Life Changes Among Homeless Persons With Mental Illness: A Longitudinal Study of Housing First and Usual Treatment

The sampling method for the larger RCT is described elsewhere (13). Institutional Research Ethics Board approval was obtained at all five sites. A total of 2,255 participants were recruited and randomly assigned to treatment conditions at each of the five sites. Recruitment and initial interviews began in October 2009 and ended in June 2011, and follow-up interviews ended in June 2013. A subsample was selected from the larger trial for narrative interviews (“narrative subsample”). For the first few interviews, one of every ten participants per treatment condition was selected to be interviewed. As subsample selection progressed, sites shifted to more purposeful selection to ensure that the subsample was representative of the larger sample.

At baseline, the narrative subsample (N=219), 10% of the total sample, was compared on more than 50 demographic, diagnostic, and outcome measures with participants who were not selected to participate in a narrative interview (N=2,036). In the subsample, the proportion of persons who identified as female or transgender (32%, N=219) was significantly greater than in the larger sample (28%, N=2,036) (χ²=7.47, df=2, p=.02). Also, a significantly smaller portion of participants in the subsample (32%, N=214) had three or more symptoms on a measure of substance use compared with the larger sample (36%, N=1,940) (χ²=9.80, df=2, p=.01). Finally, participants in the subsample had a significantly higher level of income in the past month (mean=$781.00±$839.70) than those in the larger sample (mean=$681.00±$660.30) (t=2.07, df=2,253, p=.02). Given that only three significant differences were found for more than 50 variables, the subsample appeared to be representative of the larger sample.

Overall, attrition rates were low: 90% of participants were retained at the 18-month follow-up interview (attrition rates across sites ranged from 0% to 20%). Reasons for attrition included inability to locate the participant and participant refusal, incarceration, or death. A total of 197 participants completed both the baseline and 18-month follow-up narrative interviews (Housing First, N=119; treatment as usual, N=78). No significant differences were found between those who completed the 18-month interview and those who did not on a number of demographic, diagnostic, and outcome variables at baseline.

Baseline qualitative interviews focused on life before enrollment in the study, and the 18-month narrative interviews focused on changes that participants had experienced since the baseline interview (17). A common protocol was used at each site for both the baseline and 18-month interviews. [The protocols are available in an online supplement to this article.] Open-ended questions were used to elicit data about 13 domains: life changes, typical day, education, work, general medical health, mental health, substance use, relationships, housing and living situation, finances and material situation, mental health services, other services, and hopes for the future (11). Participants provided informed consent for each interview and were reimbursed between $20 and $50 Canadian for each interview (rates varied by site). Forty-five interviews (23%) were conducted in French, and 152 interviews (77%) were conducted in English. All interviews were digitally recorded and transcribed.

Following methods used by Padgett and colleagues (11), each transcript was coded on 51 categories of change that were grouped within 13 larger change domains. Coders examined each participant’s baseline and 18-month follow-up interviews. On the basis of this analysis, each participant was given a code of positive, mixed-neutral, or negative change for each of the 13 domains. Each domain was weighted equally. An example of a positive housing change was: “This is the first time, you know, that I’ve had a home . . . that I actually feel, like I’ve had supportive housing before, but I didn’t feel like I was safe. And, this is the first place like I . . . feel like I love to go home. . . . I feel so safe. And . . . being safe is a major issue for me, you know?” An example of a negative change was: “They discharged me to a hotel. I left the next day. It was noisy, bug-infested, full of drugs.” An example of a mixed-neutral code was: “That’s what life is, cause it’s just like I said, like picking up, losing it all, picking up, losing it all, picking up, losing it all.”

