The high rates of HIV infection among black men who have sex with men (BMSM) in large cities across the United States suggest the urgent need to evaluate prevention and access to care in this underserved population. According to the Centers for Disease Control, in 2006, 35% of MSM with new HIV infection were black, and among MSM aged 13 to 29 years, the number of new infections in black MSM was nearly twice that of white MSM.1 In Chicago, this disparity is even higher—young black men have 7 times the rate of HIV infection as young white men with similar sexual behavior and substance use patterns.2 The high rates of HIV infection among BMSM and the concomitant lack of HIV status awareness3,4 motivate the present analysis of utilization of HIV services for this population.
Increased disparities in HIV prevalence rates among BMSM are because of a number of factors including limited utilization of HIV prevention and treatment services.3,4 Although a few studies consider the individual-level factors associated with MSM receiving HIV testing or visiting a health care provider,5 other studies move beyond the individual-level factors associated with access to care, and instead they focus on organizational-level factors of health care centers in an effort to explain differing levels of access.6,7 In a study of MSM in Massachusetts, individual factors, such as a higher level of social support and having health insurance, were associated with access to primary care.5 Individual-level factors associated with BMSM that report no recent history of HIV testing include, lower levels of attained education, higher reported sexual risk-taking behavior, and not having visited a health care provider in the past 12 months.5
Other studies focus on organizational-level aspects of health care centers that enhance or diminish access among minority MSM.6,7 Some studies stress the need for health care settings that are multifaceted and offer not only a medical component but also case management and supportive programs.7 A study of retention of young HIV seropositive minority MSM suggests that the most important element in linkage and retention in care is whether participants felt respected in the clinic environment.6
Taken together, findings from these strands of research raise concerns about the possibility that different subgroups of at-risk populations selectively affiliate with different health centers; behavior that may or may not be aligned with HIV-prevention services offered at these respective centers. Further, differential affiliation may both give public health authorities a misleading picture of the at-risk population and give clients a skewed understanding of the kinds of services that are available to them. Centers that have different profiles of clients will inevitably receive different information about clients' needs and may develop different service priorities in response, just as clients' specific affiliations will shape the information they receive about specific resources and expertise. For example, public health authorities might make assumptions about the population's needs based on which clients are selecting into a specific clinic and act according to these assumptions (e.g., shifting resources to substance use counseling). Knowledge of the pattern of HIV health center utilization could also be useful for efficient appropriation of HIV-prevention programming at select centers. Without such data, appropriation might be relegated to all centers, or worse yet, to those most able to effectively lobby for their inclusion.
Our objective is to describe specific individual- and organizational-level factors that contribute to utilization patterns of health care centers offering HIV prevention and/or treatment services among BMSM. We use a 2-mode8 affiliation network analysis of health center utilization patterns to answer the following questions: (1) Which types of HIV prevention and care centers do BMSM use? (2) What individual-level factors are associated with health care utilization patterns among BMSM? (3) What individual-level and organizational-level factors are associated with BMSM's utilization of these same health centers?
MATERIALS AND METHODS
In 2010, 204 BMSM were recruited into the study using respondent-driven sampling.9,10 The protocol was approved by relevant institutional review boards.
Candidate “seed” participants were eligible for the study if they (1) self-identified as black, (2) identified as male, (3) aged 18 years or older, and (4) reported anal intercourse with a man in the past 12 months. Subsequent recruits were deemed eligible by the same criteria.
Respondent-driven sampling (RDS) has been widely applied to study hard to reach populations including MSM.11–14 RDS is an efficient method that uses respondents' network connections to generate a sample that approximates a probability sample, allowing for valid statistical analysis when its assumptions are met.11,15,16 Twenty-one seeds were recruited from 4 diverse venues across the city. Seeds were recruited from the following venues: (1) a Federally Qualified Health Center that provides HIV primary care; (2) a community-based organization that provides HIV prevention services; (3) a substance use treatment program; and (4) an LGBT (Lesbian Gay Bisexual Transgender) health care center. Seeds were asked to refer up to 4 recruits who were MSM from their social networks (by definition, it could include all networks), using vouchers, with each subsequent recruit doing the same. All study participants were provided with a $50 gift card for participation.
