Cross-sectional and Longitudinal Association Between Sleep and HIV Prevention and Care Behaviors Among Transgender Women of Color: The TURNNT Cohort Study Authors: Dustin T Duncan, Alexander Furuya, Asa Radix, Adam Whalen, Jenesis Merriman, Denton Callander, Nour Makarem Published: 2026 ABSTRACT AbstractWe investigated the association between sleep health and HIV care and prevention outcomes among transgender women of color (TWOC). We used data from TWOC living in New York City collected from 2020 to 2022. Our exposures of interest were short sleep (sleeping for less than seven hours per night), poor-quality sleep (self-rated quality of sleep as “very bad” or “fairly bad”), and long sleep onset latency (taking at least 30 min to fall asleep). We asked participants about their HIV care and prevention outcomes, including HIV/STI testing, condom use, PrEP use, and HIV viral load suppression. We used Targeted Maximum Likelihood Estimation to estimate the association between sleep and these outcomes and included age, education, income, US-born nativity, and hormone replacement therapy use as potential confounders. Among the 314 participants, 54.5% had short sleep, 35.0% had poor-quality sleep, and 24.8% experienced long sleep onset latency. Among people living with HIV (51.6%), those who had short amounts of sleep were less likely to always use a condom (Adjusted Prevalence Ratio [aPR]: 0.67; 95% CI: 0.51–0.87) and be virally suppressed (aPR: 0.87; 95% CI: 0.79–0.96), and those who had poor-quality sleep were less likely to be virally suppressed (aPR: 0.83; 95% CI: 0.72–0.96). Among people living without HIV (46.2%), experiencing long sleep onset latency decreased the likelihood of always using condoms (aPR: 0.36; 95% CI: 0.15–0.85). Improving sleep health among TWOC could improve HIV care and prevention outcomes and reduce health inequities.Keywords: Gender minority, Sleep health, HIV/AIDS, Epidemiology Introduction Sexual minority populations (e.g., gay, lesbian, or bisexual individuals) and gender minority populations (e.g., transgender individuals) may experience disproportionately more sleep issues due to minority stress and gender minority stress—particularly for transgender individuals [1, 2]. Despite extant inequities in sleep factors by sexual orientation and gender identity [3, 4], sleep remains understudied among sexual and gender minority (SGM) populations [5-7]. The limited existing research on sleep among transgender populations demonstrates that sleep health is a public health priority for the population. For example, a study of transgender women and transgender men in Germany identified poor sleep quality as the strongest predictor of poor quality of life, with about 80% of respondents across both groups reporting a sleep score reflective of poor sleep quality [8]. A New York City study found that 54.8% of transgender women of color reported having shorter sleep duration (i.e., less than 7 h per night), and 29.0% of them viewed their sleep quality as poor (i.e., “fairly bad” or “very bad”) [5]. Insufficient or poor-quality sleep may play a significant role in determining both physical health and the effectiveness of preventive health behaviors. In recent years, researchers have begun to investigate how short sleep contributes to a spectrum of HIV-related health outcomes [9-11]. A handful of studies have shown that sleep has been associated with HIV risk behaviors, including in SGM populations [10, 12, 13]. For example, two studies by Duncan et al. and Millar et al. examined associations between poor sleep health and HIV-related risk behaviors among men who have sex with men (MSM). Duncan et al. conducted a cross-sectional survey with 202 MSM in London, finding that poor sleep was significantly associated with condomless receptive anal intercourse with more partners [12]. Millar et al. cross-sectionally surveyed 559 MSM in Paris and found that indicators such as poor sleep quality and long sleep onset latency were linked to increased numbers of condomless anal intercourse partners and substance use before or during sex [13]. Both studies suggest that sleep health is an important but underrecognized factor in HIV prevention efforts among MSM. Additionally, Sari et al. cross-sectionally found that poor sleep quality contributed to higher sexual risk-taking in a pathway mediated by substance use among LGBTQ+ individuals [14]. Taken together, these findings suggest that sleep may serve as a potential biobehavioral pathway linking structural inequalities to health, whereby structural determinants, such as intersectional trauma, may influence HIV-related risk through sleep-mediated mechanisms [15]. However, none of these cross-sectional studies included or analyzed data specific to transgender populations. The literature’s current