Secondary Logo

Journal Logo

Epidemiology and Prevention

A Data-Driven Simulation of HIV Spread Among Young Men Who Have Sex With Men

Role of Age and Race Mixing and STIs

Beck, Ekkehard C. MS*; Birkett, Michelle PhD; Armbruster, Benjamin PhD*; Mustanski, Brian PhD

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: October 1, 2015 - Volume 70 - Issue 2 - p 186-194
doi: 10.1097/QAI.0000000000000733

Abstract

INTRODUCTION

Between 2006 and 2009, the number of new HIV infections in the United States attributable to young men who have sex with men (YMSM) aged 13–29 years rose by 34%, resulting in 27% of all new HIV infections in 2009 attributed to YMSM.1,2 In contrast, the overall number of new HIV infections has remained stable at an estimated 50,000 cases per year throughout the same period.2 Additionally, there are substantial racial/ethnic disparities in HIV among YMSM, with black YMSM accounting for more than 50% of new HIV infections among YMSM in 2009,1,2 and racial and ethnic minority YMSM have a higher incidence of HIV compared with white YMSM.3 For example, black YMSM are estimated to have a 3- to 6-fold increased annual HIV incidence compared with white YMSM.3,4

Our understanding of the high incidence and racial/ethnic disparities in HIV among YMSM is limited.5,6 There is evidence that some individual-level mechanisms, such as sexually transmitted infections (STIs), contribute to an elevated HIV risk among YMSM.5 In particular, rectal infections of gonorrhea (NG) and chlamydia (CT) have received attention as possible drivers4,7 because of (1) the biological evidence that NG and CT increase the susceptibility and transmissibility of HIV,8–11 (2) the empirical evidence of a 2–3 times greater rectal prevalence of NG and CT compared with urethral prevalence in YMSM,4,12 and (3) the estimates of rectal testing rates being 7 and 9 times lower compared with urethral testing rates among MSM.7 However, it remains unclear to what extent urethral and rectal NG and CT infections contribute to the high HIV incidence among YMSM.13,14

However, individual-level mechanisms alone do not adequately explain the observed racial/ethnic disparities in HIV incidence among YMSM.13,15,16 Several network-level and contextual mechanisms have been hypothesized, but most evidence is inconclusive.5 Among these, age-assortative and race-assortative mixing are commonly hypothesized to contribute to racial/ethnic disparities and high incidence among YMSM.17–22 Particularly, partnerships of YMSM with older MSM are assumed to be significant drivers of racial/ethnic disparities and higher HIV incidence in YMSM because of the elevated HIV prevalence and differences in HIV prevalence among older MSM partners.19,20,23 However, only a limited number of studies focus on understanding these network-level mechanisms,19,20,24 and results are mixed and inconclusive,5 thus underscoring the need for additional research.

Traditional epidemiological and statistical study designs may not be sufficient to fully explain and understand the complex HIV epidemic among YMSM.4,25 Decomposing this complex problem into several distinct analyses of hypothesized mechanisms not only may result in an inaccurate estimate of single mechanisms but also could potentially fail to detect important interactions between these mechanisms.25 Epidemic modeling, in particular simulation-based approaches, provides the opportunity to study such complex systems and examine both main effects and interactions of hypothesized mechanisms.4,25 Despite their utility for examining the HIV epidemic among YMSM, no epidemic model has yet been developed to study the impact of the broad range of hypothesized mechanisms and their interactions on HIV spread among YMSM.

In this study, we developed a data-driven agent-based dynamic network simulation model to study HIV spread among YMSM. Using this novel model, we studied the impact of age-assortative and race-assortative mixing and NG and CT infections on HIV incidence and racial disparities among YMSM. We parameterized the simulation model using the data of an ongoing longitudinal cohort study of YMSM in Chicago.

METHODS

Data-Driven Simulation Model

We developed a discrete-time stochastic agent-based dynamic network simulation model26,27 to study HIV spread among a YMSM population aged 16–21.8 years over 15 years. The simulation model consisted of 2 major components: the partnership formation and dissolution model simulating the sexual partnership behavior of YMSM and the disease transmission model simulating the transmission of HIV, NG, and CT across sexual partnerships. The online Appendix provides a detailed description of the underlying empirical study (see online Appendix section Supplemental Digital Content 1, http://links.lww.com/QAI/A706), the design and parameterization of the simulation model (see online Appendix section Supplemental Digital Content 2, http://links.lww.com/QAI/A706 and section Supplemental Digital Content 3, http://links.lww.com/QAI/A706), the design of the population size and race mix, the aging-in, death, and aging-out processes (see online Appendix section Supplemental Digital Content 4, http://links.lww.com/QAI/A706), the implementation and validation (see online Appendix section Supplemental Digital Content 5, http://links.lww.com/QAI/A706), the counterfactual scenarios (see online Appendix section Supplemental Digital Content 6, http://links.lww.com/QAI/A706), and further details on the results presented in this article (see online Appendix section Supplemental Digital Content 7, http://links.lww.com/QAI/A706).

