Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 1.
Published in final edited form as: Arch Sex Behav. 2020 Mar 30;49(6):2057–2068. doi: 10.1007/s10508-020-01663-5

Clusters of HIV risk and protective sexual behaviors in Agincourt, rural South Africa: Findings from the Ha Nakekela population-based study of ages 15 and older

Brian Houle 1,2,3,*, Shao-Tzu Yu 1, Nicole Angotti 2,3,4, Enid Schatz 2,3,5, Chodziwadziwa W Kabudula 2,6, Francesc Xavier Gómez-Olivé 2,6, Samuel J Clark 2,6,7, Jane Menken 2,3, Sanyu A Mojola 8
PMCID: PMC7321875  NIHMSID: NIHMS1580689  PMID: 32232623

Abstract

Understanding sexual behavior patterns in distinct population sub-groups along the life course is critical for effective targeting and tailoring of HIV prevention messaging and intervention activities. We examined interrelatedness of sexual behaviors and variation between men and women across a wide age range in a rural South African setting with a high HIV burden. Data come from the Ha Nakekela population-based survey of people aged 15–85-plus drawn from the Agincourt Health and Socio-Demographic Surveillance System. We used latent class analysis of six sexual behavior indicators to identify distinct sub-group sexual behavior patterns. We then examined associations between class membership and socio-demographic and other behavioral risk factors and assessed the accuracy of a reduced set of sexual behavior indicators to classify individuals into latent classes. We identified three sexual behavior classes: (1) single with consistent protective behaviors; (2) risky behaviors; and (3) in union with lack of protective behaviors. Patterns of sexual behaviors varied by gender. Class membership was also associated with age, HIV-status, nationality, and alcohol use. With only two sexual behavior indicators (union status and multiple sexual partners), individuals were accurately assigned to their most likely predicted class. There were distinct multidimensional sexual behavior patterns in population sub-groups that varied by sex, age, and HIV status. In this population, only two brief questions were needed to classify individuals into risk classes. Replication in other situations is needed to confirm these findings.

Keywords: clustering, sexual behavior, HIV, South Africa

INTRODUCTION

Information about sexual behaviors across the life course and how behaviors vary by subgroups is critical to informing the targeting and prioritization of at-risk groups and designing prevention messaging and interventions to improve sexual health. Increasing evidence shows that individuals are at risk of HIV acquisition throughout the life course – including at older ages (Houle et al., 2018; Rosenberg et al., 2017). However, most existing studies examine individual risk and protective factors, such as condom use at last sex, instead of considering how sexual behaviors cluster together and vary across the life course by distinct subgroups in the population, such as gender. Studies suggest that patterns of health behaviors are driven by events and decisions made at different life stages, as well as social experiences, contexts and disparities (Cockerham, 2005). Structural and social factors influence sexual behaviors – with differential effects across age and gender (Mojola, Williams, Angotti, & Gómez-Olivé, 2015) – including gender norms, migration, and socioeconomic changes (Wellings et al., 2006). Similarly, inequalities in HIV burden persist due to social contexts including healthcare access and gender-inequalities (Joint United Nations Programme on HIV/AIDS (UNAIDS), 2014). Greater evidence is needed to understand the patterns of sexual behaviors in distinct population sub-groups, along with the social drivers of these patterns and their link to HIV risk.

Most studies that have examined patterns across multiple sexual behavior indicators have been in high income countries (Akers et al., 2016; Cochran, de Leeuw, & Mays, 1995; Danielson et al., 2014; Dariotis et al., 2008; Davies et al., 2014; Hallfors, Iritani, Miller, & Bauer, 2007; Halpern et al., 2004; Haydon, Herring, & Halpern, 2012; Mackesy-Amiti et al., 2014; Masters, Beadnell, Morrison, Hoppe, & Wells, 2013; Masters et al., 2015; McMahon, Stanforth, Devieux, & Jean-Gilles, 2016; Mustanski et al., 2013; Noor, Ross, Lai, & Risser, 2014; Pflieger, Cook, Niccolai, & Connell, 2013; Vasilenko, Kugler, Butera, & Lanza, 2015; Waller et al., 2006; Wu, Witkiewitz, McMahon, & Dodge, 2010) (See Supplementary Figure 1 and Supplementary Table 1). Of the studies in Africa, most have focused on adolescents (Cederbaum, Gilreath, & Barman-Adhikari, 2014; Tibbits, Caldwell, Smith, Vergnani, & Wegner, 2016) and young adults (M. A. Carrasco, Nguyen, & Kaufman, 2018; Wechsberg et al., 2012) or men or women only (M. A. Carrasco et al., 2018; Wechsberg et al., 2012; Wechsberg et al., 2017). To the best of our knowledge, no studies have examined sexual behavior patterns in a high HIV prevalence setting in a large sample of both men and women across a wide age range.

