Secondary Logo

Journal Logo

Epidemiology

Temporal Variation in One-Time Partnership Rates Among Young Men Who Have Sex With Men and Transgender Women

Janulis, Patrick PhDa,b; Goodreau, Steven M. PhDc; Birkett, Michelle PhDa,b; Phillips, Gregory II PhDa,b; Morris, Martina PhDd; Mustanski, Brian PhDa,b; Jenness, Samuel M. PhDe

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: July 1, 2021 - Volume 87 - Issue 3 - p e214-e221
doi: 10.1097/QAI.0000000000002679

Abstract

INTRODUCTION

Men who have sex with men (MSM) continue to account for most new HIV infections in the United States,1 despite making up less than 5% of the population and the availability of highly effective HIV treatment and prevention tools.2 Current federal initiatives for ending the HIV epidemic3 prioritize efforts to identify, predict, and prevent new HIV infections among key populations such as young MSM, Black MSM, and transgender women who continue to experience extremely high HIV incidence.4 Yet, our ability to reduce HIV incidence in these populations relies on high-quality data to understand transmission risk and to design and deploy preventive interventions.5

Rates of sexual partnership acquisition have proven critical for understanding the spread of HIV and other sexually transmitted infections (STIs). Although between-person heterogeneity in rates have been extensively studied among MSM,7–9 a smaller number of studies have examined temporal variation in partner contact rates. For example, seasonality in contact rates has been documented and may be influenced by health (eg, STI diagnoses) or personal factors,10 specific events such as travel,11 and prevention behaviors such as initiating pre-exposure prophylaxis (PrEP).12,13 There is also evidence of variation in contact rates over longer periods of time14,15 that may reflect both episodic change as well as developmental and historical trends.16,17 Given evidence documenting volatility in sexual contact rates, it is also important to understand the impact of this variation on HIV transmission risk.

Epidemic models have found within-person variation in partnership rates increases the population prevalence of HIV18 and increases the number of acute-stage HIV transmissions.19 This may be partially due to individuals transitioning from low to high contact rates, who are less likely to be HIV positive, face a greater probability of serodiscordant partners as contact rates increase, and are more likely to transmit HIV within the acute-stage during these periods.20 Dynamic variation in risk may also decrease the effectiveness of ‟test-and-treat” strategies for HIV elimination18 but increase the effectiveness of PrEP.21 Yet, many of these studies rely on behavioral data from earlier phases of the HIV epidemic15,22 or that do not represent key populations to the current US HIV epidemic 18,21 such as Black MSM and young MSM. Accordingly, understanding of this phenomenon, among priority populations in the United States, remains limited.

This study aims to estimate variation in one-time partnership contact rates using data from a large longitudinal cohort of young MSM and transgender women from Chicago. We focus our analysis on one-time partners because the short duration of one-time partnerships leads to a strongly right-skewed distribution (ie, a small number of individuals with a large number of one-time partners). Accordingly, longitudinal data on one-time partnership rates are vitally important to accurately understand stability or volatility in contact rates, including among the small number of high contact rate individuals who may have outsized impact on the spread of HIV. This information will be useful in both clinical settings in which risk assessment is conducted and for parameterizing epidemic transmission models.

METHODS

Participants and Procedures

Data come from the RADAR study,23 an accelerated longitudinal24 cohort study of young MSM and transgender women. All participants in the RADAR study were required to meet the following criteria: 16–29 years old at enrollment, male assigned at birth (transgender women and other noncisgender identified individuals were eligible), English speaking, reported a sexual encounter with a man in the previous year or identified as gay or bisexual, and able to attend in-person research visits in Chicago. Recruitment for this study came from 2 prior cohorts of sexual and gender minority adolescents and young adults, Project Q225 and Crew 450,26 enriched with new cohort members recruited through venue-based, online, peer recruitment, and recruitment of significant partners. Expanded details regarding recruitment have been previously described.23 Participants were eligible for one study visit every 6 months. Data for the current analysis come from the beginning of data collection (February 2015) through February 2020. At the time of analysis, there were 1053 active cohort participants enrolled in the RADAR study. However, we limited our analysis to participants who had completed their first 6 study visits using data from the second through the sixth visit, representing roughly 2 years of follow-up starting 6 months after enrollment. There were 944 participants eligible to have completed their sixth visit with 804 completing the sixth visit, indicating a completion rate of 85.2%.

