Experiencing homelessness and progression through the HIV cascade of care among people who use drugs

Abstract

Objective:

To investigate the longitudinal association between periods of homelessness and progression through the HIV cascade of care among people who use drugs (PWUD) with universal access to no-cost HIV treatment and care.

Design:

Prospective cohort study.

Methods:

Data were analyzed from the ACCESS study, including systematic HIV clinical monitoring and a confidential linkage to comprehensive antiretroviral therapy (ART) dispensation records. We used cumulative link mixed-effects models to estimate the longitudinal relationship between periods of homelessness and progression though the HIV cascade of care.

Results:

Between 2005 and 2019, 947 people living with HIV were enrolled in the ACCESS study and 304 (32.1%) reported being homeless at baseline. Homelessness was negatively associated with overall progression through the HIV cascade of care (Adjusted Partial Proportional Odds Ratio [APPO] = 0.56, 95% confidence interval [CI]: 0.49–0.63). Homelessness was significantly associated with lower odds of progressing to each subsequent stage of the HIV care cascade, with the exception of initial linkage to care.

Conclusions:

Homelessness was associated with a 44% decrease in the odds of overall progression through the HIV cascade of care, and a 41–54% decrease in the odds of receiving ART, being adherent to ART and achieving viral load suppression. These findings support calls for the integration of services to address intersecting challenges of HIV, substance use and homelessness among marginalized populations such as PWUD.

INTRODUCTION

Advancements in antiretroviral therapy (ART) potency, tolerability and convenience have allowed the life expectancy of many key populations of people living with HIV to approach that of the general population in high-income countries [1–3]. Through optimal ART adherence, plasma HIV-1 RNA viral load (VL) can be suppressed to undetectable concentrations and almost entirely eliminate the risk of vertical, sexual or parenteral HIV transmission [4–6].

Among members of structurally-marginalized populations—such as people living with HIV who use drugs (PWUD)—the effectiveness of ART has been demonstrated at the individual and community level [7–9]. However, people living with HIV (PLWH) who use drugs continue to experience barriers to ART access and adherence and exhibit higher rates of suboptimal HIV virologic outcomes and more rapid HIV disease progression [10–14]. The Rhodes risk environment framework has conceptualized individual (e.g., substance use), social (e.g., HIV-related stigma) and structural risk factors (e.g., incarceration, unstable housing) that intersect to shape HIV treatment engagement and disease progression among PWUD [15–17]. Empirical data from PWUD have shown that factors including addiction-related instability, medical mistrust, criminalization and social marginalization are significant barriers to ART access and adherence [18–21]. In addition, 20–30% of PWUD report experiencing homelessness in the past year and homelessness in particular has been identified as a critical structural factor linking upstream economic and social determinants of health to proximal physical and social exposures that influence general health and HIV care [22, 23]. Homelessness has also been linked to increases in the intensity of alcohol and illicit drug use that can exacerbate HIV risk, interfere with ART adherence and hasten HIV disease progression [18, 24–26]. Previous systematic reviews have found strong evidence that exposure to marginal housing environments was associated with HIV risk behaviour, HIV acquisition, suboptimal HIV care and increased risk of mortality [22, 27–29].

Engagement and progression through the stages of HIV care has been measured through the HIV cascade (or continuum) of care, which includes five distinct steps: diagnosis; linkage to care; retention in care; adherence to ART and viral suppression [30–33]. The cascade facilitates individual assessment of HIV care progression as well as the monitoring and evaluation of community-level HIV care strategies [34]. Despite the common acknowledgement of the deleterious impact of homelessness on measures of HIV treatment and disease, we are unaware of any existing studies that have sought to investigate the longitudinal association between homelessness and progression through the HIV care cascade. The impacts of these intersecting structural vulnerabilities on HIV treatment progression warrant further investigation to identify areas of intervention and improve HIV care among PWUD [18, 21, 27, 34]. Given this knowledge gap, we undertook the present study to evaluate the longitudinal association between homelessness and global progression through the HIV care cascade, as well as the impact of homelessness on transitions between individual stages of the cascade, using data from a prospective cohort of PWUD in a setting with universal no-cost HIV treatment and care.

