired apartments. TF participants report that they are less inclined to develop new relationships with peers or staff members due to the time-limited nature of the TF programs.

Conclusions

These findings suggest that HF participants are not more socially isolated than those in traditional care. Implications for practice, policy and future research are discussed.

Housing First (HF) is an evidence-based practice for adults experiencing homelessness and serious mental illnesses. HF is effective at ending homelessness (Goering et al., 2014; Tsemberis, Gulcur, & Nakae, 2004) and is consistent with basic human rights (Byrne & Culhane, 2011). HF refers to the provision of immediate access to permanent housing with ongoing consumer-driven support services. HF stands in contrast to a traditional clinician-driven staircase model that requires temporary or transitional housing and treatment placements before accessing permanent housing—an approach also known as treatment first (TF). Given that strong, effective social support is a critical element of mental health recovery (Davidson, 2005; Perese & Wolf, 2005; Ridgway, 2001), social isolation is an important treatment consideration. Specifically, it is important to consider whether HF tenants become more socially isolated and disconnected from social networks as compared with those in TF, who are initially placed in a group or shared living arrangement (i.e., congregate placements) before they move onto permanent housing.

Research has shown that social support from family and friends can decrease the symptoms and debilitating impact of mental illness on functioning (Albert, Becker, McCrone, & Thornicroft, 1998; Davidson et al., 2001; Erickson, Beiser, & Iacono, 1998; Fisher, 1994; Jacobson & Greenley, 2001; Mead & Copeland, 2000; Ridgway, 2001; Young & Ensing, 1999), help reduce substance use (Alverson, Alverson, & Drake, 2000; Laudet, Magura, Vogel, & Knight, 2000; Mueser, Bellack, & Blanchard, 1992), and enhance activities of daily living for people with dual diagnoses (Bellack & Mueser, 1986; Clark, 1996, 2001; Clark & Drake, 1994). However, developing and maintaining supportive relationships is often difficult for people who are homeless and have dual diagnoses. In fact, these individuals have some of the most attenuated social networks (Albert et al., 1998; Blankertz & Cnaan, 1994; Drake, Brunette, & Mueser, 1998; Trumbetta, Mueser, Quimby, Bebout, & Teague, 1999) and many report having no friends (e.g., Blankertz & Cnaan, 1994; Luhrmann, 2008; Savage & Russell, 2005). Relationships with service providers are often the primary source of social engagement and feeling connected to others for those who are homeless and dually diagnosed (Buck & Alexander, 2006; Ware, Tugenberg, & Dickey, 2004).

The importance of a social network is related not only to the size of the network but also to the quality of the network relationships. Several first-person narratives of mental health recovery have indicated how only one or two carefully chosen individuals can make a difference by providing exceptional support and acceptance (Ridgway, 2001). On the other hand, large networks composed of intense relationships can create conflict, impose demands, and contribute to relapse (O’Farrell, Hooley, Fals-Stewart, & Cutter, 1998; Savage & Russell, 2005). Indeed, relationships are often mixed, such as when individuals engage with friends by using substances that contribute to impaired functioning (Alverson, Alverson, & Drake, 2001). Therefore, investigating differences in social relationships between those who enter HF versus TF requires consideration of personal network size, composition, and quality.

The research literature on HF’s use of scatter-site housing has shown that, as compared with those in TF programs, HF tenants experience a variety of better outcomes, including reporting greater satisfaction (Siegel et al., 2006) and perceived choice (Greenwood, Schaefer-McDaniel, Winkel, & Tsemberis, 2005; Nelson, Sylvestre, Aubry, George, & Trainor, 2007); are more engaged in treatment (Tsemberis, 1999), and experience longer periods of stable housing (Tsemberis & Eisenberg, 2000). However, scatter-site housing can also mean that individuals are housed in unfamiliar neighborhoods where they have no social ties, leading them to feel rootless and isolated (Yanos, 2007). Despite the assertive community treatment typically provided to HF tenants, such as intensive case management in which providers make frequent home visits, often HF tenants will still experience social distancing from other types of relationships. HF tenants often cope with unwelcome requests, nonreciprocal relationships, lack of fit in the neighborhood, fear of discrimination, victimization, and relapse triggers (Stefancic, 2014). For example, one study that addressed social integration in an HF program found that most HF tenants were successful in developing social connections once housed; however, for some tenants, their adjustment to housing and building social connections was more difficult because of their social isolation and loneliness (Yanos, Barrow, & Tsemberis, 2004). The few studies that have examined social integration among formerly homeless individuals with severe mental illness have generally demonstrated that these individuals make limited progress, even after receiving permanent housing and comprehensive support services (Henwood, Matejkowski, Stefancic, & Lukens, 2014; Tsai, Mares, & Rosenheck, 2012; Yanos, Stefancic, & Tsemberis, 2012).