focus, therefore, limits the generalizability of findings to other SGM populations and overlooks a key population disproportionately affected by HIV and sleep-related health inequities. Our aim was to examine the cross-sectional and longitudinal association between sleep (i.e., sleep quality, sleep duration, and sleep onset latency) and HIV prevention and care behaviors (i.e., HIV testing, condom usage, PrEP adherence, and viral suppression) among transgender women of color in New York City. We hypothesized that various sleep traits (e.g., sleep quality and sleep duration) would be associated with HIV prevention and care behaviors in the sample due to mechanisms involving stress and the fact that poor sleep reduces self-regulation and self-efficacy, including those related to decision-making [16, 17]. Additionally, poor sleep has been shown to have numerous cognitive effects, including reduced cognitive function, attention, and learning, which could impact medication adherence [18]. Methods Study Design We used longitudinal data from 314 transgender women of color from the Trying to Understand Relationships, Networks and Neighborhoods Among Transgender Women of Color (TURNNT) Cohort Study [19]. The research team recruited participants from August 31, 2020, to November 4, 2022. The study was approved by the Institutional Review Board at Columbia University Irving Medical Center (IRB-AAAS8164); the study protocol can be found elsewhere [20]. At the initial enrollment interview, a consent form describing the study background, requirements, timeline, confidentiality procedures, and participant rights was reviewed with interested individuals. Participants gave consent voluntarily by agreeing to the terms of the study and signing the consent form. Participants were also asked to sign an optional release of information form to allow for the sharing of sexually transmitted infection (STI) and HIV testing information with the study team. The TURNNT cohort and study design were developed in close collaboration with community leaders from the transgender women of color community and service organizations that support transgender health [19]. We applied the Health Equity Research Production Model by including transgender community members in our research and authorship team, as well as people of color and others from marginalized communities [21]. Since the inception of the study, we have centered community leaders from the transgender women of color community and service organizations that support transgender health in designing the TURNNT study. From these collaborative efforts, the TURNNT study team and members of the transgender women of color community developed a comprehensive survey together. Of note, we engaged community members in developing the measures used for the present study. After completion of the manuscript draft, we shared the paper with all of the members of the research team to aid in the interpretation of the results. Participants We recruited participants using social media marketing, print advertising, event-based recruitment, referrals from organizational outreach and partnership, and snowball sampling via referrals from enrolled participants. Eligible participants (1) identified as transgender women of color, (2) were between 18 and 55 years old at enrollment, (3) lived in the New York City metropolitan area (including northern New Jersey, Long Island, and the lower Hudson Valley), and (4) spoke either English or Spanish. The study sample comprised both individuals with and without HIV [22]. Procedures After providing informed consent, participants of the TURNNT Cohort Study were asked to respond to an extensive survey that aimed to collect information on housing, perceptions of neighborhood characteristics, stigma and discrimination, PrEP use, gender affirming processes, sexual practices, sex work, drug use, and social capital. The research team initially interviewed the participants in person; however, due to the COVID-19 epidemic, we pivoted and interviewed participants over Zoom. There were three waves of data collection: baseline (wave 1), six-month follow-up (wave 2), and 12-month follow-up (wave 3). For this study, we used data collected at baseline to measure the exposure, outcome, and covariates; we also used outcome data collected at the six-month follow-up. We used questions from the Pittsburgh Sleep Quality Index (PSQI) to assess aspects of sleep health. The sleep exposures for the present analysis were sleep duration, sleep quality, and sleep onset latency; sleep duration and sleep quality were measured at wave 1 and wave 2; sleep onset latency was only measured at wave 2. For sleep duration, participants reported a numerical response to the following question: “In the past month, on average, how many hours of actual