Measures

At each study visit, participants completed an inventory of their sexual, drug use, and social networks, a psychosocial survey, and biomedical specimen collection (HIV test, STI tests, and drug screening). Sexual partnership data were collected during an interviewer-assisted network inventory. Partnerships were captured using ‟name generators” that asked participants to name individuals they “had sex with in the past 6 months.” After elicitation of partners, detailed information regarding sexual partners was collected, including serious partnership status, date of first and last sex (within the past 6 months), and number of anal sex acts with and without a condom. Data on sexual positioning were not collected on these partners.

For the current analysis, we classified partnerships into 3 categories, as is typical for other MSM partnership studies27,28: main (ie, serious), casual, and one-time partners. Main partners were those with whom the participant indicated they were “currently in a serious relationship.” Participants were asked to interpret ‟serious relationship”; however, they would like and told, “this could be someone you call a boyfriend or girlfriend, significant other, partner, or husband or wife.” Casual partners were partners with whom the participant indicated having sex on more than 1 day but were not indicated a main partner. One-time partners were partners whom the participant indicated the same date of first and last sex. One-time partners were considered separate from casual partners because of their strongly right-skewed distribution that may differentially impact transmission dynamics and because condom rates may differ across these 2 partner types. For casual and one-time partners, partner counts reflect the cumulative number of partners in the 6 months before the study visit.

We also include time-fixed and time-varying exposures variables in our analysis. The time-fixed exposures variables were race/ethnicity and gender identity. Time-varying exposure variables included age, HIV status during the previous observation, the number of current casual partners, the existence of a current main partner, number of casual partners during the previous observation, and existence of main partners during the previous observation and the number of one-time partners during the previous observation (ie, autocorrelation). The prior visit HIV status was used to capture the participant's perceived HIV status during the observation period and was measured using Alere Determine fourth Generation HIV-1/HIV-2 Ab/Ag Combo rapid test and confirmatory testing according to CDC guidance.

Analysis

We examined the extent of variation in one-time partnership throughout the observed study period using descriptive statistics and visualizations. First, we examined the number of participants in 4 overlapping groups based on their maximum or minimum number of one-time partners reported throughout the entire study. Second, we plotted the counts of one-time partners reported during a single visit, the number of one-time partners by age, and the number of one-time partners across study visits (ie, wave).

In our primary analysis, we used a series of random effects models to examine correlates of one-time partnership using a negative binomial distribution. Specifically, we used a ‟hybrid” random effects models29,30 to disaggregate within and between variation in our exposure variables (ie, number of casual partners and having a main partner). In these models, person-centered coefficients estimated within-person associations, whereas person-mean coefficients estimated between-person associations.

Modeling took place in 3 stages. First, an intercept-only model was examined using only a random intercept to estimate the extent to which one-time partnership rates vary within and between individuals. An intraclass correlation coefficient (ICC) was estimated adjusting for the negative binomial distribution.31 Next, a model including all exposure variables except time-lagged exposure variables was examined. Finally, a model with all previous variables and time-lagged exposure variables was examined including lagged variables for casual, serious, and one-time partners. Estimating models with lagged-dependent variables is challenging because of the intrinsic correlation between the random intercept and the person-time error (ie, endogeneity) violates a major assumption of linear models29,32 and can downwardly bias estimates of other correlates.33 To examine for bias because of endogeneity, we (1) first examined a model free of lagged terms so we can compare coefficients across models with and without a lagged dependent variable and (2) tested for autocorrelation in the residuals using a Breusch–Godfrey test, which has shown in simulations34 to be a robust indicator for detecting bias. Finally, we performed a sensitivity analysis examining the robustness of our results against potential drops in the number of sexual partners after the first visit by estimating the first 2 models (ie, without lagged variables) with data from the first visit included.