METHODS

The data for this analysis were collected from an ongoing open cohort of people living with HIV who use drugs: the AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS) [35]. ACCESS enrolls individuals who tested seropositive for HIV and reported having used an unregulated drug (other than or in addition to cannabis) in the past month at enrolment. Additional inclusion criteria include living in the Greater Vancouver region, aged ≥18 years old and provision of written informed consent. Study participants were recruited through community-based methods including extensive street outreach and self-referral. Recruitment has been ongoing since 1996 in the Downtown Eastside neighbourhood of Vancouver, Canada, which is an area that has experienced high rates of homelessness, substance use and HIV outbreaks associated with injection drug use [36].

At baseline and every six months thereafter, participants complete an interviewer-administered questionnaire and provide blood samples for HIV disease monitoring. The questionnaire collects socio-demographic data, substance use behaviours, healthcare access, and related social and structural exposures. Participants are remunerated CAD $40 at each study visit as compensation for their time. All work was conducted in accordance with the Declaration of Helsinki and written informed consent was been obtained for all participants. The study protocol has been reviewed and approved by the University of British Columbia/Providence Healthcare research ethics board on an annual basis.

In the province of British Columbia, all people living with HIV are provided with HIV treatment and care, including medications and HIV clinical monitoring, free of charge. Among ACCESS study participants, a confidential data linkage with the British Columbia Centre for Excellence in HIV/AIDS’ Drug Treatment Program, the province’s centralized ART dispensary and HIV treatment registry, provides a complete retrospective and prospective clinical profile including each participant’s VL and CD4+ cell count tests and details of all of their ART dispensations. Thus, the study is able to incorporate measures derived from their semi-annual study visits and in the course of their regular clinical care in the community.

For the present study, we included all ACCESS participants who enrolled and completed ≥ 1 study interview between December 2005 and November 2019. The primary outcome of interest was progression through the HIV cascade of care. Using data from the confidential linkage to the Drug Treatment Program, we categorized participants into their stage of the HIV cascade of care using a priori definitions from previous studies [28, 33, 34]. At each semi-annual study visit, participants were assigned to one of five potential stages of the HIV cascade of care: (1) unlinked to HIV care, included participants without HIV VL and CD4+ cell count measurements conducted in the community and without any ART dispensations in the six months prior to the study visit; (2) linked to HIV, care included participants with at least one community-based HIV VL or CD4+ cell count measurement and 0 days of ART dispensation in the last six months; (3) on ART, which included participants with at least one day of ART dispensation in the last six months; (4) adherent to ART, included participants with at least 171 days of ART dispensation, corresponding to 95% adherence in the previous six months; and (5) VL suppressed, which included participants with less than 50 copies/mL plasma at the most recent HIV VL analysis [33]. If participants did not have any available VL assessments in the 90 days preceding their most recent study visit, they were classified as unsuppressed unless their ART records indicated dispensation of medications covering ≥95% of the previous six months. Participants were assigned to the highest stage of the HIV care cascade that they achieved at each follow-up visit and individuals could be assigned to higher stages without achieving all prior stages. For example, if a participant did not have recently available VL or CD4+ test, yet was ART adherent, and achieved viral suppression, they were classified as VL suppressed.