sleep did you get at night? This may be different from the number of hours you spent in bed. If you are unsure, please estimate.” Based on their responses, participants were dichotomized into sufficient sleepers (7 or more hours) vs. short sleepers (less than 7 h). For sleep quality, the response to the following question was used: “During the past month, how would you rate your sleep quality overall?” Participants could respond, “Very good,” “Fairly good,” “Fairly bad,” and “Very bad, and those who responded, “Very good” and “Fairly good” were classified as having good sleep quality, whereas those who responded, “Fairly bad” and “Very bad” were classified as having poor sleep quality. To ascertain sleep onset latency, responses to the following question were used: “During the past month, how often have you not been able to fall asleep within 30 min?” Those who reported taking longer than 30 min to fall asleep three or more times a week in the past month were classified as having long sleep onset latency; this cutoff was selected as it was the most similar to the International Classification of Sleep Disorders definition of insomnia disorder [23]. We assessed several HIV prevention and care behaviors as our outcomes of interest. For consistent condom use, we asked participants if they always used condoms during penetrative sex in the past 6 months. For STI testing that did not include HIV testing, we asked all participants if they had obtained an STI other than an HIV test in the past year. For HIV testing, we asked the subset of those who were seronegative if they had obtained an HIV test in the past year. For current PrEP use, we asked those who were seronegative and sexually active if they were currently taking PrEP. Finally, for HIV viral load suppression, we asked those with HIV whether they currently had a detectable or undetectable HIV viral load according to their most recent viral load test. We measured sociodemographic characteristics of our study population at the baseline survey. We defined education according to a three-level categorical variable: less than a high school education, a high school diploma or general education diploma, and greater than a high school education. Additional data were collected on baseline age (categorized as 18–24 years, 25–34 years, 35–44 years, and 45–55 years), individual income (no income, $1 to $9,999, $10,000 to $29,999, $30,000 to $49,999, ≥ $50,000, or missing), US-nativity (US-born or foreign-born), and hormone replacement therapy (ever use or never use). Statistical Analysis We conducted descriptive analyses to characterize the sociodemographic composition of the study sample as well as the distribution of the exposure and outcome. To estimate the association between sleep health and HIV care and prevention outcomes, we used a doubly robust method, Targeted Maximum Likelihood Estimation (TMLE). Using doubly robust methods accounts for confounding by modeling both the exposure model and the outcome model; even if one is incorrectly specified, the algorithm can still get an unbiased estimate of the association [24]. Furthermore, TMLE is apt for machine learning approaches and can therefore incorporate super learner algorithms to further reduce bias arising from model misspecification. To conduct TMLE analysis, we used the LTMLE package in R, and we took a complete case approach for our analyses [25]. All quantitative analyses were conducted using R/R Studio (Version 2024.09.1, + 394). For the cross-sectional analyses, we used outcome data measured at the same wave as the exposure; for sleep duration and sleep quality, we used data from Wave 1, and for sleep onset latency, we used data from Wave 2. For the longitudinal analyses, we used exposure data measured at Wave 1 and outcome data measured at Wave 2; because sleep onset latency was measured only at Wave 2, we did not analyze it for the longitudinal analyses. When we estimated these associations, we accounted for age, education, income, US-nativity, and hormone replacement therapy. For the longitudinal analysis, we also accounted for the prior outcomes, which were measured at the same time as the exposure (Wave 1). Long sleep (more than 10 h) can be detrimental to health [26]; however, only a few participants in our study self-reported being long sleepers. As a sensitivity analysis, we repeated all aforementioned analyses of sleep duration, excluding long sleepers, to assess whether the results were meaningfully different. Results In total, 314 participants were included in this analysis. Table 1 details the characteristics of the study population. More than half (54.5%) had short sleep, 35.0% had poor-quality sleep, and 24.8% experienced long sleep onset latency in the past month. More than half were people with HIV (51.6%). We found that 