Health Insurance and Initiation of Direct-Acting Antivirals for Hepatitis C in US Women With Human Immunodeficiency Virus

Abstract

Background

Direct-acting antiviral (DAA) therapy for hepatitis C virus (HCV) is well tolerated, cost-effective, and yields high sustained virologic response rates, yet it has remained financially inaccessible to many patients.

Methods

Participants of the Women's Interagency HIV Study (an observational US cohort) with human immunodeficiency virus (HIV) and HCV (RNA+) reporting no prior hepatitis C treatment were followed for DAA initiation (2015–2019). We estimated risk ratios (RRs) of the relationship between time-varying health insurance status and DAA initiation, adjusting for confounders with stabilized inverse probability weights. We also estimated weighted cumulative incidences of DAA initiation by health insurance status.

Results

A total of 139 women (74% Black) were included; at baseline, the median age was 55 years and 86% were insured. Most had annual household incomes ≤$18 000 (85%); advanced liver fibrosis (21%), alcohol use (45%), and recreational drug use (35%) were common. Across 439 subsequent semiannual visits, 88 women (63%) reported DAA initiation. Compared with no health insurance, health insurance increased the likelihood of reporting DAA initiation at a given visit (RR, 4.94; 95% confidence limit [CL], 1.92 to 12.8). At 2 years, the weighted cumulative incidence of DAA initiation was higher among the insured (51.2%; 95% CL, 43.3% to 60.6%) than the uninsured (3.5%; 95% CL, 0.8% to 14.6%).

Conclusions

Accounting for clinical, behavioral, and sociodemographic factors over time, health insurance had a substantial positive effect on DAA initiation. Interventions to increase insurance coverage should be prioritized to increase HCV curative therapy uptake for persons with HIV.

Among US women with human immunodeficiency virus and hepatitis C virus, health insurance had a substantial positive effect on the initiation of direct-acting antiviral therapy.

Yearly spending on medications for hepatitis C virus (HCV), even outside of costs for the treatment of its sequelae including liver cancer and cirrhosis [1], totals billions of dollars in the United States [2]. Recent estimates project that more than 2.2 million US adults are living with HCV infection [3]. HCV therapies that contain only direct-acting antiviral (DAA) agents, available for the past decade, are taken for fewer months and better tolerated than regimens with interferon and ribavirin [4]. Sustained virologic response (SVR) rates following completion of DAA-based regimens are greater than 90% for various genotypes and populations including people with human immunodeficiency virus (HIV) [4], who historically have been less likely cured by interferon-based treatment [5].

Except when short life expectancy is not remediable, all adults with HCV are recommended for hepatitis C treatment by the American Association for the Study of Liver Diseases–Infectious Diseases Society of America (AASLD–IDSA) HCV guidance panel [6]. In 2015, it was estimated that at least 260 000 people would need to be treated annually to attain HCV elimination in the United States by 2030 [7]. However, hepatitis C treatment declined between 2015 and 2020 when an estimated 171 000 people were treated per year, well below the target for elimination [8]. DAAs have remained out of reach for many due to high cost [9–12]. Recent data suggest that the baseline cost of DAA-based treatment ranges from $39 600 to $94 500 [13] independent of additional associated fees, such as those for clinic visits and laboratory testing [14]. Health insurance coverage may facilitate access to DAAs by dramatically reducing out-of-pocket costs not only for medication but also for other required services. While several studies have found that having health insurance is associated with receipt of hepatitis C treatment [15, 16], evidence on how treatment uptake is affected by specific types of insurance coverage remains varied [3, 9–12, 17].

The primary aim of this study was to evaluate the relationship between health insurance status and DAA initiation in a prospective cohort of low-income US women with HIV and HCV between 2015 and 2019. Given its potential for robust cure rates [4], a better understanding of the determinants of DAA initiation is critical for people with HIV and HCV at risk for accelerated progression of liver disease [18]. Recognizing that our study population was composed mostly of Medicaid beneficiaries and the substantially higher prevalence of HCV among Medicaid enrollees compared with those with commercial insurance [19], we also examined the impact of Medicaid on DAA initiation among those who had no other coverage type during the follow-up period.

