Cassels, Susan PHD*; Pearson, Cynthia R. PHD†; Walters, Karina PHD†; Simoni, Jane M. PHD‡; Morris, Martina PHD§
There are an estimated 4.4 million American Indians and Alaska Natives (AI/AN) in the United States (US) with 562 federally and over 70 state recognized tribes.1–3 Despite their wealth of tribal diversity and cultural traditions, AI/AN in the United States suffer pervasive health disparities.3,4 In particular, epidemiologic evidence points to excessively high rates of sexually transmitted infections (STI) among AI/AN compared with the general population, including the second highest rates of chlamydia and gonorrhea. Through December 2007, a cumulative total of 3492 AIDS cases among AI/AN were reported to the Center of Disease Control, primarily among men who have sex with men and injection drug users. In addition, 29% of AI/AN diagnosed with AIDS are women or girls, a percentage second only to blacks (35%).4 Along with blacks, AI/AN were the only racial group to experience an increase in HIV incidence between 1990 and 2004.5,6 To avoid an escalation of the epidemic, researchers need to examine more closely STI transmission processes among AI/AN.
HIV and STI epidemiologists have begun to articulate the importance of sexual network dynamics, including concurrent partnerships and selective mixing, as key determinants of HIV racial disparities.7–10 Concurrency–an individual's having sexual partnerships that overlap in time–can dramatically increase the speed and magnitude of HIV spread. At the individual level, concurrency increases the likelihood that a person will transmit infection by removing the protective effect of partner sequencing. With concurrent partners, a person can relay infection from one partner to the other in either direction, in contrast to serial monogamy, where transmission can only occur from earlier partners to later partners in the sequence.11 At the population-level, concurrent partnerships link individuals together to create large connected “components” in a network; such connected components allow a pathogen to travel rapidly and efficiently by reducing the time lost in isolated dyads. The longer the duration of overlap for concurrent partnerships, the more opportunity the pathogen has to use this connectivity for rapid spread.
Selective mixing patterns also can influence the spread of HIV. Assortative mixing (i.e., individuals choosing socio-demographically similar sexual partners) leads to segregated networks that can contain infection. A combination of assortative mixing with differential levels of concurrency across groups makes it possible to sustain long-term prevalence differentials across populations. This kind of network pattern may help to explain the persistent racial differentials in HIV and other STI in the United States.12–15 Disassortative mixing (i.e., individuals choosing socio-demographically dissimilar sexual partners) can lead to bridging across groups. This may facilitate the spread of HIV when sexual partnerships occur between high-risk groups and low-risk groups. Among AI/AN men who have sex with men (MSM) and women who have sex with women (WSW), disassortative mixing based on partner racial characteristics or gender characteristics (i.e., men who have sex with men and women (MSMW) and women who have sex with women and men (WSWM) may be a proxy measure of “bridging” between 2 sexual networks.16
Among AI/AN, MSM and MSMW may be disproportionately affected by HIV.17 Although there is a presumption of lower HIV transmission risk among Native WSW and WSWM, the disproportionately high rates of STIs and other risk factors for potential HIV exposure suggest that Native WSW, and, in particular WSWM, warrant closer examination as well.18
In the current study, we analyzed data from the first national study of gay, lesbian, bisexual, and transgender (GLBT) AI/AN with individual respondents grouped according to the sex of their last 3 partners. Our goal was to examine prevalence of concurrency, patterns of racial mixing that link these groups to groups outside the AI/AN community, and levels of risk across these different partnership contexts.
Respondent Recruitment and Interviewing
Respondents were recruited as part of a multisite cross-sectional survey of AI/AN GLBT persons from 7 metropolitan areas in the United States: Seattle-Tacoma, San Francisco-Oakland, Los Angeles, Denver, Tulsa-Oklahoma City, Minneapolis-St. Paul, and New York.19,20 To participate, individuals were required to: (1) self-identify as American Indian, AK Native, or First Nations and either be enrolled in their tribal nation or report having at least 25% total American Indian blood, (2) self-identify as gay, lesbian, bisexual, transgender, or two-spirit or have engaged in same-sex sexual behavior in the past 12 months, (3) be 18 years of age or older, (4) speak English, and (5) reside, work, or socialize in 1 of the urban study sites.
