People who are homeless are at higher risk for health problems, such as malnutrition, stress, communicable diseases, and violence [1]. The elevated health risks among people who are homeless contribute to significantly higher mortality rates, shorter life expectancy, and more frequent hospitalizations and acute care services utilization, compared with people in the general population [2-4]. In particular, depression is approximately three times more prevalent among adults who are homeless (20-25%), compared with 8.1% among the general adult population of the United States (US) [2, 5]. Moreover, over a third of homeless individuals experience alcohol and drug problems [6]. Some evidence suggests that social support is negatively associated with both depression and substance use problems among people who are homeless [7-12]. Meanwhile, permanent supportive housing (PSH), which combines a housing voucher with supportive services, has been recognized as an effective model for stabilizing the mental and physical needs of homeless adults [1, 13]. Hence, the number of PSH beds in the US has increased by 380% or 144,000 more since 2007 [14]. Despite this progress, a crucial knowledge gap remains: little empirical evidence exists to guide program decisions after homeless adults enter PSH. In particular, few studies have examined the longitudinal associations of social support with either depression or substance use problems among PSH residents with a history of homelessness.
Permanent Supportive Housing and Social Support
Though PSH residents are affected by many of the same issues that homelessness individuals experience, PSH may provide the opportunity to begin new and steady social relationships and to gain support through newly formed connections [15]. By providing stable housing and supportive case management services, PSH can be instrumental in breaking the negative reciprocal cycle that often exists between unstable housing, mental health, and substance use problems [16-19]. However, at the same time, PSH may also introduce new challenges to people who were once homeless [20, 21]. For example, some PSH residents experience social isolation after being housed in unfamiliar locations, while others feel stigmatized in environments without access to former peers [21]. However, very little longitudinal information on social support exists among the PSH residents. Furthermore, PSH residents can be quite heterogeneous in terms of their demographics, prior experiences and needs. Therefore, it remains an empirical question whether perceived social support increases in this population over time and whether changes in social support are related to health outcomes.
Social Support and Depression
In general, research conducted with PSH residents suggests that depression is negatively related to social support [22-24]. However, most of the available evidence is either cross-sectional [23, 24], or focuses on concurrent temporal associations without examining the individual variability in underlying growth curves [22]. A growth perspective over time, as well as its limits or individual differences in the growth, may provide more helpful information for the purpose of improving care services. Given the heterogeneity in the course of adaptation and outlook for PSH residents, it may be important to study how changes in social support over time are associated with depression trajectories, while simultaneously estimating individual trajectories to assess their boundaries of promotive effects of one behavior over the other. To our best knowledge, no prior study has examined this longitudinal association among PSH residents. Note that we use latent growth curve analysis interchangeably with trajectory analysis in the current study.
Social Support and Substance Use Problems
Similarly, there is a dearth of information on the longitudinal relationship between social support and substance use problems among PSH population. A recent randomized controlled trial of a “Housing First” PSH program on substance use problems among homeless individuals in Canada found an inconsistent effect of the housing intervention (vs. the treatment as usual group) on substance use problems over time [25]. Some aspects of substance use problems (e.g., interpersonal relationships) decreased over time, while other aspects (e.g., tolerance) did not decrease over the two years following their housing placement [25]. The scant research on this population suggests that there are important knowledge gaps that need to be addressed – that is, whether substance use problems tend to decrease over time and whether a concurrent increase in social support plays a protective role.
Findings from a few available studies suggest that substance use problems may continue even after housing has been provided [26, 27], and that more specific efforts to address stress and social support may be needed to help reduce substance use problems, especially if residents use substances to cope with stress [26]. For instance, it is possible that some PSH programs may inadvertently discourage residents from quitting alcohol and drugs if the program tolerates substance use, as discussed in a recent review [28]. The limited and inconsistent evidence on the natural course of substance use problems highlights a need for studies that can help provide evidence-based guidance for this at-risk population. Documenting the related trajectories of social support, depression, and substance use over time would be the first step.
Measurement Limitations of Prior Research
In addition to the scant longitudinal evidence, the existing research on PSH residents has often encountered measurement limitations. Some of the measures used in previous studies were adapted from scales originally developed for other purposes without validation for this specific population. When used among people with more serious mental health conditions, this can cause interpretation challenges because of potential ceiling or flooring effects. Thus, it is possible that some of the inconsistent findings across existing studies may be attributed to inappropriate or weak measures. For example, the measure of social support used by Durbin et al. [26] consisted of subscales from three different instruments designed for community residents with chronic mental illness, hospitalized patients with chronic psychiatric disorders, and the general US population [29-31]. A similar approach was employed in Kirst et al. [25], where they measured the substance use problems with items from one subscale of the Global Appraisal of Individual Needs Short Screener (GAIN-SS) [32], a screening tool designed for general populations. A recent psychometric study showed that this subscale might not be suited for people at low and high levels of severity [33].
The current study addresses these measurement weaknesses and provides a more rigorous test of the longitudinal relationships of social support, depressive symptoms, and substance use problems. Using a latent variable modeling approach, we simultaneously test measurement models for these three constructs over time and examine their parallel trajectories. Additionally, by simultaneously tackling measurement models and growth models in one analysis, we more efficiently use all available data and gain precision in estimation, while accounting for measurement errors, testing measurement invariance over time, and handling missing data.
The Current Study
The current study used data from a technology-assisted health coaching program called “Mobile Community Health Assistance for Tenants” (m.chat). This program provided in-person health coaching to PSH residents as part of the Regional Healthcare Partnership (RHP 10) Medicaid Waiver program in the state of Texas [34]. In addition to the usual housing and case management services offered by the PSH programs, m.chat provided in-person health coaching to encourage PSH residents to adopt healthy behaviors, such as improved diet, exercise, recreation activities, or reducing substance use. Coaching visits were typically conducted in public locations, such as recreation centers, fast food restaurants, or the project office. Participants completed approximately one coaching visit per month for up to 18 months, and each visit lasted 52 minutes on average.
We used a growth modeling approach to better capture the extent of change over time, as well as to model individual differences at baseline and over time. We first conducted a confirmatory factor analysis for the measures of social support, depression and substance use problems and then utilized unconditional latent growth curve models [35] to examine the trajectories of social support, depression, and substance use problems over 18 months. We then used a parallel growth modeling approach [36, 37] to examine how social support trajectories were associated with corresponding trajectories of depression and substance use problems in two separate, bivariate longitudinal analyses. We hypothesized that in the context of monthly health coaching, perceived social support would increase over time, whereas both depressive symptoms and substance use problems would decrease over time. We further hypothesized that individuals with lower levels of perceived social support at baseline would have higher levels of depression and substance use problems at baseline and show slower rates of growth in social support over time, which would, in turn, predict slower rates of decline in depressive symptoms and substance use problems.