The effect of various supportive housing models on ART adherence among persons living with HIV in supportive housing

Abstract

Background:

Providing permanent supportive housing to chronically homeless persons living with HIV (PLH) contributes to improved HIV outcomes, including adherence to antiretroviral therapy (ART). This study seeks to understand whether certain components of housing, namely intensity of case management and specialized HIV housing programs, affects ART adherence for persons living with HIV in supportive housing.

Methods:

From 2015-2019 we conducted quantitative assessments with 157 persons living with HIV in supportive housing at baseline, 6-, 12-, and 18-months post baseline to identify factors associated with ART adherence. General Estimating Equations for repeated measures were performed to assess bivariate and multivariate measures.

Results:

Two-thirds of PLH in supportive housing reported 95% or greater adherence to ART. Multivariate analyses indicate that neither intensity of case management services nor specialized housing for PLH were associated with greater ART adherence. Greater time since diagnosis was positively associated with ART adherence. Greater depressive symptoms and African American race were negatively associated with ART adherence.

Conclusions:

Study findings reveal that although prior research has established the importance of receipt of housing for homeless PLH, the type or intensity of case management services associated with that housing may not be as important as simply being housed. Our results highlight the importance of considering mental health and more recent HIV diagnosis when developing treatment and case management plans to enhance residents’ ART adherence.

The United States (US) Department of Housing and Urban Development (HUD) estimates there are over 550,000 homeless individuals in the US, 10,000 of whom are living with HIV/AIDS.¹ Approximately one in twelve people living with HIV (PLH) in the US have an unmet need for housing assistance² and at least half of all PLH experience homelessness or housing instability at some point in their lives.³ Homeless or unstably housed PLH are at risk for numerous poor health and social outcomes. Homeless PLH are more likely to receive a late-stage diagnosis and have lower CD4 counts and higher viral loads at diagnosis.⁴ Additionally, they are less likely to be prescribed and adhere to antiretroviral therapy (ART)^5,6 or achieve viral suppression,^7,8 which is essential to prolonging survival and reducing risk of transmitting HIV to others.^9-11

Permanent supportive housing is an evidence-based, structural-level intervention and has been shown to produce significant health improvements among formerly homeless PLH.¹² Supportive housing combines permanent, federally subsidized, independent housing with social support and case management services¹³ to increase independence and self-sufficiency among formerly homeless individuals.¹⁴ Supportive housing residents pay up to 30% of their income toward the rent and reside independently in either market-rate housing in the community (scattered-site) or in apartment buildings owned and operated by supportive housing agencies (fixed-site). Eligibility criteria for programs vary, but chronically homeless individuals, or those individuals with long histories of homelessness and disabling physical and mental health conditions, including HIV are often prioritized.¹⁵ Most supportive housing programs take a Housing First approach, wherein programs minimize potential barriers to housing (e.g. credit histories, criminal background, substance use), providing a rapid pathway to housing.¹⁶ The supportive services offered in conjunction with housing are a critical component of supportive housing; for many chronically homeless individuals, housing in the absence of supportive services may not be sustainable.^17-19 Engagement in case management and other supportive services are often understood to be essential in improving housing stability and contributing to success in supportive housing.²⁰

Yet, supportive housing encompasses numerous types and approaches, and differences in housing programs may influence the extent to which residents achieve optimal HIV outcomes. For example, supportive housing programs vary in the intensity with which they provide services, including ART adherence. While commonly, case managers have a caseload of 15-25 and provide services weekly to monthly depending on clients’ needs, other programs provide low-intensity services in which case managers have caseloads of up to 60 residents and see their clients quarterly for home visits.²¹ There is also a continuum of services provided by programs. Some housing programs offer employment and legal services, operate health clinics, and employ behavioral health specialists, while in other programs, case managers refer out and help clients to navigate access to community services. The effects of the type and intensity of case management services on housing stability and improved health outcomes remains unclear.¹² Additionally, some housing agencies specialize in housing and social services for persons living with HIV through federal funding streams like Housing Opportunities for Persons with AIDS (HOPWA). Other PLH, however, are housed through programs for a broad range of people with chronic illnesses, including but not limited to HIV. ^20,21 It is unclear how specialization in housing for PLH influences HIV outcomes.

