Despite recent decreases in HIV incidence in adolescents and young adults, youth remain less likely than older adults to know their HIV diagnosis, link to and be retained in HIV care, receive antiretroviral therapy (ART) and achieve virologic suppression. Sexual transmission of HIV may occur whenever individuals with viremia engage in sex not protected by condoms or whenever partners are not on HIV preexposure prophylaxis (PrEP). Although suppression of the HIV viral load can virtually eliminate the risk of sexual HIV transmission [1,2], recent data from the Centers Disease Control and Prevention (CDC) demonstrate that only 27% of youth living with HIV (YLWH) in the United States achieve virologic suppression . In adults, prior studies have identified a positive association between younger age and secondary transmission risk [4,5]. Age disparities in the HIV care continuum translate into not only poor health outcomes for YLWH but also a greater risk for secondary HIV transmission.
In order to decrease population HIV incidence, there is a critical need to identify modifiable risk factors for secondary HIV transmission among YLWH and target these factors for intervention. Prior data demonstrate the roles that comorbid conditions including depression and substance abuse play in transmission risk [5–8]. However, little data exist examining the potential role of health services, such as insurance, in reducing HIV transmission risk. Insurance coverage has been associated with increased rates of ART prescription, testing and treatment for sexually transmitted infections (STI) and improved retention in care for people living with HIV. However, insurance coverage rates are lower for YLWH compared with older adults . Uninsured YLWH may have limited access to ART, which is critical to achieving HIV suppression. In addition, lack of insurance may lead to delayed diagnosis and treatment of STIs, thereby increasing genital tract inflammation and HIV RNA concentrations and potentiating transmission [10,11].
For YLWH in the United States, coverage for ART and HIV-related medical care may be provided through private insurance, public insurance (Medicaid, Medicare) or HIV-specific limited coverage through the Ryan White HIV/AIDS Programs. Ryan White is a federally funded program that has served as a safety net for low-income uninsured and underinsured individuals for over 25 years. Over 50% of people living with HIV in the United States receive benefits through Ryan White, either indirectly through funding to clinics for subsidized medical care or directly by no-cost ART access through the AIDS Drug Assistance Program (ADAP) . If ART access is a critical driver of transmission, we hypothesized that pharmaceutical-only coverage, as well as traditional insurance, should have a protective effect against the risk of secondary HIV transmission.
Therefore, our primary aims were to identify the proportion of patients meeting criteria for high HIV transmission potential within a cohort of YLWH from 2002 to 2015 and to assess associations between insurance coverage and transmission potential in this cohort.
We conducted a retrospective cohort study of YLWH receiving care at the Children's Hospital of Philadelphia (CHOP) Adolescent Initiative HIV clinic from January 2002 through June 2015.
Participants and setting
The Adolescent Initiative clinic is constituted of a multidisciplinary research and clinical team, providing care to YLWH, aged 12–26 years. Patients are seen for visits approximately every 3 months including medical assessments, HIV RNA quantification, STI screening, ART adherence counseling and intensive medical case management. The clinic's registry contains data from all individuals (n = 436) receiving care during the study period. Individuals were included in the study if they attended at least two visits, were prescribed ART and had at least one plasma HIV RNA quantification after ART start. The current analysis included all visits with plasma HIV RNA quantification after the first documented date of ART in the registry. The study was approved by the CHOP Institutional Review Board, with a waiver of informed consent.