The primary explanatory variable of interest was homelessness. Consistent with previous studies [29, 37], homelessness was defined as living on the street with no fixed address at any time in the six-month period preceding the follow-up interview. We also included secondary explanatory variables as covariates that we hypothesized to confound the association between homelessness and progression through the HIV cascade of care based on risk environment frameworks developed by Rhodes and Knowlton, which delineate how social, structural and environmental factors produce or mitigate health-related risks and health service utilization among PWUD [20, 21]. We also included covariates based on relevant empirical evidence from previous studies [18, 19, 29, 38]. These exposures included: sex at birth (male vs. female); age (per year older); race and ethnicity (White vs. Black, Indigenous and people of colour [BIPOC]) [39]; education (high school education or greater achieved vs. less the high school level education); employment (reporting regular, temporary, or self-employed work vs. none); non-injection drug use (≥daily vs. <daily); injection drug use (≥daily vs. <daily); hazardous alcohol use (yes vs. no); engagement in sex work (yes vs. no); incarceration (yes vs. no); and time since baseline visit (per year longer). Hazardous alcohol use was defined based on the National Institute on Alcohol Abuse and Alcoholism (NIAAA) definition of risky alcohol use: >14 drinks/week or >4 drinks on 1 occasion for men <65 years of age, and >7 drinks/week or >3 drinks on 1 occasion for all women and men ≥65 years of age [40]. Variable definitions were consistent with previous studies and behavioural variables are time-updated and refer to the six month period preceding the most recent follow-up visit [29, 35].

First, we analyzed the distribution of the outcome stratified by the explanatory variables at baseline, testing for difference using the Chi-square test for binary variables and the Wilcoxon’s rank sum test for continuous variables. Next, the crude associations between each of the explanatory variables and progression through the HIV cascade of care were analyzed longitudinally using cumulative link mixed effects models (CLMM) with random intercepts. This method accounts for the ordinal nature of the study outcome and inter-individual variability. The likelihood ratio test was used to assess the proportional odds assumption and non-proportional odds were estimated for covariates that violated this assumption. We included all explanatory variables in the multivariable analysis. We also conducted sub-analyses to test the association between homelessness and each individual transition between stages of the HIV cascade of care. These stages were operationalized using the same criteria as the primary analysis. Given that a HIV treatment as prevention (TasP) program was launched in 2010 in British Columbia, Canada under the Seek and Treat for Optimal Prevention of HIV/AIDS initiative (STOP), we also conducted sub-analyses of the association between homelessness and HIV cascade progression during 2005–2010 and 2011–2019. As a final step, we analyzed recent transitions in housing status (transitioning from non-homeless to homeless in the last six months) and overall progression through the HIV cascade of care. These multivariable models also included all explanatory variables and followed the same model building procedure as the primary analysis (CLMM with random intercepts). Since each of the variables in the analysis were time-updated every six months and a low proportion of values was missing during the study period (<4% of observations per variable), we excluded missing observations from the analysis. All statistical analyses were performed using SAS software version 9.3 (SAS, Cary, NC, USA) and all tests of significance were two-sided with a significance threshold of p <0.05.

RESULTS

Between December 2005 and November 2019, a total of 947 participants were enrolled in the ACCESS cohort, completed ≥ one study interview, and were eligible to be included in the present study. At the baseline interview, the median age of the sample was 42.7 years (interquartile range [IQR]= 36.4–48.3), 318 (33.6%) were female, 521 (55.0%) self-reported White race and ethnicity and homelessness was reported by 304 (32.1%) participants. Of the 947 participants, 85 (9.0%) contributed only one interview during the study period and the remainder completed at least two study visits. The distribution of participants in each stage of the HIV cascade of care at baseline included 117 (12.4%) unlinked to care, 224 (23.7%) linked to care, 157 (16.6%) on ART, 153 (16.2%) adherent to ART and 296 (31.3%) virally suppressed. Table 1 shows the baseline characteristics of the study participants stratified by homelessness at baseline.

Table 1.

Baseline characteristics stratified by homelessness in the past six months among the ACCESS cohort (n=947).

