Risk factor studies often examine characteristics of those receiving or acquiring HIV infections and sometimes they examine characteristics of relationships. Yet risk factor studies seldom examine characteristics of partners who might be transmitting the infection.1,2 Individual characteristics and behaviors and characteristics of partners and relationships are helpful in considering prevention strategies.
Youth in sub-Saharan Africa (SSA) bear a heavy burden of HIV; nearly 3.8 million 15- to 24-year-olds or approximately 76% of the world's HIV-positive youth population live in SSA.3 Extensive research has documented individual-level risk factors for HIV infection among heterosexual youth in SSA, including age, gender, use of alcohol, number of sexual partners, sexual concurrency, sexually transmitted infections, patterns of condom use, and type of sexual acts.4 In turn, prevention efforts have often focused on individual-level behavior change, such as increasing condom use with all partners, promoting fidelity, and avoiding new or multiple partners.5 Previous studies in SSA have found an association between older partner age, partners' multiple sexual partnerships, substance abuse, travel, and intimate partner violence and with HIV.6–12 Still, less is known about how partner characteristics influence youth risk of HIV acquisition.
A recent review identified some key gaps in knowledge on the influence of partner characteristics on HIV risk among youth.1 First, many studies of partner characteristics associated with HIV infection come from high-income/developed countries; fewer have been conducted in contexts with generalized HIV epidemics. Second, select partner factors—such as partner age disparity and partner's concurrency—have received the most attention in studies of HIV risk in low- and middle-income country contexts. Third, much of the research relies on self-reported HIV status and is unable to link select partner characteristics with biomedically confirmed HIV status. Finally, most studies assessed partner characteristics associated with prevalent HIV rather than HIV acquisition. Studies of prevalent HIV cannot assess the temporal relationship between partner characteristics and HIV acquisition.
Our study investigates a range of sexual partner characteristics associated with HIV acquisition among youth in the rural Rakai district of southwestern Uganda. Uganda has a mature and generalized HIV epidemic with a national prevalence of 7.3%.13 Although Uganda experienced substantial declines in HIV prevalence after 1990, recent serobehavioral surveys indicate small increases in prevalence among young people and adults.13 Understanding risk characteristics of sexual partners might provide insights for developing more effective HIV prevention programs aimed at youth.
This study builds on an earlier analysis from Rakai, Uganda on HIV incidence among youth that found the risk of HIV acquisition was associated with individual-level factors, including gender, age, multiple sexual partners, sexual concurrency, alcohol use, and STI symptoms.14 The current study extends these analyses to examine risk characteristics from multiple sexual partners as reported by young women and young men and how these characteristics independently contribute to HIV acquisition, after controlling for individual-level factors.
Rakai Community Cohort Study
We use data from the Rakai Community Cohort Study (RCCS), a longitudinal population-based cohort that has been described more fully elsewhere.15 Briefly, 50 communities are surveyed approximately annually. At each survey round, a household census is conducted, and eligible, consenting participants aged 15–49 years are administered an interview and are asked to provide biological specimens for HIV and STI testing. The RCCS questionnaire elicits sociodemographic information, behaviors, health status, and HIV knowledge. Respondents also provide detailed information about their most recent sexual partners. Household characteristics are derived from the censuses.
Study Design and Sample
This study focuses on the partner characteristics reported by male and female respondents aged 15–24 years. We limit the analysis to 4 rounds of RCCS data collection (2005–2011) as these survey rounds captured detailed partner-related information on up to 4 sexual partners in the past year.
HIV status of the respondent was determined by 2 separate enzyme-linked immunosorbent assays confirmed by Western blot, as previously described.14 HIV acquisition was defined as having a positive HIV test after a negative HIV test, and it was assumed infection occurred at the midpoint between the last negative and first positive test. To ascertain HIV acquisition, we restricted our analyses to initially HIV-negative youth who were followed up at 1 or more study visits with no more than 1 survey round missing between them.13 Youth were excluded from analysis, if they were not tested in the previous 2 rounds, were sexually inexperienced, or did not have at least 1 sexual partner in the last year.
