The Centers for Medicare and Medicaid Services (CMS) is the largest single provider of health insurance in the United States, with 90 million Americans benefiting through its programs.1 Under the Patient Protection and Affordable Care Act (ACA), Medicaid coverage was expanded to 138% of the federal poverty level in states that opted for expansion.2 As of April 2021, Medicaid expansion was yet to be adopted by 11 states, including Texas; 2 others (O.K., M.O.) had adopted but not yet implemented Medicaid expansion.3

The most populous city in the Southern United States, Houston, TX, is home to over 2.3 million people. A racially/ethnically diverse population with a growing influx of immigrants and younger cohorts,4 Houston is the principal city of Harris County, which includes 4.7 million residents,5 and is a microcosm of Texas, the state with the highest uninsured rate in the country: of Texas adults under 65 years of age, nearly 25% are uninsured and <7% are covered by Medicaid.6 Despite proposals to create a state-tailored expansion, legislative debate in 2021 was short-lived,7 leaving Texas and major cities such as Houston with no end in sight to massive coverage gaps that disproportionately burden people of color.8 This disparity in medical access intersects with existing health disparities in disease burden and health inequities, a potential driver for sustained transmission of conditions such as human immunodeficiency virus (HIV) among groups that have been marginalized.

As of a recent CDC report, the Houston metropolitan area ranked ninth nationally in rate of new HIV diagnoses9 and reports lower rates of retention in HIV care and viral suppression than the United States overall.10 Of all new diagnoses in 2019, men who have sex with men (MSM) are believed to have constituted nearly 71% of diagnoses in Houston/Harris County. Among MSM 18–34 years of age, the highest burden was among young, Black men who have sex with men (YBMSM) who made up 19% of all new diagnoses per internal calculations from the Houston Health Department (HHD) where missing risk group information was imputed. Contributing factors to continued HIV transmission, especially in the South, may include intersectional stigma of race and sexual identity in geographic regions perceived as oppressive and the medical distrust that results from this and other forces.11 Expectations around masculinity and religious ideals that are common among Black communities lead some YBMSM to experience “potentially irreconcilable beliefs surrounding race and its interplay with… sexual identity.”12 Social isolation resulting from such perceptions and experiences of homophobia dissuade engagement in HIV prevention and care services. A further barrier to services includes insurance coverage; MSM in nonexpansion states are more likely to be uninsured and less likely to utilize pre-exposure prophylaxis (PrEP) than their counterparts residing in expansion states.2 Furthermore, research demonstrates that lacking health insurance is associated with lengthy delays in the initiation of PrEP care among interested YBMSM13 and may be related to lower PrEP care retention.14–18 Medicaid expansion has been shown to improve engagement in both the antiretroviral treatment (ART) and PrEP continua.19–21 A number of empirical and simulation studies have projected declines in downstream HIV incidence because of improved ART and PrEP continuum engagement.22

However, little is known about how Medicaid expansion is likely to impact downstream HIV incidence, particularly in Houston, which remains among the urban epicenters of HIV in the Southern United States. Estimating the population health impacts of policy reforms is difficult because the effects of such policies may be realized over time periods that are longer than the duration of most empirical studies. Second, these impacts occur at a multitude of scales, starting from the microlevel individual, to the mesolevel networks, and the macrolevel population or community impacts. Computational modeling allows for the investigation of the impact of policy interventions before their real-world implementation and provides rigorous methods to quantify uncertainties given that data on relevant parameters are sometimes parse, uncertain, or unavailable.

This study takes a simulation approach to investigate the impact of Medicaid expansion strategies on HIV incidence among YBMSM in the context of Houston, TX. This simulation study is grounded on the modeling framework of a previously developed and calibrated agent-based network model (ABNM) and relies on empirically defined parameter estimates.22 ABNMs are a complex systems modeling technique that provide the flexibility to model individual persons and members of their community networks as “agents” and the social network structures that connect them as “ties.” The co-evolution of agents and networks provides practical insight into the impacts of health policy implementations.23 The goal is to study the population health impacts of Medicaid expansion strategies that have shown promise at the individual-level in improving engagement across the ART and PrEP continua.19–21,24 To our knowledge, the impact of the resulting population-level improvements in ART and PrEP continua engagement on downstream HIV incidence have not yet been estimated.