For all models, multiple imputation was used to account for 369 (9.2%) visits with any missing data. Missing observations were not associated with participant level characteristics (ie, race, gender identity, or age) but were more common over time. We used multiple imputation by chained equations using predictive mean matching for count variables and logistic regression for dichotomous variables35,36 with 20 imputed data sets. Aggregate model results are presented using averaged parameter estimates and standard errors.37

RESULTS

Table 1 presents the demographics characteristics of participants (n = 804) and descriptive statistics regarding one-time partners across baseline and all study follow-up visits. The sample was predominantly Black/African American (34.8%) or Hispanic (30.3%) with a mean age at the second visit of 21.7 years (SD = 3.0). In addition, this table includes descriptive statistics regarding categories of participants according to their number of one-time partners across the entire 6 waves of observation. The mean number of partners in a single observation was 0.54 (SD = 1.22) for one-time, 0.91 (SD = 1.59) for casual, and 0.35 (SD = 0.49) for serious. Notably, a sizable minority of participants (42.2%) did not report any one-time partners during the entire study. Furthermore, a small (6.1%) number of participants reported at least a single one-time partner at each visit during the entire follow-up period.

TABLE 1. - Baseline Demographics and One-Time Partnership Characteristics
Demographics n (%)
Race/ethnicity
 Black/African American 280 (34.8)
 White 189 (23.5)
 Hispanic/Latino 244 (30.3)
 Others 91 (11.3)
Gender
 Male 744 (92.5)
 Transgender female 42 (5.2)
 Other 18 (2.2)
Age* 21.7 (3.0)
Perceived HIV status
 HIV positive 131 (16.3)
 HIV negative or unknown 673 (83.7)
One-time partnerships
 No one-time partners 339 (42.2)
 Never >1 539 (67.0)
 Never >2 656 (81.5)
 Never >3 717 (89.2)
 Never <1 49 (6.1)
 Never <2 11 (1.4)
*Age represents mean (SD) at the second visit.
HIV status represents results of HIV at first visit.

Figure 1 shows the distribution of one-time partnerships during a single visit. As indicated, the most common number of partners during a 6-month period is zero with 72.6% of all visits in this category. Figures 2 and 3 use a random sample of 200 to display variation in one-time rates over age and wave. Figure 2 shows the trajectory of one-time partners across age with the mean number for each age group shown in blue. Figure 3 displays a heat map of one-time partners across each observation wave, with each horizontal line representing a single participant. These figures demonstrate that, despite the overall low average number of one-time partners, participants do exhibit time variation in the number of one-time partners. In addition, Supplemental Digital Content (see Figures, https://links.lww.com/QAI/B638) presents the distribution of one-time partners across rates of casual (see Figure, Supplemental Digital Content, https://links.lww.com/QAI/B638) and main partners (see Figure, Supplemental Digital Content, https://links.lww.com/QAI/B638).

F1
FIGURE 1.:
Number of one-time partners during a single visit.
F2
FIGURE 2.:
Number of one-time partners across age. This figure represents data from a random sample of 200 participants for purposes of interpretability. The blue line represents the mean number of one-time partners across age.
F3
FIGURE 3.:
Number of one-time partners across visits. This figure represents data from a random sample of 200 participants. Horizontal changes in color across a single point on the y-axis represent temporal variation in the number of one-time partners.

The intercept-only model (Table 2) indicated that one-time partners varied both within and between individuals (ICC = 0.46), indicating 46% of the total variance in one-time partners occurred between individuals. The second model with exposure variables indicated several associations with one-time partners. Having more casual partners was associated with a higher number of one-time partners for both person-centered {incidence rate ratio (IRR) = 1.17 [95% confidence interval (CI): 1.09 to 1.26]} and person-mean [IRR = 1.64 (95% CI: 1.50 to 1.79)] exposure variables. By contrast, having a main partner was associated with having a lower number of one-time partners for both the person-centered [IRR = 0.54 (95% CI: 0.44 to 0.66)] and person-mean [IRR = 0.47 (95% CI: 0.33 to 0.68)] exposure variables. Finally, the third model with lagged exposure variables indicated that the prior number of one-time partners was positively associated with the current number of one-time partners [IRR = 1.09 (95% CI: 1.03 to 1.16)], whereas all other associations remained in the same direction with similar magnitude. The Breusch–Godfrey tests for the final models were not significant for any of the 20 imputed data sets (P-value range = 0.114–0.903), indicating these models did not have significantly autocorrelated residuals. The sensitivity analysis indicated no differences in results after including data from the first visit.