Despite the strengths of the existing literature, most prior research has studied ART adherence by comparing formerly homeless supportive housing residents to homeless individuals, examining the effect of receipt of housing on HIV outcomes.^22-24 The current study examined factors associated with ART adherence, including intensity of case management services and HIV-specific housing programs, among PLH already housed in supportive housing programs. By identifying the types or components of supportive housing that facilitate ART adherence, supportive housing programs may be better able to help PLH achieve optimal health outcomes.

Methods

This research was conducted from 2015-2019 in Chicago, IL by a community-academic partnership between [Blinded for review]. We used a stratified sampling strategy to recruit 888 residents from two configurations of supportive housing and three supportive service models, identified during the formative phase of this study.²¹ Housing configurations included fixed-site (residents residing in a single building) and scattered-site (residents lived in fair-market housing within the community). Supportive service models included low-intensity case management, intensive case management, and behavioral health models. Stratified sampling methods were used to recruit a sufficient number of participants in each category; target numbers varied depending on number of potential participants in each housing category to ensure no housing program or agency was overrepresented and to allow for meaningful comparisons.

Recruitment occurred in partnership with supportive housing agencies across Chicago. Case managers were given recruitment materials to distribute to housing residents and posters to hang for common living or service areas. Interested participants called a toll-free number where they were provided with more information about the study and screened for eligibility. Participants were recruited by word of mouth from other study participants.

Eligible individuals were 18 years or older, living in a participating supportive housing program, and did not have a cognitive impairment that would preclude informed consent. Eligible participants were scheduled for an appointment at supportive housing agencies and public libraries across the city. All participants provided written informed consent, which included permission to contact their supportive housing agency or case manager to verify their enrollment in supportive housing. We verified 855 participants as living in a supportive housing program and eligible for the study; the remaining 33 participants were not living in supportive housing at the time of the baseline survey. The current study examined the 157 study participants who self-reported living with HIV. All study protocols were approved by the Institutional Review Board of [Blinded for Review].

Survey Procedures

After collecting informed consent, research assistants administered the Mini International Neuropsychiatric Interview (MINI) using the online platform, completed only at baseline. Participants then completed audio-assisted computer self-interview assessments at baseline, 6-, 12-, and 18-months post-baseline. Participants received $40 and two transportation vouchers for each assessment.

Assessment Measures

ART Adherence was assessed by showing participants a percentile scale and asking them to estimate the percentage of total doses of ART they took during the past 30 days, an assessment tool that has demonstrated validity among PLH.^25,26 Adherence was dichotomized as greater or less than 95% adherence in the prior month. Optimal virologic success declines significantly in patients taking fewer than 95% of prescribed doses.^27,28

Predictors

Demographic Background included age, gender, race, ethnicity, sexual orientation, partnership status, education, income, employment, number of lifetime homeless episodes, number of homeless episodes during the three years prior to entering the supportive housing program, duration of longest homeless episode, number of incarcerations, and duration of longest period of incarceration.

Housing Status.

Participants reported the name and location of their supportive housing program, as well as their case manager’s name and contact information. Case managers confirmed the date an individual was housed and type of housing program. Housing configuration (fixed site, scattered site) and service provision model (low-intensity, high-intensity, behavioral health) were coded. For these analyses, we focused on the intensity of supportive housing services, which was comprised of three categories. In low intensity case management programs, the case manager to client ratio was greater than 20:1 and residents received face-to-face case management services provided less than monthly. Caseloads for programs in this study ranged from 25:1 to 60:1. Intensive case management programs were defined as programs in which the case manager to client ratios were no greater than 20:1, with a range from 8:1 to 20:1, and where residents received face-to-face case management services provided at least monthly. Finally, behavioral health programs were characterized as those with case manager to client ratios of less than 20:1 within programs that specialize in housing and case management for individuals with behavioral health challenges.²¹ We also examined housing agencies specializing in housing for PLH. This was a dichotomous measure that included supportive housing programs specifically designed or intended for PLH as compared to more general supportive housing programs for other homeless individuals, which may include persons living with HIV, substance use disorders, mental illness, and other chronic conditions.