In contrast to HF, a TF approach begins by placing adults in congregate settings that are governed by program rules regarding abstinence from substance use; participation in treatment; and restrictions regarding visitors, off-site passes, and curfews. Such congregate-living situations offer the potential for peer support and frequent interaction with program staff members who are often on-site 24 hours a day, but this setting can also limit social contacts outside of the program. Congregate-living situations have been described as a trade-off between autonomy and belonging to a community in the program (Whitley, Harris, & Drake, 2008), and can lead to disengagement from services either by choice or because of an infraction of program rules (Stanhope, Henwood, & Padgett, 2009).

In this study, we explored the important social relationships of 75 dually diagnosed homeless adults over a 12-month period after enrolling in either an HF or TF program. Based on the existing literature, we sought to address the following research questions:

• In what ways do the network size, composition, and quality of important social relationships differ (if at all) between those enrolled in an HF versus TF program?

• How do the social networks of people who disengage from programs differ (if at all) from those who remain enrolled in programs?

• As compared with TF participants, are HF participants more socially isolated or disconnected from important social relationship networks?

To answer these questions, this study used a sequential mixed-methods design as recommended and described by Morse (1991) as a QUAL → quant → qual design. In this study, participant interview responses from a large qualitative study were quantified to obtain counts of the various types of relationships consumers experienced after enrolling in either an HF or TF program. Quantitative group comparisons were contextualized by returning to the qualitative data to provide a nuanced understanding of group differences in social relationships.

Methods

Sampling and Data Collection

A sample of 75 individuals was drawn from new enrollees of four programs for dual-diagnosed homeless adults in New York City that offered treatment or housing services or both. Of this sample, 48 participants were recruited from three programs that followed a TF approach. These programs included two sites that provided temporary congregate shelter and services, usually for 6 to 12 months, until a less-structured placement (e.g., a single-room occupancy or supported apartment) could be secured by the client’s case manager. The third program site was a day program serving dually diagnosed homeless adults whose temporary housing placement was provided by other programs that also followed a TF approach. The remaining 27 participants were recruited from an HF program that immediately placed clients into independent, scattered-site supported apartments throughout the city and provided assertive community treatment services. All four recruitment sites served homeless adults with serious mental illness, and all shared the same low-threshold intake process, consisting of self-referrals and referrals from street outreach workers, shelters, jails, or hospitals. As reported elsewhere (Padgett, Stanhope, Henwood, & Stefancic, 2011), there was no significant difference in referral source between HF and TF participants.

Staff members at the programs invited all new clients meeting eligibility criteria to participate in the study, until a target sample of 80 persons was reached. To be eligible for study inclusion, clients had to have a DSM-IV Axis I diagnosis, history of substance abuse, and history of homelessness. After providing informed consent, enrollees participated in three in-depth interviews (baseline, 6 months, and 12 months). In total, 84 eligible people were invited to participate in the study, and 83 consented to participate. Of these potential participants, 75 completed the year-long study, but eight potential participants were lost to follow-up (i.e., 6 month interview because they had moved from the state, had a disconnected telephone, were institutionalized, or some combination of these factors. Interviews were conducted by four trained interviewers, all of whom had research experience with this population. The interviews, took place at the study’s offices or the participant’s residence. Interviews had an average length of 75 minutes, and participants received a $30 incentive for completing each interview. Study participation was not tied to program participation, and interviews were conducted even if the individual disengaged from the program. Basic demographic information for participants was obtained from program intake records. The sample description (see Table 1) includes details of participants’ mental disorder status, history of substance abuse, history of homelessness, family socioeconomic status, and education.

Table 1.