Data on BMSM's awareness and utilization of health centers offering HIV services were used to construct the key dependent variables in the study. Before recruitment of the main RDS sample, 12 preparatory in-depth qualitative interviews of BMSM from 2 geographic areas in Chicago were conducted to identify health centers associated with either HIV prevention and/or treatment services. There was considerable overlap of center awareness between the participants in this qualitative study, resulting in a list of 9 health centers. Centers that were mentioned only once were not included in this final center roster. Health centers were classified by type of center (e.g., community-based organization [CBO], public clinic, or private clinic) and area of the city (e.g., north, south, or west side of Chicago). Services provided in the past 2 years were determined by discussions with HIV program staff for each of the 9 health centers. All centers provided HIV voluntary counseling and testing. Centers that included any of the following services were categorized as providing HIV-prevention services: effective behavioral interventions, preexposure or postexposure prophylaxis, and prevention for positives programming. Availability of condoms, education/awareness only, or referral to STI testing was not considered as prevention services in the current analysis. Centers that provided antiretroviral-based HIV treatment were categorized as providing HIV treatment services. The 9 centers included on the final roster reflect a full spectrum of venues along these classifications.
To assess awareness of health centers offering HIV services, participants were presented with the roster of 9 Chicago health centers and were asked, “Which of the following centers have you heard of?” To assess utilization, these participants were also asked, “Which of the following centers have you been to?” The result of this measure was an adjacency matrix; each cell in the matrix contains the value (0 = no; 1 = yes) indicating whether one person (row) is aware of or uses a health center (column).8
The dependent variable differs from a count of the number of organizations with which a given participant was affiliated; rather, the dependent variable is a count of the number of health centers that a given pair of participants list in common (204 × 203 = 41,412 potential pairs of participants). For example, if person 1 listed Centers A, C, and D, and person 2 listed Centers A, C, and E, then 2 of their affiliations overlap.
The goal of this analysis is to determine which socio-demographic and risk-related factors are associated with the likelihood that 2 participants share similar profiles of health center affiliation. If we were to find that HIV status is significantly associated with this dependent variable, then it would mean that HIV-positive and HIV-negative BMSM tend to affiliate with different centers. In other words, the analysis is designed to detect a split or partitioning in the network that signals differential affiliation with certain health centers based on individual attributes.
Several socio-demographic and risk-related factors could lead to differential health center affiliation within the sample of BMSM. Questions assessing socio-demographic characteristics, substance use, risk behavior, and HIV status were adapted from the Centers for Disease Control's National HIV Behavioral Surveillance System, MSM Cycle,17 and the visit 51 Core Behavioral survey of the MACS (Multicenter AIDS Cohort Study) (available at: http://www.statepi.jhsph.edu/macs/Questionnaires/mas_index.html). Unprotected anal intercourse and sex-drug use18 were measured in previous work.19 Group sex was measured as having sex with 2 or more individuals at the same time. MSM social network size was measured by asking participants to estimate the number of MSM in their community who they know well and with whom they are likely to have contact in the next 2 weeks.