Multiple sampling strategies were used to minimize selection bias, including venue-based recruitment, convenience sampling, and respondent-driven sampling (RDS) techniques. Each site coordinator proposed 6 to 8 diverse (by gender and age) first wave “seeds” (n = 36) of which 33 participated. A second wave of RDS generated 58 nominees, of whom 50 participated. Volunteer respondents also were solicited through newsletters, brochures, posters, and word-of-mouth; of 469 volunteers who were solicited, 368 participated. We achieved a total response rate of 80.1%. There were no significant differences between RDS (seeds and nominees) and volunteer respondents for the cohort overall or by site on key socio-demographic variables (i.e., gender, education, employment, income, or housing).
Each respondent received $65 for completing a 3 to 4 hour computer-assisted self-interview. A total of 451 respondents were interviewed between July 2005 and March 2007. Of these, 4 respondents were later excluded due to ineligibility, leaving a total of 447 participants.
The interview included, among other items, sections on socio-demographics and health status, sexual risk behavior, and a sexual network module21 for the last 3 sexual partners, which provides an egocentric sample of sexual network.
Respondents indicated their current gender identity (coded as male, female, or transgendered), whether they had a penis, sexual orientation, age, and Native blood quantum (combining all tribes degree of AI/AN blood (100%, 75%–99%, 50%–74%, 25%–49%, 24%, or less). Socio-economic status variables included education level and income.
Multiple closed-ended items queried sexual behaviors over the respondent's lifetime as well as the last 12 months. Respondents indicated whether they had ever traded sex, whether they had a current partner, their HIV status (positive, negative, or unknown) and the number of their lifetime sexual partners. The number of times the respondent had sex without a condom or other barrier (never, once, 2–10, 11–100, or >100) was collected for each of the respondent's last 3 partnerships.
Sexual Partner Characteristics.
For each of up to 3 most recent sexual partners (reported having oral, anal, or vaginal sex), respondents provided partner's sex; race/ethnicity (coded as AI/AN, non-Hispanic white, non-Hispanic black, other); type of partner (main or steady partner, not a main or steady partner but one with whom respondent had sex more than once, one-time sexual partner, or transactional partner with whom respondent traded sex for money or other favors). Respondents also were asked whether they had sex with anyone else while they were sexually involved with each of their partners, and whether they believed their partners had additional partners during the course of their partnership.
Sex of Partner Groupings.
Based on respondents' physical anatomy (i.e., men were defined as those answering yes to the item about whether they had a penis) and reported sex of the last 3 partners, we divided respondents into 4 mutually exclusive groups: men who had sex with only men/1 man (MSOM), men who had sex with at least 1 man and at least 1 woman (MSMW), women who had sex with only women/1 woman (WSOW), and women who had sex with at least 1 woman and at least 1 man (WSWM). We choose to group respondents based on sexual behavior rather than self-identified sexual orientation because the data show reported behavior different from reported sexual orientation.
Concurrent partnerships have a number of attributes based on their timing and duration. We calculated 3 aspects of concurrency for the present analyses: point prevalence, cumulative prevalence, and overlap duration. Point prevalence is defined as having more than 1 ongoing sexual partnership on the day of the interview. The variable is based on questions that asked whether the respondents expected to have sex again with each of the partners they reported. Cumulative prevalence is defined as reporting any overlapping partnerships among the last 3 partners. We report 2 aspects of cumulative prevalence: the first restricted to overlaps that occurred during the past 12 months and the second unrestricted by time. These are based on the reported dates (mo/yr) of first and last sex with each partner. If one partnership ended and another began in the same month, we conservatively assumed no concurrency. Our findings, therefore, represent a lower bound estimate of cumulative prevalence. As seen in Figure 1, cumulative concurrency = 0 in 2 cases: (1) when the respondent practices serial monogamy, which requires each partnership to end before the next one begins, and (2) when respondents have fewer than 2 partners in their lifetime. Concurrency = 1 when 2 partnerships overlap only with each other, concurrency = 2 when there are 2 pairs of overlapping partnerships, and concurrency = 3 if all 3 partnerships overlap at the same time. Time spent in concurrent partnerships is defined as the number of months that the respondent has more than 1 partner. Figure 2 illustrates the sample and the differences between the analytic sample among those who reported sexual activity and partnership data.