Sample Demographics

	Total (N = 75)		TF (n = 48)		HF (n = 27)		Engaged (n = 46)		Disengaged (n = 29)
	n	%	n	%	n	%	n	%	n	%
Gender (male)	51	68.00	34	70.83	17	62.96	26	56.51	25	86.21
Agea	41.28	10.18	39.60	9.81	44.07	10.42	42.71	9.91	39.00	10.38
Race/ ethnicity
Caucasian	12	16.00	6	12.50	6	22.22	9	19.57	3	10.34
African   American	39	52.00	30	62.50	9	33.33	20	43.48	19	65.52
Otherb	11	14.67	6	12.50	5	18.52	8	17.39	3	10.34
Hispanic	13	17.33	6	12.50	7	25.93	9	19.57	4	13.79
Education
< High school	40	53.33	25	52.08	15	55.56	22	47.83	18	62.07
High school   diploma	8	10.67	4	8.33	4	14.81	6	13.04	2	6.90
Equivalency   certificate	10	13.33	7	14.58	3	11.11	3	6.52	7	24.14
Some college	13	17.33	10	20.83	3	11.11	8	17.39	5	17.24
Associate’s   degree or   other	4	5.33	4	8.33	0	0.00	2	4.35	2	6.90
Family SES
Low-income	24	33.33	18	39.13	6	23.08	13	29.55	11	39.29
Working class	26	36.11	15	32.61	11	42.31	18	40.91	8	28.57
Middle- or   upper-class	17	23.61	9	19.57	8	30.77	10	22.73	7	25.00
Unknown	5	6.94	4	8.70	1	3.85	3	6.82	2	7.14
Primary diagnosis
Schizophrenia	20	28.57	12	27.91	8	29.63	13	30.95	7	25.00
Bipolar disorder	22	31.43	14	32.56	8	29.63	12	28.57	10	35.71
Schizoaffective	13	18.57	5	11.63	8	29.63	10	23.81	3	10.71
Major depression	14	20.00	12	27.91	2	7.41	6	14.29	8	28.57
Other	1	1.43	0	0.00	1	3.70	1	2.38	0	0.00

Note. HF = Housing First; HS = high school; SES = socioeconomic status; TF = treatment first.

^aAge statistics are means (M) and standard deviation (SD)

^bIncludes Asian, Pacific Islander, mixed, or unknown

All interviews followed a semi-structured interview guide that addressed recent life events, service experiences, current needs, substance abuse, mental health, and social contacts. Although the study guide was not intended to generate an exhaustive list of participants’ social networks at each time point, we believe that prolonged engagement with study participants during the course of a year provided an extensive representation of important social relationships. As part of the interview guide, participants were asked about their relationships, including who was important in their lives, on whom they could count and who counted on them, their family ties, and their relationships with other clients in their program. If participants used nonspecific references such as “people in my 12-step group,” interviewers would probe for specific individuals. All interviews were digitally audio recorded, transcribed verbatim, and entered into Atlas.ti software to facilitate qualitative analysis. All study protocols were approved by a university institutional review board.

Mixed-Method Data Analysis

Quantitative analysis

To examine study participants’ important social relationships in terms of network size, composition, and quality, each participant’s important relationships were logged in network case summaries for each of the three time points (i.e., baseline, 6 months, and 12 months). Important relationships were defined as those individuals mentioned when participants were asked to identify on whom they could count on or who counted on them, in addition to anyone else named as a source of regular, purposeful contact. Only identified individuals were included; that is, nonspecific references (e.g., “people in my 12-step group”) were not included unless the participant identified specific individuals in such groups. To ensure consistency, two team members independently reviewed the transcripts, and then compared their findings. A high level of agreement was reached in terms of identifying which of the participant contacts were important and what information about those contact should be included. The few disagreements were settled through case discussions.

Next, every important relationship was rated along two dimensions: overall quality of the relationship, and change in the relationship during the study duration (based on the participant’s perspective). Overall relationship quality was categorized as negative, mixed, or positive. Examples of negative relationships included relationships with drug users, estranged ex-partners, and family members who made excessive demands. Mixed relationships included those with elements of conflict and support. Examples of positive relationships included primarily supportive relationships with people who the participant consistently described as helpful and reliable, such as attentive service providers, siblings or partners who offer help, and biological offspring or nieces and nephews who provide strong motivation for progress. In addition, three categories were used to describe changes in relationships during the study period (1 year): grew distant, no change, or grew closer. These categories were based on the extent of contact the participant reported having with a specific network member, and the extent to which those contacts were described as meaningful. Two team members independently rated each relationship for more than half of the participants, and then compared findings. Because the two raters had almost no disagreement, the ratings for the remaining relationships were based on the assessment of one team member.