87.6% reported being sexually active, and among those sexually active, 50.1% reported always using condoms during sex. Among the study population, 78.7% of the participants took an STI test within the past 6 months. Among seronegative individuals, 87.6% reported having taken an HIV test in the past 6 months, and 42.8% were currently on PrEP. Among people with HIV, 87.7% reported having an undetectable viral load. Table 1Characteristics of the study population, the TURNNT cohort studyCharacteristicProportion (n = 314) (%)Age 18 to 2410.2 to 3429.6 35 to 4437.9 45 to 5521.0 Missing1.3Education Less than High School33.4 High School28.3 More than High School38.3US-Born Yes58.3 No41.1 Missing0.6Income No Income16.2 $1 to $9,99939.2 $10,000 to $29,99922.0 $ 30,000 to $49,9997.0 $50,000 or more4.8 Missing10.8Hormone Replacement Therapy Use Yes88.2 No10.5 Missing1.3Sleep Duration Long Sleep (≥ 7 h)45.5 Short Sleep (< 7 h)54.5 Missing0.0Sleep Quality Very Good18.1 Fairly Good45.9 Fairly Bad24.5 Very Bad10.5 Missing1.0Long Sleep Onset Latencya Not during the past month43.7 More than once a month16.2 Once or twice a week14.4 Three or more times a week24.8 Missing0.9HIV Status Positive51.6 Negative46.2 Missing2.2Past 6 Month STI (non-HIV) Testing Yes78.7 No21.0 Missing0.3Always Condom Use Yes43.9 No43.7 Missing or Not Applicable12.4Past 6 Month HIV Testingb Yes87.6 No12.4 Missing0.0Current Pre-Exposure Prophylaxis Useb Yes42.8 No53.8 Missing3.4Current Undetectable Viral Loadb Yes87.7 No9.3 Missing3.0This table describes the sociodemographic characteristics of the study population, as well as the distribution of sleep health variables and HIV care and prevention outcomesaLong sleep onset latency was first measured at six-month follow-up (n = 222)bAmong people living with HIVcAmong people not living with HIV Characteristics of the study population, the TURNNT cohort study This table describes the sociodemographic characteristics of the study population, as well as the distribution of sleep health variables and HIV care and prevention outcomes Long sleep onset latency was first measured at six-month follow-up (n = 222) Among people living with HIV Among people not living with HIV The results of the TMLE analysis can be seen in Tables 2, 3, 4 and 5. TMLE calculates the statistical estimand of interest (such as the prevalence ratio or relative risk) and calculates standard errors for the confidence interval. When the null value for the measure of association (i.e., prevalence ratio equals 1.0 or relative risk equals 1.0) is not within the confidence interval, then we consider the association to be statistically significant. Here, we report on associations found to be statistically significant. In terms of cross-sectional associations, we found that, among people with HIV, short sleepers were less likely to always use a condom (Adjusted Prevalence Ratio [aPR]: 0.67; 95% CI: 0.51–0.87), compared to those who reported sufficient sleep. Short sleepers, compared to those who reported sufficient sleep, were also less likely to be virally suppressed (aPR: 0.87; 95% CI: 0.79–0.96). Those who had poor-quality sleep were less likely to be virally suppressed (aPR: 0.83; 95% CI: 0.72–0.96) compared to those who had good-quality sleep. Among people without HIV, experiencing long sleep onset latency was inversely associated with consistent condom use (aPR: 0.36, 95% CI: 0.15–0.85). Table 2Cross-sectional association between sleep health and HIV care and prevention outcomes among people living with HIVAlwaysCondomUsePast 6MonthsSTITestingUndetectableViralLoadSufficient SleepRef.Ref.Ref.Ref.Short SleepAdjusted Prevalence Ratio [95% Confidence Interval] 0.67 [0.51, 0.87] 1.02 [0.87, 1.19] 0.87 [0.79, 0.96] Good Quality SleepRef.Ref.Ref.Ref.Poor Quality SleepAdjusted Prevalence Ratio [95% Confidence Interval]0.99 [0.77, 1.28]0.99 [0.84, 1.18] 0.83 [0.72, 0.96] Normal Sleep Onset LatencyRef.Ref.Ref.Ref.Long Sleep Onset LatencyAdjusted Prevalence Ratio [95% Confidence Interval]1.02 [0.70, 1.50]1.00 [0.83, 1.21]0.98 [0.87, 1.12]Using doubly-robust methods, we estimated the prevalence ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, and hormone replacement therapy. Bolded cells indicate statistically significant associations Cross-sectional association between sleep health and HIV care and prevention outcomes among people living with HIV Using doubly-robust methods, we estimated the prevalence ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, and hormone replacement therapy. Bolded cells indicate statistically significant associations Table 3Cross-sectional association between sleep health and HIV care and prevention outcomes among people not living with HIVAlways CondomUsePast 6MonthsSTITestingCurrentPrEPUsePast 6MonthsHIVTestingSufficient SleepRef.Ref.Ref.Ref.Ref.Short