Housing satisfaction was measured using a modified version of the 18-item SAMSHA Housing Satisfaction Scale with a five-point Likert scale (strongly disagree to strongly agree). There are four subscales:²⁹ choice, safety, privacy and proximity, all with reliability exceeding .83 (for the current sample, alpha=.88). Perceived neighborhood crime and violence was measured using the Neighborhood Risk Assessment,³⁰ seven items that ask participants to rate how often (never, rarely, sometimes, often, very often) various problems are present in their community, including gangs, killings, drug dealing, and drug use (alpha=.93).³⁰

HIV/Health Care and Service Utilization variables included: time since HIV diagnosis; number of HIV clinic visits in the past 6 months; location of HIV or other medical care (hospital clinic, neighborhood clinic, or private physician’s office); receipt of treatment for a psychological or emotional issue in the past 6 months; and current health insurance status (yes/no coverage) and type of insurance (e.g. private, Medicare, Medicaid, VA). We measured participants’ perceived access to health care using the Access to Care scale; six items on a five-point Likert scale (strongly agree to strongly disagree, example “If I need medical care, I can get admitted without any trouble”).³¹ Reliability for the main sample was marginal (alpha=.67). We also assessed barriers to health care within the last month. Participants responded agree or disagree to 10 items that assessed reasons they may not have gotten needed medical care in the previous month such as, “I did not have child care” (alpha=.60).³²

We assessed Mental Health using baseline MINI diagnoses (dichotomous indicators) of severe mental illness (SMI), substance use disorder (SUD), and dual SMI and SUD diagnoses. At baseline and follow-up assessments, depressive symptoms were assessed, using the Center for Epidemiological Studies of Depression (CES-D 20),³³ and anxiety symptoms, using the Generalized Anxiety Disorder Assessment (GAD-7).³⁴ Reliability of both indices for the main sample was good (alpha = .88 and .91, respectively). Social support included five items to measure availability of various types of support (e.g. “someone to help with transportation”, “someone to love and make you feel wanted”) and was measured with a five-point Likert scale (none of the time to all of the time) (alpha=.77).^35,36

Analytical Procedures

We examined the effects of case management service intensity on ART adherence among a sample of PLH in supportive housing at baseline, 6-, 12-, and 18-months post-baseline. We hypothesized that 1) ART adherence among PLH receiving high-intensity case management would be greater compared to PLH receiving low-intensity case management services; and 2) ART adherence among PLH in permanent supportive housing programs specializing in housing PLH would be greater compared to traditional permanent supportive housing programs. We controlled for demographics, health care service utilization, and mental health factors. In our exploratory analysis, we examined the bivariate associations of adherence with housing service intensity, along with a variety of potential covariates from the variable domains mentioned previously. General Estimating Equations (GEE) for repeated measures (binomial distribution) were performed to assess the bivariate relationships. To avoid rejecting variables that were not nominally significant when considered alone, predictors achieving a threshold alpha set at p<.15 in the bivariate analysis were provisionally included in a multivariate GEE repeated measures analysis. The main intervention effect of service intensity was estimated over time (assessment); the significance of any change found was determined by the interaction of service intensity and time. To preserve statistical power given our limited sample size, covariates were entered in the GEE model in a backward stepwise algorithm. Predictors were removed, one at a time, in order of least statistical significance, until only covariates with a p-value<.05 remained. At baseline and each subsequent follow-up, we computed estimated marginal mean proportions of participants who achieved at least 95% adherence. The same procedures were repeated, substituting the indicator variable for housing agencies specializing in PLH as the main factor. Estimates of partial odds ratios, 95% confidence intervals, and p-values were also computed for service intensity or agency type and each covariate.

Results

Sample characteristics for the 157 PLH supportive housing residents are presented in Table 1. The majority of PLH (67.5%; n=106) resided in high-intensity case management programs. Almost 46% (n=72) of the participants were housed with an agency that specialized in serving PLH. Fifty-seven percent of PLH (n=89) reported ART adherence 95% of the time or greater within the past month. There were no significant differences in adherence by housing type or intensity of case management services at baseline (results not shown).

Table 1.