Next, using quantified relationship data, new variables were created that reflected the sum of the different types of identified relationships for each participant. Sum variables were created for type of relationship (friend, family, staff, or other), quality of relationship (good, bad, or mixed), relationship trajectory (grew distant, grew closer, or no change), type of relationship by quality (e.g., staff mixed relationship), and type of relationship by trajectory (e.g., family grew closer). All sum variables were divided by the number of relationships identified for each participant, creating standardized proportion variables for each category. We used t tests to assess group differences based on program type and whether each participant had remained in the program.

Qualitative analysis

Rather than seeking to corroborate the results that emerged from the quantitative group comparisons, we used a subset of 42 participants to contextualize our quantitative findings for the purpose of completeness and expansion of perspectives (Flick, 2004; Palinkas et al., 2011). To accomplish this step and to ensure sufficient and similar amounts of data across groups, we used data from 15 HF participants, 15 TF participants who were engaged in services, and 12 TF participants who had disengaged from services. These selected participant transcripts were coded through a process of selective open coding conducted by five members of the research team and included social relationship information. For the first stage of the coding process, each team member worked independently to read several transcripts, and then the five team members met to discuss emergent patterns (Patton, 2002). This iterative process of coding and consensus building continued for 2 months, with team members analyzing additional transcripts and attending weekly meetings. During these meetings, the team reviewed open-coded transcripts to enhance intercoder agreement while generating queries and discussion. Data for 16 study participants (48 transcripts) were initially coded and reviewed in this intensive manner, with the team reviewing each coded passage and its respective code. This process helped ensure that all relevant social interactions were accurately captured and identified. Two team members produced analytic memos during this process. After finalizing the codebook and establishing a high degree of intercoder agreement, transcripts for an additional 26 study participants (77 transcripts) were coded by one team member. Although future plans for this line of research include a broader thematic analysis of these data, this study used comparative methods to review the coded materials—specifically regarding social relationships—to better understand and illuminate differences in the quality or trajectory of the relationships of participants from each group (i.e., HF, TF-engaged, TF-disengaged).

Results

Table 1 includes the background characteristics of the sample (N = 75). The sample was primarily male (68 %) and African American (52%). Most participants had less than a high school education (53.33%). The mean age for the sample was 41.28 years (SD = 10.18). Family social economic status was evenly distributed across poor, working class, and middle- or upper- class. The most common primary diagnosis was bipolar disorder was (31.42%), followed by schizophrenia (28.57%), major depression disorder (20.00%), and schizoaffective disorder (18.57%). As reported elsewhere (Padgett et al., 2011; Stanhope et al., 2009), TF clients were more likely to be African American (OR = 3.33, p < .05) and less likely to have a primary diagnosis of schizoaffective disorder (OR = −0.28, p < .05) as compared with HF clients. Disengaged clients were more likely to be male (OR = 4.80, p < .01) than those who remained engaged.

The 75 study participants identified 732 important relationships. Table 2 describes the composition of personal networks during the year after enrolling in a HF program (n = 27) or a TF program (n = 48). HF participants had statistically significantly larger networks. On average, HF participants identified 12.67 (SE = 7.76) people in their network compared with 8.13 (SE = 3.58) for participants enrolled in TF. Differences in mean proportions of relationship variables across groups are reported in Table 2. HF participants had a greater proportion of program staff members in their networks. TF participants had a greater proportion of mixed relationships and family relationships that did not change during the study period.

Table 2.

Differences in Networks of Housing First (HF) and Treatment First(TF) Participants

	TF (n = 48)		HF (n = 27)
	M	SE	M	SE	t
Total important relationships identified	8.13	3.58	12.67	7.77	2.87**
Relationship type
Family	0.54	0.22	0.47	0.25	−1.09
Friend	0.24	0.19	0.21	0.16	−0.56
Staff	0.13	0.18	0.23	0.19	2.15*
Other	0.09	0.12	0.06	0.09	−1.29
Relationship quality
Good	0.67	0.19	0.72	0.20	1.09
Bad	0.10	0.13	0.12	0.15	0.48
Mixed	0.19	0.19	0.11	0.13	−2.15*
Relationship trajectory
Closer	0.45	0.25	0.46	0.27	0.17
No change	0.30	0.19	0.23	0.20	−1.57
Distant	0.18	0.16	0.15	0.17	−0.78
Relationship type and quality
Family good	0.36	0.24	0.31	0.20	−0.96
Family bad	0.05	0.08	0.06	0.09	0.49
Family mixed	0.10	0.13	0.07	0.12	−0.81
Friend good	0.16	0.16	0.17	0.15	0.23
Friend bad	0.05	0.08	0.06	0.09	0.49
Friend mixed	0.04	0.09	0.02	0.04	−1.63
Staff good	0.11	0.13	0.22	0.19	2.59
Staff bad	0.00	0.02	0.00	0.02	0.22
Staff mixed	0.02	0.08	0.01	0.02	−0.73
Relationship type and trajectory
Family closer	0.29	0.22	0.26	0.26	−0.55
Family no change	0.16	0.17	0.07	0.13	−2.54*
Family distant	0.09	0.16	0.11	0.19	0.47
Friend closer	0.08	0.11	0.69	0.97	0.51
Friend no change	0.08	0.11	0.05	0.07	−1.57
Friend distant	0.06	0.09	0.05	0.09	−0.71
Staff closer	0.04	0.10	0.02	0.05	−1.19
Staff no change	0.04	0.06	0.08	0.11	1.68
Staff distant	0.04	0.10	0.02	0.05	−1.19