Because the observations used to create network variables are not independent, traditional regression techniques are not sufficient. To study the association between individual-level factors and patterns of affiliation with health centers, we used a 2-mode network analytic method—Multiple Regression Quadratic Assignment Procedure (MRQAP).20 Like multiple regression analysis, independent variables predict the dependent variable in MRQAP. However, MRQAP is distinguished from traditional multiple regression analysis in that the independent and dependent variables are matrices describing relationships between pairs of participants, which facilitate an analysis in which the dyad (pairs of participants) constitutes the unit of analysis. Therefore, each MRQAP model related the entire affiliation matrix (containing data on the dependent variable) to each of the matrices of independent variables. Because the observations are not independent of each other, MRQAP generates random permutations of the independent matrix and then computes the regression and saves the resulting r2 values and all coefficients. Low proportions of random permutations, such as <0.05, suggest that there is a low likelihood that the relationship between the matrices of interest occurred by chance, and thus a significant relationship is suggested.21,22 Regression coefficients with 95% confidence intervals were reported for all models, with P values <0.05 considered statistically significant. UCINET (Harvard, MA) version 6.174 was used for all MRQAP analyses.20
In addition, we constructed 2-mode affiliation network graphs8 to visualize how health center affiliation is associated with each of the individual-level factors that were associated with utilization patterns in the MRQAP analysis. For instance, because HIV status is significantly associated with patterns of utilization according to the MRQAP analysis, the 2-mode networks of health center utilization were stratified by HIV status for comparison of utilization patterns. All 2-mode affiliation networks were visualized using NetDraw 2.083.20
At the final stage, we identified health centers that emerge as particularly integral to specific subgroups within the BMSM population. To compare the stratified network graphs, we computed degree centrality and conducted a faction analysis in UCINET. Degree centrality is defined as the number of ties to a person or setting,8 and in this instance, the number of participants tied to a particular health care setting. This makes it possible to identify the most used health care setting among participants sharing certain individual-level characteristics, such as HIV status. A faction analysis further divided the network into factions based on the number of links between nodes.23 Nodes (e.g., health care centers) with strong indirect connections to each other are placed in the same faction, in which an indicator is used to represent each faction in the network. Health centers that share a significant number of participants are placed in the same faction, whereas health centers that do not share a significant number of participants are placed in separate factions.
Characteristics of study participants (n = 204) are displayed in Table 1. We enrolled 204 BMSM, including 21 seeds, and an average of 23 individuals across 9 RDS waves. Sixty-eight percent of the sample aged <30 years, 23% had previously enrolled in an HIV-prevention program, and 44% self-reported HIV-infected status. Characteristics of the 9 Chicago health centers offering HIV services are described in Table 2. Three of the centers were classified as public clinics, 1 as a private clinic, and 5 as community-based organizations. Four centers were located on the south side, 3 on the north side, and 2 on the west side of the city.
The MRQAP analyses presented in Table 3 identifies individual-level factors associated with differential affiliation with the 9 health centers. Individual-level factors significantly associated with the affiliation network of utilization of Chicago health centers included age (coeff., 0.13; P < 0.05), HIV status (coeff., 0.27, P < 0.001), and size of MSM social network (coeff., 0.20; P < 0.01). In other words, BMSM who are in the same age-group, who have the same HIV status (positive or negative), and/or who have similarly sized social networks, are significantly more likely to use the same health centers as of each other than are BMSM who are from different age-groups, who have different HIV statuses, or who have differently sized social networks.
To illustrate the utilization patterns of the overall network, the network of health center affiliations for all 204 participants is graphed in Figure 1. Nodes C, E, and H (in black) represent north side health centers and are located within the same faction according to the faction analysis. The triangle encompasses these health centers from the same faction. Nodes F, G, and J represent health centers on the south side and comprise the remaining 2 factions. This suggests that BMSM affiliate with health centers based on a geographical distribution (North, South, or West).