South Africa, with the world’s largest HIV burden and antiretroviral therapy program (UNAIDS, 2016), represents a crucial setting for greater knowledge of interrelated sexual behaviors and HIV risk. Evidence from former apartheid homeland areas, such as the Agincourt Health and Socio-Demographic Surveillance site (Agincourt) in rural northeast South Africa, highlights the links between social contexts and inequalities in HIV burden and circumstances supporting safer sexual behaviors (Mojola et al., 2015). Agincourt has been impacted by high AIDS-related mortality (Kabudula et al., 2017) and persistent socioeconomic mortality differentials (Kabudula et al., 2016). Norms of hegemonic masculinity have been linked to increased HIV risk, multiple sexual partnerships, and interpersonal violence (Gorbach, Drumright, & Holmes, 2005; Hunter, 2007; Varga, 2003). Limited employment opportunities have led to high levels of labor migration (Clark, Collinson, Kahn, Drullinger, & Tollman, 2007; Kahn et al., 2012), thereby impacting relationships (Lurie et al., 2003; Mojola et al., 2015).

We used the Ha Nakekela population-based survey of men and women aged 15–85-plus in Agincourt to examine the interrelatedness of sexual behaviors, variation in behavior patterns between men and women, and how sexual behavior patterns are influenced across age groups by socio-demographic and behavioral factors, as well as HIV sero-status.

METHODS

We used latent class analysis (LCA) to investigate how individuals reporting any sexual partners in the past two years fall into distinct subgroups/classes with different sexual behavior patterns and evaluated subgroup differences between men and women. We then controlled for socio-demographic and behavioral covariates and examined the effect of gender, age, and HIV status on class membership. Finally, for future screening purposes, we explored how accurately a reduced set of sexual behavior indicators identified class membership.

Setting and data sources

Data come from the Agincourt Health and Socio-Demographic Surveillance System (AHDSS). The AHDSS conducts an annual census of the population, collecting information on all vital events as well as sociodemographic statuses (Kahn et al., 2012). At the time of the study, the population under surveillance comprised approximately 90,000 people in 27 villages.

The Ha Nakekela survey was conducted in 2010–11. It included measures of HIV, noncommunicable disease risk factors and biomarkers, and lifetime and recent (past two years) sexual behaviors and partners (Gómez-Olivé et al., 2013; Houle et al., 2018). The study targeted a random, sex-age stratified sample of 7662 individuals, including an oversample of older adults, who were resident in 2009 in the AHDSS, of whom 4362 constituted the final analytic sample and 3541 reported at least one sexual partner in the past two years.

Measurements

Sexual behavior indicators.

We included respondent formal or informal union status (based on the 2009 AHDSS census, where formal union includes those who are married with lobola (bridewealth) and informal union includes those couples living together without lobola) and self-reports of ever having been diagnosed with a sexually transmitted infection (STI, excluding HIV) and, in the past two years: total number of sexual partners; proportion of partners with whom a condom was used ‘mostly/always’; proportion of partners whose HIV status was known the first time they had sex; and whether any partner was casual/anonymous (compared to regular).

Socio-demographic and behavioral factors.

We included sociodemographic factors of: gender, age, education level, tertile of household socioeconomic status, and nationality (South African or Mozambican/Other) from the 2009 census (Kabudula et al., 2016). (About a quarter of the population in the 2009 census was Mozambican, former refugees of the Mozambican civil war.) We also included respondent’s HIV status from the survey biomarker test, self-report of ever having an HIV test, and alcohol consumption in the past month.

Statistical analysis

In each of our LCAs, we fit a series of models that included 2–6 classes. We evaluated model fit using a variety of criteria, focusing primarily on the Bayesian information criterion (BIC) to select a parsimonious model, as it penalizes model complexity more than other measures (Nylund, Asparouhov, & Muthén, 2007). To identify the number of latent classes, we selected the best model according to the BIC and examined the resulting classes for substantive differences and interpretation. For each respondent and class, LCA estimated the probability that the respondent was a member of that class. It also provided a variable assigning the respondent to the most likely class.

We began by fitting a series of models of the six sexual behavior indicators defined above to the total population. We next fit sex-specific models to test if men and women were best represented by the same number of classes. Then, to test for differences between men and women, we fit a multi-group LCA model. This model allowed us to test for measurement invariance between men and women. To test for full measurement invariance, we compared the fully unconstrained model (all item-response probabilities (the probability of responding yes to each item given class membership) could vary by sex) with the fully constrained model (all item-response probabilities constrained to be the same for men and women) using a chi-square difference test (Satorra & Bentler, 2010). To test for partial measurement invariance, we substantively reviewed differences between men and women on item-response probabilities for each indicator in the fully unconstrained model, subsequently allowing gender-specific direct effects if there was substantive variation across gender and class and holding the remaining probabilities equal by gender. We selected the resulting (partial) measurement invariant model as the final LCA model.

We included the sociodemographic and behavioral variables defined above in a one-step, complete case LCA multinomial logistic regression with direct effects by gender to predict the probability of latent class membership. We tested two-way interactions between gender and age, gender and HIV status, and age and HIV status. Since the conditional item-response probabilities may be driven by inclusion of covariates and direct effects, we compared the conditional model (with covariates and direct effects) with the unconditional model to ensure the model-building process did not change the substantive meaning of class formation.