HIV transmission potential
The primary outcome of high HIV transmission potential, was defined as time-concurrent plasma HIV RNA quantification more than 1500 copies/ml (log10 >3.17) and incident bacterial STI. As STIs lead to genital tract inflammation and may increase HIV RNA in genital tract secretions, STI functioned as a biomarker of both condomless intercourse and heightened transmission risk [10,11]. We required both viremia and bacterial STI as the endpoint of interest because STI in the absence of viremia has not been demonstrated to confer an increased transmission risk [2,13]. The HIV RNA threshold was selected based on prior studies demonstrating increased transmission risk in serodiscordant heterosexual couples at this level [14,15]. To avoid bias from self-reported sexual behavior, we used laboratory-confirmed incident bacterial STI as our proxy measure for high-risk sexual behavior. Per clinic protocol, patients were screened at baseline and every 3 months, irrespective of symptoms, for syphilis via rapid plasma regain (RPR) and for gonorrhea and chlamydia by urine, oral and rectal nucleic acid amplification tests (NAAT). Patients may have refused testing at specific anatomic sites based on sexual risk behavior. As the clinic was Ryan White funded, no additional expense was incurred by uninsured individuals for having multiple sites screened or having STI testing more frequently than every 3 months.
Incident bacterial STI was defined per CDC STI treatment guidelines  as any of the following after the baseline visit: positive NAAT of urine, rectal and/or oral fluid for Neisseria gonorrhea or Chlamydia trachomatis, newly positive RPR with confirmatory positive treponemal test, four-fold rise in RPR titer and/or failure to achieve four-fold decrease in RPR titers within 12 months of treatment. We did not include herpes simplex virus (HSV) in the definition of transmission potential, as we were unable in the retrospective data to distinguish incident HSV infection, which would be a marker of condomless sex, from reactivation disease, which may have no relationship to sexual behavior. In addition, most diagnoses of HSV were clinical, and not laboratory-confirmed either at baseline or subsequent visits. For visits wherever STI testing was obtained and HIV RNA was not, we imputed the value for HIV RNA results obtained within 180 days of the STI testing. If HIV RNA assay results were available within both the 180 days before and after the visit, the mean of the two log10 HIV RNA values was imputed to the STI-testing visit.
The primary exposure was insurance coverage, measured at each visit based on appointment registration data. Prior research examining the impact of insurance on HIV-related outcomes has typically defined insurance as public or private insurance and considered ADAP pharmacy-only coverage as uninsured [9,17]. However, in our causal pathway toward transmission potential, we hypothesized that pharmacy-only coverage would operate differently than total lack of coverage (neither medical nor pharmacy benefits) by allowing ongoing access to ART. Therefore, in primary analyses, we defined private insurance, public insurance (Medicaid) and ADAP pharmacy-only coverage as insured and total lack of coverage as uninsured. To test that our findings were robust to varying definitions of coverage, we performed a secondary analysis utilizing a dichotomized insurance coverage definition wherein ADAP pharmacy-only coverage was considered uninsured.
Retention in care
Prior studies have demonstrated associations between insurance and retention, insurance and virologic failure and retention and virologic failure [18,20]. Ryan White pharmacy benefits in Pennsylvania require re-application every 6 months. Therefore, we hypothesized that retention in care would function as a time-dependent confounder in the relationship between insurance and HIV transmission potential by allowing case managers to assist with insurance applications and by its other positive effects on virologic outcomes (e.g. increased medication adherence counseling) [18,19]. We measured retention in care with the time-varying visit adherence ratio, calculated at each visit as the total number of completed visits divided by the total number of scheduled visits to date. Visits were defined as visits with a medical provider, as well as combined laboratory and case management visits. The visit adherence ratio has the strength of capturing missed visits, a known predictor of both virologic failure and mortality in HIV [20,21]. The ratio also accounts for disease severity, as patients with more advanced disease or psychosocial complexity are scheduled at more frequent intervals . As an additional retention measure, we identified presence or absence of gaps in care more than 6 months immediately preceding each visit.
Gender was defined by self-reported gender at the final visit, in order to account for developmental trajectories in which youth may identify gender differently throughout the study period [23,24]. Mode of HIV acquisition was dichotomized as behavioral (via sexual behavior or injection drug use) or nonbehavioral (perinatally acquired). Race was self-reported. Baseline STI status was determined by self-report of STI prior to HIV diagnosis or a laboratory-confirmed STI at the baseline clinic visit.