Partnership Formation and Dissolution Model

The design of the partnership formation and dissolution model was informed using data from the Crew 450 study.28,29 We modeled partnerships by first splitting sexual partner type by those who were one-night partnerships (“one-night partners”) and those who were extended partners. Then, we further divided extended partners into both partners who were also YMSM (“within partners”) and partners who were not YMSM (“outside partners”). Therefore, there are 3 major types of partnerships: “one-night partnerships,” “outside-partnerships,” and “within-partnerships” (see online Appendix section Supplemental Digital Content 2, http://links.lww.com/QAI/A706). Despite being a study of YMSM, including outside partners (those older than 21.8 years at baseline or female) increases the accuracy of our model vs. others,30 as these relationships are prevalent within our sample and differ in associated risks. For example, 27.3% of the 421 YMSM within the Crew 450 cohort identified their sexual orientation as something other than “homosexual-only” or “mostly homosexual,” and 11.3% of all sex contacts named by YMSM at data collection waves T1 and T2 were female (see online Appendix section Supplemental Digital Content 1, http://links.lww.com/QAI/A706).

One-night partnerships, outside-partnerships, and within-partnerships are each modeled in 1 of 2 ways within our simulation. Within-partnerships are modeled as a tie in the network of YMSM, but both one-night partnerships and outside-partnerships are not. Instead, these partnerships are modeled as attributes of YMSM in the network and are dependent on the individual attributes and the sexual momentary degree of the individual YMSM. We chose to model the networks of within-partners as ties because the strongest empirical data available to us were around YMSM and their partners who were also YMSM (see online Appendix section Supplemental Digital Content 2.2.1, http://links.lww.com/QAI/A706).

We assumed that each YMSM had a sexual tendency shaped by their self-reported sexual orientation, their desired sex role, and their desired sex frequency (see online Appendix section Supplemental Digital Content 2.4, Figure 7, http://links.lww.com/QAI/A706). In our model, the sexual orientation of a YMSM impacts his choice of forming a partnership either with a man or a woman. Additionally, because the desired sex role and sex frequency might differ from the actual sex role behavior and sex frequency in a partnership, a novel approach was used in which we modeled the desired sex role and sex frequency as latent variables that influence the actual sex role behavior and sex frequency in a partnership. Actual sexual behavior was then modeled as a function of the sexual tendency of each individual YMSM, the sexual tendency of his partner, and the overall sexual behavior among the YMSM cohort (see online Appendix section Supplemental Digital Content 2.4, http://links.lww.com/QAI/A706).

After the partnership formation, partnership attributes such as oral-sex only, seriousness, mean length and propensity of unprotected anal and vaginal intercourse are determined in sequence using probability estimates derived from multivariate regression models (see online Appendix sections Supplemental Digital Content 2.2.3, 2.2.4, and 2.2.5, http://links.lww.com/QAI/A706). We assumed that outside- and within-partnerships dissolve at each time step with a probability determined by the mean duration for each partnership31 (see online Appendix section Supplemental Digital Content 2.3, http://links.lww.com/QAI/A706).

Disease Transmission Model

We model simultaneous HIV, NG, and CT spread among YMSM (see online Appendix section Supplemental Digital Content 3, http://links.lww.com/QAI/A706) where sexually active YMSM could become infected with HIV, NG, and CT having either penile-vaginal or insertive anal intercourse in female–male partnerships or receptive or insertive anal intercourse in male–male partnerships. Transmission through oral sex was not considered because of the very low transmission risk for HIV (ie, 0.04% per sex act32) and missing evidence about the pharyngeal-to-urethral transmission risk for NG and CT.33

The level of infectiousness of an HIV-positive individual differed by time since infection,34 use of antiretroviral therapy (ART),35 and full or partial viral suppression,30,36 all of which were stratified by race. HIV-infected YMSM initiated ART only if they tested positive and an appropriate amount of time had passed since their exposure to the HIV infection, reflecting current access to treatment and treatment levels30,37 (see online Appendix section Supplemental Digital Content 3.2, http://links.lww.com/QAI/A706). We assumed that on average all outside partners stratified by race and sex have the same HIV prevalence.31,32 Thus, outside partners were randomly assigned to be HIV-positive or HIV-negative based on their race and gender. HIV, NG, and CT prevalence of outside male partners was updated over time due to aging-out of YMSM (see online Appendix sections Supplemental Digital Content 3.2.2 and Supplemental Digital Content 7.1, http://links.lww.com/QAI/A706). The infectiousness of HIV-infected outside partners was also stratified by sex and race.