TABLE 2. - Results of Random Effects Models of One-Time Partners
Parameters Intercept Only Exposure Variables Lagged Exposure Variables
IRR (95% CI) P IRR (95% CI) P IRR (95% CI) P
Intercept 0.28 (0.24 to 0.32) <0.001 0.29 (0.18 to 0.45) <0.001 0.30 (0.19 to 0.46) < 0.001
Casual partners
 Centered 1.17 (1.09 to 1.26) <0.001 1.17 (1.09 to 1.26) < 0.001
 Mean 1.64 (1.50 to 1.79) <0.001 1.56 (1.43 to 1.71) < 0.001
 Centered T-1 0.93 (0.88 to 0.99) 0.022
Serious partners
 Centered 0.54 (0.44 to 0.66) <0.001 0.53 (0.43 to 0.65) < 0.001
 Mean 0.47 (0.33 to 0.68) <0.001 0.50 (0.35 to 0.70) < 0.001
 Centered T-1 1.04 (0.86 to 1.27) 0.689
One-time partners (T-1) 1.09 (1.03 to 1.16) 0.002
Age 0.96 (0.93 to 0.99) 0.017 0.96 (0.93 to 0.99) 0.016
HIV positive 0.80 (0.57 to 1.12) 0.186 0.79 (0.58 to 1.09) 0.155
Gender
 Cisgender man (ref)
 Transgender woman 0.53 (0.28 to 0.99) 0.047 0.54 (0.30 to 0.99) 0.046
 Other 0.91 (0.48 to 1.73) 0.777 0.92 (0.50 to 1.70) 0.797
Race
 Black (ref)
 White 1.75 (1.34 to 2.29) <0.001 1.69 (1.31 to 2.17) <0.001
 Hispanic/Latino 1.56 (1.22 to 2.00) <0.001 1.52 (1.20 to 1.93) 0.001
 Others 2.01 (1.46 to 2.76) <0.001 1.91 (1.41 to 2.58) <0.001
Residual variance 4.52 (3.40 to 6.00) <0.001 2.13 (1.77 to 2.57) <0.001 1.83 (1.50 to 2.24) <0.001
ICC 0.46 0.30 0.26
AIC/BIC 7316/7335 6960/7048 6951/7058
Bold indicate P < 0.05.

DISCUSSION

In this study, we found that young MSM and transgender women reported both between and within-person variation in one-time partners over time. Roughly half of the variance in one-time partners was accounted for by individual differences and half by within-person change. To the extent that one-time partnership rates help explain the spread of HIV and other STIs, these findings suggest focus on high-risk individuals alone will not fully capture the dynamic nature of this risk.

Descriptive statistics indicated a majority of participants (57.8%) reported a one-time partner at least once during the study, but participants also reported no one-time partners during most visits (72.6%). This suggests one-time partners are common but are not a consistent feature of most participants' sexual networks. This finding could have important implications for the spread of HIV as individuals who have low contact rates are more likely to acquire HIV when they transition to high contact rate.20 Although our focus on one-time partner contact rates makes direct comparisons with previous studies challenging because they have mainly examined aggregate risk scores16,38,39 or aggregated across all partner types,14,15 our main findings broadly concur with these studies that although MSM vary in their contact rates and level of risk over time, there remains a preponderance of low-risk behavior. For example, one study39 found that a majority of participants were ‟monogamous” or ‟risk minimizers,” whereas another study found a majority of participants were ‟low risk” in their trajectory,38 despite both studies indicating change in risk over time.

