The Effect of Homelessness on the HIV Care Continuum in an Underserved Metropolitan Area of the South : Potential Implications for Ending the HIV Epidemic in America

Vladimir Berthaud (  vberthaud@mmc.edu ) Meharry Medical College https://orcid.org/0000-0001-5585-4853 Livette Johnson Meharry Medical College Ronda Jennings Meharry Medical College Maxine Chandler-Auguste Meharry Medical College Abosede Osijo Meharry Medical College Marie Baldwin Meharry Medical College Paul Juarez Meharry Medical College Patricia Matthews-Juarez Meharry Medical College Derek Wilus Meharry Medical College Mohammad Tabatabai Meharry Medical College

partnerships where the index participants had <350 CD4 cells/μL, but unmeasured confounders remain a signi cant issue in these cohort studies. The prospective, observational PARTNER (Partners of People on ART-A New Evaluation of the Risks) study was conducted at 75 clinical sites in 14 European countries. It enrolled 1,166 HIV serodifferent couples (HIV-positive partner taking suppressive ART) who reported condomless sex (September 2010 to May 2014). The study population included 548 heterosexual (61.7%) and 340 MSM (38.3%). Eligibility criteria for inclusion of couples' years of follow-up were condomless sex and plasma viral load less than 200 copies/mL. In the preliminary results, among serodifferent heterosexual and same-sex couples in which the HIV-positive partner was using suppressive ART and who reported condomless sex, there were no documented cases of within-couple HIV transmission during median follow-up of 1.3 years per couple (upper 95%con dence limit, 0.30/100 couples' years of follow-up) 4 . For same-sex couples, the nal results of the PARTNER study provide a similar level of evidence on viral suppression and HIV transmission risk as previously reported for heterosexual couples. These ndings suggest that the risk of HIV transmission is indeed zero in same-sex couples through condomless sex when the HIV-seropositive partner maintains HIV viral load suppression 5 .
The more recent implementation of pre-exposure and post-exposure HIV prophylaxis (PrEP and PEP) in standard practice has expanded HIV prevention tools for heterosexual and same-sex partners. Large randomized controlled international clinical trials have established the substantive e cacy of PrEP at preventing infection in high-risk heterosexual men and women, as well as MSM. A randomized, doubleblind, placebo-controlled trial enrolled 4,410 HIV-1 partners of heterosexual serodifferent couples in Kenya and Uganda in a three-arm study of daily oral intervention including tenofovir and tenofovir/emtricitabine. The combined results showed that adherence as measured by detection of tenofovir in plasma samples was associated with an 85% relative risk reduction in HIV-1 transmission for the tenofovir arm and 93% for the tenofovir/emtricitabine arm (p<0.001) 8,9 . Based on the model proposed in the Preexposure Prophylaxis Initiative (iPrEx) study of 2,499 MSM couples, two doses of tenofovir disoproxil fumarate plus emtricitabine per week correspond to 76% protection and 4 doses per week to 96% protection against HIV infection.
The drastic reduction of vertical transmission also illustrates this principle. As the result of scaled up prevention of maternal-to-child transmission interventions, the United States has prevented an estimated 21,956 (95% CI : 20,191-23,759) cases of vertical transmissions since 1994 6  "Ending the HIV Epidemic: A Plan for America" and The National HIV/AIDS Strategy. government has developed the "Ending the HIV Epidemic: A Plan for America," spearheaded by the White House and aiming "to achieve the important goal of reducing new HIV infections to less than 3,000 per year by 2030." Started in 2020, the initial phase targets localities with the highest incidence of HIV transmission. The Plan for America promotes and implements four pillars to reduce HIV transmissions substantially: Diagnose, Treat, Prevent, and Respond 10 . However, the Plan for America does not speci cally address social determinants of health.