We assumed increased HIV susceptibility and HIV transmissibility in case of an infection with NG or CT, stratified by site of infection (urethra or rectum)8–11 (see online Appendix section Supplemental Digital Content 3.3.3, http://links.lww.com/QAI/A706). Because of missing or limited biological evidence, we assumed NG and CT infections to be independent of each other, as well as of HIV infection and ART, and that these factors would not impact the spread and course of NG and CT. YMSM and outside partners could have HIV, NG, and CT infections simultaneously. For NG and CT, we assumed that only 1 site, ie, urethra or rectum, could be infected because of the unknown pharyngeal-to-urethral transmission risk and the low prevalence of dual-site infections in particular for CT7,38 and that the newly infected site is complementary to the infected body site of the sex partner (ie, having sex with a rectally infected can only result in an urethral infection). The course of rectal and urethral NG and CT infections was stratified by type of infection (symptomatic vs. asymptomatic), treatment-seeking behavior, and the decision to cease sex while being infected. Individuals with an asymptomatic infection could only receive treatment if they tested positive (see online Appendix sections Supplemental Digital Content 3.3.1 and 3.3.2, http://links.lww.com/QAI/A706).

Parameterization

Longitudinal Cohort Study: Crew 450

Empirical data were used from an ongoing longitudinal study of 450 Chicago YMSM, with study recruitment starting in December 2009 and ending in February 2013. After baseline (T1), data were collected every 6 months. Retention between waves was high, with 86.7% of participants completing the assessment at T2. An individual was eligible for participation if they were between the ages of 16 and 20 years at baseline, birth sex male, spoke English, reported a sexual encounter with a male or an identity of gay/bisexual, and was available for 2 years of follow-up. Participants were recruited through a modified form of respondent-driven sampling28,39 (see online Appendix section Supplemental Digital Content 1, http://links.lww.com/QAI/A706).

Model Parameterization

We simulate a YMSM population of size n = 4484 YMSM over 15 years where the total population size increases by 0.264% each year.40 Size and race mix were chosen such that the simulated population is representative of the YMSM population aged 16–21.8 years in Chicago. YMSM age-in, die, or age-out of the simulated population over time (see online Appendix section Supplemental Digital Content 4.2, http://links.lww.com/QAI/A706). The partnership formation and dissolution model was parameterized using data of n = 421 YMSM enrolled in the Crew 450 at T1 and 6-month follow-up (T2). The fixed age range of the simulated YMSM population was determined by the age range of these YMSM across T1 and T2. Multivariate regression analysis was performed, and significant regression covariates were used to predict partnership formation rates and partnership attributes such as the seriousness or mean length of a partnership. Other partnership attributes including race and sex of the partner mixing probabilities (ie, for one-night partnerships and outside-partnership) were calculated using available partnership data from T1 and T2 (see online Appendix sections Supplemental Digital Content 2.2.3, 2.2.4, and 2.2.5, http://links.lww.com/QAI/A706).

We parameterized our disease transmission model using biomedical testing data from Crew 450, data from the Chicago Department of Public Health (CDPH),41,42 and other publicly available surveillance data,12,43,44 as well as estimates published in the literature (see online Appendix section Supplemental Digital Content 3, http://links.lww.com/QAI/A706).

Validation

To validate the model, we compared the biomedical findings of the Crew 450 study with our simulated results. Figure 18 in the online Appendix section Supplemental Digital Content 5.2, http://links.lww.com/QAI/A706 shows the simulated HIV incidence per 100 person-years compared with the empirical estimates of the Crew 450 study after 3.5 years. The simulated results are within the 95% confidence intervals (CIs) of the empirical results. Further details of the validation are available in online Appendix section Supplemental Digital Content 5.2, http://links.lww.com/QAI/A706 where besides sensitivity analysis, the outcomes of the NG and CT transmission and partnership formation model are compared with estimates of the Crew 450 study and other published findings. The comparison of simulated biomedical, partnership formation, and network topology measurements with empirical findings as well as the results of the sensitivity analysis show an appropriate validation of our simulation model.