We examined the influence of a broad range of partner characteristics grouped into 3 categories—partner's sociodemographic characteristics, sexual activity characteristics, and partner's HIV risk–on–HIV acquisition among youth respondents (Fig. 1). We grouped sexual partnerships into marital and nonmarital categories (including boyfriend/girlfriend and casual partnerships).
The partner's sociodemographic characteristics included the partner age, living situation, and occupation. The partner age variable was classified as older, younger, or the same age as the respondent. We also assessed whether the partner lived in the same household as the respondent, and the distance between where the partner lived and the respondent's place of residence for those not living in the same household. Partner's main occupation was a categorical variable of 3 main occupation types: trucker, bar worker, or student. In previous studies, truckers and bar workers have been associated with higher HIV prevalence,16–18 while being a student was protective against HIV acquisition.14
We examined a set of sexual activity characteristics. We examined how long respondents had known each other before they had sex the first time—grouped as less than 1 month, 1–2 months, and more than 2 months. We examined partner's alcohol use before sex at the last sexual encounter. We generated an annual coital frequency variable. We also assessed condom use and consistency with each sexual partner.
To assess partner's risk of HIV, we examined the respondent's assessment of the partner's sexual concurrency (number of other sexual partners in the last year), the partner's HIV risk, and whether the partner had shared his or her HIV test results with the respondent in the past 12 months.
Statistical Methods and Analysis
We summarized partner characteristics using contingency tables, stratified by the gender of the respondent and marital or nonmarital relationship status. We used χ2 tests to determine whether the characteristics of partners were significantly different between female and male respondents, and between marital and nonmarital partners.
HIV incidence rate was estimated per 1000 person-years. To analyze the association between partner-related risk factors and HIV acquisition of the index respondent, we used a Poisson regression model to estimate the incidence rate ratio (IRR) and 95% confidence intervals. To control for multiple partnerships with a single index respondent, we made 2 assumptions. One is a common effect assumption, that is, if 2 partners have the same characteristics, their contributions to the HIV acquisition of the index respondent are equivalent. Second, we modeled the risk of HIV acquisition per partner, and if a respondent had multiple partners with a particular high-risk characteristic, then the risk of HIV acquisition was multiplicative.
To illustrate the model, we denote Yij as the indicator of HIV acquisition. Yij = 1 if the i-th respondent changed HIV positive in his/her j-th visit, and Yij = 0 otherwise. We then denote Xijk as a risk factor of the k-th partner of the i-th respondent on the j-th visit. If X is a categorical variable with L levels, then Xijk is a vector with (L − 1) binary indicators. We also define dij as the exposure time in years of the i-th respondent on the j-th visit since the (j − 1)-th visit. With these notations, we build the following adjusted Poisson regression model.
Under equation (1), exp(β0) is the incidence rate per person year if all of the partners are at the reference level and exp(β1) is the relative HIV incidence risk associated with risk factor X. We used generalized estimation equations with autoregressive working correlation to accommodate the within-subject correlation from the longitudinal data structure. We used Wald tests to determine statistical significance.
Since we observed big differences between marital and nonmarital partners, we further adjusted the model by including the marital status. The expanded model writes:
where Mijk is the marital status (marital vs. nonmartial) of the k-th partner of the i-th respondent on his/her j-th visit. Exp(β2) in the expanded model is the relative risk for HIV acquisition adjusting for the martial status of the partners. Likelihood ratio tests were used to test potential interactions. The comparison between equations (1) and (2) addressed the question, whether the contribution was directly associated with the risk factor itself or because of the underlying marital status. We further expanded the model to include other partner-related risk factors and individual risk behaviors/characteristics. The forward model selection with Akaike Information Criterion (AIC) was used to identify the final multivariate model.19 Specifically, we started with the most predictive bivariate model and added variables sequentially, to arrive at a set of variables that best predict the respondent's HIV incidence risk.