The ABNM is deployed here as a tool to estimate how Medicaid expansion might impact HIV incidence in Houston in the next decade. Quantifying any projected gaps between the estimated declines in HIV incidence and the levels required to achieve a “functional zero” incidence (ie, a level at which the projected number of new HIV infections is below the threshold needed to sustain the epidemic) can help guide intervention planning tailored for the YBMSM population at this important juncture.25

METHODS Agent-Based Network Model Development

The ABNM combines an agent-based simulation approach with modeling of a sexual network structure and includes a number of processes that impact HIV transmission (described below). The sexual network structure was estimated using exponential random graph models;26 this approach is consistent with the methodology deployed in recent work across US urban areas.27–29 The ABNM presented here was implemented using the statnet30 suite of packages in the R programming language to simulate dynamic networks. The ABM components were developed with the C++-based Repast HPC ABM toolkit.31–33

Demographic, Network, Behavioral, and Biological Data

The demographic, clinical, and behavioral parameters of the ABNM was estimated using data sources that were representative of YBMSM in Houston, TX. Primarily, these were: a population-based cohort data from the Young Men’s Affiliation Project (YMAP)34 and the National HIV Behavioral System (NHBS).35 The biological parameters were obtained from published data sources; data sources for all key parameters are provided in Table A.3, Supplemental Digital Content 1, https://links.lww.com/MLR/C523. Procedures and protocols were approved by relevant institutional review boards.

Baseline Model

Baseline HIV transmission was simulated to capture existing epidemic features among adolescents and young adults (age: 18–34 y), populated with 10000 individuals at the start of the dynamic simulations. Combining information on HIV prevalence from a YMSM cohort and surveillance data on BMSM from Houston, a prevalence target of 32% was identified.34,35 An empirical HIV incidence rate of ∼4 per 100 person years was computed using the same cohort data (though this estimate was not published). Simulations proceeded in daily time steps. The substantive model components included agent arrival into the model (“aging in”), departures (agents “aging out” or experiencing mortality), dynamic sexual network structure, the temporal evolution of CD4 counts and HIV RNA (“viral load”), HIV testing and diagnosis, dynamics of ART and PrEP use, external HIV infections, and HIV transmission dynamics. These processes are described in greater detail in Section A.4 of the Appendix, Supplemental Digital Content 1, https://links.lww.com/MLR/C523.

The baseline model was simulated 30 times to assess to assess the inherent (aleatory) variability in model outputs for each parameter set (Appendix Section A.6, Supplemental Digital Content 1, https://links.lww.com/MLR/C523). The mean HIV prevalence rate across the 30 runs was 30.5 (SD: 0.80) and the mean HIV incidence rate was 4.1 per 100 py (SD: 0.37). For computational feasibility and since the replicates did not differ meaningfully from each other, we chose 1 of the 30 replicates for the subsequent analyses to assess the difference between the baseline model and the Medicaid Expansion intervention scenarios.

Modeling Medicaid Expansion Interventions

The Medicaid Expansion scenario was modeled by assuming a 5% increase in HIV testing rates, in accordance with recent data,19 a 2% increase in the proportion of HIV-negative persons using PrEP,20,36 and a 17% improvement in ART adherence.19,37 Using this scheme, three Medicaid expansion policy scenarios were simulated: (a) Medicaid Expansion (ME1), as described above (Tables 1, 2); (b) ME1 plus a 20% increase in PrEP uptake (ME2); and (c) ME1 plus a 30% increase in PrEP uptake and a 15% increase in ART uptake (ME3). The ART and PrEP uptake expansion rates in scenarios (b) and (c) were selected based on the study team’s results on achieving HIV elimination targets when applying this model to a YBMSM population in another large US city.25 The PrEP increase was simulated to uniformly achieve the targets uptake level over the 10-year intervention period. Improvement in the HIV testing, time between diagnosis and ART initiation, and ART adherence and linkage parameters were set at the start of the intervention period, and the consequent improvement in ART uptake was achieved over the 10 years of the intervention.