The current study also found that the previous number of one-time partners was weakly associated with the current number of one-time partners, indicating a small but statistically significant positive autocorrelation. This concurs with a previous study which also found a small positive autocorrelation in sexual contact rates among HIV-negative MSM.15 The small magnitude of this autocorrelation combined with substantial within-person variation and high prevalence of low contact rates with one-time partners provides important nuance to existing research on the ‟core” group theory of STI transmission.40,41 This theory posits that a small number of individuals with high levels of turnover in their sexual contacts have a large impact on STI transmission. Although we did find a small group (6.1% of participants) that consistently reported at least a single one-time partner every 6 months, we also found a relatively modest level of autocorrelation and substantial within-person variation over time. Accordingly, these findings could suggest a ‟core group” of individuals is unlikely to maintain a consistently high level of contact rates among one-time partners.

In addition, these results indicated casual and main partnerships play an important role in understanding variation in one-time partner contact rates. We found that having main partners, both within and between-person, was associated with having fewer one-time partners. This finding also concurs with previous studies indicating an inverse association between having a main partner and having nonmain partners among MSM9 as would be expected given MSM report high rates of monogamous partnership agreements with main partners.42 By contrast, the number of casual partners was positively associated with the number of one-time partners, again for both within and between persons. The within-person association (ie, person centered) between casual partners and one-time partners supports the notion of ‟seasons” of higher levels of sexual contact that are associated with health, psychological, and social contextual factors identified in a number of prior studies.10,11,43 However, the between-person association (ie, person mean) suggests that individuals who tend to have higher numbers of casual partners on average also tend to report greater numbers of one-time partners. Accordingly, the highest rates of one-time partnership occur during periods of higher than average contact with casual partners (ie, compared with an individual's own average rate) among individuals who have higher levels of casual partners generally. Therefore, identifying additional correlates of these periods of high contact rates, in addition to the characteristics of the individuals that report above average levels of contacts (ie, a core group), remains an important area for future research.

Together, these findings have important implications for understanding transmission dynamics for HIV in this target population. Variation in risk behavior may substantially impact the spread of HIV and the effectiveness of interventions.15,18–20,44 Previous studies have found temporal variation may reduce the effectiveness of test and treat strategies18 while it could increase the effectiveness of PrEP.21 Accordingly, the observed variation could impact the ideal combination of local HIV interventions.45 These findings also suggest efforts to target high-risk individuals to prevent STIs46 may not yield the expected prevention benefits if these individuals do not maintain consistent levels of risk. However, to the extent that these temporary periods of high-risk can be understood and predicted, tailored interventions could be implemented during these periods. In fact, several studies have found that MSM report selectively using preventions services during periods of high risk and report growing interest for ‟on-demand” PrEP.11,47–51

However, the detailed implications of these findings on the spread of HIV require additional study. Epidemic models that explicitly account for the observed variation are likely to provide the most accurate impact of these behaviors. For example, we found a preponderance of observations with zero one-time partners and the impact of transitioning from zero to one partner on transmission may be particularly important. Epidemic models are well positioned to estimate the impact of this temporal variation on intervention effectiveness,51,52 optimizing prevention policies,53,54 helping efficiently target resources,55,56 and estimating the cost-effectiveness of prevention strategies.57,58 Given the foundational work that has already examined the impact of episodic variation in contact rates on HIV,15,20,44 incorporating new parameters based on the current data to these models provides a promising future direction to contextual these findings and leverage them to inform public health action.

LIMITATIONS

Limitations of the current analysis include the nonprobability sample of RADAR participants. However, this large, recent, and racially/ethnically diverse sample of young MSM provides much needed data examining the extent of temporal variation in contact rates. Furthermore, the focus of the current study on one-time partners makes it challenging to compare our findings with previous studies, especially given the rarity of one-time partners in these samples. Nonetheless, these partners remain important to understand given their strong right skew and limited existing data. Finally, these data had notable missing values that we attempted to correct for using multiple imputation.

