The Study of Treatment and Reproductive Outcomes (STAR) is an ongoing study of cisgender women ages 18–45 years with and without HIV enrolled in 6 Southern US sites (Atlanta, Georgia; Birmingham, Alabama; Jackson, Mississippi; Chapel Hill, North Carolina; Miami, Florida; and Washington, District of Columbia), encompassing a region in which PWH experience limited access to care and adverse health outcomes compared with people in other US regions.^21–23 STAR collects longitudinal clinical, behavioral, and laboratory data every 6 months through interviewer-administered questionnaires, biological specimens, and medical record abstraction. For this analysis, we included baseline data from participants enrolled in March 2021 through August 2023. STAR is approved by a single Institutional Review Board at the University of Miami; all participants provided informed consent in English or Spanish.

Housing status was characterized using results of 3 survey items: (1) Where are you living now? (own house/apartment, parent’s house, someone else’s house/apartment, rooming/halfway house, shelter/welfare hotel, on the street(s)/beach, jail/other correctional facility, residential drug/alcohol treatment facility, car, assisted living, other place); (2) Is the housing where you are living now subsidized? That is, are you paying lower rent because the Federal, State, or local government is paying for part of your rent? (yes, no, not applicable); and (3) Are you currently on a waiting list for subsidized housing? (yes, no, not applicable). Additional variables included length of stay at current location and number of adults and children living with participants.

For analyses, the primary independent variable was housing status. Participants who reported living in their own house or apartment were categorized as “stably housed.” Participants who reported living in rooming/halfway house, shelter/welfare hotel, on the street(s)/beach, jail/other correctional facility, residential drug/alcohol treatment facility, car, assisted living, or other place were categorized as “unstably housed.” Participants who reported living in their parent’s house or someone else’s house/apartment were grouped into a third category entitled “dependently housed,” based on prior literature. Because STAR data could not differentiate between voluntary and involuntary dependent housing, both groups comprise the “dependently housed” category. For logistic regression models, unstably and dependently housed participants were grouped into one “unstably housed” category.

We explored the following self-reported demographic variables: age, race/ethnicity, pregnancy status, marital status, highest level of education, employment status, and annual household income. We also created a combined four-level socioeconomic status variable based on data provided from the education, employment, and income variables (Table 1). Psychosocial covariates included history of incarceration, history of transactional sex, depressive symptoms,²⁴ hazardous drinking,²⁵ tobacco use, marijuana use in the past year, and illicit drug use in the past year.

Sense of community and neighborhood safety were also assessed due to their previously explored influence on health outcomes.^26,27 Psychological sense of community (PSOC) includes 4 components: needs fulfillment (a perception that “members’ needs will be met by the community”), group membership (a feeling of belonging), influence (a sense that one can make a difference in their community), and emotional connection (a feeling of attachment).²⁸ Participants completed a modified, eight-item Brief Sense of Community Scale (BSCS) to assess the four dimensions of PSOC using a 4-point Likert scale (“not at all” true to “completely” true).²⁷ For neighborhood safety, we used questions aggregated by Mujahid et al^29,30 that measure dimensions of safety and violence. For both BSCS and neighborhood safety, scores were calculated using averaged responses to each question overall and for each dimension.

Our primary outcome of interest was health care utilization (health care visit in the past year). We also described health insurance status and barriers to health care (time without health insurance, lack of care due to unaffordability, or delayed care due to worry about cost in the past year). For WWH, additional outcomes were: (1) missed HIV care visit in the past year (self-reported) and (2) viral nonsuppression, defined as HIV RNA ≥200 copies/mL (using laboratory results collected at the baseline or the 6-month follow-up visit). We additionally described current ART use, ART treatment interruption (off treatment ≥1 week in the past 6 months), ART adherence (self-reported as almost always/always, sometimes/usually, never/rarely), and CD4⁺ lymphocyte count.

We summarized and compared demographic, psychosocial, and housing characteristics by HIV and housing status using t-tests or ANOVA for continuous variables, Wilcoxon rank-sum tests or Kruskal–Wallis tests for non-normally distributed continuous variables, χ² tests for categorical variables, and Fisher exact tests when expected cell counts were <5. Mean values for BSCS, neighborhood safety/violence, and age were summarized by HIV and housing status and compared using Kruskal–Wallis tests. We used χ² and Fisher exact tests to examine the association between housing status and each health outcome, overall and stratified by HIV status. For WWH, we examined the association between housing status and HIV outcomes using the same methods. Finally, we evaluated the association between housing status and health care visit attendance in the past year using a logistic regression model, adjusting for age, race, socioeconomic status, depressive symptoms, and illicit drug use (variables previously associated with housing status and health outcomes^2,17). Among WWH, we evaluated the association between housing status and (1) missed HIV care visits, and (2) viral nonsuppression, using logistic regression models that adjusted for these same variables. Given the associations in the literature between both unstable and dependent housing and adverse health outcomes, along with the sample size limitations based on the relatively small group reporting unstable housing, these categories were analyzed separately in unadjusted analyses but combined in for the multivariable logistic regression models. Statistical analysis was performed in SAS v9.4 with α = 0.05 for all analyses.

Among 759 women, the mean age was 36 years, and 74.9% self-identified as non-Hispanic Black (Table 1). Most were unmarried/unpartnered (71.1%) and were not currently/recently pregnant (77.1%). Approximately half were unemployed (50.1%) and reported their annual household income as >$24,000 (57.3%). More than one-third (34.5%) reported a history of incarceration and 9.3% reported transactional sex. Overall, 30.4% reported currently using tobacco, 13.6% reported hazardous drinking, 44.9% reported marijuana use, and 11.2% reported illicit drug use in the past year; 36.3% reported depressive symptoms (CES-D ≥16).