TABLE 1 - Evidence Supporting Parameterization of Medicaid Expansion (ME) Scenario I Parameter References Study Setting Summary of Conclusions Impact on Medicaid Expansion (ME) Model Comments on Validity HIV testing rates Simon et al38 Behavioral Risk Factor Surveillance System (BRFSS) data from all 50 states and District of Columbia, restricted to adults below 65 y of age with household incomes below 100% federal poverty level; compared states with ME vs. states without ME ME increased the probability of receiving an HIV test by 5% HIV testing rate improved by 5% in the presence of ME BRFSS “designed to be representative of the non-institutionalized adult population in the United States”; difference-in-differences methodology used with differences in outcomes before/after ME computed and comparison made of differences between control and treatment states; sensitivity analyses included no violation of parallel trends assumption (ie, without treatment, outcomes in treatment group follow same trend as control) and robustness of results to variation in sample and model specification ART adherence Furl et al37 and Adamson et al19 AIDS Drug Assistance Program participants recruited for insurance enrollment in Nebraska ACA enrollment associated with a 17.2% increase in proportion of days covered by HIV medication refill (refill ratio/year) 17% improvement in ART adherence under ME Electronic medical records, pharmacy, and Ryan White program databases utilized; pharmacy database for claims used to validate proportion of days covered PrEP uptake Farkhad et al20 32 US states and District of Columbia with ME vs. 18 states without ME ME increased the rate of PrEP use in the general population by a small amount (2.643/100000 population) PrEP use expansion is negligible in terms of increase in proportion of HIV-negative persons using PrEP with ME Difference-in-differences methodology used with differences in outcomes before/after ME computed and comparison made of differences between control and treatment counties; sensitivity analyses included state-specific time trends and no violation of parallel trends assumption (ie, without treatment, outcomes in treatment group follow same trend as control) Siegler et al36 National data set of PrEP users data from claims data aggregator used to develop county-level estimates of PrEP use in 2018. Regression used to explore associations of state Medicaid expansion policies and PrEP assistance programs with rates of PrEP use PrEP prevalence increases by 25% in states with ME. PrEP use rate in our population would increase to about 25% with ME (compared with 20% without, see Table 2). We took the average of the two studies (Farkhad et al20 and Siegler et al36) to estimate an ∼2% increase in PrEP uptake with Medicaid Expansion PrEP use method determination validated in previous studies,39,40 including use of national pharmacy data regardless of payer source and validated against medical claims; sensitivity analysis accounting for missing data used to select national best estimate

ACA indicates Affordable Care Act; ART, antiretroviral treatment; PrEP, pre-exposure prophylaxis.

TABLE 2 - Parameters in Policy Scenarios Medicaid Expansion Scenarios Parameters ME1 ME2 ME3 HIV Testing Frequency 20% decline in proportion of persons with least frequent HIV testing (1–2 tests in previous 2 y)* 30% decline in proportion of persons with least frequent HIV testing (1–2 tests in previous 2 y)* Same as ME2 Time between HIV diagnosis and ART initiation Same as baseline 60% increase in HIV treatment initiations within 1 week after diagnosis* Same as ME2 Distribution of ART adherence 17% increase in always adherent persons* Same as ME1 15% increase in ART uptake relative to ME1 Number of PrEP users 2% increase relative to Baseline* 20% increase relative to ME1, distributed uniformly over the 10 y of intervention* 30% increase relative to ME1, distributed uniformly over the 10 y of intervention* Baseline: control scenario, no Medicaid expansion. Parameters listed in Table 1.

ME1 indicates Medicaid Expansion, including a 2% increase in PrEP uptake and a 17% improvement in viral suppression; ME2, ME1 plus a 20% increase in PrEP uptake and a 15% increase in ART uptake; ME3, ME1 plus a 30% increase in PrEP uptake and a 15% increase in ART uptake; PrEP, pre-exposure prophylaxis.

Outcomes

The control setting and interventions were each simulated 30 times. The primary outcome was the HIV incidence rate 10 years after the start of implementation, averaged over the 30 model simulations. The annual incidence rate in the tenth year under the ME1, ME2, and ME3 scenarios were computed, and compared with the control scenario where levels of engagement in the HIV prevention and treatment continua were held constant at levels consistent with the Baseline Model throughout the 10-year intervention period. In addition, the cumulative number of new infections averted in each year of ME1, ME2, and ME3 are reported, relative to the control scenario.

Uncertainty in the incidence projection estimates was quantified by using bootstrap estimates derived through simulation. To do this, the 30 simulation runs for each policy scenario at each time point were sampled 1000 times with replacement. The mean for each of the resampled datasets was computed, and the 2.5% and 97.5% quantile of these means were taken to obtain the 95% bootstrap simulation interval around the mean.