Simulation Studies

To determine the impact of age-assortative mixing on HIV spread among YMSM, we first examine the overall HIV incidence per 100 person-years stratified by partnership type. Second, we determine the HIV incidence per male–male partnership-years because male–male partnerships are the main mode of transmission and HIV incidence per 100 person-years equals the HIV risk on the dyad-level, ie, per partnership-years, multiplied with the actual number of partnerships (see online Appendix section Supplemental Digital Content 7.2, http://links.lww.com/QAI/A706). Race-assortative mixing is assumed to maintain and potentially increase racial disparities in HIV among YMSM because of significant differences in HIV prevalence among races.17,45 To isolate the impact of the HIV risk of a partnership attributable to race-assortative mixing from the impact of racial differences in HIV prevalence, we use 2 counterfactual scenarios: one with no race-assortative mixing and one with no racial differences in HIV prevalence (see online Appendix section Supplemental Digital Content 6, http://links.lww.com/QAI/A706). Additionally, to quantify the impact of HIV transmission through older MSM, we use a counterfactual assuming no HIV transmission occurs in outside-partnerships (ie, partnerships with older MSM or females) and another assuming a 50% reduction in HIV transmission risk for YMSM in outside-partnerships. Finally, we quantify the overall impact of NG and CT infections on HIV spread among YMSM by comparing the total HIV incidence of a counterfactual where there is no increased HIV transmissibility and susceptibility due to NG and CT infections with the base-case (see online Appendix sections Supplemental Digital Content 3.3.3 and Supplemental Digital Content 5.4, http://links.lww.com/QAI/A706).

Results shown in the following section are statistically significant (ie, nonoverlapping 95% CIs). Results are expressed as mean values. The half-width of the 95% CI is ≤1.5% of the mean, unless otherwise stated.

RESULTS

HIV Epidemic Over Time

Black YMSM experienced the highest HIV prevalence and incidence compared with Latino and white YMSM (Figs. 1A, B), but only a moderate increase in HIV prevalence and incidence across the modeled 15 years (ie, HIV incidence increased 1.59-fold). However, Latino and white YMSM experienced steeper increases in HIV prevalence and incidence (ie, HIV incidence increased 1.97-fold for Latino YMSM and 2.03-fold for white YMSM). A total of 3076 YMSM were newly infected with HIV over 15 years (ie, 1220 black YMSM vs. 836 Latino and 770 white YMSM) (see also Table 37 in Supplemental Digital Content 5.2, http://links.lww.com/QAI/A706).

FIGURE 1
FIGURE 1:
Simulated HIV prevalence (A) and incidence per 100 person-years (B) stratified by race over 15 years. B, shows the mean HIV incidence per 100 person-years per time step (ie, 0.5 months) with the half-width of the 95% CI ≤3.5% of the mean except for other YMSM.

Age and Race Mixing

Overall, 44.4% of all new HIV infections among YMSM occurred in within-partnerships, 34.5% in outside-partnerships, and 21.1% in one-night partnerships as shown in Figure 2. These proportions varied marginally across races and time, ie, except for other YMSM deviations to the above fractions were within 5 percentage points.

FIGURE 2
FIGURE 2:
Simulated new HIV infections per 100 person-years over 15 years among YMSM stratified by race and relationship type. “Total” shows the simulated number of total HIV infections that occurred in all sexual relations. “One-night”: simulated new HIV infections that occurred in one-night partnerships; “Outside”: simulated new HIV infections that occurred in outside-partnerships, ie, with older male or female partners. HIV infections from females were rare. In outside-partnerships with females, 0.0049 HIV infections per 100 person-years (95% CI: 0.0040 to 0.0057) occurred. In one-night partnerships with females, 0.0035 (95% CI: 0.0029 to 0.0041) occurred. “Within”: simulated new HIV infections that occurred in within-partnerships, ie, partnerships with other YMSM.

We examined male–male partnerships as the main mode of HIV transmission among YMSM with a specific focus on black YMSM given their high incidence; HIV incidence per 100 male–male partnership-years was highest for black–black and non-black–(older) black outside-dyads (Fig. 3A). Without race-assortative mixing (ie, partners are selected without regarding their race; Fig. 3B), HIV incidence decreased for black–non-black outside-partnerships and for all within-partnerships, whereas HIV incidence for non-black–non-black outside-partnerships increased. Assuming the same HIV prevalence for all male outside partners (17.2%) and the same HIV prevalence for all YMSM (5.6%) at baseline t = 0 (Fig. 3C), differences in HIV incidence for male–male outside-partnerships across racial combinations almost vanished with the HIV incidence for black–non-black outside-partnerships being marginally higher compared with black–black and non-black–non-black outside-partnerships. In a counterfactual scenario with both (Fig. 3D), we observe an increase in HIV incidence for non-black–non-black outside-partnerships and a decrease in HIV incidence for all within-partnerships compared with the counterfactual shown in Figure 3C.