Overall, 90.7% of participants reported having a health care visit in the past year (see Table 1, Supplemental Digital Content, http://links.lww.com/QAI/C541). Most (96.1%) reported having health insurance. However, 15.5% reported time in the past year with no health insurance and 13.5% reported not getting care because of unaffordability in the past year.

Among WWH, 96.6% reported a HIV care visit in the past year. Nearly all (98.5%) endorsed ART use, and 87.3% reported that they “almost always/always” take their prescribed HIV medications. Overall, 91.9% of participants had a CD4 count >200 c/mL and 80.5% were virally suppressed; however, 70 of 473 had missing CD4 data and 227 of 473 had missing HIV RNA data.

Overall, 64.4% of women had stable housing, 28.9% had dependent housing, and 6.9% had unstable housing; this did not significantly vary by HIV status (Table 2, P = 0.2995). Among dependently housed participants, 13.6% lived in their parents’ house and 15.3% lived in someone else’s house/apartment. Among unstably housed women, the most common housing reported was living in a shelter/welfare hotel (41.2%). On average, women reported living with 1 adult and taking care of 1 child. In total, 26.6% of participants reported living in subsidized housing and 12.6% reported being on a waiting list for subsidized housing. WWH reported significantly higher rates of subsidized housing and status on a waiting list than WWoH (P < 0.0001). Overall, 40.6% reported living at their current address for less than 5 years.

Age, marital status, employment, annual household income, socioeconomic status, incarceration history, history of transactional sex, depressive symptoms, hazardous drinking, tobacco use, and drug use varied by housing status (all P < 0.05, Table 3 and Table 2, Supplemental Digital Content, http://links.lww.com/QAI/C541).

The average response for all 8 items on the BSCS was 2.3 (SD = 0.8), indicating that the average participant felt a moderately low sense of community in their neighborhood. The mean values for each dimension of the BSCS included 2.5 (SD = 1.0) for needs fulfillment, 2.4 (SD = 1.1) for group membership, 2.0 (SD = 1.0) for influence, and 2.3 (SD = 1.0) for emotional connection; these did not differ by HIV or housing status (see Table 3, Supplemental Digital Content, http://links.lww.com/QAI/C541).

The mean score for neighborhood safety was 2.6 (SD = 1.2), indicating that the average participant either felt neutral or agreed with positive statements about safety in their neighborhood. The mean score for neighborhood violence was 3.4 (SD = 0.8), indicating that the average participant rarely experienced fights, gang fights, sexual assault/rape, or robbery/mugging in their neighborhood during the past 6 months, and this did not differ by HIV or housing status.

There were significant associations between housing category and health care utilization and barriers. A lower proportion of unstably housed participants had seen a health care provider in the past year (82.4%) compared with dependently (89.4%) and stably (92.3%) housed participants (Table 3). Unstably housed participants reported time in the past year with no health insurance (30.0%), not getting care because of unaffordability in the past year (29.4%), and delayed seeking of care because of worry about cost in the past year (23.5%) more frequently than dependently or stably housed participants (Table 3).

In the logistic regression model, women who were unstably or dependently housed were significantly less likely to have a health care visit in the past year (odds ratio [OR] 0.44, 95% confidence interval [CI]: 0.21 to 0.96); this association persisted after adjusting for sociodemographic and psychosocial variables (OR 0.40, Cl: 0.18 to 0.90, Table 4).

When stratifying by HIV status (see Table 4, Supplemental Digital Content, http://links.lww.com/QAI/C541), lack of care because of unaffordability in the past year was the only health care barrier that varied significantly by housing status among WWoH; unstably housed WWoH reported lack of care because of unaffordability more than dependently and stably housed participants. Among WWH, lack of care because of unaffordability in the past year and delayed seeking of care because of worry about cost in the past year varied significantly by housing status (all P < 0.05). Unstably housed WWH reported time in the past year with no health insurance more than dependently and stably housed WWH as well as lack of care because of unaffordability in the past year more than dependently and stably housed WWH. Unstably and dependently housed WWH reported similar rates of delayed seeking of care that were higher than stably housed WWH.

A higher proportion of unstably housed WWH missed at least 1 HIV visit in the past year (30.3%) compared with dependently (18.8%) or stably (14.5%) housed participants. In addition, viral suppression was lower among WWH who were unstably housed (57.1%) compared with women who were dependently (83.3%) or stably housed (82.2%). Adherence to HIV medications, status off-treatment ≥1 week in the past 6 months, CD4 cell count, and viral suppression also significantly varied by housing status (all P < 0.05, Table 3). Unstably housed WWH were less likely to report “almost always/always” taking their prescribed HIV medicines (78.8%) than dependently (94.3%) and stably (85.3%) housed WWH. They were also more likely to report being off treatment in the past 6 months (48.5%) than dependently (29.0%) and stably housed (25.1%) WWH. CD4 <200 c/mL was more common among unstably (25.8%) versus dependently (6.6%) and stably housed (6.7%) WWH.

In a logistic regression model, WWH who were unstably or dependently housed were significantly more likely to have a missed HIV care visit in the past year (OR 2.43, CI: 1.11 to 5.34, Table 4); this association persisted after adjusting for sociodemographic and psychosocial variables (OR 2.35, CI: 1.02–5.41). WWH who were unstably or dependently housed were also significantly more likely to have viral nonsuppression (OR 3.58, CI: 1.41 to 9.07), which persisted after adjusting for sociodemographic and psychosocial variables (OR 2.92, Cl: 1.08 to 7.87).