RESULTS

Table 3 provides the mean 10th year incidence rate (per 100 person years), the mean number of HIV infections in the tenth year, and the mean HIV prevalence at the end of the average annual incidence rates, Figures 1 and 2 display simulated data for the HIV incidence rate (per 100 py) and the annual mean number of new HIV infections under all 4 scenarios, with color bands that demonstrate the bootstrap simulation intervals. Table 4 provides the cumulative number of new infections averted in each year of ME1, ME2, and ME3, relative to the control scenario.

TABLE 3 - Mean Annual Incidence Rate and Annual HIV Incidence for Each Policy Scenario in the Tenth Year of the Simulation Policy Scenario Incidence Rate (Per 100 Persons Years) New HIV Infections Absolute (%) Difference Between New Infections in Tenth Year and Functional Zero* Target Baseline 4.96 (4.82,5.11)† 367 (356, 379) 167 (83.5) ME1 4.09 (3.99,4.19) 313 (305, 320) 113 (56.5) ME2 2.86 (2.76, 2.94) 229 (222, 236) 29 (14.5) ME3 2.54 (2.45, 2.62) 205 (199, 213) 5 (2.5)

Baseline: control scenario, no Medicaid Expansion.

ME1 indicates Medicaid Expansion, including a 2% increase in PrEP uptake and a 17% improvement in viral suppression; ME2, ME1 plus a 20% increase in PrEP uptake and a 15% increase in ART uptake; ME3, ME1 plus a 30% increase in PrEP uptake and a 15% increase in ART uptake; PrEP, pre-exposure prophylaxis.

*The “functional zero target” is 200 new infections per year.

†The uncertainty is given by the bootstrap simulation interval (SI) across 1000 replicates.

FIGURE 1:

Simulated annual mean HIV incidence rates (top) and number of new HIV infections (bottom), under different policy scenarios. Baseline: control scenario, no Medicaid expansion. HIV indicates human immunodeficiency virus; ME1, Medicaid expansion, including a 2% increase in PrEP uptake and a 17% improvement in viral suppression; ME2, ME1 plus a 20% increase in PrEP uptake and a 15% increase in ART uptake; ME3, ME1 plus a 30% increase in PrEP uptake and a 15% increase in ART uptake; PrEP, pre-exposure prophylaxis.

TABLE 4 - Cumulative Infections in Each Year Averaged Across the 30 Simulations Year Base ME1 ME2 ME3 1 365 359 348 342 2 736 717 676 665 3 1114 1067 991 978 4 1485 1413 1286 1267 5 1852 1747 1574 1547 6 2218 2084 1844 1807 7 2587 2415 2107 2047 8 2959 2730 2364 2273 9 3332 3045 2608 2492 10 3699 3358 2837 2698 Absolute (%) difference in cumulative infections between baseline and the Medicaid expansion scenarios 1 — 5.79 (1.6) 11.52 (3.2) 22.7 (6.2) 2 — 19.06 (2.6) 41.01 (5.6) 71.77 (9.7) 3 — 46.68 (4.2) 76.55 (6.9) 136.23 (12.2) 4 — 71.63 (4.8) 126.69 (8.5) 218.07 (14.7) 5 — 104.38 (5.6) 172.95 (9.3) 304.78 (16.5) 6 — 134.36 (6.1) 240.07 (10.8) 411.75 (18.6) 7 — 171.82 (6.6) 307.96 (11.9) 539.52 (20.9) 8 — 229.02 (7.7) 366.07 (12.4) 685.39 (23.2) 9 — 287.10 (8.6) 436.57 (13.1) 839.39 (25.2) 10 — 341.75 (9.2) 520.31 (14.1) 1001.11(27.1)

Baseline: control scenario, no Medicaid Expansion.

ME1 indicates Medicaid Expansion, including a 2% increase in PrEP uptake and a 17% improvement in viral suppression; ME2, ME1 plus a 20% increase in PrEP uptake and a 15% increase in ART uptake; ME3, ME1 plus a 30% increase in PrEP uptake and a 15% increase in ART uptake; PrEP, pre-exposure prophylaxis.

A considerable decline in the HIV incidence rate and the projected mean number of new HIV infections is observed under all 3 Medicaid expansion scenarios. The ME1 scenario resulted in a 17.5% reduction in the HIV incidence rate and about a 14.9% decline in the number of infections in the tenth year of the intervention versus baseline. ME2 yielded 10th-year declines of 42.3% in the HIV incidence rate and 37.8% in mean number of new infections. ME3 yielded reductions of 48.7% in the HIV incidence rate and 44.0% in the number of new infections in the tenth year.