As mental health has been demonstrated to impact both virologic suppression and transmission risk, we included a covariate in the model for a lifetime history of mental illness [7,8,25]. This definition included self-report of prior mental health treatment at baseline, or chart-documented diagnosis of mental illness or suicide attempt a by a medical provider (MD, nurse practitioner or social worker) during their time in HIV care. In addition, as homelessness may impact engagement in HIV care, insurance status and treatment outcomes , we included a covariate for history of unstable housing after ART initiation. Unstable housing was defined as not having a primary residence, and included staying with friends, transient housing, group home or foster care, incarceration, or homelessness. Whenever mental health and unstable housing variables were not recorded, we included the risk factor as being absent. In order to assess the robustness of this approach, we performed a secondary analysis in which we only included participants for whom these data were complete.
Chi-squared and Wilcoxon rank sum tests compared baseline characteristics and overall retention in care between participants with and without at least one episode of high transmission potential.
To determine the association between insurance status and high HIV transmission potential for all clinic visits, we utilized marginal structural modeling adjusting for baseline covariates and time-varying retention in care [27,28]. To minimize the hypothesized time-dependent confounding among insurance, retention in care and transmission potential, we used an inverse probability treatment-weighted (IPTW) model [27,29]. To estimate the stabilized weights for insurance coverage, we utilized a multivariable logistic regression model examining the probability of participants receiving their observed insurance status at each visit given their time-varying visit adherence ratio, and baseline race, age, sex, transmission route and baseline STI status. We then performed logistic regression analyses on the weighted dataset with clustering by subject to determine the odds of high transmission potential at the visit level by insurance status, adjusting for visit adherence, race, sex, age at and time from ART initiation and baseline STI. We excluded five individuals from the weighted analysis in which ART start date could not be definitively established.
In secondary analyses, we repeated the modelling, utilizing an alternate definition of insurance coverage, wherein ADAP pharmacy-only coverage was considered uninsured. Lastly, as the virologic threshold for transmission through anal intercourse is unknown and may be less than 1500 copies/ml, we performed a sensitivity analysis in which the outcome was defined as co-occurring quantitative HIV RNA greater than 200 copies/ml (log10 >2.03) and incident STI.
Statistical analyses were conducted using the software program Stata 14.1 (StataCorp, College Station, Texas, USA) and statistical significance was set at P < 0.05.
There were 240 participants (55% of total cohort, Fig. 1) who met our inclusion criteria and were followed for a median of 22 months after ART initiation (IQR 8.1–49.0), contributing 2661 visits to the analysis. Seventeen percent (n = 458) of visits had imputed quantitative HIV RNA results because of time discordant testing of HIV RNA and STIs.
Participants were predominately African-American men and transgender women who have sex with men, with a median age at entry to care of 19 years (IQR 17–21; Table 1). Nearly half the participants had an STI at or before entry to care. Participants were insured at 84% (n = 2241) of visits, with the breakdown of coverage type being 18% private, 59% Medicaid, 7% ADAP pharmacy-only coverage and 16% uninsured (no medical or pharmacy benefits). Although 228 (95%) of participants obtained some form of insurance during the study period, loss of coverage was experienced at least once by 65 participants (27%). The median visit adherence ratio was 0.72 (IQR 0.55–0.88), with 28% of participants experiencing at least one gap in care at least 6 months.
With respect to HIV transmission risk, incident STI was detected at least once in 123 (51%) participants after ART initiation, with at least one incident STI occurring at 286 (11%) of visits. There were 144 episodes of chlamydia, 154 episodes of gonorrhea and 37 episodes of incident syphilis. Multisite infection with a single pathogen occurred in 8% of chlamydial infections and 18% of the gonococcal infections. The most common sites of gonococcal infection were oropharyngeal (n = 85) and rectal (n = 75). The most common sites of chlamydial infection were rectal (n = 94) and urethral (n = 41). Multiorganism infection occurred commonly, with n = 48 (17%) of STI episodes involving more than one pathogen. Viremia more than 1500 copies/ml was detected at least once in 110 (46%) participants and 762 (32%) visits during the study period.