Note. HF = Housing First; TF = treatment first.

^*p < .05.

^**p < .01

Table 3 displays differences in the network composition of individuals who left the program they were enrolled in (i.e., disengaged) compared with those who remained in the program (i.e., engaged). Of the participants who disengaged (n = 29), only three were from the HF program; 90% (n = 26) of participants who disengaged were enrolled in TF programs. Participants who disengaged from services had similar network sizes as those who stayed in the program. Those who disengaged had a greater proportion of mixed relationships and relationships that grew distant during the study period. Engaged participants had a smaller proportion of relationships with family members that did not change. Engaged participants had a greater proportion of relationships with staff members that did not change but a smaller proportion of staff relationships that grew distant compared with participants who disengaged. These findings formed the backdrop for the qualitative findings reported in the next section.

Table 3.

Differences in Networks of Disengaged and Engaged Participants

	Disengaged (n = 46)		Engaged (n = 29)
	M	SE	M	SE	t
Total relationships	10.67	6.42	8.31	4.52	1.73
Relationship type
Family	0.50	0.24	0.53	0.22	−0.43
Friend	0.23	0.17	0.23	0.20	0.09
Staff	0.18	0.15	0.15	0.24	0.78
Other	0.09	0.12	0.07	0.11	−0.92
Relationship quality
Good	0.72	0.19	0.64	0.19	1.60
Bad	0.12	0.15	0.09	0.11	1.13
Mixed	0.12	0.16	0.22	0.19	−2.48*
Relationship trajectory
Closer	0.49	0.28	0.39	0.22	1.71
No change	0.25	0.20	0.31	0.19	−1.29
Distant	0.12	0.13	0.23	0.19	−2.82**
Relationship type and quality
Family good	0.35	0.20	0.34	0.26	0.22
Family bad	0.06	0.09	0.03	0.07	1.41
Family mixed	0.07	0.13	0.12	0.11	−1.80
Friend good	0.17	0.15	0.15	0.17	0.50
Friend bad	0.06	0.09	0.03	0.07	1.41
Friend mixed	0.03	0.08	0.04	0.07	−0.26
Staff good	0.17	0.14	0.12	0.18	1.50
Staff bad	0.01	0.03	0.00	0.00	1.23
Staff mixed	0.00	0.02	0.03	0.10	−1.35
Relationship type and trajectory
Family closer	0.31	0.25	0.22	0.20	1.58
Family no change	0.08	0.11	0.20	0.20	−3.03*
Family distant	0.09	0.16	0.12	0.20	−0.72
Friend closer	0.09	0.13	0.08	0.10	0.62
Friend no change	0.07	0.09	0.07	0.11	0.11
Friend distant	0.05	0.09	0.07	0.10	−1.23
Staff closer	0.07	0.13	0.05	0.13	0.42
Staff no change	0.07	0.09	0.02	0.08	2.35*
Staff distant	0.01	0.03	0.07	0.13	−2.73**

^*p < .05.

^**p < .01.

Qualitative Between-Group Differences

Qualitative analysis was used to complement quantitative results by addressing the nature of group differences, including the presence of more relationships featuring a greater proportion of staff members and mixed-quality relationships among HF participants as compared with TF participants. In addition, as compared with participants who remained enrolled in programs, participants who left programs had a greater number of mixed relationships and more relationships that grew distant during the study period, particularly relationships with program staff members.