Characteristics of 157 PLH Participants at Baseline

Percent of ART medications taken in the past 30 days—Mean (SD)	79.3 (34.14)
95%-plus ART adherent—% (n)
Less than 95% adherent	42.7 (67)
95%-plus adherent	56.7 (89)
Not ascertained	0.6 (1)
Predictors:
Demographic Background:
Age (in years)—Mean (SD)	49.7 (10.59)
Gender—% (n)
Male	72.6 (114)
Female	22.9 (36)
Transgender/other	4.4 (7)
Race/Ethnicity—% (n)
African-American	81.5 (128)
White	6.4 (10)
Other non-Hispanic	2.5 (4)
Hispanic	9.6 (15)
Relationship Status—% (n)
Married/partnered	7.6 (12)
Widowed/divorced/separated	17.8 (28)
Single, never married	61.8 (97)
Other	12.7 (20)
Education—% (n)
Less than complete high school	35.7 (56)
High school graduate	32.5 (51)
Some higher education	31.8 (50)
Monthly income—Mean (SD)	633.7 (420.14)
Sexual orientation—% (n)
Heterosexual	45.9 (72)
Gay/lesbian	32.5 (51)
Bisexual	12.1 (19)
Other	6.4 (10)
Not ascertained	3.2 (5)
HIV/Health Background and Beliefs:
Years since HIV diagnosis—% (n)
Less than 1 year	1.3 (2)
1 to 4 years	11.5 (18)
5 to 9 years	14.6 (23)
10-plus years	70.1 (110)
Not ascertained	2.5 (4)
Has health insurance—% (n)	91.7 (144)
Perceived access to healthcare—Mean (SD)	25.0 (4.63)
Perceived barriers to healthcare—Mean (SD)	1.3 (1.5)
Mental Health Factors:
Meets criteria for SMI diagnosis (MINI)—% (n)	38.2 (60)
Meets criteria for SUD diagnosis (MINI)—% (n)	34.4 (54)
Meets criteria for dual SMI/SUD diagnosis (MINI)—% (n)	24.2 (38)
CESD-20 depression—Mean (SD)	17.2 (9.90)
GAD-7 anxiety—Mean (SD)	5.4 (5.43)
Social support—Mean (SD)	7.2 (4.81)
Health Care Service Utilization variables:
Number of HIV clinic visits in past 6 months—Mean (SD)	2.4 (1.82)
Saw someone for psychological or emotional issue in past 6 months—% (n)	27.4 (43)
Where received medical care, including for HIV in past 6 months—% (n)
Hospital clinic	38.9 (61)
Community/neighborhood clinic	32.5 (51)
Private physician’s office	10.2 (16)
Housing Characteristics and Background:
Supportive housing services received—% (n)
Low-intensity case management	18.5 (29)
Intensive case management	67.5 (106)
Behavioral health	14.0 (22)
Housing program configuration—% (n)
Fixed site	35.0 (55)
Scattered site	65.0 (102)
Agencies specializing in PLH services—% (n)
General services only	54.1 (85)
PLH specialized services	45.9 (72)
In supportive housing less than 2 years—% (n)	40.1 (63)
Longest homeless period was > 1 year—% (n)	58.0 (91)
Housing satisfaction—Mean (SD)	56.8 (9.61)
Perceived neighborhood crime and violence—Mean (SD)	13.4 (8.23)

Bivariate Analysis.

Bivariate associations are shown in Table 2. High-intensity housing services did not predict ART adherence. Specialized housing for PLH was a marginally significant predictor of greater ART adherence (OR=1.60, p=.053). Housing satisfaction was significantly and positively associated with ART adherence (OR=1.04; 95% CI=1.01,1.07). African-Americans and participants who did not complete high school had lower rates of ART adherence. Participants with long-term HIV diagnoses (>10 years) had higher rates of adherence, as did those who utilized a community or neighborhood clinic for HIV care. Greater perceived access to healthcare was significantly associated with greater ART adherence. Finally, the bivariate analysis revealed that greater depression and anxiety symptomology, as measured by the CES-D and GAD-7, were significantly linked with lower levels of ART adherence.

Table 2.