		Homelessness
Characteristic	Total n (%)	Yes 304 (32.1%) n (%)	No 636 (67.2%) n (%)	p - value
HIV cascade of care
Unlinked to HIV care	117 (12.4)	53 (17.4)	60 (9.4)	0.230
Engaged in HIV care	830 (87.6)	251 (82.6)	576 (90.6)	0.073
On ART	606 (64.0)	159 (52.3)	445 (70.0)	0.024
Adherent to ART	449 (47.4)	101 (33.2)	346 (54.4)	0.018
Viral suppression	296 (31.3)	56 (18.4)	238 (37.4)	0.010
Age
Median	42.7	40.2	44.0	<0.001
IQR	(36.4 – 48.3)	(33.7 – 45.8)	(37.7 – 49.5)
Sex
Male	628 (66.3)	203 (66.8)	422 (66.4)	0.897
Female	318 (33.6)	101 (33.2)	214 (33.6)
Race and ethnicity
White	521 (55.0)	166 (54.6)	354 (55.7)	0.745
BIPOC	415 (43.8)	135 (44.4)	275 (43.2)
Employment a
Yes	170 (18.0)	50 (16.4)	119 (18.7)	0.398
No	777 (82.0)	254 (83.6)	517 (81.3)
Non-injection drug use a
≥Daily	508 (53.6)	183 (60.2)	323 (50.8)	0.007
<Daily	438 (46.3)	121 (39.8)	313 (49.2)
Injection drug use a
≥Daily	239 (29.9)	84 (38.0)	154 (26.7)	0.002
<Daily	561 (70.1)	137 (62.0)	422 (73.3)
Hazardous alcohol use a
Yes	123 (13.0)	49 (16.1)	72 (11.3)	0.039
No	822 (86.8)	254 (83.6)	564 (88.7)
Sex trade involvement a
Yes	147 (15.5)	52 (17.1)	93 (14.6)	0.334
No	798 (84.3)	252 (82.9)	541 (85.1)
Incarceration a
Yes	125 (13.2)	70 (23.0)	52 (8.2)	<0.001
No	814 (86.0)	230 (75.7)	581 (91.4)

Notes:

^aRefers to activities in the 6 months prior to the follow-up interview, Ref=reference category, IQR=interquartile range, BIPOC=Black, Indigenous and people of colour, Bold text refers to P-values <0.05, Not all cells may add up to 947 as participants may choose not to answer sensitive questions.

The unadjusted and adjusted CLMM analyses of progression through the HIV cascade of care are presented in Table 2. In the adjusted analysis, homelessness (Adjusted Partial Proportional Odds Ratio [APPO]=0.56, 95% CI: 0.49–0.63), non-injection drug use (APPO=0.82, 95% CI: 0.74–0.91), injection drug use (APPO=0.64, 95% CI: 0.57–0.72) and hazardous alcohol use (APPO=0.78, 95% CI: 0.67–0.92) were negatively associated with progression the HIV care cascade. Older age (APPO=1.07, 95% CI: 1.06–1.08), male sex (APPO=1.27, 95% CI: 1.02–1.58) and longer time since baseline (APPO=1.14, 95% CI: 1.12–1.15) were positively associated with progression through the HIV care cascade. Recent transitions to becoming homeless were not significantly associated with HIV cascade of care progression (APPO=1.07, 95% CI: 0.92–1.23).

Table 2.

Bivariable and multivariable cumulative linked mixed-effects models (CLMM) analysis of factors associated with progression through the HIV cascade of care.

	Unadjusted			Adjusted
Characteristic	OR (95% CI)	p-value	t-value	OR (95% CI)	p-value	t-value
Homelessness a
(yes vs. no)	0.39 (0.34, 0.44)	<0.001	−15.121	0.56 (0.49, 0.63)	<0.001	−8.854
Age
(OR per year older)	1.08 (1.07, 1.09)	<0.001	15.270	1.07 (1.06, 1.08)	<0.001	11.942
Sex
(male vs. female)	1.63 (1.30, 2.05)	<0.001	4.207	1.27 (1.02, 1.58)	0.034	2.127
Race and ethnicity
(White vs. BIPOC)	1.12 (0.90, 1.40)	0.312	1.012	0.88 (0.72, 1.09)	0.244	−1.165
Education
(≥High school diploma vs. less)	1.03 (0.82, 1.29)	0.789	0.267	0.90 (0.73, 1.10)	0.288	−1.062
Employment a
(yes vs. no)	1.07 (0.95, 1.21)	0.267	1.109	0.98 (0.87, 1.11)	0.768	−0.294
Non-injection drug use a
(≥daily vs. <daily)	0.63 (0.57, 0.70)	<0.001	−9.090	0.82 (0.74, 0.91)	0.002	−3.716
Injection drug use a
(≥daily vs. <daily)	0.64 (0.57, 0.71)	<0.001	−7.690	0.64 (0.57, 0.72)	<0.001	−7.467
Hazardous alcohol use a
(yes vs. no)	0.89 (0.76, 1.04)	0.133	−1.503	0.78 (0.67, 0.92)	0.003	−3.026
Sex trade involvement a
(yes vs. no)	0.62 (0.51, 0.74)	<0.001	−5.230	1.05 (0.87, 1.27)	0.601	0.553
Incarceration a
(yes vs. no)	0.54 (0.45, 0.65)	<0.001	−6.579	0.98 (0.81, 1.18)	0.795	−0.260
Time since baseline
(per year longer)	1.13 (1.13, 1.14)	<0.001	16.927	1.14 (1.12, 1.15)	<0.001	21.788