Based on the above, we present 4 sets of regression results (Tables 3 and 4): (1) unadjusted IRR between individual characteristics (of the partner and the respondent) and HIV acquisition, (2) adjusted IRR for the type of relationship with each partner (marital or nonmarital), (3) separate multivariate models among respondent and partner-related risk factors, and (4) the final mixed model to take into account both significant partner-related risk factors and characteristics of the respondents previously found to be associated with HIV acquisition among youth in Rakai.13
There were 4646 partner characteristics for 1969 male youth respondents (ie, 2.4 partners per respondent) and 4416 partner characteristics for 2826 female youth respondents (ie, 1.6 partners per respondent). Among them, there were 45 male HIV incident cases and 96 female incident cases detected in the 4 rounds of data collection.
Respondent characteristics previously found to be significantly associated with HIV acquisition among Rakai youth13 are summarized in Table 1. The median ages for female and male respondents were 22 and 21 years, respectively. Respondents were mostly rural. Sixty-two percent of young men had never been married, whereas 67% of young women were currently married. Ninety-five percent of female youth reported only 1 sexual partner in past 12 months. Among male youth respondents, 26% reported 2 partners and 13% reported 3 or more partners in the past 12 months. Symptoms of sexually transmitted disease were not uncommon amongst all youth, but female youth were more likely to report genital ulcers. Finally, 22% of male respondents and 20% of female respondents reported alcohol use before sex.
Description of Partner Characteristics
Partner characteristics varied among male and female respondents (Table 2). For male respondents, 91% reported nonmarital partners (n = 4226) and 9% were marital partners. In contrast, among the partners reported by female respondents, 63% were nonmarital and 37% were marital partners.
Several sociodemographic characteristics of marital and nonmarital partners were different for male and female respondents. Male respondents generally had partners who were younger, whereas most female respondents had partners who were older, regardless of the type of partnership. Almost all marital partners lived in households with respondents. Fifty-one percent of nonmarital partners of female youth respondents lived in the same household, whereas only 13% of nonmarital partners of male youth respondents shared the same household with the respondents. Forty-seven percent of nonmarital partners of male respondents were students, whereas only 2% of marital partners were students. In comparison, only 7% of female respondents reported nonmarital partners who were still in school.
Sexual activity characteristics also varied among marital and nonmarital partners for male youth respondents but less so for females youth respondents. With 37% of nonmarital partners, male respondents engaged in sexual activity within 1 month of knowing their partner compared to an interval <1 month for 17% of marital partners. This differential between marital and nonmarital partners was not observed in female respondents. Compared to male respondents, female respondents waited for significantly longer time before initiating sexual activity with their partners. Coital frequency was higher with marital than nonmarital partners for male and female youth respondents. Partners' alcohol use before sex was not common among male respondents, whereas female respondents reported that 29% of their nonmarital and 31% of their marital partners consumed alcohol before last sex. Condom use was very low for young men and young women. Both young men and young women reported more consistent condom use with nonmarital partners than marital partners (60% vs. 7% for male respondents and 31% vs. 2% for female respondents).
Similarly, partners' HIV risk varied between marital and nonmarital partners for male and female youth respondents. Young men and young women reported that more nonmarital partners were engaged in other sexual relationships than their marital partner (33% vs. 4% for men and 49% vs. 38% for women). However, when assessing their partner's HIV risk, young men attributed significantly higher perceived HIV risk for their nonmarital partners, whereas young women perceived HIV risk of their marital and nonmarital partners as approximately equal. Male respondents knew the HIV status of 49% of their nonmarital partners and 32% of their marital partners. In comparison, female respondents knew the HIV status of 56% of their nonmarital partners and marital partners. Marital partners were more likely to receive couple counseling with the respondents than nonmarital partners.
Partner Characteristics Associated With HIV Acquisition
Tables 3 and 4 present the IRRs for HIV acquisition among female and male respondents. IRRs reflect the risk of HIV acquisition per partner. Per our model assumptions, if a young woman has 2 nonmarital partners, her relative risk of HIV incidence is 1.602 or 2.56 compared to a young woman who only has a marital partner.