From another perspective, the 5-year plan of the National HIV/AIDS Strategy 2020 focuses on four overarching goals: reduce new HIV infections; increase access to care and improve health outcomes for PLWH; reduce HIV-related health disparities and health inequities; and achieve a more coordinated national response 11 . The HIV care continuum illustrates program performance at different stages that include HIV diagnosis, linkage to care, retention in care, and viral suppression, the key performance measure 12 . Incomplete viral suppression could increase the risk of HIV transmission and jeopardize "Ending the HIV Epidemic: A Plan for America." Homelessness may erode outcome performance throughout the four stages of HIV care and hinder the four goals of the National HIV/AIDS Strategy in underserved communities.
Homelessness and health outcomes. The  Moreover, HIV-seropositive persons are disproportionately overrepresented among the homeless population. While the homeless population continues to grow exponentially, the National Alliance to End Homelessness estimated that HIV-seropositive persons made up 3.4% of homeless population in the United States in 2007, while they represented 0.4% of the general population. Later, in 2016, the CDC reported that 8.4% of people in HIV medical care were homeless 16 . To examine the effect of housing status on different health outcomes, Chad Leaver and his group undertook a systematic review of 29 studies. Their results suggest that housing instability represents a strong predictor of non-adherence to antiretroviral therapy, HIV risk behaviors, utilization of health and social services, and health status 17 . A second systematic review of 152 studies involving 139,757 participants found substantial evidence that inadequate housing represents a signi cant barrier to reduction of HIV risk behaviors, access to HIV medical care, treatment adherence, and sustained viral suppression 18 . A third qualitative review of 17 published papers examining the effect of homelessness on health status, HIV treatment adherence, and health outcomes showed that homelessness is highly prevalent among PLWHA and strongly associated with poorer health status, lower adherence to antiretroviral therapy, and worse CD4 cell count and viral load outcomes 19 . Homelessness in Nashville, Tennessee.
During the past ve years, in Nashville, Tennessee, the trend of homelessness paralleled the economic growth curve that propelled a booming housing market and aggressive gentri cation. Nashville o cials located approximately 2,300 homeless people on an overnight count in 2017. However, local advocates estimate the number of people in Nashville living on the streets, in cars, camps, motels or in shelters at 20,000 20 . As one of the fastest growing cities in the United States, Nashville reported 30,000 new residents in 2018, while average rents increased from $882 in 2013 to $1,428 today citywide in 2018. One year later, in the aftermath of the March 3, 2020 devastating tornado and the ongoing coronavirus pandemic, the number of homeless in Nashville is certainly much higher. The majority of the patients attending Meharry Community Wellness Center (MCWC) resides in North Nashville, an area directly hit by the tornado. Subsequently, many of them lost their jobs in places such as hotels, restaurants, and factories due to local and state government-imposed "Stay-at-Home" and social distancing orders related to mitigation of the COVID-19 pandemic. The national economic downturn could exacerbate the homelessness crisis in the United States and widen the gaps in care and HIV health disparities.