FIGURE 3
FIGURE 3:
HIV infections per 100 partnership-years, male–male partnerships only. A, Base-case scenario corresponding to Figures 1, 2. HIV infections per 100 male–male partnership-years in case of one-night partnerships denote HIV infections per average number of one-night partnerships per year. For within-partnerships, the number of partnership-years is the sum of the number of susceptible-infected partnership-years plus 2 times the number of susceptible-susceptible partnership-years (see online Appendix section Supplemental Digital Content 7.2 for details). B, Counterfactual scenario with no race-assortative mixing, ie, YMSM select partners independent of race. C, Counterfactual scenario with no difference by race in initial HIV prevalence (5.6% at baseline) and among outside partners (set to 17.2%). D, Counterfactuals of (B) and (C) combined. In case of outside-partnerships, ie, partnerships of YMSM with older MSM, black–non-black partnerships denote both partnerships of black YMSM with older non-black MSM and partnerships of non-black YMSM with older black MSM (see also Figure 27 in online Appendix section Supplemental Digital Content 7.2, http://links.lww.com/QAI/A706).

As shown in Figures 3A–D, HIV incidence in outside-partnerships was always higher than that in within-partnerships. With no HIV transmissions occurring in outside-partnerships, total HIV incidence would decrease by 61.9% (95% CI: 61.57 to 62.19) (Fig. 4D) and HIV prevalence be close to steady state after 5 years (Fig. 4C). If HIV transmission risk is reduced by 50% in outside-partnerships, HIV infections would decrease by 26.6% (95% CI: 26.13 to 27.06) (Figs. 4B, D). In both counterfactuals scenarios, racial disparities decreased but remained significant.

FIGURE 4
FIGURE 4:
HIV prevalence stratified by race and total number of new HIV infections over 15 years (D) for the base-case scenario (A) corresponding to Figures 1, instead of a counterfactual scenario where no HIV transmission occurs in outside-partnerships of YMSM with older MSM or females (C); and for a counterfactual scenario where HIV transmission risk in outside-partnerships is reduced by 50% compared with the base-case scenario (B). HIV infections from females were rare, see also the caption of Figure 2.

Gonorrhea and Chlamydia

Using a counterfactual where NG and CT do not affect HIV transmission, we found that the fraction of HIV infections attributable to NG or CT was 14.6% (95% CI: 14.1 to 15.2). In the base-case, 66.4% of these HIV infections were attributable to rectal NG or CT infections. Also, 41.7% of all HIV infections attributable to NG or CT in this scenario were attributable to increased HIV susceptibility due to NG or CT infection of an HIV-negative individual.

DISCUSSION

We developed a novel data-driven simulation model of HIV spread among YMSM. Our study focused on the impact of age and race mixing and STIs on HIV incidence and racial disparities and was motivated by the limited understanding about the impact of these mechanisms on HIV spread among YMSM.5

Over 15 years, the HIV epidemic among YMSM continued to rise with an estimated 3076 new HIV infections. Racial disparities also continued to persist, but increases in HIV prevalence and incidence differed by race, with Latino and white YMSM facing greater increases. These data map onto YMSM specific data from the CDPH, which show an overall increase in HIV diagnoses among YMSM and steeper increases in HIV diagnoses and prevalence46 in Latino/white YMSM vs. black YMSM (see online Appendix section Supplemental Digital Content 5.2, http://links.lww.com/QAI/A706). Furthermore, our estimates of the total number of HIV infections among YMSM are within the 95% CI of the CDPH HIV incidence estimates of 15- to 24-year-old YMSM adjusted for age range and the fraction of HIV-infected YMSM being unaware of their HIV infection (see online Appendix section Supplemental Digital Content 5.2, http://links.lww.com/QAI/A706). These results alone suggest that HIV will continue to place a heavy burden on the YMSM population. While black YMSM will continue to be disproportionately impacted, our model suggests that racial disparities in this group will decrease because of increasing incidence in white and Latino YMSM, but unfortunately not because of declining incidence among black YMSM.