CONCLUSIONS

This study found within-person and between-person variation in one-time partner contact rates among a large and diverse sample of young MSM and transgender women. Results indicated a small positive autocorrelation in one-time partnership rates, an inverse correlation with main partners, and a positive association with casual partners. These data provide much needed nuance to our understanding of temporal variation in contact rates in these priority populations. Future studies should continue to explore dynamic variation in partnership rates, how this variation impacts HIV/STI transmission, and leverage these insights to inform prevention activities.

REFERENCES

1. Matthews DD, Herrick AL, Coulter RW, et al.; POWER Study Team. Running backwards: consequences of current HIV incidence rates for the next generation of black MSM in the United States. AIDS Behav. 2016;20:7–16.
2. Sidibé M, Loures L, Samb B. The UNAIDS 90-90-90 target: a clear choice for ending AIDS and for sustainable health and development. J Int AIDS Soc. 2016;19:21133.
3. Fauci AS, Redfield RR, Sigounas G, et al. Ending the HIV epidemic: a plan for the United States. JAMA. 2019;321:844–845.
4. Crepaz N, Hess KL, Purcell DW, et al. Estimating national rates of HIV infection among MSM, persons who inject drugs, and heterosexuals in the United States. AIDS. 2019;33:701–708.
5. Delva W, Leventhal GE, Helleringer S. Connecting the dots: network data and models in HIV epidemiology. AIDS. 2016;30:2009–2020.
6. Pellis L, Ball F, Bansal S, et al. Eight challenges for network epidemic models. Epidemics. 2015;10:58–62.
7. 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.
8. Millett GA, Peterson JL, Flores SA, et al. Comparisons of disparities and risks of HIV infection in black and other men who have sex with men in Canada, UK, and USA: a meta-analysis. The Lancet. 2012;380:341–348.
9. Rosenberg ES, Sullivan PS, DiNenno EA, et al. Number of casual male sexual partners and associated factors among men who have sex with men: results from the national HIV behavioral surveillance system. BMC Public Health. 2011;11:189.
10. Carlo Hojilla J, Koester KA, Cohen SE, et al. Sexual behavior, risk compensation, and HIV prevention strategies among participants in the san francisco PrEP demonstration project: a qualitative analysis of counseling notes. AIDS Behav. 2016;20:1461–1469.
11. Elsesser SA, Oldenburg CE, Biello KB, et al. Seasons of risk: anticipated behavior on vacation and interest in episodic antiretroviral pre-exposure prophylaxis (PrEP) among a large national sample of U.S. Men who have sex with men (MSM). AIDS Behav. 2016;20:1400–1407.
12. Liu AY, Vittinghoff E, Chillag K, et al. Sexual risk behavior among HIV-uninfected men who have sex with men participating in a tenofovir preexposure prophylaxis randomized trial in the United States. J Acquir Immune Defic Syndr. 2013;64:87–94.
13. Montaño MA, Dombrowski JC, Dasgupta S, et al. Changes in sexual behavior and STI diagnoses among MSM initiating PrEP in a clinic setting. AIDS Behav. 2019;23:548–555.
14. Lim SH, Christen CL, Marshal MP, et al. Middle-aged and older men who have sex with men exhibit multiple trajectories with respect to the number of sexual partners. AIDS Behav. 2012;16:590–598.
15. Romero-Severson EO, Volz E, Koopman JS, et al. Dynamic variation in sexual contact rates in a cohort of HIV-negative gay men. Am J Epidemiol. 2015;182:255–262.
16. Basten M, Heijne JC, Geskus R, et al. Sexual risk behaviour trajectories among MSM at risk for HIV in Amsterdam, The Netherlands. AIDS. 2018;32:1185–1192.
17. Swann G, Newcomb ME, Crosby S, et al. Historical and developmental changes in condom use among young men who have sex with men using a multiple-cohort, accelerated longitudinal design. Arch Sex Behav. 2019;48:1099–1110.