Housing First as the basis for relationships versus treatment first transient relationships

For HF participants, having a permanent home may have provided a foundation to reconnect or restore broken relationships. One woman who had lost contact with her family while homeless had been searching for her adult son for several years. After moving into an apartment, she had a chance encounter with her former sister-in-law in the neighborhood, whose assistance helped her reconnect with her son, as did the fact that the participant had a fixed address and telephone number. Participants also described relationships in the context of normal, everyday situations afforded by having their own residence. For example, one woman described how her partner helped her in her apartment.

We get along great. He’s a big help for me. He mops. I don’t even ask him to.... He does most of the cooking. I try to cook and he gets angry. He wants to cook because he likes to cook. He’s a big help to me. A couple of times I was sick, I had swollen ankles.... I suffer from edema sometimes, poor circulation. So he was a big help for me then.

Participants also mentioned the important role that HF staff members played in their lives, with one person noting,

They give you love, those people, they give me love. I can feel it. They really care about me.... Even when I was in the hospital, they used to call me to see how I’m doing.

However, despite having a greater number of important people in their network as compared with TF participants, HF participants did not necessarily grow closer to people over time. The comments of HF participants in interviews indicated these participants might have been guarding themselves against possible exploitation, particularly with respect to their newly acquired apartments. Along with the increased status and opportunities that came with having an apartment, these participants were also worried about jeopardizing a valuable resource. Participants said they sometimes struggled to keep network members, such as friends from the streets or newfound acquaintances, from moving into their apartments or otherwise taking advantage of their housing. The following quote is from a participant who chose to abruptly leave his apartment after he could not evict a guest.

I choose my friendships, but ... it’s not like I didn’t choose this friendship, but, it was like I was going out with his cousin and then he came in for a day and needed some money. Then I let him stay for the night. So then it moved into another day, and from there, and so on.

Thus, HF participants were often guarded about becoming closer with others, even positive network members, given that even the best of relationships or intentions could sour or lead to being taken advantage of, which in turn, could potentially lead to losing their apartment. Although providing a friend with a place to stay or entertaining guests at home was viewed as beneficial in many ways, participants also said they appreciated their newfound home and privacy, and that safeguarding their home was foremost in their minds.

Although individuals in traditional TF programs were able to develop new relationships, the time-limited nature of these programs led many to say they believed that these relationships, whether with peers or staff members, were temporary. One woman explained,

Respondent: It’s just to get through the time. A lot of times you make friends and you think when they get their places, you think you could go. But they don’t want no part of the [program] any more.

Interviewer: Will you want to be part of [the program] when you’re out?

Respondent: If I made friends, I wouldn’t put them out. I wouldn’t do that. But you know, you have to.... They are personal, you know, they don’t want to feel that part again, go through that again.

This respondent expressed the value she places on friendship while acknowledging that program participants share an understanding of the program context.

Despite having greater exposure to peers through congregate living, TF participants often reported hesitation to connect with program peers due to mistrust or drug use.

I think that a lot of these dudes really don’t have anything going on. And I don’t want to surround myself with a person who doesn’t want anything ... who don’t want to pursue anything. That kind of rubs off on me. And the next thing you know I become lazy, and don’t want to do anything, and just there for a meal and a [public transportation pass].

Regardless of whether TF participants remained enrolled or left the program during the course of the study, they said they approached connecting with peers cautiously.

Benefits of program enrollment versus mixed effects of program disengagement

For individuals remaining in either an HF or TF program, being in a program was indicative of two factors that facilitated sustained relationships: a commitment to improving personal circumstances and residential stability. For HF participants, remaining in the program was reflected by having obtained an apartment that served both as a marker of accomplishment and as a normative setting in which to engage in routine, daily activities with network members. For individuals in traditional programs, markers of progress included quitting drug use or improving physical health, and their residences afforded opportunities for social engagement with peers. In contrast, participants who left their programs did not have these same indicators of recovery progress and had few normative settings in which to develop and maintain positive relationships; therefore, these participants had a greater number of relationships that grew distant.

Being in a program often gave participants the support and respect of network members. HF participants said having an apartment was an immediate marker of success. “They [family members] look up to me [now] that I got my apartment. I’m free and I got opportunities.” Individuals enrolled in traditional programs also said network members recognized their accomplishments, although such recognition was usually not related to housing:

I never knew that I have the love that I have.... My daughter, my niece. People on the block when I go around and they see me and I’m lookin’ good. The parents of the kids that I grew up with ... they grown, they got kids like I got kids now. So when they see me doin’ good, lookin’ good, you can see the admiration in their face. They’re like “Ah, what’s up, man? How you feelin’?” That’s a good feelin’.