SleepAdjusted Prevalence Ratio [95% Confidence Interval]1.38 [0.92, 2.06]1.03 [0.89, 1.19]0.84 [0.58, 1.20]0.95 [0.87, 1.04]Good Quality SleepRef.Ref.Ref.Ref.Ref.Poor Quality SleepAdjusted Prevalence Ratio [95% Confidence Interval]1.33 [0.89, 1.96]1.06 [0.91, 1.23]1.24 [0.86, 1.79]0.99 [0.90, 1.09]Normal Sleep Onset LatencyRef.Ref.Ref.Ref.Ref.Long Sleep Onset LatencyAdjusted Prevalence Ratio [95% Confidence Interval] 0.36 [0.15, 0.85] 1.02 [0.83, 1.26]0.75 [0.45, 1.24]0.99 [0.85, 1.16]Using doubly-robust methods, we estimated the prevalence ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people not living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, and hormone replacement therapy. Bolded cell indicates statistically significant association Cross-sectional association between sleep health and HIV care and prevention outcomes among people not living with HIV Using doubly-robust methods, we estimated the prevalence ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people not living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, and hormone replacement therapy. Bolded cell indicates statistically significant association Table 4Longitudinal effect of sleep health on care and prevention outcomes among people living with HIVAlwaysCondomUsePast 6Months STITestingUndetectableViralLoadSufficient SleepRef.Ref.Ref.Ref.Short SleepAdjusted Risk Ratio [95% Confidence Interval]0.99 [0.71, 1.36]0.96 [0.83, 1.11]1.04 [0.95, 1.13]Good Quality SleepRef.Ref.Ref.Ref.Poor Quality SleepAdjusted Risk Ratio [95% Confidence Interval]1.01 [0.75, 1.36]1.01 [0.87, 1.17]1.01 [0.95, 1.08]Using doubly-robust methods, we estimated the risk ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, hormone replacement therapy, and prior outcome Longitudinal effect of sleep health on care and prevention outcomes among people living with HIV Using doubly-robust methods, we estimated the risk ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, hormone replacement therapy, and prior outcome Table 5Longitudinal effect of sleep health on care and prevention outcomes among people not living with HIVAlwaysCondomUsePast 6MonthsSTITestingCurrentPrEPUsePast 6MonthsHIVTestingSufficientRef.Ref.Ref.Ref.Ref.Sleep Short SleepAdjusted Risk Ratio [95% Confidence Interval]0.96 [0.60, 1.54]1.19 [0.98, 1.45]1.32 [0.91, 1.91]1.02 [0.88, 1.18]Good Quality SleepRef.Ref.Ref.Ref.Ref.Poor Quality SleepAdjusted Risk Ratio [95% Confidence Interval]0.86 [0.52, 1.44]1.05 [0.87, 1.27] 1.51 [1.07, 2.14] 1.03 [0.89, 1.19]Using doubly-robust methods, we estimated the risk ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people not living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, hormone replacement therapy, and prior outcome. Bolded cell indicates statistically significant association Longitudinal effect of sleep health on care and prevention outcomes among people not living with HIV Using doubly-robust methods, we estimated the risk ratio for the effect of sleep health measurements on HIV care and prevention outcomes among people not living with HIV. We also calculated the 95% confidence intervals by estimating standard errors using the doubly-robust methods. We adjusted for age, education, income, US-nativity, hormone replacement therapy, and prior outcome. Bolded cell indicates statistically significant association In the longitudinal analyses, among those without HIV, poor quality sleep was associated with a greater likelihood of reporting current PrEP use (Adjusted Risk Ratio: 1.51; 95% CI: 1.07–2.14). We did not find any statistically significant longitudinal association between sleep health and HIV care and prevention outcomes among people with HIV. For analyses involving sleep duration, we excluded long sleepers (more than 10 h; n = 6) as a sensitivity analysis. We found that all of the results were similar when excluding long sleepers. Discussion This study contributes to the research scholarship by assessing the effects of sleep health on HIV prevention and care behaviors of transgender women of color. In our study among transgender women of color in New York City in the TURNNT cohort, we show that more than half reported less than seven hours of sleep (54.5%), 35.0% reported having poor-quality sleep, and a quarter (24.8%) experienced long sleep onset latency in the past month. Generally, transgender women of color who experienced worse sleep health were more likely to have negative HIV care and prevention outcomes, though we found some nuanced relationships from our analyses. Several potential mechanisms may provide an explanation for our results, which show that there is a connection between sleep health and HIV care and prevention outcomes. Insufficient sleep is associated with prefrontal cortex dysregulation, which itself is associated with impaired