Analysis of Longitudinal Predictors of 95% ART Adherence among 157 HIV-Positive Participants in Permanent Supportive Housing^a

Variable Domain/Predictor	Odds Ratio	95% Confidence Interval	p
Demographic and Socioeconomic Background:
Age (in years)	1.01	0.45, 1.49	.519
Gender (female)	0.78	0.44, 1.37	.384
Race (African-American)	0.44	0.23, 0.84	.013
Hispanic ethnicity	1.73	0.76, 3.94	.191
Education (< complete high school)	0.60	0.37, 0.96	.033
LGBT orientation	0.95	0.60, 1.49	.817
HIV/Health Background and Beliefs:
Time since HIV diagnosis (> 10 years)	1.82	1.14, 2.92	.013
Has health insurance	1.34	0.65, 2.80	.414
Perceived access to healthcare scale	1.11	1.04, 1.19	.003
Perceived barriers to healthcare scale	0.74	0.61, 0.88	.001
Mental Health Factors:
SMI (MINI) diagnosis at baseline	0.88	0.54, 1.41	.588
SUD (MINI) diagnosis at baseline	1.02	0.62, 1.67	.930
Dual SMI & SUD diagnoses at baseline	0.89	0.52, 1.53	.684
Social support scale	0.98	0.92, 1.04	.484
CESD-20 depression scale	0.96	0.93, 0.98	.001
GAD-7 anxiety scale	0.93	0.89, 0.98	.009
Housing Characteristics and Background:
Housing service type:
Behavioral health	1.04	0.45, 2.42	.491
Intensive case management	0.78	0.43, 1.44	.434
Low-intensity case management	0b
Housing agencies specializing in PLH service	1.60	0.99, 2.58	.053
In supportive housing less than 2 years	1.15	0.69, 1.89	.597
Longest homeless period was > 1 year	1.11	0.70, 1.75	.668
Housing satisfaction scale	1.04	1.01, 1.07	.005
Perceived neighborhood crime and violence scale	1.00	0.97, 1.03	.941
Service Utilization Variables:
Where received med care, past 6 mos.
Hospital clinic	1.51	0.86, 2.65	.148
Community/neighborhood clinic	1.78	1.01, 3.13	.046
Private physician’s office	1.86	0.94, 3.68	.073
Saw someone for psychological or Emotional issue	1.11	0.73, 1.68	.630
Number of HIV clinic visits in past 6 months	0.98	0.89, 1.07	.593

^aART adherence was measured as taking 95% or more of anti-retroviral medication doses in the past month. The relationship between this outcome and each predictor was assessed individually in a General Estimating Equation for repeated measures (binomial outcomes) between the baseline and 18-month follow-up. Predictors achieving p-values less than .10 were then tested in a subsequent multivariate model.

^bReference category.

Multivariate Analysis.

The results of our multiple GEE models are presented in Tables 3 and 4. Contrary to our hypotheses, case management service intensity was not significantly associated with ART adherence (Table 3). African American residents had lower odds of ART adherence compared to White residents (AOR: 0.39, p=.006). Additionally, time since diagnosis was a significant predictor of ART adherence; individuals who had received their diagnoses greater than ten years ago had greater ART adherence (AOR: 1.71, p=.027). Finally, higher levels of depression as measured by the CES-D were associated with lower rates of adherence (AOR: 0.96, p=.002). When we examined ART adherence by agencies specializing in housing for PLH, we found no significant association between adherence and agency type (Table 4). That is, when controlling for other predictors, participants who lived in housing programs specializing in housing PLH did not have greater adherence as compared to those living in more general supportive housing programs intended for the general homeless population. Similar to what we found when examining ART adherence by service type, ART adherence was significantly associated with African American race (AOR: 0.39, p=.006), time since HIV diagnosis (AOR: 1.66, p=.036) and depression (AOR: 0.96, p=.002).

Table 3.