Our results indicate that approximately 45% of all HIV infections among YMSM occurred in partnerships between YMSM. While in a partnership, we find that the risk of HIV infection is highest across partnership types for black YMSM with an older black MSM. This is consistent with findings from studies using other methods.18,20,47,48 Differences in HIV prevalence among races, particularly the high HIV prevalence among black MSM, drive the high HIV risk of a black YMSM–black (older) MSM partnership compared with other non-black–(older) non-black partnerships, confirming the hypotheses of multiple studies.5,17,22,45 While 34.5% of all HIV infections among YMSM in the simulation occur in partnerships with older MSM or females, a hypothetical scenario where no HIV transmissions happen in such partnerships decreases HIV infections among YMSM by 61.9%. Thus, each YMSM transmission avoided from an older MSM partner will also prevent an additional 0.8 HIV infections among YMSM. Therefore, prevention should target the reduction of HIV transmission in age-disassortive partnerships.

Simulating the simultaneous spread of HIV, NG, and CT among YMSM, we determined the fraction of HIV infections attributable to NG or CT to be 14.6%, a proportion within the range of the few estimates reported. Chesson and Pinkerton49 used a simple modeling approach to estimate the fraction of HIV infections attributable to NG or CT in the adult heterosexual US population to be between 4.6% and 9.2%. Another modeling study among MSM in the Netherlands50 estimated the fraction of HIV infections solely attributable to CT to be 15.2%. Among the 14.6% of all HIV infections attributable to NG or CT in our study, 66.4% and 41.7% were due to rectal infections and increased susceptibility, respectively. Rectal testing of NG and CT is rare within the population,7 and these data suggest further evaluation of HIV and STI testing policies is necessary to determine a holistic and cost-optimal testing strategy.7,14

This study has several limitations. First, we simulated the HIV spread among a cohort of Chicago YMSM over time, which limits the generalizability of our results to YMSM populations in other US cities or older age groups. However, both the comparability of the empirical estimates from the Crew 450 study with the estimates of other studies3,4 and the validation of the model suggest the applicability of our findings to other YMSM populations. Furthermore, as the input parameters of the disease transmission model are based on few data sources, they may be biased due to sampling error; especially, in the case of NG and CT, parameter estimates were difficult to obtain and estimates varied widely,51 highlighting the need for more accurate parameter estimates of NG, CT, and their interaction with HIV in this context. Furthermore, modeling only within-partnerships as a network did not allow us to examine whether potential network effects on HIV transmission observed in within-partnerships also apply to one-night and outside-partnerships. Finally, to parameterize the partnership formation model, we used multivariate regression analysis where only significant parameters were used to predict partnership formation rates and partnership attributes. Thus, certain effects hypothesized in other studies were observable but not significant and therefore not included in the model. This could contribute to the fact that racial/ethnic disparities had a lower magnitude in our simulated results within the first 3.5 years and thus might also influence racial differences in the increase in HIV prevalence and incidence over 15 years.

Using an agent-based dynamic network simulation model of HIV spread among YMSM, we demonstrated first that the HIV epidemic among YMSM continues to rise especially among Latino and white YMSM; second, racial disparities in HIV risk per partnership are mostly driven by differences in HIV prevalence among older MSM partners; third, YMSM and in particular black YMSM having an older black MSM partner are at highest risk; and fourth, NG and CT, particularly rectal infections, account for a sizeable portion of all HIV infections. These results emphasize the need for HIV prevention efforts targeting all YMSM, holistic HIV and STI testing strategies and suggest that prevention interventions focusing on transmission between YMSM and older MSM might be highly effective.

ACKNOWLEDGMENTS

The authors thank the Chicago Department of Public Health for providing detailed data about HIV diagnoses among 15- to 24-year old YMSM in Chicago from 2009 to 2013; the reviewers for their constructive and helpful comments and ideas; Gregory L. Philips II, Alexander Gutfraind, Richard D'Aquila, Hank Seifert, Uri Wilensky, and Hendricks Brown for helpful comments; and Daniel Ryan for his assistance in data analysis.