Methods
Study setting. Exclusion criteria. From this study analysis, we excluded prison inmates of Tennessee Department of Corrections, seen through our telemedicine program and jail inmates who had face-to-face clinic encounters.
Data sources. Data elements extracted from CAREWare electronic patients' charts included the following: age group, race/ethnicity, gender, HIV risk factor, federal poverty level (FPL), type of medical insurance, housing status, rate of clinic visits per month, last recorded CD4 cell count and plasma HIV viral load. MCWC staff entered all data elements in CAREWare at each patient visit. We performed data quality check, for accuracy, duplication, missing, and unknown elements before generating the study database. No personal identi ers were included. Nearly all viral load and CD4 cell count results were imported from our contractual commercial lab web portal and very few of them came from external providers' reports.

Measures.
Primary outcome variable. Viral load suppression, de ned as plasma HIV viral load below the detection limit of 20 copies/mL measured by Real Time PCR assay, constituted the primary outcome variable. Of note, our viral suppression threshold is lower than commonly used 200 copies/mL in published reports.
Predictive variables. We chose a pool of predictive variables from patient's electronic medical records (CAREWare 6, version 58). Demographic variables consisted of age group (18-24, 25-44, 45-64, and 65+), gender (male, female and transgender), and race/ethnicity (Black or African American, non-Hispanic white, and other including Hispanic/Latinx). Variables of social determinants of health comprise HIV risk category (heterosexual, MSM, and/or bisexual men, injection drug use or IDU, and other), housing status (permanently housed and homeless), FPL, and type of medical insurance (no insurance, Medicaid/Medicare, private insurance and other). The homeless category consisted of patients living in the streets, shelters, transition homes, and those in temporary residence. We did not use the classi cation of stable/unstable housing. We entered modi ed adjusted gross income (MAGI) as the percentage of FPL, computed automatically by CAREWare, and we analyzed it as a continuous variable. As a payor of last resort, the Ryan White Program provided ambulatory/outpatient medical insurance coverage to eligible, low-income, uninsured clients. We determined retention in care as the rate of monthly medical visits. The last recorded CD4 cell count and plasma viral load were collected for this study. We treated CD4 cell count and viral load as dichotomous variables (<500 or > 500 cells/mm 3 ) and (≤20 or > 20 copies/mL). We considered plasma viral load ≤20 copies/mL as viral suppression.
Statistical analysis.
First, we performed exploratory data analysis and checked for data quality, distribution, and assumption violations. We applied Pearson's χ2 test to determine the association between pairs of categorical variables. Then, we used a multivariable logistic regression to nd out the effect of homelessness on viral load suppression, controlling for demographic, social, and clinical variables: age group, gender, race/ethnicity, type of medical insurance, FPL, HIV risk factor, rate of outpatient and HRSA visits, and CD4 cell count. Finally, we plotted a receiver operating characteristic (ROC) curve to corroborate the result.

Results
Population characteristics. lived below 100% of the FPL. The hallmark of our patient population was the alarming number of homeless: 192. In fact, this number increased three-fold higher than the national average for HIVseropositive people (8.4%) 16 and twice the average proportion of homeless Ryan White patients nationwide (12.9%) 21 . In the aftermath of the devastating tornado of March 3, 2020 and the ongoing COVID-19 pandemic, the proportion of homeless may skyrocket in year 2020 and beyond. As a payor of last resort, the Ryan White program covered all eligible, low-income, uninsured patients, under either Part B-subsidized Insurance Assistance Program (IAP), or Ryan White Part A Transitional Grant Area or C Early Intervention Services. Thus, the Ryan White programs covered 260 patients (37.6%). Other publicly insured health plans, Medicaid and Medicare, covered 23.1% and 13.4%, respectively. Fourteen patients (2%) were inmates of county jails and excluded from the study. The group of privately insured patients (23.8%) encompassed Affordable Care Act (ACA) customers (14.8%) and a small number of medically insured clients through employer-sponsored health plans (9%). In relation to HIV risk factor, 46% of the patients reported as heterosexual in 2019, a sharp decrease from 62.4% in 2014, partly attributed to the proportional increase in MSM, from 34.8% in 2014 to 42.8% in 2019. This re ects the current epidemic trend in the South. The proportion of patients who considered IDU as their HIV risk category decreased from 11.4% in 2014 to 9.4% in 2019, re ecting the epidemic shift in opioid addiction, from inner cities to suburban and rural areas, and from Black/African American to White/Caucasian communities. This observation correlates with the overall decrease in the number of clients tested with chronic hepatitis C infection from 74% in 2014 to 57% in 2019.