High HIV transmission potential occurred at least once in 37 (15%) participants and 3% of all visits (Fig. 1). In the bivariate analysis (Table 1), there were no significant differences between those with or without an episode of high transmission potential with respect to age at entry to care, baseline insurance status, race, gender, lifetime history of mental illness, history of unstable housing or retention in care.
The results of the marginal structural model examining the impact of insurance coverage on transmission potential are displayed in Table 2. After adjusting for race, age at and time from ART initiation, sex and baseline STI status, the odds of high HIV transmission potential was reduced by more than half among participants with insurance coverage (public, private or ADAP-only) compared with those without insurance. Those with a history of STI diagnosed at or before entry to HIV care, however, had more than three-fold higher adjusted odds of high HIV transmission potential during the period in care, whereas retention in care was not associated with high transmission potential. In our sensitivity analysis, wherever missing data was approached as a complete case analysis, there were neither substantive changes in the point estimates or the P values of the associations. The results of the secondary analysis, wherein ADAP pharmacy-only coverage was considered as lack of coverage, are displayed in Table 3. Insurance coverage remained protective against high HIV transmission potential, although the association did not achieve statistical significance.
In our sensitivity analysis using a lower virologic threshold for high transmission potential, 47 participants (19%) met the definition of high transmission potential at least once in the study period. Insurance coverage remained protective against high HIV transmission potential [adjusted odds ratio (aOR) 0.55, 95% confidence interval (CI) 0.30–0.99] and baseline STI status remained highly associated with high HIV transmission potential (aOR 3.11, 95% CI 1.59–6.09) in these analyses.
Our findings demonstrate a protective effect of insurance, whenever defined as both conventional and ADAP pharmacy-only coverage, on HIV transmission risk among a cohort of YLWH over a 13-year period. These findings underscore the critical role that insurance coverage, including ADAP benefits, plays in controlling the ongoing US HIV epidemic. At a time whenever access to health insurance is at risk in the United States , maintaining funding for HIV-related medical and pharmaceutical coverage provides an essential public health safety net, not only promoting the health of people living with HIV but also preventing incident HIV infections.
The mechanism for the protective effect of insurance on transmission potential is likely multifactorial. In addition to providing ART coverage, insurance may keep patients linked to care and provide ongoing access to harm reduction counseling, both of which may lower transmission risk. As substance abuse and mental illness can be barriers to care, insurance may improve HIV-related outcomes by providing treatment for these conditions, thus reducing sexual risk behavior and improving ART adherence. Interestingly, a 2015 study of individuals in a re-linkage to care program in Washington by Dombrowski et al. found that insurance was the most common barrier cited by patients who had been lost to care, despite near-universal access to HIV care through Ryan White-funded clinics in the study region. These data suggest that people living with HIV may not be aware that they can continue accessing HIV care after insurance loss.
Notably, when we categorized ADAP pharmacy-only coverage as uninsured, the association between insurance and transmission potential was weaker and no longer statistically significant. This suggests that no-cost ART coverage may be a primary mechanism by which insurance decreases transmission risk. This finding is supported by a 2016 study of Medical Monitoring Project data by Bradley et al., which identified higher rates of virologic suppression in patients with Ryan White-only coverage than those with Medicaid.
Our findings emerge at a time when the future of universal coverage for ART and HIV-related medical care is uncertain. Federal Ryan White funding for ADAP is proposed to be maintained at current levels; however, if there were to be significant reductions in Medicaid access for people living with HIV, the burden placed on Ryan White to provide essential medical and pharmaceutical services would likely grow, making the current level of funding insufficient .