Egocentric network data can be analyzed at multiple levels: respondent, partnership, and partnership configuration.21 This article reports results of analyses that use the respondent as the unit of analysis. Respondents were grouped according to sex of their last 3 partners (i.e., MSOM, MSMW, WSOW, WSWM) and descriptive statistics were used to compare sexual network patterns and risk across these 4 groups. Statistical significance was assessed using χ2 tests (for categorical variables) or ANOVAs (for continuous variables). When these suggested significance, we followed-up with χ2 and post hoc Bonferroni t tests to pinpoint group differences.
Table 1 summarizes the socio-demographic characteristics and sexual risk behaviors of the 330 respondents who reported at least 1 partner in the network module, by grouping based on the sex of their last 3 sexual partners with no time limit. As seen in the table, there were some important differences beyond the traditional variations in gender, anatomy, sexual orientation, and income.
Over 20% of the sample self-reported being HIV-positive, with an additional 11% unsure of their status. Percentage was highest for MSOM (36%), followed by MSMW (19%), WSWM (15%), and WSW (2%). Over one-third of the sample (38%) reported trading sex, which was highest among MSOM (51%) and WSWM (38%). Just over half of each group (57%–58%) reported a current, steady partner except for MSOM (41%). About two-third of each group (66%–69%) reported 10 or more lifetime partners except for MSOM (82%). Per each age category 46.9% of the 18 to 25 years, 78.9% of the 26 to 34 year olds, 78.6% of the 35 to 44 years old, and 74.7% of the 45 years and over reported more than 10 partners.
Table 1 presents findings on type of partner, partnership duration, and condom use among respondents' last 3 partners (no time limit) by grouping based on the sex of their last 3 sexual partners with no time limit. A total of 728 partnerships were reported. About partner type, men were less likely than women to report at least 1 of their last 3 partners was a steady partner (55%–63% vs. 84%–85%), and more likely to report a nonsteady partner with whom they had had sex more than once (55%–56% vs. 45%–48%), a 1-time partner; (34%–37% vs. 12%–25%), and a transactional partner (i.e., one with whom they traded sex; 3%–5% vs. 1%–2%).
Median partnership duration was 26 months, and 33% of respondents reported at least 1 partnership lasting less than 1 month. MSOM (41%) and WSWM (34%) were most likely to report a partnership of less than 1 month.
Consistent condom use with all partners was very low: only 9% of the respondents reporting having always used condoms with all partners. Consistent condom use for anal sex ranged from 13% to 22%, which was generally higher than for vaginal sex (0%–14%).
Concurrency findings, based on the subsample of 144 respondents with “complete” partnership data, are reported in Table 2. Of those that reported complete partnership data 3 of the 144 respondents had 1 partner only: 1 was MSOM, 1 was WSOM, and 1 was MSMW. The point prevalence of concurrency was 24%, with 17% reporting 2 ongoing partners, and 7% reported 3 ongoing partners on the day of the interview. Concurrency varied by sex of partners, and was generally higher among MSOM and WSWM.
Cumulative prevalence of concurrency in the last 12 months was 40%, with 15% reporting 1 concurrency, 13% reporting 2, and 12% reporting 3. The upper limit, assuming concurrency if one partnership ended and another began in the same month, was 45%. Again, cumulative prevalence was highest for MSOM and WSWM. Removing the time restriction, cumulative prevalence of concurrency rose to 59% (upper limit was 70%); 27% reported 1 concurrency; 16% reported 2; and 16% reported 3. Prevalence was significantly higher for MSOM.