Impact of incarceration on viral suppression.
Many studies have shown that incarceration is associated with poor health outcomes among HIVseropositive adults including greater use of emergency room visits and hospital admissions, and lower prevalence of viral suppression 28,29 . According to the Brookings Institution, North Nashville has an incarceration rate of 14%, the highest in the country by far, and 93% of those incarcerated are Blacks. In other words, one in seven people who were born in the primary zip code of North Nashville between 1980 and 1986 went to jail or prison at some point in their lives 30 . Twenty ve percent of our patients reside in that neighborhood and 41% of them belong to this age group. The surge in unemployment related to the COVID-19 pandemic exacerbates homelessness, depression/anxiety, and drug use and could fuel the cycle of poverty, incarceration and recidivism. These poor outcomes not only impose a nancial burden on the strained healthcare system but also magnify the risk of HIV transmission and jeopardize the plan to end the HIV epidemic and HIV disparities.
Impact of homelessness on mortality.
Homelessness affects not only viral suppression but also the survival of PLWHA and the occurrence of cardiovascular comorbidity and mortality. In a large study of all-cause mortality among 6,558 people living with AIDS in San Francisco, including 641 homeless, 67% of the homeless persons survived ve years compared with 81% of their housed counterparts (p<0.0001). After adjustments for potential confounders, the adjusted relative hazard (aRH) for homelessness was 1.20 (95% CI 1.03-1.41).
Supportive housing lowered the risk of death (aRH 0.20 (95% CI 0.05-0.81)) 31 . A two-year observational study conducted in New York City underlines the signi cant role of temporary pattern of homelessness and sporadic incarceration on higher mortality risk from all-cause, drug-related, and HIV as compared with continuous homelessness and persistent incarceration 32 . Homelessness and incarceration may accelerate HIV disease progression and mortality. A longitudinal study assessing differences in causes of death among housed and homeless people diagnosed with HIV in San Francisco revealed that at the time of death, homeless HIV-seropositive persons tend to be younger, the proportion of AIDS-related mortality tend to decrease, while heart disease and mental disorders are emerging causes of death 33 .
Potential impact of the COVID-19 pandemic on homelessness in Nashville. However, African Americans and Latinx are certainly at a much higher risk for coronavirus disease severity and mortality, as already reported in New York, Michigan, and elsewhere 37,38,39  Summary.
In spite of the different viral suppression thresholds and homelessness nomenclature used, all published studies concur that homelessness is associated with lower prevalence of viral suppression in PLWHA and many of them show a trend in overall worse health outcomes. Several randomized controlled trials have demonstrated substantial bene ts of housing interventions on viral suppression, morbidity, and mortality.
We need to mobilize our resources and create innovative programs that respond to the needs of homeless PLWHA in the era of social distancing related to the COVID-19 pandemic.

Strengths and Limitations.
Strengths of our study include use of a population disproportionately affected by homelessness and HIV/AIDS, adequate sample size and follow-up duration, and collection of data using validated measures. Third, the study does not account for non-infectious comorbidities that could have contributed to higher rates of hospitalizations and emergency room visits, more frequent drug interactions, intolerance, and side effects, leading to sub-optimal treatment adherence and lower rate of viral load suppression, and higher mortality. These critical issues deserve utmost attention because HIV infection may contribute to chronic, sub-clinical in ammation, which promotes the development of cardiovascular and metabolic complications. Notwithstanding the study limitations, multiple researchers have reported the ndings that homelessness may be associated with sub-optimal viral suppression.

Conclusions
The scienti c literature provides substantial evidence of homelessness-related poor health outcomes. Our study has weaknesses inherent in observational data, but we nd it reassuring that our results arrive at the consistent message that homelessness contributes to sub-optimal viral suppression in PLWHA. These ndings should alert public policy makers on the urgent need to mitigate the deleterious consequences of gentri cation and the shortfall of affordable housing on the health outcomes of underserved, low-income HIV-seropositive individuals residing in large metropolitan areas of the United States. Stigmatization and marginalization of homeless individuals and HIV-seropositive persons, implicit bias and discrimination, antiquated housing laws, and unfair criminal justice system deserve renewed attention of federal, state, and local authorities. The United States has already missed the CDC goal to reduce the current percentage of persons receiving HIV medical care who are homeless (8.4%), to no more than 5% in 2020 16 .
The successful implementation of "Ending the HIV Epidemic: A Plan for America" and The National HIV/AIDS Strategy will require a multi-pronged approach to HIV prevention and care, integrated housing services and novel policies and programs. As a public health intervention, housing services align with national medical priorities such as disease prevention and unfettered access to cost-effective and quality health care. In the context of COVID-19 crisis, skyrocketing unemployment, stay-at-home and social distancing measures will likely propel a surge in anxiety/depression, fear of homelessness, substance use, and poorer treatment adherence, leading to virologic failure, increased risk of cardiovascular events,