To distinguish further between the utilization patterns of BMSM with specific individual-level characteristics, we constructed additional 2-mode graphs based on variables that were statistically significant in the MRQAP analysis. We constructed sets of 2-mode matrices that were (1) stratified by HIV status; (2) stratified by HIV status limited to participants practicing unprotected anal intercourse within the past 12 months; and (3) stratified by social network size (Fig. 2, panels A–F). In the graph of HIV-positive participants (Fig. 2A), nodes C, D, and G are CBOs offering prevention services only, including prevention for positives, and are located within the same faction on the periphery of the graph (triangle encompasses this faction) and are with lower-degree centrality (fewer nominations). The nodes composing the other 2 factions are located at the core of the network and represent health centers offering treatment services. In the graph of HIV-negative participants (Fig. 2B), node D (squared), a CBO offering prevention services, is the most popular health center according to the in-degree centrality measure (0.63). Patterns of affiliation based on computation of faction membership (HIV health centers [HHCs] sharing same participants) and degree centrality (number of HHC nominations received) become apparent, demonstrating that subgroups of BMSM may not be using centers with prevention services appropriate for their statuses. For example, in Figure 2D, a public clinic that does not offer HIV-prevention services has the highest-degree centrality (most popular) for HIV-negative participants who engage in high-risk behavior. In the graph of participants with a small MSM social network (Fig. 2E), a public clinic that provides only treatment services has the highest-degree centrality, and faction analysis does not distinguish between health centers. In Figure 2F, the graph of participants with a large MSM social network, node E, a CBO has the highest-degree centrality within the affiliation network and provides both prevention and treatment services. The faction analysis indicates no specific utilization patterns based on organizational-level characteristics of health centers.
In response to the number of studies that underscore the contribution of limited health care access to the HIV disparity among black MSM,24–26 we aimed to take a new approach to clarify multiple competing individual and organizational-level factors of utilization of health centers in a large urban setting. Although a few researchers have applied an affiliation network approach to describe the role that venues play in disease transmission,27 or tobacco control,28 no published studies that we are aware of have used affiliation network analysis to identify factors associated with patterns of health service utilization. Using this approach, we were able to identify several important individual- and organization-level factors that shape patterns of contact of BMSM with a range of health centers offering HIV services. We found that, in aggregate, BMSM depict a pattern of HHC affiliation based on their age, HIV status, and social network size. Analysis of health center grouping, however, demonstrates that a subgroup of high-risk uninfected men affiliate less with centers providing HIV-prevention services. Similarly, BMSM who are HIV positive and who have smaller social networks use public or private health clinics only offering treatment services. HIV-prevention centers that provide specialized HIV prevention for positive programming, however, were not affiliated with these HIV-infected participants.
HIV Status and Utilization
In contrast to the utilization patterns of HIV-uninfected participants, HIV-infected BMSM in our study reported limited engagement with health centers offering HIV-prevention programs, including prevention for HIV-positive participants. In the affiliation network of HIV-uninfected participants, a CBO offering prevention services has the highest-degree centrality and is thus the most used center in this subset of participants. Notably, health centers offering prevention services comprise the periphery of the affiliation network of HIV-infected participants, whereas health centers offering treatment services compose the core of the affiliation network. This finding is consistent with previous literature documenting that HIV-infected BMSM are less likely to participate in behavioral interventions.29
Role of Social Support in Health Center Utilization
Our analysis suggests that CBOs play a significant role in health center-seeking behaviors of participants with larger MSM social networks. The central location of a CBO in the affiliation network of participants with large MSM social networks suggests that this group of participants use CBOs in addition to more traditional health care settings. In contrast, the peripheral position of 3 CBOs in the affiliation network of participants with smaller MSM social networks suggests that men with smaller networks of social support rely on public and private clinics. Study participants reporting a larger social network of MSM were more likely to have used more health centers than study participants with a smaller social network of MSM. This finding is consistent with previous work documenting that individuals with higher levels of social support are more likely to access health care services.30 Social network effects (e.g., friend influence) on health care utilization also occurs across a broad range of non-HIV–related settings and clients including participation in influenza vaccine clinics, use of herbal preparations among clients of antenatal clinics, tuberculosis care seeking, and intrauterine device acceptors in a gynecological clinic.