REFERENCES

1. Prejean J, Song R, Hernandez A, et al.. Estimated HIV incidence in the United States, 2006-2009. PLoS One. 2011;6:e17502.
2. Centers for Disease Control and Prevention. CDC Fact Sheet: Estimates of New HIV Infections in the United States, 2006–2009. Atlanta, GA; Centers for Disease Control and Prevention: 2011:1–6.
3. Balaji AB, Bowles KE, Le BC, et al.. High HIV incidence and prevalence and associated factors among young MSM, 2008. AIDS. 2013;27:269–278.
4. Sullivan PS, Peterson J, Rosenberg ES, et al.. Understanding racial HIV/STI disparities in black and white men who have sex with men: a multilevel approach. PLoS One. 2014;9:e90514.
5. Maulsby C, Millett G, Lindsey K, et al.. HIV among black men who have sex with men (MSM) in the United States: a review of the literature. AIDS Behav. 2014;18:10–25.
6. Mustanski BS, Newcomb ME, Du Bois SN, et al.. HIV in young men who have sex with men: a review of epidemiology, risk and protective factors, and interventions. J Sex Res. 2011;48:218–253.
7. Patton ME, Kidd S, Llata E, et al.. Extragenital gonorrhea and chlamydia testing and infection among men who have sex with men—STD Surveillance Network, United States, 2010-2012. Clin Infect Dis. 2014;58:1564–1570.
8. Bernstein KT, Marcus JL, Nieri G, et al.. Rectal gonorrhea and chlamydia reinfection is associated with increased risk of HIV seroconversion. J Acquir Immune Defic Syndr. 2010;53:537–543.
9. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect. 1999;75:3–17.
10. Jin F, Prestage GP, Imrie J, et al.. Anal sexually transmitted infections and risk of HIV infection in homosexual men. J Acquir Immune Defic Syndr. 2010;53:144–149.
11. Koblin BA, Husnik MJ, Colfax G, et al.. Risk factors for HIV infection among men who have sex with men. AIDS. 2006;20:731–739.
12. Hotton A, Gratzer B. Howard Brown Health Center: STI Annual Report, 2011. Chicago, IL: Howard Brown Center; 2012.
13. Millett GA, Peterson JL, Wolitski RJ, et al.. Greater risk for HIV infection of black men who have sex with men: a critical literature review. Am J Public Health. 2006;96:1007–1019.
14. Vriend HJ, Lugner AK, Xiridou M, et al.. Sexually transmitted infections screening at HIV treatment centers for MSM can be cost-effective. AIDS. 2013;27:2281–2290.
15. Millett GA, Flores SA, Peterson JL, et al.. Explaining disparities in HIV infection among black and white men who have sex with men: a meta-analysis of HIV risk behaviors. AIDS. 2007;21:2083–2091.
16. Clerkin EM, Newcomb ME, Mustanski B. Unpacking the racial disparity in HIV rates: the effect of race on risky sexual behavior among black young men who have sex with men (YMSM). J Behav Med. 2011;34:237–243.
17. Berry M, Raymond HF, McFarland W. Same race and older partner selection may explain higher HIV prevalence among black men who have sex with men. AIDS. 2007;21:2349–2350.
18. Joseph HA, Marks G, Belcher L, et al.. Older partner selection, sexual risk behaviour and unrecognised HIV infection among black and Latino men who have sex with men. Sex Transm Infect. 2011;87:442–447.
19. Mustanski B, Birkett M, Kuhns LM, et al.. The role of geographic and network factors in racial disparities in HIV among young men who have sex with men: an egocentric network study. AIDS Behav. 2015;19:1037–1047.
20. Newcomb ME, Mustanski B. Racial differences in same-race partnering and the effects of sexual partnership characteristics on HIV risk in MSM: a prospective sexual diary study. J Acquir Immune Defic Syndr. 2013;62:329–333.
21. Raymond HF, McFarland W. Racial mixing and HIV risk among men who have sex with men. AIDS Behav. 2009;13:630–637.
22. Tieu HV, Murrill C, Xu G, et al.. Sexual partnering and HIV risk among black men who have sex with men: New York City. J Urban Health. 2010;87:113–121.
23. Hurt CB, Matthews DD, Calabria MS, et al.. Sex with older partners is associated with primary HIV infection among men who have sex with men in North Carolina. J Acquir Immune Defic Syndr. 2010;54:185–190.
24. Johnson BT, Redding CA, DiClemente RJ, et al.. A network-individual-resource model for HIV prevention. AIDS Behav. 2010;14(suppl 2):204–221.
25. Maglio PP, Sepulveda MJ, Mabry PL. Mainstreaming modeling and simulation to accelerate public health innovation. Am J Public Health. 2014;104:1181–1186.
26. Bonabeau E. Agent-based modeling: methods and techniques for simulating human systems. Proc Natl Acad Sci U S A. 2002;99:7280–7287.
27. Grimm V, Railsback SF. Individual-Based Modeling and Ecology. Princeton, NJ: Princeton University Press; 2005.