METHODS

Source Population, Eligibility, and Follow-up

Our data source was the Women's Interagency HIV Study (WIHS), a longitudinal, prospective cohort of US women with or at risk for HIV. WIHS procedures and participant characteristics have been previously detailed [20]. In brief, individuals attended semiannual visits that included collection of biological specimens and self-reported medical and sociodemographic information. All women in the WIHS consented to contribute their data to research, with approval granted by applicable institutional review boards.

Starting in April 2015, women who reported ever being diagnosed with hepatitis C completed a detailed hepatitis C treatment assessment, including questions on whether they were interested in treatment, had talked to a provider about treatment, had been referred for treatment, had been recommended for treatment, had initiated treatment, had completed treatment, and had been told that treatment was successful, that is, a “cure” or SVR. Women were also asked whether they had taken any medications for hepatitis C. If “yes,” they were asked about individual drugs on a list including all DAA classes (NS3/4A protease inhibitors, NS5B nucleoside polymerase inhibitors, NS5B nonnucleoside polymerase inhibitors, and NS5A protein inhibitors [21]) and other therapies such as interferon and ribavirin. HCV laboratory testing was not systematic in the WIHS except for universal antibody (Ab) screening at “enrollment” (ie, first 3 visits of WIHS participation) reflexed to HCV RNA if positive. Other HCV Ab/RNA testing throughout the study occurred at irregular intervals.

For this analysis, we restricted the sample to women with HIV and HCV (RNA+; ie, active infection) who reported no history of treatment in the detailed hepatitis C assessment. The first semiannual WIHS visit between April 2015 and September 2018 at which no history of treatment was reported was defined as the baseline visit. Additionally, to allow women the opportunity to report hepatitis C treatment, they needed to have attended at least 1 subsequent visit by March 2019. Eligible participants were followed through their final visit before the end of March 2019.

Variables

Health insurance was time-varying and categorized as “yes” if a woman reported medical coverage of any type since their last visit. We used 3 mutually exclusive health insurance categories based on Kaiser Family Foundation groups [22], assigned in the following order: Medicaid, private (employer-provided or purchased through health insurance marketplaces), and other (including Medicare, TRICARE, Veteran's Administration, city/county, and others). At each visit, hepatitis C treatment initiation was categorized as “yes” if a woman selected a DAA from the drug list or reported treatment initiation, completion, or success in the detailed assessment. As this study occurred in the DAA era (post-2010) and because DAAs were chosen by all women who reported treatment in the detailed assessment and also selected a hepatitis C medication from the drug list, “hepatitis C treatment” is synonymous with “DAA initiation” in this analysis.

Baseline covariates were age (continuous), race (Black or other), and education (less than high school or at least high school). Time-varying binary covariates were US region of residence (South or other), annual household income (≤$18 000 or >$18 000), alcohol use since last visit, use of injection/noninjection recreational drugs since last visit (any of methamphetamines, crack, cocaine, heroin, marijuana, hallucinogens, club drugs, or injected narcotics), AIDS Drug Assistance Program participation, and advanced liver fibrosis defined as either aspartate aminotransferase to platelet ratio index [23] ≥1.5 or Fibrosis-4 index [24] ≥3.25. Time-varying continuous covariates were HIV viral load and CD4 cell count.