The overall median duration of overlap in concurrent partnerships was 4 months, with 28% of respondents reporting an overlap of 1 month or less. Among MSOM, 41% reported at least 1 overlap in partners with duration of less than 1 month, compared with 14% to 33% for the other groups. Median duration of overlap was 2.25 months for MSOM, compared with 1 to 15 months for the other groups.
Among respondents who reported at least 1 concurrency (no time limit), 72% reported unprotected sex with more than 1 of their partners. Most respondents (79%) believed their concurrent partners also were having sex with other partners.
Concurrency (no time limit) was related to gender, HIV, number of partners, and reported sex of the last 3 partners, but not self-reported sexual orientation, education, age, income, percent Native blood, and whether the respondent had ever traded sex or has a current partner (data not shown in table). Specifically, mean concurrency was significantly higher among respondents who were male (n = 60) than female or transgendered (n = 84; M[SD] = 1.33 [1.13] vs. = 0.88 [1.04], P < 0.01). In fact, 70% (42/60) of the males reported concurrency of at least one. Concurrency also was significantly higher among HIV-positive respondents (n = 21) compared with HIV-negative or HIV-unknown respondents (n = 123; M[SD] = 1.62[1.25] vs. 0.98[1.05], P < 0.01). Respondents with greater than 10 lifetime partners compared to those with less than 10 (M[SD] = 1.27[1.15] vs. 0.74[0.93], P < 0.01) reported more concurrent partners compared with those with less than 10 (M[SD] = 1.27[1.15] vs. 0.74[0.93], P < 0.01).
Partnership Mixing Across Race
Table 3 shows the breakdown of the racial composition of respondent's networks by sex of partner groupings. More than half (68%) of the sample reported all of their last 1 to 3 partners were of the same race/ethnicity: all Native (18%), all non-Hispanic black (3%), all non-Hispanic white (28%), or all of “other” race/ethnicity (19%). MSOM were least likely to report only Native partners (11% vs. 22%–24%) and most likely to report only white partners (40% vs. 18%–22%). About one-third of the sample reported that they had had partners of various races (by definition, these respondents reported more than 1 partner). Of the 239 respondents who reported the race of 2 or more partners, 57% reported partners of the same race, with 25 of 239 (10%) reporting all Native partners. Forty-three percent of the respondents with more than 1 partner reported partners of various races (data not shown in table).
Results from this study, the first national survey of GLBT AI/AN, revealed an alarmingly high prevalence of HIV (22%), indicating that this population merits attention in terms of STI prevention efforts. Rivaling HIV prevalence reported for black MSM and sub-Saharan Africans, MSOM in our study had the highest HIV prevalence (36%), followed by MSMW (19%) and WSWM (15%).
Further analyses suggested the high levels of HIV may be due to high numbers of lifetime partners, multiple short partnerships, and limited condom use. Men were more likely to have had a nonsteady sexual partner more than once, a one-time partner, and a transactional partner than women. Nevertheless, nearly half of the women reported having multiple partnerships. Additionally, consistent condom use with all partners was very low.
Concurrent sexual partnerships may be another critical factor in spreading HIV in GLBT AI/AN populations. Prevalence of concurrency in this population was quite high (24% point prevalence, 40% cumulative over last 12 months), with MSOM and WSWM reporting the highest levels (55% and 35%) of cumulative prevalence of concurrency in the past year. These levels are higher than the prevalence of concurrency in the past year among black populations (11%–21%)22 and are consistent with levels found in sub-Saharan Africa (e.g., Lesotho-55% and Cote d'Ivoire-36%)8 and among MSM in China (33%).23 Of the 85 respondents who reported at least 1 concurrency, 72% reported unprotected sex with more than 1 of their partners. Moreover, more than 75% of MSOM and 84% of the WSWM believed that at least 1 of their sexual partners also had concurrent partners. This combination of concurrency and inconsistent condom use may enhance the spread of HIV at the network level.