18. Rozhnova G, van der Loeff MF, Heijne JC, et al. Impact of heterogeneity in sexual behavior on effectiveness in reducing HIV transmission with test-and-treat strategy. PLoS Comput Biol. 2016;12:e1005012.
19. Romero-Severson EO, Alam SJ, Volz E, et al. Acute-stage transmission of HIV. Epidemiology. 2013;24:516–521.
20. Zhang X, Zhong L, Romero-Severson EO, et al. Episodic HIV risk behavior can greatly amplify HIV prevalence and the fraction of transmissions from acute HIV infection. Stat Commun Infect Dis. 2012;4:1041.
21. Rozhnova G, Heijne JC, Basten M, et al. Impact of sexual trajectories of men who have sex with men on the reduction in HIV transmission by pre-exposure prophylaxis. Epidemics. 2019;28:100337.
22. Romero-Severson EO, Alam SJ, Volz EM, et al. Heterogeneity in number and type of sexual contacts in a gay urban cohort. Stat Commun Infect Dis. 2012;4:4.
23. Mustanski B, Morgan E, D'aquila R. Individual and network factors associated with racial disparities in HIV among young men who have sex with men: results from the RADAR cohort study. J Acquir Immune Defic Syndr. 2019;80:24–30.
24. Duncan SC, Duncan TE, Hops H. Analysis of longitudinal data within accelerated longitudinal designs. Psychol Methods. 1996;1:236–248.
25. Mustanski B, Garofalo R, Emerson EM. Mental health disorders, psychological distress, and suicidality in a diverse sample of lesbian, gay, bisexual, and transgender youths. Am J Public Health. 2010;100:2426–2432.
26. Mustanski B, Johnson AK, Garofalo R, et al. Perceived likelihood of using HIV pre-exposure prophylaxis medications among young men who have sex with men. AIDS Behav. 2013;17:2173–2179.
27. 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.
28. Weiss KM, Goodreau SM, Morris M, et al. Egocentric sexual networks of men who have sex with men in the United States: results from the ARTnet study. Epidemics. 2019;24:100386.
29. Falkenström F, Finkel S, Sandell R, et al. Dynamic models of individual change in psychotherapy process research. J Consulting Clin Psychol. 2017;85:537–549.
30. Firebaugh G, Warner C, Massoglia M. Fixed effects, random effects, and hybrid models for causal analysis. In: Morgan SL, ed. Handbook of Causal Analysis for Social Research. Netherlands: Springer; 2013:113–132.
31. Nakagawa S, Johnson PC, Schielzeth H. The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. J R Soc Interf. 2017;14:124.
32. Wilkins AS. To lag or not to lag?: re-evaluating the use of lagged dependent variables in regression analysis. Polit Sci Res Methods. 2018;6:393.
33. Achen CH. Why lagged dependent variables can suppress the explanatory power of other independent variables. 2000. Annual Meeting of the Political Methodology Section of the American Political Science Association, ; July 20-22, 2000; Los Angeles, CA.
34. Keele L, Kelly NJ. Dynamic models for dynamic theories: the ins and outs of lagged dependent variables. Polit Anal Annu Publ Methodol Section Am Polit Sci Assoc. 2006;14:186–205.
35. Enders CK, Mistler SA, Keller BT. Multilevel multiple imputation: a review and evaluation of joint modeling and chained equations imputation. Psychol Methods. 2016;21:222–240.
36. van Buuren S, Groothuis-Oudshoorn K. Mice: multivariate imputation by chained equations in R. J Stat Softw. 2010;45:1–68.
37. Rubin DB. Multiple Imputation for Survey Nonresponse. New York, NY: Wiley; 1987.
38. Pines HA, Gorbach PM, Weiss RE, et al. Sexual risk trajectories among MSM in the United States: implications for pre-exposure prophylaxis delivery. J Acquir Immune Defic Syndr. 2014;65:579–586.
39. Wilkinson AL, El-Hayek C, Fairley CK, et al. Measuring transitions in sexual risk among men who have sex with men: the novel use of latent class and latent transition analysis in HIV sentinel surveillance. Am J Epidemiol. 2017;185:627–635.