We also identified that history of STI at or before entry to HIV care was associated with an approximately three-fold increase in the odds of high HIV transmission potential. This finding was stable across models irrespective of our HIV RNA threshold or definitions of insurance coverage. Prior data by Cope et al. demonstrate that having an STI prior to entering HIV care was predictive of subsequent STI incidence. Our data support this finding, and underscore that baseline STI status may be a powerful biomarker for HIV transmission risk. These data support the need for integrated case management services for sexual health and ART adherence, also provided by Ryan White funds, with special attention to those YLWH who enter care with a history of STI.
Finally, whereas the focus of the current study is HIV transmission risk, our findings also have implications for the health of YLWH. Over half of our sample experienced at least one episode of viremia and/or incident STI. Each event represents a preventable episode of morbidity for YLWH. Regardless of transmission risk, these findings are sobering with respect to the potential for disease progression, ART resistance and STI-related health complications, as well as increasing STI and HIV-related stigma among YLWH .
Our study is subject to limitations. Our model was highly conservative, as individuals may have been at high risk of HIV transmission via co-incident viremia and any sexual behavior not protected with condoms or PrEP, even if it did not result in an STI. Because of nonconcordance of HIV RNA and STI testing, we imputed proximal quantitative HIV RNA results to the time of STI testing visits. Given the correlation between adherence and retention , this likely resulted in imputation of suppressed HIV RNA results to visits wherever individuals were not suppressed. Therefore, our findings likely underestimate the proportion of participants with high HIV transmission potential. However, as self-reported condomless sex may be biased, our use of STI as a biomarker for sexual risk acts to minimize outcome misclassification. In addition, we included syphilis in our definition of incident bacterial STI, yet this infection may be transmitted in the absence of condomless sex (i.e. via oral contact). However, as syphilis has been highly associated with HIV transmission risk , we felt that exclusion of syphilis from the outcome definition could result in an underreporting of transmission potential in our already conservative model. Lastly, we did not include HSV in our definition of high transmission potential because of our inability to distinguish between primary infection and reactivation disease nor is it a routinely laboratory-confirmed diagnosis, which also could have lead to an under-estimation of the prevalence of incident STI in the cohort.
This study was performed exclusively among YLWH linked to care and prescribed ART at a single US clinic, and therefore results may not be generalizable to the larger US HIV epidemic or youth at different states of the HIV care continuum. However, the sociodemographic characteristics of the cohort are reflective of the current US youth HIV epidemic. Finally, the marginal structural model assumption of ignorability requires that there is no unmeasured confounding in the model. Although it is possible that there are characteristics of participants that may influence transmission potential that are not measured within our data, such as substance abuse, these variables are highly associated with retention in care , and our retention measure may thus act as a proxy for these covariates. Although substance use data was included in the database, the proportion of missing data precluded inclusion of this covariate. However, in sensitivity analyses addressing the missing substance use values after ART initiation, there was minimal effect on the point estimate for the association between insurance coverage and transmission risk. Thus, it is unlikely that unmeasured substance use accounts for a significant amount of the variation in transmission risk within this cohort.
In conclusion, we identified a protective effect of insurance coverage, including HRSA Ryan White ADAP pharmacy-only coverage, on HIV secondary transmission risk among a cohort of YLWH. These findings underscore the need to maintain access to health insurance and universal ART coverage for YLWH in order to improve health outcomes and decrease transmission risk.
All authors (S.W., S.R., C.G., S.L., N.D. and R.G.) met criteria for authorship including study conception and design; acquisition (N.D., S.L.), analysis (S.W., S.R.) and interpretation of data (all authors); and drafting and critical revision of the work for important intellectual content (all authors). All authors approved the final version for publication and agree to be accountable for all aspects of the work. S.W. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The following individuals contributed to the data collection process: Isabella Lopez, Derek Standlee, David Yusavitz, Papia Paul, Tali Warburg and Paul Nguyen.