Additionally, an examination of partnership mixing by race indicated potential network patterns that could contain the epidemic within the small GLBT community (i.e., concurrency and assortative mixing) and bridge to other populations (disassortative mixing). Purely assortative mixing by race (i.e., Native partners only) occurred most frequently among women (WSOW and WSWM) and MSMW (22%–24%). Additionally, almost half of the respondents in these groups reported at least 1 Native partner. The uniformly high rate of AI/AN partner selection creates the potential for amplification of disease spread within this small community. Disassortive mixing by race occurred mostly among MSOM. Thus, MSOM who have high levels of unsafe sex, short partnership durations, and high prevalence of concurrency are also the ones partnering the most with non-AI/AN. Combined with high rates of selecting partners of other races, MSOM create the potential for effective bridging to other groups in the transmission network.
Bridging also can occur between the sexes. Although the MSMW subsample is fairly small and the findings should be interpreted with caution, there are a few trends in the data that are consistent with non-AI/AN MSMW populations.24–27 Specifically, our findings demonstrate that MSMW had a high number of sexual partners, engaged in high rates of unprotected anal sex with male and female partners, and engaged in unprotected vaginal sex with female partners. Thus, MSMW may be an effective bridge population across sex. The combination of high rates of concurrency and low condom use also makes WSWM an effective bridge population across race and sex. Specifically, WSWM in our study had very high rates of concurrency, second only to MSOM, and they reported 22% consistent condom use for anal sex and only 3% for vaginal sex.
There were several limitations to our study. As with other self-reports of sensitive information, our data are subject to the possible influences of social desirability and recall bias. Although we used computer-assisted self-interviewing to reduce inhibitions about disclosing, the accuracy of participants' responses cannot be determined. It is likely that any inaccuracies were related to underreporting of risk behaviors, however, meaning our findings are likely conservative estimates. Although the overall sample was rather large, some of the groups and specific pairings included small numbers of respondents. Additionally, a programming error, possibly exacerbated by the lengthy interview, led to some missing data on partnerships. This was less of an issue in the partnering analysis (n = 330, in which the analytic subsample was very similar to the overall sample) than the concurrency analysis (n = 144, in which the analytic subsample included fewer males, was more educated, was younger, and had higher incomes than those in the total sample). Also, the concurrency analyses were based only on respondents with “complete” data on in the network module, which could have biased it in either direction. Those with incomplete data who were excluded might include both individuals in long-term monogamous partnerships who could not recall details about early partners (which would inflate our estimate of concurrency) or highly sexually active individuals with multiple and concurrent partnerships who did not know the details about their partners (which would lower our estimate of concurrency). We did not include as concurrent partnerships that started and ended in the same month, which would also lower our estimate of concurrency. Finally, the network module only inquired about the last 3 partners. For many respondents, this was a small subset of their total sexual partners. For others currently in long-term partnerships, these may represent partners from many years ago for which their memory may be less reliable and their behaviors less representative of their current situation.
Despite these limitations, the findings provide some of the first insights into sexual networks and concurrency among GLBT AI/AN populations, a group that has been virtually ignored in HIV prevention efforts. The data demonstrate that GLBT AI/AN men and women deserve attention in HIV prevention efforts at individual, dyadic and population levels.
1. U.S. Census Bureau. Census 2000 American Indian and Alaska Native Summary File (AIANSF), 2007.
2. Freeman C, Fox M. Status and trends in the education of American Indians and Alaska natives In: U.S. Department of Education NCES: Washington, DC: U.S. Government Printing Office, 2005.
3. World Health Organization: The World Health Report 2002 - Reducing Risks, Promoting Healthy Life, 2002.
4. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report, 2007, vol 19, Atlanta, GA: US Department of Health and Human Services, CDC, 2009.