It remains an open question whether individual characteristics lead to selective affiliating (e.g., friends who are similar to each other lead each other to the same centers, or people simply select into centers that seem to cater to their own “type” or their specific needs/problems) or whether the fact that 2 people attending the same centers increases the likelihood that they will form a social connection to each other, which then somehow influence each other such that they become similar to each other. The fact that from MRQAP analysis, risky behavior profiles are not related to persons' health center affiliation profiles, whereas HIV status is interesting. It might suggest, for example, that health centers that are more associated with HIV-infected BMSM are just as useful targets for interventions aimed at reducing risk behavior as are health centers that are more associated with HIV-negative BMSM, as risk behavior is not associated with health center affiliation in the same way that HIV status is.
There are several limitations. The data are cross-sectional, and we cannot determine the impact that exposure to particular health centers may have had on the behavior of study participants. Study participants reported their awareness and utilization of a predetermined list of health centers offering HIV care, which may have excluded other HHCs. However, we selected centers based on discussions with BMSM during focus groups. Additionally, although we describe utilization patterns of health centers and prevention services used by participants, we do not explore the quality, timing, or frequency of visits to health centers. However, the results provide critical insight into contact that BMSM have with a range of health centers. Our narrow definition of HIV prevention might have limited some centers' inclusion as prevention centers even if they provided standard of care risk-reduction counseling for HIV-infected clients. Finally, the study took place in one city potentially limiting its generalizability. However, health care centers described in this analysis were categorized by typology, and they are similar to those in other urban settings.
Our findings suggest that patterns of health care utilization among BMSM in Chicago are structured by several individual-level characteristics. Subcategories of BMSM in this sample affiliated with HIV health centers that may not provide appropriate HIV-prevention services. Given the strong affiliation of HIV-infected BMSM with public and private clinics offering treatment services, future studies should consider the degree to which HIV-infected BMSM receive behavioral interventions in public and private clinic settings (prevention for positives), and whether CBOs should shift focus to provide such programming to target-specific clients with, for example, smaller social networks. Public health authorities could better match prevention services to both HIV-infected and -uninfected BMSM need. Lack of knowledge of the networked pattern of HIV health center utilization could allow resource allocation to all HHCs, and matching appropriate HHC services to specific clients is critical in an environment of constrained resources.
1. Prejean J, Ruiguang S, Hernandez A, et al.. Estimated HIV incidence in the United States, 2006–2009. PLoS One 2011; 6: e17502.
2. CDPH. HIV Prevalence and Unrecognized Infection among Men Who Have Sex with Men—Chicago. Chicago, IL: Chicago Department of Public Health, 2008.
3. Millett GA, Ding H, Marks G, et al.. Mistaken assumptions and missed opportunities: Correlates of undiagnosed HIV infection among Black and Latino men who have sex with men. J Acquir Immun Defic Syndr 2011; 58:64–71.
4. 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.
5. Mimiaga MJ, Reisner SL, Cranson K, et al.. Sexual mixing patterns and partner characteristics of black MSM in Massachusetts at increased risk for HIV infection and transmission. J Urban Health 2009; 68:602–623.
6. Magnus M, Jones K, Phillips G II, et al.. Characteristics associated with retention among African American and Latino adolescent HIV-positive men: Results from the outreach, care, and prevention to engage HIV-seropositive young MSM of color special project of national significance initiative. J Acquir Immune Defic Syndr 2010; 53:529–536.
7. Johnson RL, Botwinick G, Sell RL, et al.. The utilization of treatment and case management services by HIV-infected youth. J Adolesc Health 2003; 33(suppl 2):31–38.
8. Wasserman S, Faust K. Social Network Analysis: Methods and Applications, 1st ed. Cambridge, United Kingdom: Cambridge University Press, 1994.
9. Heckathorn D. Respondent-driven sampling: A new approach to the study of hidden populations. Soc Probl 1997; 44:174–199.
10. Heckathorn DD. Respondent-driven sampling II: deriving valid population estimates from chain-referral samples of hidden populations. Soc Probl 2002; 49:11–34.