Source of funding: Funding for the study's data collection and management was provided through NIMH K23MH102128 (PI: N.D.). Funding for study design, data analysis and interpretation was provided through NIMH F32 MH111341 (PI: S.W.) and NIMH K23MH102128 (PI: N.D.) and supported through Core services provided through the Penn Center for AIDS Research (NIH P30 AI 045008, PI: Dr Collman) and the Penn Mental Health AIDS Research Center (P30 MH 097488, PI: Dr Evans).
Conflicts of interest
There are no conflicts of interest.
An abstract of these data was presented on 25 July 2017 at the Ninth International AIDS Society Conference on HIV Science.
1. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al. HPTN 052 Study Team. Prevention of HIV-1 infection with early antiretroviral therapy
. N Engl J Med
2. Rodger AJ, Cambiano V, Bruun T, Vernazza P, Collins S, van Lunzen J, et al. PARTNER Study Group. Sexual activity without condoms and risk of HIV transmission in serodifferent couples when the HIV-positive partner is using suppressive antiretroviral therapy
3. Centers for Disease Control and Prevention. HIV Continuum of Care, U.S., 2014, overall and by age, race/ethnicity, transmission route and sex. 27 July 2017 [online].
4. Gowda C, Coppock D, Brickman C, Shaw P, Gross R. Determinants of HIV transmission risk among HIV-infected persons engaged in care
. Conference on Retroviruses and Opportunistic Infections
; 22–26 February 2016; Boston.
5. Kalichman SC, Cherry C, Kalichman MO, Washington C, Grebler T, Merely C, et al. HIV sexual transmission risks in the context of clinical care: a prospective study of behavioural correlates of HIV suppression in a community sample, Atlanta, GA, USA
. J Int AIDS Soc
6. Gowda C, Coppock D, Brickman C, Shaw PA, Gross R. Determinants of HIV transmission risk among HIV-infected persons engaged in care
. AIDS Educ Prev
7. Millar BM, Starks TJ, Grov C, Parsons JT. Sexual risk-taking in HIV-negative gay and bisexual men increases with depression: results from a U.S. National Study
. AIDS Behav
8. O’Cleirigh C, Newcomb ME, Mayer KH, Skeer M, Traeger L, Safren SA. Moderate levels of depression predict sexual transmission risk in HIV-infected MSM: a longitudinal analysis of data from six sites involved in a ‘prevention for positives’ study
. AIDS Behav
9. Yehia BR, Fleishman JA, Agwu AL, Metlay JP, Berry SA, Gebo KA. HIV Research Network. Health insurance coverage for persons in HIV care, 2006-2012
. J Acquir Immune Defic Syndr
10. Cohen MS. Sexually transmitted diseases enhance HIV transmission: no longer a hypothesis
1998; 351 (suppl 3):5–7.
11. Cohen MS, Hoffman IF, Royce RA, Kazembe P, Dyer JR, Daly CC, et al. Reduction of concentration of HIV-1 in semen after treatment of urethritis: implications for prevention of sexual transmission of HIV-1. AIDSCAP Malawi Research Group
12. United States Department of Health and Human Services. Improving health outcomes through data utilization: 2016 Ryan White HIV Program Highlights. https://hab.hrsa.gov/sites/default/files/hab/data/biennialreports/2016_HRSA_Biennial_Report.pdf
[Accessed November 2017]
13. Bavinton B, Grinsztejn N, Phanuphak F, Jin F, Zablotska I, Prestage G, et al. HIV treatment prevents HIV transmission in male serodiscordant couples in Australia, Thailand and Brazil
. Ninth International Conference on HIV Science
; July, 2017; Paris.