Longitudinal Analysis of 95 Percent ART Medication Adherence, by Supportive Housing Service Type, Controlling for Demographic, Contextual, and Health-Related Covariates (n=157)^a

	Estimated Mean Proportion of Clients by Assessment:
Factor/Covariateb	Baseline	6-Months	12-Month	18-Month	OR	95% CI	p-value
Assessment	0.66	0.63	0.66	0.64			.890
Supportive Housing Service Type:
Behavioral health	0.71	0.68	0.71	0.69	1.35	0.55, 3.29	.512
Intensive case management	0.61	0.58	0.62	0.59	0.87	0.46, 1.65	.868
Low-intensity case management	0.65	0.62	0.65	0.63	0c
Race (African-American)					0.38	0.19, 0.75	.005
Time since HIV diagnosis (> 10 years)					1.75	1.09, 2.81	.021
CESD-20 depression scale					0.96	0.93, 0.99	.002

^aThe analysis was a Generalized Estimating Equation for repeated measures (binomial outcomes), factorial design, with covariates. The main effect for supportive housing service type was estimated over time (assessment). The interaction of service type and assessment was not significant and was omitted. Other covariates (predictors) with a p-value < .10 in the univariate analysis were entered in the model of main factors and interactions and removed one at a time, in order of least significance, until only those with a p-value < .05 remained, except for PLH housing agencies, which was retained in the model. The analysis sample was limited to the number of HIV positive participants at baseline.

^bEstimated marginal mean scores (proportions) of the dependent variable are presented by assessment for service intensity, along with partial estimates and p-values. Grand means and a summary significance value are presented for the assessment factor. Covariates that were not significant and were removed from the backward stepwise model included, education, perceived access to healthcare, perceived barriers to healthcare, GAD-7 anxiety scale, and housing satisfaction scale.

^cReference category.

Table 4.

Longitudinal Analysis of 95 Percent ART Medication Adherence, by Specialized PLH Housing Agencies, Controlling for Demographic, Contextual, and Health-Related Covariates (n=157)^a

	Estimated Mean Proportion of Clients by Assessment:
Factor/Covariateb	Baseline	6-Months	12-Month	18-Month	OR	95% CI	p-value
Assessment	0.64	0.61	0.64	0.62			.894
Housing agencies that specialize in PLH:
Specialized agencies	0.67	0.64	0.67	0.65	1.30	0.80, 2.09	.288
Other agencies	0.61	0.58	0.61	0.58	0c
Race (African-American)					0.39	0.20, 0.77	.006
Time since HIV diagnosis (> 10 years)					1.66	1.03, 2.67	.036
CESD-20 depression scale					0.96	0.94, 0.99	.002

^aThe analysis was a Generalized Estimating Equation for repeated measures (binomial outcomes), factorial design, with covariates. The main effect for specialized housing agencies for PLH clients was estimated over time (assessment). The interaction of PLH agencies and assessment was not significant and was omitted. Other covariates (predictors) with a p-value < .10 in the univariate analysis were entered in the model of main factors and interactions and removed one at a time, in order of least significance, until only those with a p-value < .05 remained. The analysis sample was limited to the number of HIV positive participants at baseline.

^bEstimated marginal mean scores (proportions) of the dependent variable are presented by assessment for service intensity, along with partial estimates and p-values. Grand means and a summary significance value are presented for the assessment factor. Covariates that were not significant and were removed from the backward stepwise model included, education, perceived access to healthcare, perceived barriers to healthcare, GAD-7 anxiety scale, and housing satisfaction scale.

^cReference category.

Discussion

ART adherence is critical to improving the health outcomes of PLH and reducing the public health impact of HIV.²⁸ Although estimated rates of ART adherence among homeless individuals vary, individuals who are homeless or socially disadvantaged have approximately a 45% reduced odds of achieving optimal ART adherence.³⁷ In this study, nearly sixty percent (57.1%) of PLH residing in permanent supportive housing reported optimal ART adherence in the previous month.

Contrary to our hypotheses, we did not find a significant relationship between intensity of case management services and ART adherence. Previous research has suggested that intensive case management may be an important aspect of increasing ART adherence among PLH in supportive housing.² Yet, the majority of this research has compared supportive housing with homelessness or housing instability, demonstrating the impact of receipt of supportive housing on ART adherence.² Similarly, we found that individuals housed with agencies specializing in housing PLH did not have greater ART adherence as compared to PLH housed in more general supportive housing programs. Although PLH-specific housing programs may offer specialized services and case management teams more knowledgeable about HIV treatment,^20,38 this may not necessarily translate into improvements in ART adherence or other HIV health outcomes. However, it is important to note that despite services being available, the extent to which participants utilized or were satisfied with those services was not clear. For example, in Housing First programs, which represent the majority of programs in this study, services are voluntary and flexible and so residents may have received different services although in the same programs

In our bivariate analysis, housing satisfaction was positively associated with ART adherence, a relationship that has not been adequately examined in the literature. Housing First models have been shown to increase resident satisfaction in some early studies.¹⁸ Similarly, greater choice in where one lives is associated with greater housing satisfaction.³⁹ Additional research with a larger sample of PLH may help better elucidate the relationship between housing satisfaction and ART adherence, particularly when considering the various configurations of housing and services.