Cumulatively, the 30 simulation runs of the control scenario produced a mean of 3699 new infections across the tenth year. In contrast, ME1, ME2, and ME3 respectively yielded 3358, 2837, and 2698 cumulative infections, corresponding to declines of 9.2%, 14.1%, and 27.1%, relative to the control scenario.

DISCUSSION

In this study, we explored the impact that Medicaid expansion might have on HIV incidence among YBMSM in Houston. We observe that Medicaid expansion can make a sizeable impact on HIV incidence. In addition, increasing PrEP and ART uptake beyond the projected increases under Medicaid expansion, can yield further decreases in the rate of new cases. An increase in ART and PrEP uptake, beyond the levels projected under Medicaid expansion, might be necessary to eliminate HIV transmission in the next decade. For example, drug assistance programs in combination with Medicaid expansion, have been associated with 99% increase in PrEP use,36 and may prove effective in increasing uptake. While Medicaid expansion alone may not get Houston to the goal of zero incidence, expanding PrEP and ART interventions in combination may help achieve this goal.

Previous work in Chicago, with a comparably sized HIV epidemic, has suggested that increasing ART and PrEP uptake to 30% above current levels would lower incidence to about 200 new infections per year among YBMSM, reaching a “functional zero” transmission rate.25 Simulated data presented in this study indicate some similarity between the 2 cities, with comparably sized populations of YBMSM, where about a 30% increase in PrEP uptake and a 15% increase in ART use is projected to result in 205 new HIV infections per year among YBMSM in Houston.

With the Texas legislature’s decision not to expand Medicaid in April 2021,7 there is little evidence that widespread coverage gaps will soon be mitigated through this avenue. However, Houston has made substantial strides to increase PrEP and ART uptake that may bring the city closer to HIV Ending the Epidemic goals. In 2019, the HHD initiated a locally driven awareness campaign with residents from the LGBTQ community serving as ambassadors. Branded I am Life, the campaign focused on PrEP and treatment as prevention (TasP) messages among the populations most impacted in Houston—Black and Hispanic MSM and transgender individuals.41 Complementing work to increase demand for PrEP, the HHD simultaneously increased availability by expanding hours/days of operation at PrEP-specific clinics.

In order to promptly disrupt onward transmission, Houston also continues to expand rapid initiation of ART. Just 1 of 15 grantees, a Houston community-based organization was awarded new federal funding in 2020 to build capacity for rapid ART implementation.42,43 The Houston jurisdiction’s Ryan White Grant Administration additionally scaled up rapid start as part of their Ending the HIV Epidemic initiative. Over the period of 2020–2021, they have provided over $3.7 million to initiate or enhance rapid start in 5 provider locations in the Houston area (personal communication, Carin Martin).

There are several limitations in this study. The external transmission from Black MSM older than 34 years, women, and non-Black MSM were not included in the model. This is because of limited evidence that supports infections among YBMSM originating from these populations and high levels of racial homophily among Black MSM,44,45 although results are somewhat mixed.46 Ongoing extensions to the model are expanding the representation of race/ethnicities, gender identities, and age groups included. Another limitation is that the target for a functional zero is not precisely defined yet. Research on a precise statistical definition of a functional zero is underway, and such a definition will help in planning better focused policy prescriptions. We also did not have data on the frequency of anal sex among YBMSM in Houston. We did, however, have comparable information for a similar population (younger Black MSM between 18 and 34 years of age) in Chicago. This frequency of anal sex produced calibration outcomes that were consistent with our Houston targets. Further research on estimating this parameter in Houston could help address this gap. In addition, our model does not take into account expanded coverage that may exist from local safety-net programs, such as financial assistance or sliding scale medical care through Federally Qualified Health Centers (FQHC) or Harris Health System’s Financial Assistance Program.47

Improved ART and PrEP engagement have shown promise in reducing HIV transmission in a number of US populations that bear a disproportionate burden of the epidemic. Expanding Medicaid, as a number of states have already done, may help substantially reduce the burden of HIV among underserved populations in Houston and throughout Texas. Recently announced, the federal mandate that nearly all medical insurers must eliminate cost sharing for PrEP,48 may prompt substantial acceleration toward the goal of HIV elimination; however, without concerted efforts to reach the uninsured, this change has the potential to increase rather than to decrease the disparities in HIV incidence that disproportionately affects Black MSM. Modeling studies can continue to provide insight as updates to policies are planned and implemented.