11. Iguchi MY, Ober AJ, Berry SH, et al.. Simultaneous recruitment of drug users and men who have sex with men in the United States and Russia using respondent-driven sampling: Sampling methods and implications. J Urban Health 2009; 86:S5–S31.
12. Johnston LG, Khanam R, Reza M, et al.. The effectiveness of respondent driven sampling for recruiting males who have sex with males in Dhaka, Bangladesh. AIDS Behav 2008; 12:294–304.
13. Ramirez-Valles J, Heckathorn DD, Vázquez R, et al.. From networks to populations: The development and application of respondent-driven sampling among IDUs and Latino gay men. AIDS Behav 2005; 9:387–402.
14. Mimiaga MJ, Goldhammer H, Belanoff C, et al.. Men who have sex with men: Perceptions about sexual risk, HIV and sexually transmitted disease testing, and provider communication. Sex Transm Dis 2007; 34:113–119.
15. Gile KJ, Handcock KS. Respondent-driven sampling: An assessment of current methodology. Sociol Methodol 2010; 40:285–327.
16. Goel S, Salganik MJ. Respondent-driven sampling as Markov chain Monte Carlo. Stat Med 2009; 28:2202–2229.
17. Sanchez T, Finlayson T, Drak A, et al.. Human immunodeficiency virus (HIV) risk, prevention, and testing behaviors–United States, National HIV Behavioral Surveillance System: Men who have sex with men, November 2003-April 2005. MMWR Surveill Summ 2006; 55:1–16.
18. Ostrow DG, Stall RC. Alcohol, tobacco, and drug use among gay and bisexual men. In: Wolitski RJ, Stall R, Valdiserri RO, eds. Unequal Opportunity: Health Disparities Affecting Gay and Bisexual Men in the United States. Oxford, United Kingdom: Oxford University Press, 2008:121–158.
19. Ostrow DG, Plankey MW, Cox C, et al.. Specific sex drug combinations contribute to the majority of recent HIV seroconversions among MSM in the MACS. J Acquir Immune Defic Syndr 2009; 51:349–355.
20. Borgatti SP, Everett MG, Freeman LC. UCINET for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies, 2002; 2006.
21. Borgatti SP, Everett MG, Freeman LC. UCINET VI for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies, 1999.
22. Labianca G, Brass DJ, Gray B. Social networks and perceptions of inter-group conflict: The role of negative relationships and third parties. Acad Manag J 1998; 41:55–67.
23. Borgatti SP, Everett MG. Network analysis of 2-mode data. Soc Networks 1997; 19:243–269.
24. Malebranche DJ. Black men who have sex with men and the HIV epidemic: Next steps for public health. Am J Public Health 2003; 93:862–865.
25. Crosby R, Holtgrave DR, Stall R, et al.. Differences in HIV risk behaviors among black and white men who have sex with men. Sex Transm Dis 2007; 34:744–748.
26. Hall HI, Byers RH, Ling Q, et al.. Racial/ethnic and age disparities in HIV prevalence and disease progression among men who have sex with men in the United States. Am J Public Health 2007; 97:1060–1066.
27. Frost SD. Using sexual affiliation networks to describe the sexual structure of a population. Sex Transm Infect 2007; 83:I37–I42.
28. Wipfli HL, Fujimoto K, Valente TW. Global tobacco control diffusion: The case of the framework convention on tobacco control. Am J Public Health 2010; 100:1260–1266.
29. Hatfield LA, Ghiselli ME, Jacoby SM, et al.. Methods for recruiting men of color who have sex with men in prevention-for-positives interventions. Prev Sci 2010; 11:56–66.
© Copyright 2012 American Sexually Transmitted Diseases Association
30. Merighi JR, Chassler D, Lundgren L, et al.. Substance use, sexual identity, and health care provider use in men who have sex with men. Subst Use Misuse 2011; 46:452–459.