Statistical Analyses

We estimated risk ratios (RRs) of the relationship between time-varying health insurance and DAA initiation reported at a given visit (ie, 6-month risk), using stabilized inverse probability weights to adjust for confounders selected through use of a literature-informed directed acyclic graph [25] (Supplementary Figure 1). Numerators and denominators for the visit- and individual-specific weights were estimated via logistic regression models, each with time-varying health insurance as the outcome. The numerator model included baseline covariates as regressors while the denominator model included time-varying covariates (lagged by 1 visit), prior visit health insurance, WIHS visit, and baseline covariates. The weights (numerator/denominator quotient) were then applied in a log-binomial regression model, with DAA initiation as the outcome and health insurance, WIHS visit, and baseline covariates as regressors, to yield inverse probability-weighted RRs. Continuous variables were specified as 3-knot restricted cubic splines [26], with missing data filled by carrying forward the most recent nonmissing values (no variable was missing for >10% of observations). If the state of residence remained missing, we assumed participants lived in the state of their WIHS site.

We used proportional subdistribution hazard regression [27] to estimate the 2-year unweighted and weighted cumulative incidences of DAA initiation by health insurance status. The weighted function incorporated weights identical to those described above, except that they were multiplied over time to yield the contrast of continuous vs never having had health insurance. Cumulative incidence estimation ended at 2 years because after that the number of uninsured women was prohibitively small (<5); we accordingly compared the functions at that time point using RRs and risk differences (RDs). Within-subject correlation induced by weighting was accounted for by use of the robust sandwich covariance estimate [28], and 95% confidence limits (CLs) were calculated using robust standard errors.

To provide more specific insight on the impact of Medicaid on DAA initiation given the preponderance of Medicaid coverage in the WIHS, sensitivity analyses repeated all modeling after removing from the population women who ever reported a health insurance coverage type in addition to Medicaid during the study period. Analyses were completed in SAS 9.4 (SAS Institute Inc, Cary, NC).

RESULTS

In total, 139 women qualified for the analysis (Figure 1); nearly three-quarters (74%) were Black. Characteristics at baseline are detailed in Table 1. The median age was 55 years (interquartile range, 50–59), and 85% had annual household incomes ≤$18 000. Most women (86%) reported having health insurance, with Medicaid covering 87% of the insured. Alcohol use (45%) and injection/noninjection recreational drug use (35%) since the last visit were common, and 21% of women had advanced liver fibrosis. Across 439 subsequent study visits, 90% with women reporting health insurance (Supplementary Figure 2), 88 women (63%) reported hepatitis C treatment. Of these 88 women, 71 (81%) also selected a DAA from the drug list, and none selected only a non-DAA medication.

Figure 1.

Women's Interagency HIV Study participants with human immunodeficiency virus and active hepatitis C infection reporting no history of hepatitis C treatment, 2015–2019 (n = 139). Abbreviations: Ab, antibody; HIV, human immunodeficiency virus; WIHS, Women's Interagency HIV Study.

Table 1.

Characteristics at Baseline of 139 Women's Interagency HIV Study Participants With Human Immunodeficiency Virus and Active Hepatitis C Infection Followed for Hepatitis C Treatment Initiation: 2015–2019

Characteristic	N (%) or Median (Interquartile Range)
Age, y	55 (50–59)
Black race	103 (74)
Less than high school education	60 (43)
Health insurance	120 (86)
Medicaid	104 (87)
Private	9 (8)
Other	7 (6)
AIDS Drug Assistance Program participation	30 (22)
Resided in the Southa	31 (22)
Combination antiretroviral therapy	112 (81)
HIV viral load, copies/mL	Undetectable (undetectable-41)
HIV viral load, undetectable	91 (65)
CD4 cell count, cells/µL	599 (369–828)
Alcohol use	63 (45)
Injection/noninjection recreational drug useb	48 (35)
Advanced liver fibrosis (aspartate aminotransferase to platelet ratio ≥1.5 or Fibrosis-4 index ≥ 3.25)	29 (21)
Annual household income ≤$18 000	118 (85)

Abbreviation: HIV, human immunodeficiency virus.