ACKNOWLEDGMENTS

The authors acknowledge the Houston Health Department’s NHBS team for data collection. The authors also acknowledge Carin Martin (Ryan White Grant Administration, Harris County Public Health) for her contributions to this study.

REFERENCES 1. Centers for Medicare and Medicaid Services. CMS Roadmaps overview. Department of Health and Human Services. 2021. Available at: https://www.cms.gov/medicare/quality-initiatives-patient-assessment-instruments/qualityinitiativesgeninfo/downloads/roadmapoverview_oea_1-16.pdf. Accessed April 24, 2021. 2. Baugher AR, Finlayson T, Lewis R, et al. Health care coverage and preexposure prophylaxis (PrEP) use among men who have sex with men living in 22 US cities with vs without Medicaid Expansion, 2017. Am J Public Health. 2021;111:4. 3. Kaiser Family Foundation. Status of State Medicaid Expansion Decisions: Interactive Map. 4. 2022 World Population by Country. World Population Review. 2022. Available at: https://worldpopulationreview.com/us-cities/houston-tx-population.Accessed June 5, 2022. 5. United States Census Bureau. Quick Facts: Houston City, Texas. 2019. Available at: https://www.census.gov/quickfacts/houstoncitytexas. Accessed June 2, 2021. 6. Tolbert J, Orgera K, Damico A Key Facts about the Uninsured Population. Kaiser Family Foundation. 2021. Available at: https://www.kff.org/uninsured/issue-brief/key-facts-about-the-uninsured-population/. Accessed June 2, 2021. 7. Harper KB Texas House votes down budget amendment aimed at giving health coverage to more uninsured Texans. The Texas Tribune. Available at: https://www.texastribune.org/2021/04/22/texas-house-medicaid-expansion-uninsured/?utm_source=articleshare&utm_medium=social. Accessed April 22, 2021. 8. Garfield R, Orgera K, Damico A The Coverage Gap: Uninsured Poor Adults in States that Do Not Expand Medicaid. Kaiser Family Foundation. 2021. Available at: https://www.kff.org/medicaid/issue-brief/the-coverage-gap-uninsured-poor-adults-in-states-that-do-not-expand-medicaid/. Accessed June 2, 2021. 9. Centers for Disease Control and Prevention. HIV Surveillance Report, 2018 (Updated); vol. 31. Available at: http://www.cdc.gov/hiv/library/reports/hiv-surveillance.html. Accessed June 2, 2021. 10. Abbas UL, Hallmark CJ, McNeese M, et al. Human immunodeficiency virus in the State of Texas of the United States: past reflections, present shortcomings, and future needs of the public health response. Open Forum Infect Dis. 2020;7:10. 11. Elopre L, McDavid C, Brown A, et al. Perceptions of HIV pre-exposure prophylaxis among young, Black men who have sex with men. AIDS Patient Care STDS. 2018;32:511–518. 12. Elopre L, Hussen SA, Ott C, et al. A qualitative study: the journey to self-acceptance of sexual identity among young, Black MSM in the South. Behav Med. 2021;47:324–334. 13. Serota DP, Rosenberg ES, Thorne AL, et al. Lack of health insurance is associated with delays in PrEP initiation among young Black men who have sex with men in Atlanta, US: a longitudinal cohort study. J Int AIDS Soc. 2019;22:10. 14. Coy KC, Hazen RJ, Kirkham HS, et al. Persistence on HIV preexposure prophylaxis medication over a 2-year period among a national sample of 7148 PrEP users, United States, 2015 to 2017. J Int AIDS Soc. 2019;22:2. 15. Doblecki-Lewis S, Liu A, Feaster D, et al. Healthcare access and PrEP continuation in San Francisco and Miami after the US PrEP Demo Project. JAIDS J Acquir Immune Defic Syndr. 2017;74:5. 16. Whitfield THF, John SA, Rendina HJ, et al. Why I quit pre-exposure prophylaxis (PrEP)? A mixed-method study exploring reasons for PrEP discontinuation and potential re-initiation among gay and bisexual men. AIDS Behav. 2018;22:11. 17. Morgan E, Ryan DT, Newcomb ME, et al. High rate of discontinuation may diminish PrEP coverage among young men who have sex with men. AIDS Behav. 2018;22:11. 