28. Kuhns LM, Kwon S, Ryan DT, et al.. Evaluation of respondent-driven sampling in a study of urban young men who have sex with men. J Urban Health. 2014;92:151–167.
29. Mustanski B, Ryan DT, Garofalo R. Associations of sexually transmitted infections with condom problems among young men who have sex with men. Sex Transm Infect. 2014;41:427–432.
30. Goodreau SM, Carnegie NB, Vittinghoff E, et al.. What drives the US and Peruvian HIV epidemics in men who have sex with men (MSM)? PLoS One. 2012;7:e50522.
31. Krivitsky P, Handock M. A separable model for dynamic networks. J R Stat Soc Series B Stat Methodol. 2014;76:29–46.
32. Vittinghoff E, Douglas J, Judson F, et al.. Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol. 1999;150:306–311.
33. Bissessor M, Tabrizi SN, Fairley CK, et al.. Differing Neisseria gonorrhoeae bacterial loads in the pharynx and rectum in men who have sex with men: implications for gonococcal detection, transmission, and control. J Clin Microbiol. 2011;49:4304–4306.
34. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission, by stage of infection. J Infect Dis. 2008;198:687–693.
35. Cohen MS, Chen YQ, McCauley M, et al.. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
36. Weintrob AC, Grandits GA, Agan BK, et al.. Virologic response differences between African Americans and European Americans initiating highly active antiretroviral therapy with equal access to care. J Acquir Immune Defic Syndr. 2009;52:574–580.
37. Swindells S, Cobos DG, Lee N, et al.. Racial/ethnic differences in CD4 T cell count and viral load at presentation for medical care and in follow-up after HIV-1 infection. AIDS. 2002;16:1832–1834.
38. Kent CK, Chaw JK, Wong W, et al.. Prevalence of rectal, urethral, and pharyngeal chlamydia and gonorrhea detected in 2 clinical settings among men who have sex with men: San Francisco, California, 2003. Clin Infect Dis. 2005;41:67–74.
39. Phillips GKL, Garofalo R, Mustanski B. Do recruitment patterns of young men who have sex with men (YMSM) recruited through respondent-driven sampling (RDS) violate assumptions? J Epidemiol Community Health. 2014;68:1207–1212.
40. U.S. Census Bureau. State and County QuickFacts. Chicago, IL; U.S. Census Bureau: 2015. Available at: http://quickfacts.census.gov/qfd/states/17/1714000.html. Accessed April 1, 2015.
41. Chicago Department of Public Health. HIV/STI Surveillance Report, 2013. Chicago, IL: City of Chicago; 2013.
42. Chicago Department of Public Health. The Heterosexual HIV Epidemic in Chicago: Insights into the Social Determinants of HIV. Chicago, IL: City of Chicago; 2011.
43. Centers for Disease Control and Prevention (CDC), National Center for Injury Prevention and Control. WISQARS (Web-Based Injury Statistics Query and Reporting System). Atlanta, GA; 2015. Available at: http://www.cdc.gov/injury/wisqars/. Accessed April 1, 2015.
44. Jones RC, Harper-Jemison DM, Clark J, et al.. Leading Causes of Death in Chicago, 2007-2009. Chicago, IL; 2013.
45. Morris M, Goodreau S, Moody J. Sexual networks, concurrency, and STD/HIV. In: Sexually Transmitted Diseases. Holmes K, ed. New York, NY: McGraw-Hill; 2007:109–126.
46. Chicago Department of Public Health. HIV Risk and Prevention Behaviors Among Men Who Have Sex With Men, Chicago, 2008 and 2011. Chicago, IL: City of Chicago; 2012.
47. Mustanski B, Newcomb ME. Older sexual partners may contribute to racial disparities in HIV among young men who have sex with men. J Adolesc Health. 2013;52:666–667.
48. Oster AM, Wiegand RE, Sionean C, et al.. Understanding disparities in HIV infection between black and white MSM in the United States. AIDS. 2011;25:1103–1112.
49. Chesson HW, Pinkerton SD. Sexually transmitted diseases and the increased risk for HIV transmission: implications for cost-effectiveness analyses of sexually transmitted disease prevention interventions. J Acquir Immune Defic Syndr. 2000;24:48–56.
50. Xiridou M, Vriend HJ, Lugner AK, et al.. Modelling the impact of chlamydia screening on the transmission of HIV among men who have sex with men. BMC Infect Dis. 2013;13:436.
51. Davies BAS, Turner K, Ward H. How robust are the natural history parameters used in chlamydia transmission dynamic models? A systematic review. Theor Biol Med Model. 2014;11:8.
Keywords:

HIV; STI; men who have sex with men; sexual behavior; sexual partners; mathematical models

Supplemental Digital Content

Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.