14. Attia S, Egger M, Muller M, Zwahlen M, Low N. Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis
15. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group
. N Engl J Med
16. Workowski KA, Bolan GA. Centers for Disease Control and Prevention. Sexually transmitted diseases treatment guidelines, 2015
. MMWR Recomm Rep
2015; 64 (RR-03):1–137.
17. Agwu AL, Lee L, Fleishman JA, Voss C, Yehia BR, Althoff KN, et al. Aging and loss to follow-up among youth living with human immunodeficiency virus in the HIV Research Network
. J Adolesc Health
18. Paz-Bailey G, Pham H, Oster AM, Lansky A, Bingham T, Wiegand RE, et al. Engagement in HIV care among HIV-positive men who have sex with men from 21 cities in the United States
. AIDS Behav
2014; 18 (suppl 3):348–358.
19. Dombrowski JC, Simoni JM, Katz DA, Golden MR. Barriers to HIV care and treatment among participants in a public health HIV care relinkage program
. AIDS Patient Care STDS
20. Zinski A, Westfall AO, Gardner LI, Giordano TP, Wilson TE, Drainoni ML, et al. The contribution of missed clinic visits to disparities in HIV viral load outcomes
. Am J Public Health
21. Horberg MA, Hurley LB, Silverberg MJ, Klein DB, Quesenberry CP, Mugavero MJ. Missed office visits and risk of mortality among HIV-infected subjects in a large healthcare system in the United States
. AIDS Patient Care STDS
22. Mugavero MJ, Westfall AO, Cole SR, Geng EH, Crane HM, Kitahata MM, et al. Centers for AIDS Research Network of Integrated Clinical Systems (CNICS). Beyond core indicators of retention in HIV care: missed clinic visits are independently associated with all-cause mortality
. Clin Infect Dis
23. Corliss HL, Wadler BM, Jun HJ, Corliss HL, Austin SB. Sexual-orientation disparities in cigarette smoking in a longitudinal cohort study of adolescents
. Nicotine Tob Res
24. Marshal MP, Dermody SS, Cheong J, Burton CM, Friedman MS, Aranda F, Hughes TL. Trajectories of depressive symptoms and suicidality among heterosexual and sexual minority youth
. J Youth Adolesc
25. Tanney MR, Naar-King S, MacDonnel K. Adolescent Trials Network for HIVAIPT. Depression and stigma in high-risk youth living with HIV: a multisite study
. J Pediatr Healthcare
26. Aidala AA, Wilson MG, Shubert V, Gogolishvili D, Globerman J, Rueda S, et al. Housing status, medical care, and health outcomes among people living with HIV/AIDS: a systematic review
. Am J Public Health
27. Hernan MA, Brumback BA, Robins JM. Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures
. Stat Med
28. Hernan MA, Robins JM. Causal inference
. Boca Raton: Chapman & Hall/CRC; 2017.
29. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men
30. Kates J, Dawson L. What is at stake in ACA repeal and replace for people with HIV? Menlo Park, California, May, 2017.
31. Bradley H, Viall AH, Wortley PM, Dempsey A, Hauck H, Skarbinski J. Ryan White HIV/AIDS program assistance and HIV treatment outcomes
. Clin Infect Dis
32. Cope AB, Crooks AM, Chin T, Kuruc JD, McGee KS, Eron JJ< ET-AL>. Incident sexually transmitted infection as a biomarker for high-risk sexual behavior after diagnosis of acute HIV
. Sex Transm Dis
33. Bonacci RA, Frasca K, Swift L, Sha D, Bilker WB, Bamford L, et al. Antiretroviral refill adherence correlates with, but poorly predicts retention in HIV care
. AIDS Behav
34. Solomon MM, Mayer KH, Glidden DV, Liu AY, McMahan VM, Guanira JV, et al. iPrEx Study Team. Syphilis predicts HIV incidence among men and transgender women who have sex with men in a preexposure prophylaxis trial
. Clin Infect Dis
* Nadia L. Dowshen and Robert Gross contributed equally to the manuscript.