South includes Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina; other includes California, Washington, DC, Illinois, Indiana, Maryland, New York, Virginia. These groups closely overlap with Medicaid expansion (1 participant lived in Kentucky where there was Medicaid expansion; 4 participants lived in Virginia where there was no Medicaid expansion).

Includes crack, cocaine, heroin, marijuana, hallucinogens, club drugs, methamphetamines, injected narcotics.

Compared with no insurance, health insurance increased the likelihood of reporting DAA initiation at a given visit (weighted RR, 4.94; 95% CL, 1.92 to 12.8; Table 2). The inverse probability weights, which had a mean of 0.99 and ranged from 0.07 to 6.93, yielded an estimate markedly stronger (but less precise) than the unweighted (RR, 1.48; 95% CL, 0.76 to 2.91). Figure 2 depicts the weighted and unweighted cumulative incidences of DAA initiation by health insurance status. At 2 years, the weighted cumulative incidence of DAA initiation was higher (RR, 14.6; 95% CL, 3.46 to 61.6 and RD, 47.7%; 95% CL, 37.8% to 57.7%) among the insured (51.2%; 95% CL, 43.3% to 60.6%) than the uninsured (3.5%; 95% CL, 0.8% to 14.6%). These contrasts between groups were of greater magnitude than when the cumulative incidences among the insured (53.9%; 95% CL, 46.2% to 62.8%) and uninsured (29.9%; 95% CL, 16.2% to 55.4%) were unweighted (RR, 1.80; 95% CL, 0.95 to 3.40 and RD, 24.0%; 95% CL, 3.8% to 44.2%).

Table 2.

Associations Between Health Insurance and Reporting Hepatitis C Treatment Initiation at a Given Visit for Women's Interagency HIV Study Participants With Human Immunodeficiency Virus and Active Hepatitis C Infection Followed for Hepatitis C Treatment Initiation: 2015–2019

Model	Primary Analysis (n = 139)		Sensitivity Analysisa (n = 104)
	Risk Ratio	95% Confidence Limit	Risk Ratio	95% Confidence Limit
Weightedb	4.94	1.92–12.8	2.63	.79–8.79
Unweighted	1.48	.76–2.91	1.01	.48–2.11

Excluding women who ever reported a health insurance coverage type in addition to Medicaid during the study period.

Weighted models adjust for baseline (age, race, education) and time-varying (US region of residence, annual household income, alcohol use since last visit, use of injection/noninjection recreational drugs since last visit, AIDS Drug Assistance Program participation, advanced liver fibrosis, human immunodeficiency virus viral load, CD4 cell count) confounders. Details on covariates and models are outlined in the Methods section.

Figure 2.

Weighted (A) and unweighted (B) cumulative incidence of direct-acting antiviral initiation by health insurance status for Women's Interagency HIV Study participants with human immunodeficiency virus (HIV) and active hepatitis C infection reporting no history of hepatitis C treatment, 2015–2019 (n = 139). Weighted models adjust for baseline (age, race, education) and time-varying (US region of residence, annual household income, alcohol use since last visit, use of injection/noninjection recreational drugs since last visit, AIDS Drug Assistance Program participation, advanced liver fibrosis, HIV viral load, CD4 cell count) confounders. Details on covariates and models are outlined in the Methods section.

After excluding women who ever reported a health insurance coverage type in addition to Medicaid during the study period (n = 34), estimates were attenuated and less precise. The weighted RR of 2.63 (95% CL, 0.79 to 8.79) suggested that Medicaid increased the likelihood of reporting DAA initiation at a given visit compared with no insurance. The distribution of weights (mean, 0.97; range, 0.07 to 6.12) mirrored that in the full population analysis. The unweighted RR was null (1.01; 95% CL, 0.48 to 2.11). When weighted, the cumulative incidence of DAA initiation appeared higher among those covered by Medicaid (52.3%; 95% CL, 44.7% to 61.2%) than among the uninsured (16.0%; 95% CL, 4.6% to 55.6%) at 2 years (RR, 3.27; 95% CL, 0.93 to 11.5 and RD, 36.3%; 95% CL, 14.7% to 57.8%). This distinction was not apparent when the Medicaid (53.2%; 95% CL, 44.3% to 64.0%) and no health insurance (44.9%; 95% CL, 26.2% to 76.8%) cumulative incidences were unweighted (RR, 1.19; 95% CL, 0.67 to 2.09 and RD, 8.4%; 95% CL, −17.7% to 34.4%).