The first column presents the unadjusted IRR of HIV acquisition based on respondent and partner characteristics. For young women (Table 3), having a partner who was older, lived outside the same household, lived further away, was not a marital partner, worked as truck driver, drank alcohol before having sex, had 1 additional sexual partner, and was at least somewhat likely to be exposed to HIV significantly increased their risk of HIV acquisition. Moreover, for female respondents, receiving couple counseling with her partner was protective against HIV acquisition. For male respondents (Table 4), having a partner who was younger or older than the respondent, a partner who was not his current wife, a partner who had sex within 1 month of knowing the respondent, a partner who had sex more than 48 times (once a week in average) in the past 12 months, a partner who used condom inconsistently, a partner who had more than 2 sexual partners, and a partner who was not perceived to be at risk of HIV significantly increased the risk of HIV acquisition.
The second column in Tables 3 and 4 provide IRRs adjusted for the type of partnership. For female youth (Table 3) and independent of the type of partnership, partner's occupation as a truck driver [adjusted IRR: 1.97 (1.12 to 3.47)], a partner who drank alcohol before sex [adjusted IRR: 1.71 (1.24 to 2.36)], and a partner who used condoms inconsistently [adjusted IRR: 2.15 (1.49 to 3.11)] increased the risk of HIV acquisition. For male youth (Table 4), having a partner who was not a student [adjusted IRR: 1.59 (1.15 to 2.18)], who had increased coital frequency [adjusted IRR: 1.50 (1.01 to 2.21)], and whose HIV risk was unknown [adjusted IRR: 3.13 (1.19 to 8.20)] significantly increased the risk HIV acquisition.
The third column in Tables 3 and 4 present the results of the multivariate analyses. Model 1 presents the significant respondent-specific risk factors for HIV acquisition. For female youth, individual-level risk factors associated with HIV acquisition included living in a trading village, being separated or divorced, having 2 or more partners in the last year, and presence of genital warts. For male youth, the same individual-level risk factors were associated with HIV acquisition, but risk was increased with being currently married.
Model 2 presents factors associated with HIV acquisition among all the partner-related risk factors. After adjustment, a nonmarital partner [IRRModel 2: 2.10 (1.54 to 2.86)], a partner who drinks alcohol before sex [IRRModel 2: 1.76 (1.24 to 2.50)], and inconsistent condom use [IRRModel 2: 2.30 (1.53 to 3.46)] were significant factors for HIV acquisition among female youth (Table 3). For male youth respondents (Table 4), after adjustment, the following partner-related factors remain significantly associated with the HIV acquisition: partner was not a student [IRRModel 2: 1.50 (1.10 to 2.30)] and the partner's HIV risk was unknown to the respondent [IRRModel 2: 2.35 (0.98 to 5.64)].
Finally, the fourth column in Tables 3 and 4 presents the IRRs of the mixed-model multivariate analyses (model 3) that assess HIV acquisition risk based on significant partner-related risk factors and adjust for the respondent's own risk factors. For young women, nonmarital partners [IRRModel 3: 1.60 (1.11 to 2.32)], partners who drank alcohol before sex [IRRModel 3: 1.57 (1.11 to 2.21)], and used condoms inconsistently [IRRModel 3: 1.99 (1.33 to 2.98)] were significantly associated with HIV acquisition (Table 3). Whereas, for young men, nonmarital partners [IRRModel 3: 1.54 (1.20 to 1.98)] were significantly associated with HIV acquisition after adjustment for individual respondent characteristics (Table 4).
Partner characteristics independently influenced HIV risk among youth in Rakai. For both young men and young women, after controlling for individual-level risk factors, sex with nonmarital partners increased the likelihood of HIV acquisition. Additionally, for women, the behaviors of their male partners—alcohol use and inconsistent condom use—enhanced the risk of HIV. Risk of HIV acquisition increased if respondents had a partner with multiple high-risk characteristics or multiple partners with 1 or more high-risk characteristics.