Notes: OR= odds ratio, CI= confidence interval, BIPOC=Black, Indigenous and people of colour,

^aRefers to activities in the 6 months prior to the follow-up interview, bold text refers to P-values <0.05.

Figure 1 shows the sub-analyses of the association between homelessness and transitioning to each subsequent stage of the HIV cascade of care. After adjusting for potential confounders, homelessness was not significantly associated with engagement in HIV care, yet was significantly associated with lower odds of being on ART, adherence to ART and achieving viral load suppression (Table 3, Figure 1). The sub-analyses of the association between homelessness and HIV cascade progression during 2005–2010 vs. 2011–2019 showed that homelessness, age, non-injection drug use, injection drug use and time since baseline were significantly associated with HIV cascade progression during both time periods, whereas sex and hazardous alcohol use were only associated with HIV cascade progression during 2011–2019 (Table 4).

Figure 1.

Adjusted cumulative linked mixed-effects model analysis of the association between homelessness and progression through the HIV cascade of care among 947 PWUD living with HIV. Notes: CI=confidence interval, ART=antiretroviral therapy, VL= plasma HIV-1 RNA viral load.

Table 3.

Multivariable cumulative linked mixed-effects models (CLMM) analysis of the associations between homelessness and each individual transition of the HIV cascade of care.

Characteristic	OR (95% CI)	p - value	t-value
Engaged in HIV care vs. Unlinked to HIV care
	0.90 (0.64, 1.27)	0.553	−0.593
On ART vs. Engaged un care
	0.58 (0.44, 0.77)	<0.001	−3.738
ART adherent vs. On ART
	0.46 (0.37, 0.59)	<0.001	−6.361
VL suppressed vs. Adherent to ART
	0.59 (0.51, 0.70)	<0.001	−6.497

Notes: OR= odds ratio, CI= confidence interval, ART= antiretroviral therapy, VL= viral load, bold text refers to P-values <0.05.

Table 4.

Multivariable cumulative linked mixed-effects models (CLMM) analysis of factors associated with progression through the HIV cascade of care during 2005–2010 and 2011–2019.

	2005–2010		2011–2019
Characteristic	OR (95% CI)	p - value	OR (95% CI)	p - value
Homelessness a
(yes vs. no)	0.44 (0.30, 0.63)	<0.001	0.55 (0.44, 0.68)	<0.001
Age
(OR per year older)	1.16 (1.12, 1.20)	<0.001	1.06 (1.05, 1.08)	<0.001
Sex
(male vs. female)	1.55 (0.86, 2.79)	0.148	1.43 (1.06, 1.93)	0.018
Race and ethnicity
(White vs. BIPOC)	0.84 (0.48, 1.48)	0.540	1.13 (0.86, 1.49)	0.391
Education
(≥High school diploma vs. less)	0.67 (0.39, 1.16)	0.150	0.90 (0.69, 1.19)	0.471
Employment a
(yes vs. no)	0.85 (0.56, 1.29)	0.449	1.17 (0.96, 1.42)	0.126
Non-injection drug use a
(≥daily vs. <daily)	0.70 (0.50, 0.97)	0.034	0.69 (0.59, 0.81)	<0.001
Injection drug use a
(≥daily vs. <daily)	0.43 (0.30, 0.63)	<0.001	0.64 (0.53, 0.78)	<0.001
Hazardous alcohol use a
(yes vs. no)	0.98 (0.57, 1.69)	0.945	0.68 (0.53, 0.87)	0.003
Sex trade involvement a
(yes vs. no)	0.77 (0.42, 1.39)	0.382	1.09 (0.80, 1.49)	0.577
Incarceration a
(yes vs. no)	0.63 (0.39, 1.02)	0.062	1.15 (0.82, 1.60)	0.411
Time since baseline
(per year longer)	1.28 (1.14, 1.45)	<0.001	1.11 (1.07, 1.14)	<0.001