DISCUSSION

In this predominantly low-income cohort of US women with HIV, we found that after accounting for clinical, behavioral, and sociodemographic factors, women who reported having health insurance were approximately 5 times more likely to have started hepatitis C treatment in the last 6 months and nearly 15 times more likely to start hepatitis C treatment over 2 years than women who were uninsured. These findings, which are concordant with prior studies highlighting the relationship between insurance and hepatitis C treatment initiation [15, 16], underscore the importance of insurance coverage as a means to reduce financial barriers to DAA access [9–12, 14].

Most women in the analysis had annual household incomes ≤$18 000. While the most common insurance type reported by study participants was Medicaid, which is unsurprising given the women's low reported incomes, the high prevalence of Medicaid among women with chronic HCV infection was also consistent with a prior analysis of large managed Medicaid and commercially insured populations [19]. Sensitivity analyses suggested that women with Medicaid were more likely to initiate DAAs than women without insurance. However, effect estimates from analyses of the full sample were of greater magnitude, indicating that those with other insurance types were more likely to start hepatitis C treatment, as noted in past studies [11, 12]. Although Medicaid theoretically covers costs associated with hepatitis C treatment, varying restrictions across state Medicaid programs [9, 29, 30] may partially explain the suboptimal DAA uptake that we noted among Medicaid recipients. Medicaid expansion has resulted in greater prescribing of hepatitis C treatment and reductions in end-stage liver disease mortality. These gains are most notable in states where treatment eligibility is less restrictive (eg, not requiring individuals to be sober or have advanced liver disease to qualify for DAAs) [31, 32]. Elimination of restrictions for hepatitis C treatment may be especially relevant given the notable disparities in treatment for people who use drugs, evidence supporting positive treatment outcomes for these individuals, and the growing prevalence of HCV in this group during the prolonged opioid crisis [33]. These realities suggest that the adoption of Medicaid expansion by all states, along with consistent standards that authorize DAAs for the broad majority of individuals with HCV, could meaningfully increase DAA uptake in populations similar to our cohort (eg, people who use drugs or alcohol, regardless of sex or HIV status). Notably, expanding Medicaid coverage for hepatitis C treatment also has spillover effects that benefit those with other insurance types, such as increased hepatitis C testing among privately insured populations in states that are relaxing Medicaid treatment restrictions [30].

Despite high health insurance coverage rates, roughly half of insured participants did not initiate hepatitis C treatment over 2 years. These findings are troubling, given that nearly everyone with HCV should receive treatment according to the AASLD–IDSA HCV guidance [6], and reflect the presence of barriers to starting DAAs, even for the insured. Additionally, women with HIV likely have more interaction with the healthcare system than the general population. Our findings reveal ongoing missed opportunities for linkage to treatment [34]. Past work with this cohort has revealed that women face multiple barriers to hepatitis C treatment initiation, not only related to coverage exclusions and cost but also to navigating healthcare administrative systems (eg, scheduling appointments) [34]. Research has also identified substantial multilevel barriers to treatment and disparities in treatment outcomes, particularly for people who have less formal education [35], as well as for people who are Black [12, 34, 36] and people who use drugs [12, 33, 34]. This highlights that although insurance coverage may be a critical factor to increase DAA uptake, more comprehensive approaches are needed to effectively reduce disparities in treatment and reach the 2030 hepatitis C elimination goals. Delayed or lack of treatment initiation may have been partially due to switching individuals to antiretroviral regimens including an integrase strand transfer inhibitor (INSTI) to avoid drug–drug interactions with DAAs [37]. This likely occurred, as the proportion of regimens that included an INSTI increased from 42% at baseline to 65% at the final visit. Among those without insurance, uptake of DAAs was minimal. While this finding was not unexpected, it emphasizes that more efforts should be made to connect qualifying individuals to safety-net systems and patient assistance programs, the likely mechanisms by which uninsured WIHS women were able to access medication [14, 38].