40. Brunham RC. Core group theory: a central concept in STD epidemiology. Venereol. 1997;10:34.
41. Liljeros F, Edling CR, Nunes Amaral LA. Sexual networks: implications for the transmission of sexually transmitted infections. Microbes Infect Inst Pasteur. 2003;5:189–196.
42. Feinstein BA, Dellucci TV, Sullivan PS, et al. Characterizing sexual agreements with one's most recent sexual partner among young men who have sex with men. AIDS Educ Prev. 2018;30:335–349.
43. Underhill K, Guthrie KM, Colleran C, et al. Temporal fluctuations in behavior, perceived HIV risk, and willingness to use pre-exposure prophylaxis (PrEP). Arch Sex Behav. 2018;47:2109–2121.
44. Henry CJ, Koopman JS. Strong influence of behavioral dynamics on the ability of testing and treating HIV to stop transmission. Scientific Rep. 2015;5:9467.
45. Nosyk B, Zang X, Krebs E, et al.; Localized HIV Modeling Study Group. Ending the HIV epidemic in the USA: an economic modelling study in six cities. Lancet HIV. 2020;7:e491–e503.
46. Giguère K, Alary M. Targeting core groups for gonorrhoea control: feasibility and impact. Sex Transm Infections. 2015;91:241–244.
47. Beymer MR, Holloway IW, Pulsipher C, et al. Current and future PrEP medications and modalities: on-demand, injectables, and topicals. Curr HIV AIDS Rep. 2019;16:349–358.
48. Cornelisse VJ, Lal L, Price B, et al. Interest in switching to on-demand HIV pre-exposure prophylaxis (PrEP) among Australian users of daily PrEP: an online survey. Open Forum Infect Dis. 2019;6:ofz287.
49. Gafos M, Horne R, Nutland W, et al. The context of sexual risk behaviour among men who have sex with men seeking PrEP, and the impact of PrEP on sexual behaviour. AIDS Behav. 2019;23:1708–1720.
50. Saberi P, Scott HM. On-demand oral pre-exposure prophylaxis with tenofovir/emtricitabine: what every clinician needs to know. J Gen Intern Med. 2020;35:1285–1288.
51. Hamilton DT, Rosenberg ES, Jenness SM, et al. Modeling the joint effects of adolescent and adult PrEP for sexual minority males in the United States. PLoS One. 2019;14:e0217315.
52. Luo W, Katz DA, Hamilton DT, et al. Development of an agent-based model to investigate the impact of HIV self-testing programs on men who have sex with men in atlanta and seattle. JMIR Public Health Surveill. 2018;4:e58.
53. Jenness SM, Goodreau SM, Rosenberg E, et al. Impact of the centers for disease control's HIV preexposure prophylaxis guidelines for men who have sex with men in the United States. J Infect Dis. 2016;214:1800–1807.
54. Kelly SL, Martin-Hughes R, Stuart RM, et al. The global Optima HIV allocative efficiency model: targeting resources in efforts to end AIDS. Lancet HIV. 2018;5:e190–e198.
55. Elion RA, Kabiri M, Mayer KH, et al. Estimated impact of targeted pre-exposure prophylaxis: strategies for men who have sex with men in the United States. Int J Environ Res Public Health. 2019;16:1592.
56. Goodreau SM, Hamilton DT, Jenness SM, et al. Targeting human immunodeficiency virus pre-exposure prophylaxis to adolescent sexual minority males in higher prevalence areas of the United States: a modeling study. J Adolesc Health. 2018;62:311–319.
57. Harmon TM, Fisher KA, McGlynn MG, et al. Exploring the potential health impact and cost-effectiveness of AIDS vaccine within a comprehensive HIV/AIDS response in low- and middle-income countries. PLoS One. 2016;11:e0146387.
58. Zhang L, Peng P, Wu Y, et al. Modelling the epidemiological impact and cost-effectiveness of PrEP for HIV transmission in MSM in China. AIDS Behav. 2019;23:523–533.
Keywords:

sexual contact rates; sexual networks; young men who have sex with men; young transgender women

Supplemental Digital Content

Copyright © 2021 Wolters Kluwer Health, Inc. All rights reserved.