decision-making and greater risk-taking behavior in the general population [27, 28]. Transgender women of color may be less likely to engage in protective HIV behaviors after receiving insufficient sleep due to reduced executive function, lower impulse control, and impaired emotional regulation; this can amplify the stress that minoritized individuals already encounter [28]. Sleep loss can also affect memory and attention, suggesting that transgender women of color who may typically engage in preventative behaviors, such as PrEP adherence or condom use, may forget to do so, or feel emotionally incapable of negotiating condom use with a partner after receiving short sleep. Additionally, there is a bidirectional and cyclical relationship between sleep and mental health; poor sleep can exacerbate mental health problems, which can, in turn, degrade sleep [29]. This is particularly concerning in the context of HIV prevention, as poor mental health has previously been found to be strongly associated with reduced adherence to preventative behavioral interventions, though interventions that incorporate a mental health component have been found to be effective in improving adherence and related outcomes [30]. Therefore, a holistic approach to HIV prevention should likely include mental health, as well as sleep health components, to improve adherence and reduce transmission or disease progression. We posit that sleep functions as a key biobehavioral pathway through which structural inequalities become biologically embodied. Structural factors (e.g., intersectional trauma) resulting from systems of oppression can disrupt sleep health, which can, in turn, shape downstream outcomes such as HIV care and prevention. Our findings are in part in line with past research on sleep and HIV risk behaviors. For example, Duncan et al. found that several poor sleep health variables were independently associated with condomless anal intercourse among a sample of MSM [12]. For example, a typical nightly sleep duration of less than 7 h was associated with condomless receptive anal intercourse with a higher number of sexual partners [12]. In bivariate analyses, Millar et al. found that all four sleep variables (i.e., sleep quality, sleep duration, problems falling asleep, and problems staying awake during wake-time activities) were associated with the three condomless anal intercourse outcomes, whereas poor sleep quality and problems falling asleep were positively associated with using substances before or during sex among a sample of MSM [13]. Most of these associations remained significant when adjusting for various socio-demographic and behavioral covariates. Additionally, our results regarding the association between sleep and HIV viral load suppression are consistent with what Borker et al. found; shorter sleep times were associated with lower CD4 cell counts [31]. However, while some studies found positive associations between worse sleep health and PrEP non-use among MSM samples [32], we found that poor quality sleep was predictive of PrEP use in our sample of transgender women of color. There may be several reasons for this: first, it is possible that those who are under increased stress (i.e., have poor sleep health) are more likely to seek healthcare or have more contact with medical institutions. Second, it is possible that there may be differential misclassification bias, in that those on PrEP may be more engaged and therefore report their sleep outcomes differently. In the future, we hope to understand the mechanisms behind this unexpected association using qualitative research or mixed-methods research. There are several limitations of this study. First, we may have unmeasured confounders that could bias our measures of association. Based on the review of the literature, we assume that we can estimate unbiased measures of association by accounting for the potential outcomes of age, income, education, US-nativity, and hormone replacement therapy [33]. While we adjusted for the aforementioned variables, there may be potential unmeasured confounding that could be biasing our results. Second, our measurements of sleep health and HIV care and prevention outcomes, and the operationalization of these measurements, may introduce measurement error or may not fully capture the complexity of these constructs. For this study, we assessed self-reported sleep and HIV prevention and care behaviors; as such, we are at risk of misclassification and same-source bias, albeit using validated screening tools that are widely recognized across fields of sleep and HIV research literature, including the well-studied PSQI. Several missed opportunities prevented us from certain analyses; for example, we did not measure sleep latency at Wave 1. Additional sleep outcomes (i.e., sleep