Several limitations should be considered when interpreting our findings. The numbers of individuals and DAA initiations were small, particularly in the sensitivity analysis, and the proportion of uninsured persons was low, limiting the precision of estimates. Because self-reported data were not confirmed by primary sources (eg, medical record review), some values may have been incorrect. As the WIHS is an interval cohort with data collection protocols and instruments not allowing for ascertainment of the exact timing or ordering of insurance coverage changes and hepatitis C treatment initiation, there is a potential for exposure misclassification. We anticipate that this affected results minimally because there were only 21 visit-to-visit insurance gains in 434 visits (<5%), with just 4 DAA initiations in those 21 visits. Given the strong association between insurance and treatment, it is unlikely that all (or any) of these occurred when women were uninsured. The extent to which this potential misclassification may have biased our results away from the null was explored by assigning women as uninsured for those 4 initiations. While the estimate for initiation at a given visit was moderated (RR, 3.44; 95% CL, 1.47–8.08) as anticipated, the same conclusion would have been reached. It is also worth noting that state Medicaid rules and restrictions related to hepatitis C treatment may have biased our estimates downward compared with an analysis that included only residents of states with broad coverage and treatment allowances (eg, not requiring prior authorization and including all types of plans, prescribers, and patients regardless of substance use or degree of liver fibrosis) since participants were drawn from a heterogenous policy landscape.

There are several noteworthy strengths of this study. We examined hepatitis C treatment over an extended period while incorporating time-varying insurance status and covariates, drawing from a rich, high-quality dataset that included key variables posited necessary for confounding control. Hepatitis C status was substantiated by laboratory testing, and the reliability of DAA initiation assignment was strengthened by longitudinal data collection in 2 interview sections. Our observational cohort was unlike most claims data where all individuals are insured, allowing for a rare contrast of health insurance to the lack thereof. Also, as included participants were mostly Medicaid beneficiaries, a parallel assessment of the impact on hepatitis C treatment of this coverage type (vs no insurance) was possible. This is also one of a limited number of analyses to investigate DAA uptake among US persons with HIV [15, 36, 39, 40] and extends our prior work examining the HCV treatment cascade and associated barriers [34]. Although this study uniquely focused on women with HIV, findings may not be generalizable to women without HIV. In generalizing our Medicaid results to other low-income populations, we postulate that the impact of coverage on treatment would be stronger as state program restrictions continue to abate.

In conclusion, our findings collectively support health insurance coverage as a critical tool in increasing access to hepatitis C treatment for low-income US women with HIV and HCV. Policy interventions to improve coverage, such as Medicaid expansion and the elimination of Medicaid coverage restrictions, should be mandated to increase uptake of HCV curative therapy for this and similar vulnerable groups, including people who use drugs, and to ultimately achieve the 2030 hepatitis C elimination goals.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