If more domains were coded positive than negative or mixed-neutral, the participant was given an overall code of positive; if more domains were coded negative than positive or mixed-neutral, an overall code of negative was assigned; and if the number of domains with positive, negative, and mixed-neutral codes were mostly equal or if mixed-neutral predominated, then an overall code of mixed-neutral was given. A second coder, who was blind to treatment conditions, independently coded 20 of the transcripts in the same way (four per site). A high level of interrater reliability was obtained after one training session (κ=.77). Next, matrix displays were created to compare positive, mixed-neutral, and negative changes by Housing First and treatment as usual for each site and overall. A Mantel-Haenszel chi square test was used to test the association between treatment group and the change codes (positive, mixed-neutral, and negative) across sites (18). Finally, factors uncovered through thematic analysis that were related to changes were examined, again by using matrix displays. Qualitative themes constituted one dimension of the matrix, and type of change (positive, mixed-neutral, and negative) was the other dimension. Several methods were used to establish the trustworthiness of the qualitative data and analysis (19), including member checking, team coding, and triangulation (16).

Table 1 summarizes the characteristics of the subsample. As shown in Table 2, the percentage of participants in Housing First who described positive changes was more than double the percentage of participants in treatment as usual (Mantel-Haenszel χ²=28.5, df=1, p=.001). Conversely, participants in treatment as usual were four times more likely to report negative changes.

Several factors were important for positive changes across the Housing First and treatment as usual groups. Stable housing, together with the hopefulness it catalyzed for most participants, was an important factor. The acquisition of stable housing gave most participants hope and confidence and provided opportunities for them to take on or reclaim valued social roles. For example, one participant emphasized the value of providing housing in a “normal environment,” as opposed to being offered a “room” in a chaotic social environment. He went on to explain how that motivated him to participate in an environmental rally because he felt “like a member of the community.”

A second factor was positive social contacts. In Toronto, participants who had social support through positive relationships with friends and reconnection with family tended to report positive changes. The same was true of Aboriginal participants in Winnipeg who connected with their cultural traditions and communities. One participant said, “Yeah, I’ve made friendships in this, in the program, new friends.” Supportive social contacts were associated with reduced substance use. “I don’t sniff any more. I don’t hang around with . . . street people, I don’t hang around with the people I used to hang around with before that, that made me unhealthy cause I was getting myself really unhealthy.” Finally, new social roles were important for positive changes across sites. Many participants began pursuing activities such as volunteering, coaching sports, working, attending school, or becoming peer support workers. These endeavors gave participants opportunities to take on valued social roles that expressed a positive social identity.

A number of factors were related to negative changes. Precarious housing—losing housing, living in shelters, poor-quality or unstable housing, or negative experiences with housing—was related to negative changes. One participant commented, “It is a very old building, and it is dirty and my window is dirty, you know. And I don’t have energy to clean it, and on the floor I see a lot of insects, you know. And since I moved here, I see a lot and I’m scared of this.”

Negative social contacts and isolation were also associated with negative changes. Although negative social contacts affected both housed and unhoused participants, isolation was typically more common among housed participants. Isolated participants lacked the supportive social contacts that are important in helping to make changes. One participant stated, “Because of my loneliness I tend to bring in strangers thinking they will be my friend and be good to me but they’re not, they’re not my friends at all. They’re trying to use me or to hurt me somehow. I think . . . I am an easy target . . . maybe it’s my own fault. I don’t know, maybe it’s the choices I am making or my loneliness. Like I get so lonely I let people in.”

Increased or continued heavy substance use was associated with negative changes and was related to participants’ remaining involved in social groups that use substances. Finally, hopelessness was an important factor linked to negative changes. Hopelessness presented across sites as the pervasive belief that life would not improve for the individual. In some cases, participants remained hopeless despite being offered housing. For others, their prevailing sense of hopelessness was rekindled after losing their housing. According to a Vancouver participant who faced eviction, “I’m an addict. I screwed up. I was clean for eight months and then I relapsed. . . . Maybe I’m not good enough to have an apartment. I’m thinking that now.”

Those with mixed-neutral changes made uneven progress, with a split of roughly equal positive and negative changes or of little to no change. Similar to those who reported negative changes, participants who reported mixed experiences showed sustained substance use and setbacks resulting from relapse. Perceived failures and disappointments were a salient factor associated with mixed changes. As with participants who reported negative changes and hopelessness, participants with mixed-neutral changes often tried to make changes but had difficulty following through when faced with setbacks. A salient example of a mixed-neutral change was for a Vancouver participant who attempted to return to school and resume contact with his family. Both pursuits did not go well, leaving the participant feeling depressed and hopeless.