Previous research has shown considerable variation in reported marital and nonmarital sexual partnerships among youth.20,21 Opio et al22 found increases in reports of nonspousal sexual activity among young men in Uganda between 2001 and 2005, at a time when Uganda was no longer experiencing declines in HIV prevalence. Young men in our sample were more likely to have nonmarital partners compared to young women. For young men, nonmarital partnerships were characterized by greater age disparities, partners from outside the household, partners who were enrolled in school, shorter relationship durations, increased sexual concurrency, stronger suspicions of HIV risk, and limited knowledge of partner HIV status compared to their marital partner characteristics. Among young women's nonmarital partners, characteristics of note included living outside the household, higher coital frequency, greater partner concurrency, and limited knowledge of the partner's HIV status compared to marital partner characteristics. We found that nonmarital partnerships added substantial risk of HIV acquisition to both young men and young women in Rakai. Greater exploration of couple dynamics within marital and nonmarital partnerships is needed.
Our study confirms previous research that partner characteristics are an important determinant of HIV risk for female youth.4 Type of partnership and the context of sexual activity with partners influenced young women's risk of HIV acquisition. Partner's use of alcohol before engaging in sexual activity enhanced risk of HIV acquisition among young women in Rakai. Similarly, inconsistent condom use significantly predicted HIV acquisition when we controlled for the type of partnership for female youth. Age discrepant relationships did not increase the likelihood of HIV acquisition when controlling for other partner-related factors, as found in previous studies.23 For young men, it was primarily the type of relationship that they were engaged in rather than their partner's sociodemographic, sexual activity, or HIV risk characteristics that increased their risk for HIV acquisition. It is likely that young women's limited ability to negotiate the risk behaviors of their partners is putting them at risk for HIV, whereas young men might have greater control over the behavior of their partners or their partner selection.
Partner risk characteristics were reported by the respondents and were subject to potential reporting errors and biases. Future research linking respondent information with marital and nonmarital partners' information could help assess potential bias, but such linked partnership data are difficult to capture. We did not have data on the exact age difference between the respondent and the partner. Previous analysis in Rakai showed that use of alcohol was associated with HIV incidence,13 although we have consistent alcohol use information on the partners, respondent's own alcohol use questions were not consistently asked across all survey rounds and were not included in this analysis.
Partner attributes contribute substantially to HIV risk among youth in rural Uganda independent of individual characteristics and risk behaviors. HIV prevention programs targeting youth need to address risk from marital and nonmarital partners. At the individual-level HIV prevention programs should emphasize knowledge of partner characteristics that increase HIV risk, in addition to current emphasis on abstinence, condom use, partner reduction, and HIV testing and disclosure. Young people need increased access to couple counseling and HIV testing. Furthermore, HIV prevention efforts might need to take into account how to tailor HIV risk and prevention messages for different types of partners. Because partner characteristics can influence HIV risk, youth also need greater access to life skills programs that help them negotiate and potentially influence the behaviors of their partners within the relationship context.
The authors would like to thank Drs Zoe Edelstein and Stephane Helleringer for their constructive feedback during the analysis process. We would also like to thank the donors that support the Rakai Health Sciences Program (RHSP), the data team at RHSP, and the cohort participants.
1. Swartzendruber A, Zenilman JM, Niccolai LM, et al.. It takes 2: partner attributes associated with sexually transmitted infections among adolescents. Sex Transm Dis. 2013;40:372–378.
2. Hargreaves JR, Morison LA, Kim JC, et al.. Characteristics of sexual partnerships, not just of individuals, are associated with condom use and recent HIV infection in rural South Africa. AIDS Care. 2009;21:1058–1070.
3. United Nations Children’s Fund. Opportunity in Crisis: Preventing HIV From Early Adolescence to Young Adulthood. New York, NY: UNICEF; 2011.
4. Napierala Mavedzenge S, Olson R, Doyle AM, et al.. The epidemiology of HIV among young people in sub-Saharan Africa: know your local epidemic and its implications for prevention. J Adolesc Health. 2011;49:559–567.