Notes: OR= odds ratio, CI= confidence interval, BIPOC=Black, Indigenous and people of colour,

^aRefers to activities in the 6 months prior to the follow-up interview, bold text refers to P-values <0.05.

DISCUSSION

We observed that homelessness was a pervasive exposure and was associated with a 44% decrease in the odds of overall progression through the cascade. While homelessness was not significantly associated with linkage to care, homelessness was associated with a 41–54% decrease in the odds of receiving ART, being adherent to ART and achieving viral load suppression. These associations were observed in multivariable models adjusted for a range of relevant demographic, behavioural and socio-structural covariates, including substance use behaviours such as injection drug use and hazardous alcohol use that were also independently associated with HIV cascade of care progression.

Although numerous studies have investigated the impact of homelessness and unstable housing on discrete HIV outcomes, including ART adherence [41, 42] and HIV viral response and rebound [43, 44], this is the first study to our knowledge to estimate the association between homelessness and progression through the HIV cascade of care among PWUD [22]. There are a number of mechanisms that may account for the deleterious associations we observed. The daily challenges of securing basic survival needs including shelter may take precedence over other health needs such as engaging in HIV care [45, 46]. People who are homeless may be more likely to engage in survival-based income generating activities such as sex work and drug dealing [47–49]. Criminalization of these behaviours, as well as intersecting HIV and substance use-related stigma, have been shown to create barriers to HIV care and ART access [17, 44, 49, 50]. People who are homeless tend to have increased psychiatric comorbidities and engage in higher intensity substance use, both of which can limit engagement with health services [51, 52]. We observed that unregulated substance use (injection and non-injection) and hazardous alcohol use were also significantly associated with HIV cascade of care progression. This suggests that there are mechanisms by which homelessness impacts HIV cascade of care progression that are independent from substance use. Our findings (Table 3, Figure 1), together with other evidence, indicate that these mechanisms significantly interfere with ART initiation, ART adherence and VL suppression [17, 22]. Transient living situations and managing material security may be barriers to engaging with HIV care and testing, as well as accessing prescribers to initiate ART [53–56]. Even after accessing HIV care and ART, the lack of a safe place to store medication and lack of privacy may interfere with ART adherence [53, 56]. Food insecurity has also been associated with lower ART adherence, incomplete viral suppression and lower CD4+ cell counts among people who are homeless [55–57]. For example, other studies have found a 70% decrease in the odds of HIV viral suppression among people who were food insecure after adjusting for ART adherence [54, 58, 59]. These mechanisms may account for our observation that homelessness was associated with a 41–54% decrease in the odds of receiving ART, being adherent to ART and achieving viral load suppression (Figure 1). We also found that recent transitions to becoming homeless were not significantly associated with overall HIV cascade of care progression. It is possible that the impact of homelessness on HIV cascade of care progression may be more significant among people who have experienced greater cumulative exposure to challenges and harms associated with material insecurity, unstable housing and transient living. This is supported by a previous investigation in our study setting showing that longer durations of homelessness were associated with a lower likelihood of achieving non-detectable plasma HIV-1 RNA viral load [60].