There is a need to better understand the effects of various models of case management services beyond service intensity to identify which components may contribute to housing stability and health improvement for PLH. In our comparison of housing type, we classified housing programs based on intensity of case management, defined using case manager to client ratios. However, our analyses do not capture other potentially important characteristics of case management. For example, although lower case manager to client ratios allow for more frequent contact, we do not have data on frequency of home visits, nor did we capture satisfaction with case manager or length of case manager-client relationships. These and other factors may influence the quality of the relationship and may be more important than caseload when considering the effects of services on client health outcomes. Finally, there may be other benefits of case management services beyond ART adherence that are important to consider. Supportive housing with intensive case management has been found to increase housing stability⁴⁰ and reduce HIV viral load,²² emergency department visits,⁴¹ and unmet health care needs⁴² as compared to individuals who are homeless or in free-market housing. Research with a larger sample of PLH is needed to examine how case management services influence health and social outcomes for PLH within various supportive housing programs.

Residents with depressive symptoms had lower levels of ART adherence. While behavioral health models of supportive housing are designed to provide care and services for people disabled by severe mental health and/or substance use disorders,²¹ depressive symptoms and episodes are common among individuals in supportive housing. However, our results showed no difference in ART adherence among those housed in behavioral health models of housing as compared to high- and low-intensity case management programs.⁴³ Regardless of service intensity, housing case managers should continue to screen for depression among PLH housing residents and provide appropriate treatment referrals to improve both mental health outcomes and ART adherence.

We also found greater ART adherence among those diagnosed ten or more years ago. Previous research on the effects of time since diagnosis on ART adherence have been mixed. In a meta-analysis examining predictors of ART adherence, researchers found that greater time since HIV diagnosis was associated with ART adherence, although effect size was small.⁴⁴ Yet, other research has not found significant associations between ART adherence and years since HIV diagnosis.⁴⁵ Our results suggest that for supportive housing residents, case managers should be particularly mindful of the needs of residents with more recent diagnoses, who may require additional supports, reminders, and techniques to enhance ART adherence.

Notably, African American residents had lower levels of ART adherence than white residents. This may be due to complex structural factors including historical systematic exclusion of African Americans from healthcare as well as contemporary discrimination and challenges accessing healthcare. Medical mistrust,⁴⁶ conspiracy beliefs,⁴⁷ and racial discrimination⁴⁸ have been found to be associated with lower ART adherence among African Americans. Racial stigma and discrimination may be further compounded by HIV stigma, homonegativity, substance use and mental illness stigma, and/or homelessness stigma.⁴⁹ For example, post-traumatic stress disorder and discrimination due to race, sexual identity, or HIV status have been found to negatively affect ART adherence among African American men.⁵⁰

There are limitations of this study. ART adherence was self-reported and thus, subject to bias. Participants may have over- or under-estimated their adherence. Additionally, we looked at an 18-month period and participants had been housed for various lengths of time upon study enrollment. However, we did examine associations between ART adherence and length of time housed and longest homeless episode. Future longitudinal research should follow supportive housing residents beyond 18 -months to gain a stronger understanding of the impact of supportive housing over several years. Finally, this was a small sample of PLH, and the majority were housed in intensive case management programs. It is possible that our sample was not large enough to detect meaningful differences. Additional research with a larger sample of PLH in a variety of supportive housing program types is needed to further unpack which housing and case manager characteristics contribute to ART adherence.

There is significant empirical evidence of the impact of permanent supportive housing on HIV treatment and prevention.^2,3 Yet, as our research demonstrates, it remains unclear which components or configurations of housing and services best support PLH and contribute to HIV treatment outcomes. Our findings highlight the importance of considering mental health and more recent HIV diagnosis when developing treatment and case management plans to help enhance residents’ ART adherence.