18. Liu AY, Cohen SE, Vittinghoff E, et al. Preexposure prophylaxis for HIV infection integrated with Municipal- and Community-Based Sexual Health Services. JAMA Intern Med. 2016;176:1. 19. Adamson B, Lipira L, Katz AB. The impact of ACA and Medicaid expansion on progress toward UNAIDS 90-90-90 Goals. Curr HIV/AIDS Rep. 2019;16:105–112. 20. Fayaz Farkhad B, Holtgrave DR, Albarracín D. Effect of Medicaid expansions on HIV diagnoses and pre-exposure prophylaxis use. Am J Prev Med. 2021;60:335–342. 21. Karletsos D, Stoecker C. Impact of Medicaid expansion on PrEP utilization in the US: 2012–2018. AIDS Behav. 2021;25:1103–1111. 22. Khanna AS, Schneider JA, Collier N, et al. A modeling framework to inform preexposure prophylaxis initiation and retention scale-up in the context of “Getting to Zero” initiatives. AIDS. 2019;33:1911–1922. 23. Bruch E, Atwell J. Agent-based models in empirical social research. Sociol Methods Res. 2015;44:186–221. 24. National Health Care for the Homeless Council. Medicaid Expansion & Criminal Justice-Involved Populations: Opportunities for the Health Care for the Homeless Community. 2013. Available at: https://www.soa.org/Files/Sections/health-MedicaidExpansion-Justice-Final.pdf. 25. Khanna AS, Edali M, Ozik J, et al. Projecting the number of new HIV infections to formulate the “Getting to Zero” strategy in Illinois, USA. Math Biosci Eng. 2021;18:3922–3938. 26. Robins G, Snijders T, Wang P, et al. Recent developments in exponential random graph (p*) models for social networks. Soc Networks. 2007;29:192–215. 27. Singleton AL, Marshall BDL, Zang X, et al. Added benefits of pre-exposure prophylaxis use on HIV incidence with minimal changes in efficiency in the context of high treatment engagement among men who have sex with men. AIDS Patient Care STDS. 2020;34:12. 28. Hamilton DT, Rosenberg ES, Sullivan PS, et al. Modeling the impact of PrEP programs for adolescent sexual minority males based on empirical estimates for the PrEP continuum of care. J Adolesc Health. 2021;68:3. 29. Jenness SM, Knowlton G, Smith DK, et al. A decision analytics model to optimize investment in interventions targeting the HIV preexposure prophylaxis cascade of care. AIDS. 2021;35:9. 30. Handcock MS, Hunter DR, Butts CT, et al. statnet: Software tools for the Statistical Modeling of Network Data. 2003. 31. Collier N, North M. Parallel agent-based simulation with repast for high performance computing. Simulation. 2013;89:1215–1235. 32. Large-scale agent-based modeling with repast HPC: a case study in parallelizing an agent-based model. Vol. 9523. Springer International Publishing; 2015:454–465. Available at: http://link.springer.com/10.1007/978-3-319-27308-2_37. Accessed May 7, 2020. 33. North MJ, Collier NT, Vos JR. “Experiences Creating Three Implementations of the Repast Agent Modeling Toolkit,”. ACM Transactions on Modeling and Computer Simulation. 2006;16:1–25; ACM, New York, NY. 2006;16:1-25. 34. Fujimoto K, Cao M, Kuhns LM, et al. Statistical adjustment of network degree in respondent-driven sampling estimators: venue attendance as a proxy for network size among young MSM. Soc Networks. 2018;54:118–131. 35. Finlayson T, Cha S, Xia M, et al. Changes in HIV preexposure prophylaxis awareness and use among men who have sex with men—20 Urban Areas, 2014 and 2017. MMWR Morb Mortal Wkly Rep. 2019;68:27. 36. Siegler AJ, Mehta CC, Mouhanna F, et al. Policy- and county-level associations with HIV pre-exposure prophylaxis use, the United States, 2018. Ann Epidemiol. 2020;45:24–31.e3. 37. Furl R, Watanabe-Galloway S, Lyden E, et al. Determinants of facilitated health insurance enrollment for patients with HIV disease, and impact of insurance enrollment on targeted health outcomes. BMC Infect Dis. 2018;18:1. 38. Simon K, Soni A, Cawley J. The Impact of Health Insurance on Preventive Care and Health Behaviors: Evidence from the First Two Years