We observed that homelessness was negatively associated with longitudinal transitions into each stage of the HIV cascade of care, except for linkage to care. The lack of association with linkage to care may be the result of the universal no-cost access to HIV care available in this setting which facilitates initial contact with HIV services, yet the challenges of homelessness are more significant barriers for achieving subsequent steps of the cascade of care. The observation that homelessness was negatively associated with each subsequent transition of the HIV cascade of care demonstrates that, despite universal access to care, additional HIV care supports are needed for PLWH who use drugs during periods of homelessness despite the presence of HIV TasP initiatives that aimed to improve widespread HIV testing and accelerate the initiation of no-cost ART [22, 61]. One approach that may address this need are Housing First services that provide permanent and subsidized housing for persons with health challenges experiencing homelessness without requiring prior treatment of sobriety [62, 63]. A systematic review of Housing First programs found that these strategies decreased homelessness by 37%, increased rates of HIV viral load non-detectability by 22% and also reduced emergency department use, hospitalization and mortality by 36–41% [64]. Other studies have found that patient navigation models that use designated care coordinators or mobile care teams are an effective approach to improve access to stable housing and support engagement and retention in HIV, mental health and substance use care [63, 65–67]. A previous patient navigation study among people living with HIV and co-occurring mental health and substance use disorders found that nearly two thirds of the participants transitioned to more stable housing and these individuals were twice as likely to be retained in care and achieve viral suppression [65].

Our findings, in conjunction with many other studies, underscore the need for policies and programs that address multiple structural determinants of health among vulnerable populations such as PLWH who use drugs [16, 22, 41]. Integrating HIV care (e.g., screening, preventative counselling, ART) with services such as harm reduction or opioid agonist therapies has shown significant improvements HIV diagnosis, linkage to care, ART uptake and adherence, viral suppression and infectious disease vaccination [33, 68, 69]. Advanced integrative care models that also include access to social services and psychiatric support were associated with a three-fold increase in the rate of viral suppression compared to HIV primary care clinics [70]. To address the burden of navigating multiple, and often siloed, health systems, street outreach programs have been developed to provide HIV counselling, case management, low-barrier medical care, and mental health and substance use treatment [69]. These programs have demonstrated improved virological outcomes and patients preferred accessing these services via mobile care teams over traditional care settings [71, 72].

This study has a number of limitations. We were unable to analyze the association between homelessness and HIV diagnosis since the ACCESS cohort includes participants who had already tested seropositive for HIV. Since homelessness was measured by self-report and was dichotomized (yes versus no), we were not able to determine the exact duration of homelessness at each study visit. Participants in the ACCESS study were recruited through community-based referral rather than random sampling and these findings may not be generalizable to people living with HIV who use drugs in other settings. HIV cascade of care measures were based on available data from the Drug Treatment Program and measuring ART adherence based on dispensation records rather than self-report may have overestimated our measures of ART adherence if medication was dispensed but not ingested. Viral load measurements were also based on available data from the Drug Treatment Program and the threshold of <50 copies/mL is lower than limits used in other studies [73]. Assessing stigmatized and criminalized behaviours such as homelessness and substance use via self-report raises the possibility of social desirability and recall biases. However, the reliability and validity of self-reported measures among PWUD has been demonstrated in previous studies [74, 75]; also, our outcome was not derived from participant self-report. Although this analysis accounted for several covariates, other important variables identified in risk environment frameworks (e.g., HIV stigma, patient-provider relationships) were not collected in this cohort and thus could not be included in the analysis [20, 21]. Lastly, it is possible that residual confounding may have influenced our findings since this was an observational study design.

The present study analyzed the association between homelessness and progression through the HIV cascade of care among a prospective cohort of PWUD with universal access to care. We found that homelessness was significantly associated with slower progression through each step of the HIV cascade of care, with the exception of linkage to care. These findings build on evidence demonstrating the deleterious impacts of homelessness on HIV care and support calls for the integration of services to address intersecting challenges of HIV, substance use and homelessness to promote health equity among marginalized populations such as PWUD.

Data availability statement

The datasets generated during and/or analyzed during the current study are not available due to the ethical agreements of the parent cohorts. Please contact M-J Milloy (bccsu-mjm@bccsu.ubc.ca) regarding access to the datasets generated and/or analysed for this study.