MACS/WIHS Combined Cohort Study (MWCCS) Principal Investigators. Atlanta Clinical Research Site (CRS) (Ighovwerha Ofotokun, Anandi Sheth, and Gina Wingood), U01-HL146241; Baltimore CRS (Todd Brown and Joseph Margolick), U01-HL146201; Bronx CRS (Kathryn Anastos, David Hanna, and Anjali Sharma), U01-HL146204; Brooklyn CRS (Deborah Gustafson and Tracey Wilson), U01-HL146202; Data Analysis and Coordination Center (Gypsyamber D'Souza, Stephen Gange, and Elizabeth Topper), U01-HL146193; Chicago–Cook County CRS (Mardge Cohen and Audrey French), U01-HL146245; Chicago–Northwestern CRS (Steven Wolinsky), U01-HL146240; Northern California CRS (Bradley Aouizerat, Jennifer Price, and Phyllis Tien), U01-HL146242; Los Angeles CRS (Roger Detels and Matthew Mimiaga), U01-HL146333; Metropolitan Washington CRS (Seble Kassaye and Daniel Merenstein), U01-HL146205; Miami CRS (Maria Alcaide, Margaret Fischl, and Deborah Jones), U01-HL146203; Pittsburgh CRS (Jeremy Martinson and Charles Rinaldo), U01-HL146208; UAB-MS CRS (Mirjam-Colette Kempf, Jodie Dionne-Odom, and Deborah Konkle-Parker), U01-HL146192; UNC CRS (Adaora Adimora and Michelle Floris-Moore), U01-HL146194.

Acknowledgments. The authors gratefully acknowledge the contributions of the study participants and dedication of the staff at the MWCCS sites.

Author contributions. Conceptualization: A. E., D. F. H., A. L. F., A. A. A. Data curation: A. E. Formal analysis: A. E. Funding acquisition: A. L. F., P. C. T., A. S., M. L. A., A. A. A. Investigation: A. E., D. F. H., J. K. E., C. R., A. L. F., P. C. T., M. P., A. S., M. A., E. C. S., K. W., M. L. A., S. A., A. A. A. Methodology: A. E., J. K. E. Project administration: A. E., D. F. H. Resources: A. E. Software: A. E. Supervision: A. E., D. F. H. Validation: A. E. Visualization: A. E. Writing the original draft: A. E., D. F. H. Reviewing and editing the original draft: A. E., D. F. H., J. K. E., C. R., A. L. F., P. C. T., M. P., A. S., M. A., E. C. S., K. W., M. L. A., S. A., A. A. A.

Data availability. Access to individual-level data from the MACS/WIHS Combined Cohort Study Data (MWCCS) may be obtained upon review and approval of a MWCCS concept sheet. Links and instructions for online concept sheet submission are on the study website.

Disclaimer. The contents presented here are solely the responsibility of the authors and do not represent the official views of the National Institutes of Health (NIH).

Financial support. This work was supported by the NIH. Data were collected by the WIHS, now the MWCCS. The MWCCS is funded primarily by the National Heart, Lung, and Blood Institute, with additional co-funding from the Eunice Kennedy Shriver National Institute of Child Health & Human Development, National Institute on Aging, National Institute of Dental and Craniofacial Research, National Institute of Allergy and Infectious Diseases, National Institute of Neurological Disorders and Stroke, National Institute of Mental Health, National Institute on Drug Abuse (NIDA), National Institute of Nursing Research, National Cancer Institute, National Institute on Alcohol Abuse and Alcoholism, National Institute on Deafness and Other Communication Disorders, National Institute of Diabetes and Digestive and Kidney Diseases, and National Institute on Minority Health and Health Disparities and in coordination and alignment with the research priorities of the NIH, Office of AIDS Research. MWCCS data collection is also supported by UL1-TR000004 (UCSF CTSA), UL1-TR003098 (JHU ICTR), UL1-TR001881 (UCLA CTSI), P30-AI-050409 (Atlanta CFAR), P30-AI-073961 (Miami CFAR), P30-AI-050410 (UNC CFAR), P30-AI-027767 (UAB CFAR), P30-MH-116867 (Miami CHARM), UL1-TR001409 (DC CTSA), KL2-TR001432 (DC CTSA), and TL1-TR001431 (DC CTSA). J. K. E. was supported by R01AI157758. D. F. H. was supported by K01DA046307. E. C. S. received additional support through the Johns Hopkins University CFAR (P30-AI-094189).