medication use and hypersomnia) were available in the six-month study visit questionnaire; however, we decided against pursuing these analyses due to concerns over low power. Furthermore, we note that our operationalization of the sleep health variables may not fully capture this multi-dimensional and complex construct. For example, sleeping for too long (i.e., 10 h or more) could be detrimental to health and could be its own meaningful category within the sleep duration variable [26]. However, only 6 participants were reported to sleep 10 or more hours; because of potential sparse cell counts and concerns about sparse cell bias, we categorized them as having sufficient sleep. As a sensitivity analysis, we re-ran all of the analyses involving sleep duration and excluded long sleepers and found similar results. In future studies, we hope to use objective tools to measure sleep and HIV care and prevention outcomes that can overcome these limitations. Third, our sample may be vulnerable to selection bias due to the convenience sampling methodology, and the results may not be generalizable. Furthermore, it is important to note that we had fewer participants who were Asian, Native Hawaiian/Pacific Islander, Native American, and Middle Eastern/Southwest Asian/North African. Taken together, the experiences of individuals in this sample may not be generalizable to all transgender women of color, all transgender women, or the overall population of New York City or the United States more broadly. We also note that data collection occurred during the COVID-19 pandemic and that the pandemic may have impacted measurements of sleep health, HIV care, and prevention outcomes, or both. These findings highlight the importance of addressing sleep health to help prevent HIV transmission among transgender women of color with HIV, to improve HIV care and prevention outcomes, and to reduce health inequalities. Future studies could explore other aspects of sleep health and determine whether the association may apply to other gender minority populations. Additionally, it would be important to understand whether specific subgroups of transgender women of color are at heightened risk for poor sleep health (e.g., race or ethnicity, socioeconomic status, and younger age), and/or are more susceptible to the effects of poor sleep on mood, cognition, and decision-making. Future research should explore potential mediating mechanisms for the observed associations between sleep and HIV outcomes, such as emotion dysregulation, impulsivity, elevated negative affect, and efforts to enhance positive affect. Acknowledgements We thank the Community Advisory Board members: Cristina Herrera of TransLatinx Network, Ceyenne Doroshow of GLITS (Gay and Lesbian in a Transgender World), Kim Watson of Community Kinship Life, Kiara St. James of New York Transgender Advocacy Group, Bianey Garcia of Make the Road NY, Morticia Godiva of Black Trans Travel Fund, Xoai Pham of Transgender Law Center, Egyptt LaBeija of the House of LaBeija, and Nala Toussaint. We also thank the Scientific Advisory Board members: Sari Reisner, ScD of Harvard Medical School, Tonia Poteat, PhD of the University of North Carolina Chapel Hill, Rachel Bluebond-Langner, MD of New York University School of Medicine, Robert Garofalo, MD of Northwestern University, Walter Bockting, PhD of Columbia University and Jae Sevelius, PhD of Columbia University. We additionally thank TURNNT study staff including: Krish J. Bhatt, MPH, Jessica Contreras, BA, Roberta Scheinmann, MPH, Jenesis Merriman, Magdalena Palavecino, Laura Staeheli, MPH, Kobe Pereira, MPH, Mia N. Campbell, MHS, Elias Preciado, MPH, Astrea Villarroel-Sanchez, and Kevalyn Bharadwaj, MPH. We further thank the following colleagues for their thoughtful input on coding racial and ethnic categories for our data: Renee M. Johnson, PhD, MPH of Johns Hopkins University, Paris “AJ” Adkins-Jackson, PhD of Columbia University, Carlos Rodriguez-Diaz, PhD of The George Washington University, as well as Howard Shih, MSE, and Ryan Vinh from AAPI Data. We also thank the participants for engaging in this research. Funding The Trying to Understand Relationships, Networks and Neighborhoods among Trans women of color (TURNNT) Cohort Study was funded through grants from the National Institute on Minority Health and Health Disparities (Grant Numbers: R01MD013554, 3R01MD013554-02S1 and 3R01MD013554-05S1; Principal Investigator: Dustin T. Duncan, ScD), and the National Institute of Allergy and Infectious Diseases under Grant T32AI114398 (Principal Investigator: Andrea Howard, MD, MS). The work described herein is the sole responsibility of the authors and does not represent the official views of NIH or its institutes. Footnotes Conflict of interest We declare no competing interests. References