5. DiClemente RJ, Salazar LF, Crosby RA. A review of STD/HIV preventive interventions for adolescents: sustaining effects using an ecological approach. J Pediatr Psychol. 2007;32:888–906.
6. Gregson S, Nyamukapa CA, Garnett GP, et al.. Sexual mixing patterns and sex—differentials in teenage exposure to HIV infection in rural Zimbabwe. Lancet. 2002;359:1896–1903.
7. Jewkes RK, Dunkle K, Nduna M, et al.. Intimate partner violence, relationship power inequity, and incidence of HIV infection in young women in South Africa: a cohort study. Lancet. 2010;376:41–48.
8. Kalichman SC, Ntseane D, Nthomang K, et al.. Recent multiple sexual partners and HIV transmission risks among people living with HIV/AIDS in Botswana. Sex Transm Infect. 2007;83:371–375.
9. Morris M, Epstein H, Wawer M. Timing is everything: international variations in historical sexual partnership concurrency and HIV prevalence. PLoS One. 2010;5:e14092.
10. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS. 1997;11:641–648.
11. Msuya SE, Mbizvo E, Hussain A, et al.. HIV among pregnant women in Moshi Tanzania: the role of sexual behavior, male partner characteristics
and sexually transmitted infections. AIDS Res Ther. 2006;3:27.
12. Woolf-King S, Maisto S. Alcohol use, high-risk sexual behavior in sub-Saharan Africa: a narrative review. Arch Sex Behav. 2011;40:17–42.
13. Ministry of Health; ICF International; Centers for Disease Control and Prevention, U.S. Agency for International Development, and WHO Uganda
AIDS Indicator Survey. Kampala, Uganda
and Calverton, MD: MOH and ICF; 2011.
14. Santelli JS, Edelstein ZR, Mathur S, et al.. Behavioral, biological, and demographic risk and protective factors for new HIV infections among youth
in Rakai, Uganda
. J Acquir Immune Defic Syndr. 2013;63:393–400.
15. Wawer MJ, Gray RH, Sewankambo NK, et al.. A randomized, community trial of intensive sexually transmitted disease control for AIDS prevention, Rakai, Uganda
. AIDS. 1998;12:1211–1225.
16. Decosas J, Kane F, Anarfi JK, et al.. Migration and AIDS. Lancet. 1995;346:826–828.
17. Bwayo J, Plummer F, Omari M, et al.. Human immunodeficiency virus infection in long-distance truck drivers in east Africa. Arch Intern Med. 1994;154:1391–1396.
18. Kapiga SH, Sam NE, Shao JF, et al.. HIV-1 epidemic among female bar and hotel workers in northern Tanzania: risk factors and opportunities for prevention. J Acquir Immune Defic Syndr. 2002;29:409–417.
19. Akaike H. A new look at the statistical model identification. Automatic Control IEEE Trans. 1974;19:716–723.
20. Harrison A, Cleland J, Frohlich J. Young people's sexual partnerships in KwaZulu-Natal, South Africa: patterns, contextual influences, and HIV risk
. Stud Fam Plann. 2008;39:295–308.
21. Harrison A, O'Sullivan L. In the absence of marriage: long-term concurrent partnerships, pregnancy, and HIV risk
dynamics among South African young adults. AIDS Behav. 2010;14:991–1000.
22. Opio A, Mishra V, Hong R, et al.. Trends in HIV-related behaviors and knowledge in Uganda
, 1989-2005: evidence of a shift toward more risk-taking behaviors. J Acquir Immune Defic Syndr. 2008;49:320–326.
23. Harling G, Newell M-L, Tanser F, et al.. Do age-disparate relationships drive HIV incidence in young women? Evidence from a population cohort in rural KwaZulu-Natal, South Africa. J Acquir Immune Defic Syndr. 2014;66(4):443–451.
Keywords:Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
partner characteristics; HIV acquisition; HIV risk; youth; Uganda