Participants who left their program did not have a similar marker to demonstrate recovery efforts and often left programs because they were not doing well, including drug relapse. Support often came in the form of mixed relationships, such as with acquaintances who also used substances. One man explained gravitating to familiar, but risky, surroundings:

One day I went to a neighborhood that I shouldn’t have been in, my old neighborhood. My family lives in the same neighborhood and I went to see them. And I ended up running into a girl.... So I ended up going back to the old neighborhood and I see the girls that I used to mess around with and it all kicks back in.

Most participants who decided to leave their program had been enrolled in TF, and left the program largely because they were unable or unwilling to follow program rules (e.g., sobriety, treatment adherence, curfew). One person who disengaged from services expressed growing weary of the monitoring and surveillance of the TF program, and voiced a desire to be more self-sufficient and competent.

I’m a grown man, first of all, and nobody’s gonna be there to hold my hand when I go into housing. And I actually don’t want nobody to be there to hold my hand. I want to be able to do it on my own.

Another TF participant expressed frustration with the staff:

I’m sick of being looked down upon like I’m being judged by them, “Well, you’re not here. You need to get there.” I’m like, OK, well, fine, in my due time.... I am where I am right now because that’s where I need to be.

For these participants, part of their explanation for leaving the program was to distance themselves from relationships with staff members.

Discussion

This study used a mixed-method design in an attempt to answer specific research questions while capturing the complexity and dynamism of social relationships among homeless adults with dual mental health diagnoses. During the course of the one year study period, HF participants identified a greater number of relationships than TF participants, suggesting that HF participants did not experience more social isolation than TF participants, which has been a general critique of supportive housing programs (Hopper, 2012). Nonetheless, HF participants did not necessarily develop closer relationships with others, even in supportive relationships that could help advance their recovery. One explanation for the lack of close relationships might be that many HF participants preferred the solitude that came with autonomy, accepting the trade-off of feeling lonely at times for a sense of safety and independence (Whitley et al., 2008). HF consumers might have avoided becoming closer with others to guard against potential risks to their stability and well-being, or to protect positive opportunities to move forward. Given this population’s common experience of exposure to trauma, breaches of trust, and strained relationships (Padgett, Hawkins, Abrams, & Davis, 2006), it is not surprising that many participants said they were wary even of positive relationships, remarking that few people could be counted as true friends or really trusted. Others said they chose to delay pursuing new or renewed relationships to avoid potentially negative influences, such as unwanted guests and drug-using acquaintances. This concept—known as positive withdrawal—has been described as an effective way of coping and creating a meaningful life in the community (Corin & Lauzon, 1992; Henwood et al., 2013).

This study’s qualitative findings suggest that just as privacy afforded by scatter-site HF does not necessarily lead to social isolation, congregate living does not necessarily provide participants with a sense of community. Similar findings were reported from a recent randomized controlled trial that included scatter-site housing, congregate housing, and a treatment-as-usual group (Patterson, Moniruzzaman, & Somers, 2014). As compared with people receiving treatment as usual (i.e., congregate housing), individuals living in scatter-site apartments scored higher only on emotional aspects of community integration. Moreover, although those in congregate housing more frequently reported knowing their neighbors, this familiarity did not translate to greater interaction or community integration (Patterson et al., 2014). These findings underscore the larger structural factors that contribute to isolation in this population, including poverty, lack of opportunities, and lack of employment (Hopper, 2012). In general, the average network size in both samples (12 for HF; 8 for TF) is similar to the network size of the general population. A study using two national surveys to examine friendship networks among Americans between the 25 and 74 years old found that in 2002 and 2007, participants had an average of 10 friends outside of their household with whom they interacted at least once a week (Wang & Wellman, 2010). It is important to note that the current study inferred network data from indirect qualitative questions. Notably, results might have differed if participants had been directly prompted to identify their network members.

Participants who disengaged from their programs (disproportionately among TF programs enrollees) generally led more turbulent lives that involved a greater number of mixed relationships. The qualitative findings suggest that involvement in mixed relationships was a product of a variety of factors: existing mixed relationships destabilized participants and led to disengagement; program enrollment created a buffer of positive support that evaporated once an individual disengaged; or disengaged participants engaged in mixed relationships with others as they returned to high-risk neighborhoods after disengaging.