5. Bertolli J, McNaghten AD, Campsmith M, et al. Surveillance systems monitoring HIV/AIDS and HIV risk behaviors among American Indians and Alaska Natives. AIDS Educ Prev 2004; 16:218–237.
7. Hudson CP. Concurrent partnerships could cause AIDS epidemics. Int J STD AIDS 1993; 4:249–253.
8. Mah TL, Halperin DT. Concurrent Sexual Partnerships and the HIV Epidemics in Africa: Evidence to Move Forward. AIDS Behav In press.
9. Morris M. Sexual networks and HIV. AIDS 1997; 11(suppl A):S209–S216.
10. Morris M, Kurth AE, Hamilton DT, et al. The Network Modeling Group: Concurrent partnerships and HIV prevalence disparities by race: Linking science and public health practice. Am J Public Health 2009; 99:1023–1031.
11. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–683.
12. Adimora AA, Schoenbach VJ. Social context, sexual networks, and racial disparities in rates of sexually transmitted infections. J Infect Dis 2005; 191:S115–S122.
13. Adimora AA, Schoenbach VJ, Bonas DM, et al. Concurrent sexual partnerships among women in the United States. Epidemiology 2002; 13:320–327.
14. Gupta S, Anderson RM, May RM. Networks of sexual contacts - Implications for pattern of spread of HIV. AIDS 1989; 3:807–817.
15. Morris M, Handcock MS, Miller WC, et al. Prevalence of HIV infection among young adults in the United States: Results from the Add Health study. Am J Public Health 2006; 96:1091–1097.
16. Gorbach PM, Drumright LN, Holmes KK. Discord, discordance, and concurrency: Comparing individual and partnership-level analyses of new partnerships of young adults at risk of sexually transmitted infections. Sex Transm Dis 2005; 32:7–12.
17. Centers for Disease Control. HIV/AIDS Surveillance Report. Department of Health and Human Services, 2007.
18. Mitchell CM, Kaufman CE, Beals J. Identifying diverse HIV risk groups among American Indian young adults: The utility of cluster analysis. AIDS Behav 2004; 8:263–275.
19. Chae DH, Walters KL. Racial discrimination and racial identity attitudes in relation to self-rated health and physical pain and impairment among two-spirit American Indians/Alaska Natives. Am J Public Health 2009; 99(suppl 1):S144–S151.
20. Lehavot K, Walters KL, Simoni JM. Abuse, mastery, and health among lesbian, bisexual, and two-spirit American Indian and Alaska Native women. Cultur Divers Ethnic Minor Psychol 2009; 15:275–284.
21. Morris M. Network Epidemiology: A Handbook for Survey Design and Data Collection. International Studies in Demography Series. Oxford: Oxford University Press, 2004.
22. Adimora AA, Schoenbach VJ, Doherty IA. HIV and African Americans in the Southern United States: Sexual networks and social context. Sex Transm Dis 2006; 33:S39–S49.
23. Choi K-H, Hudes ES, Steward WT. Social discrimination, concurrent sexual partnerships, and HIV risk among men who have sex with men in Shanghai, China. AIDS Behav 2008; 12(suppl 1):71–77.
24. Boulton M, Schramm Evans Z, Fitzpatrick R, et al. Bisexual men: Women, safer sex and HIV transmission In: Aggleton P, Davies P, Hart G, eds. AIDS: Responses, Interventions and Care. Goldsmiths' College, University of London: The Falmer Press, 1991.
25. McKirnan DJ, Stokes JP, Doll L, et al. Bisexually active men: Social characteristics and sexual behavior. J Sex Res 1995; 32:64–75.
26. Siegel K, Schrimshaw EW, Lekas HM, et al. Sexual behaviors of non-gay identified non-disclosing men who have sex with men and women. Arch Sex Behav 2008; 37:720–735.
27. Stokes JP, Vanable P, McKirnan DJ. Comparing gay and bisexual men on sexual behavior, condom use, and psychosocial variables related to HIV/AIDS. Arch Sex Behav 1997; 26:383–397.