The largest support base for all study groups consisted of relatives from the participants’ family-of-origin. This finding is not surprising but suggests that service providers can focus on these relationships because they are enduring, often complex, and can include previous abuse or trauma (Savage & Russell, 2005). It is also important to note that providers themselves are important network members. Although the relationships with staff were generally regarded as positive, the qualitative findings served as a reminder that just as a single relationship can be a source of comprehensive support and promote recovery, it can also contribute to participant disengagement from needed services. For example, the negative nature of provider relationships was the case when participants described providers who participants perceived as looking down on and judging them. The fact that HF participants had a greater proportion of relationships with program staff members than TF participants might reflect that the HF participants received team-based assertive community treatment. However, it is unclear if HF participants’ increased contact with staff members contributed to fewer relationships with others. Obscuring the nature of this difference is that frequency of contact was not specifically considered as part of the study, and TF participants were more likely to have more frequent interactions with staff members who were typically present on-site for 24 hours a day.

Because recovery is inherently a social process, providers should discuss with consumers the type of relationships they have and might want to develop. Providers and consumers could focus on engagement with individuals already in a consumer’s network as well as the quality of their consumer–provider relationship. Peer support has been acknowledged as an invaluable part of recovery, but many programs are still struggling to incorporate peer workers into existing services. Including peer support and existing network members poses additional challenges for providers, consumers, and organizations, and could require additional resources to face these challenges. In addition, it is also important to consider that some people might want to pursue a staged approach to social support, and therefore, might need more time than others to explore the types of social relationships they are interested in pursuing. Last, social networking through alternative means such as the Internet, including online groups, mailing lists, and dating services, has become an important part of larger society but is less available to many consumers because of socioeconomic factors contributing to a digital divide. Studies on the use of social networking technology among homeless youths have found that resources such as Internet access and smartphones are key factors in connecting youths to prosocial network members who are not street-based connections (Rice, Milburn, & Monro, 2011), and that maintaining these connections reduces risk behaviors and negative outcomes (Rice, 2010; Rice, Kurzban, & Ray, 2012; Rice, Monro, Barman-Adhikari, & Young, 2010). Therefore, providers should not only consider the ways in which new technologies could play a role in improving opportunities for social networking among homeless adults but also consider the resources required to access these technologies (Stefancic et al., 2013).

Strengths and Limitations

This study was designed to explore social relationships during the year following enrollment in two distinct types of programs for homeless adults with severe mental illness. Descriptions and interpretations are susceptible to well-known risks of interview data, such as the inaccuracy of self-report and memory. However, rather than an objective and accurate cataloguing of social networks, the focus of the current study was on individual perceptions of these relationships. This study quantified qualitative data to generate social network data, which is a different method as compared with more frequently used methods of social network data collection methods such as a name generator that directly ask participants to list their social network connections given certain parameters. It is possible that some connections may have been overlooked because participants were not directly asked to list important connections in their lives. However, because interviews were conducted at three time points with the same interviewer, the data most likely reflect a comprehensive understanding of participants’ personal networks. In addition, although quantifying qualitative data (as opposed to administering a uniform social network inventory) might yield inconsistencies, this method also has the potential to capture data about the nature and quality of relationships that a uniform inventory might not capture. This study did not include discrete counts of network members at each time point by group, and therefore, we could not assess how individual networks changed during the study period. Notable strategies for rigor (Padgett, 2009) included the study’s longitudinal design (which enhanced trust building and offered more than a snapshot of social engagement) and the use of multiple coders to confirm social contact information. It should also be noted that despite high levels of agreement when data were assessed by two separate researchers, the fact that approximately half of the interviews were rated by only one researcher for both quantitative and qualitative analyses is a possible limitation. Another potential limitation is that even though data were collected over time, a 12-month period is relatively short. It is possible that results could change over a longer study period and continued follow-ups. Another limitation is that participants were not randomly assigned to HF or TF groups; however, statistical tests revealed few statistically significant differences between these groups. Last, due to small sample size, only univariable statistical tests were performed, which is a limitation because analyses such as t-tests do not control for other important variables such as severity of mental illness, age, and gender. Future studies should focus on comprehensive multivariable statistical modeling.

Conclusion

In this study, we found that HF participants did not experience greater social isolation participants in traditional TF programs. Living alone does not necessarily lead to social isolation, just as congregate living does not necessarily provide participants with a sense of belonging to a community. A potential common denominator in both situations is the cumulative disadvantage and life course adversity experienced by this population that makes maintaining and developing relationships challenging (Padgett, Smith, Henwood, & Tiderington, 2012). Although both HF and TF programs can play an important role in supporting the recovery of program participants, these programs involve different considerations given each program’s unique approach.