We tested the sensitivity of our results to the inclusion of other indicators, including ever having an HIV test and indicators based on just the most recent partner, including: age difference from respondent, discussed HIV before first sex, and condom use at last sex. Finally, we examined whether behaviors clustered differently by HIV status. Our results did not change substantively in these alternative models. We preferred our final model as it included information across all sexual partners in the past two years.

Finally, to determine if a reduced set of sexual behavior indicators could accurately classify individuals into their most likely predicted class (based on the highest posterior probability), we used simplified measures of the most discriminating indicators (based on substantive review of large differences in item-response probabilities between classes and high or low item-response probabilities) to predict class membership; we then summarized the results using sensitivity and specificity estimates, and positive and negative predictive values (Akobeng, 2006). All analyses were completed using Mplus 8.

RESULTS

Sexual behavior classes

The LCA model-fitting indicated a three-class solution for the whole population according to BIC and substantive interpretation and for gender subpopulations (Supplementary Figure 2), suggesting that the latent structure did not differ by gender. Comparing the unconstrained multi-group LCA model (LL: −11088.825; df(59); BIC=22653.334) to the constrained model (LL: −11162.476; df(32); BIC=22582.950) indicated a lack of measurement invariance between men and women (p<.001). Allowing union status, then number of sexual partners, then past STI diagnosis probabilities to vary by gender achieved partial measurement invariance (LL: −11107.783; df(40); BIC=22538.064; p=.078). We chose this as our final measurement model. This model suggests that the pattern of behaviors in Class 1 is the same for both men and women, but there are some differences by gender in sexual behavior patterns in Classes 2 and 3.

Figure 1 shows the conditional probabilities for each sexual behavior indicator by most likely predicted class membership (also shown in Supplementary Table 2). We summarize each class using the following descriptive labels:

Figure 1.

Figure 1.

Open in a new tab

Conditional probabilities of each sexual behavior by Class, Agincourt, South Africa 2010 – 11 (n=3173). Results based on a three-class LCA model in which the probabilities of union status, number of sexual partners, and past STI diagnosis varied by gender for Classes 2 and 3. In union means the respondent is in a formal (those who are married with lobola (bridewealth)) or informal (those couples who are living together without lobola) union. Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Class 1 – Single with consistent protective behaviors: Class 1 was the second largest group (25%) and had near zero probability of being in union/married. Compared to the other classes, Class 1 reported the highest probability of ‘mostly/always using a condom’ with their partners (63%), and also the highest probability of always knowing their partner’s HIV status before they first had sex (17%). Class 1 also had a high probability of having mainly regular (74%) vs. casual/anonymous partners and having one sexual partner in the past two years (94%). There were no differences in the indicators between men and women.

Class 2 – Risky behaviors: Class 2 represented 15% of the sample. Class 2 included a mix of condom use behaviors, with a 12% probability of ‘mostly/always using a condom’ with all partners and a 46% probability of ‘mostly/always using a condom’ with some partners. They had a low (4%) probability of always knowing their partner’s HIV status before they first had sex, and a 71% probability of having at least one casual partner in the past two years. There were differences in some indicators by gender: compared to women, men had higher probabilities of ever being diagnosed with an STI (men=28%; women=17%), having 3 or more sexual partners in the past two years (men=34%; women=23%), and being in a union (men=31%; women=6%).

Class 3 – In union with lack of protective behaviors: Class 3 was the most common group, representing over half the sample. Class 3 had a low probability of ‘mostly/always using a condom’ with all partners and knowing their partner’s HIV status before they first had sex. They had an almost 100% probability of having only regular partners in the past two years. There were slight differences in indicators by gender: compared to women, men had higher probabilities of being in union (men=79%; women=75%), having two partners in the past two years (men=6%; women=0%), and ever having been diagnosed with an STI (men=18%; women=11%).

Associations with latent class membership

We next estimated the model including socio-demographic and behavioral covariates. An interaction between age and HIV status significantly improved model fit (ΔBIC=−35.278; p=.014), resulting in the final model.

Table 1 shows socio-demographic characteristics by most likely predicted sexual behavior class based on the final model. Table 2 shows adjusted odds ratios of being in Classes 1 or 2 compared to Class 3. Those born in South Africa were more likely to be in Classes 1 (OR=1.76 [95% CI 1.31–2.36]) or 2 (OR=1.68 [95% CI 1.24–2.29]) than Class 3, compared to non-South Africans. Those who consumed alcohol 1–3 days per month (OR=1.90 [95% CI 1.19–3.05]) or weekly (OR=2.54 [95% CI 1.40–4.63]) had a higher likelihood of being in Class 2 than Class 3, compared to those reporting no alcohol use. Those with the highest education were more likely to be in Class 3 relative to those with no education, while SES was not associated with class membership. Men were more likely to be in Classes 1 or 2 than Class 3 compared to women.

Table 1.

Socio-demographic and behavioral characteristics by most likely predicted class membership (n = 3,173).

Overall Class 1 Class 2 Class 3 p value
Population share 3,173 1,039 (33%) 474(15%) 1,660 (52%)
Gender
Female 1.850 (58%) 682 (37%) 126 (7%) 1,042 (56%) < .0001
Male 1,323 (42%) 357 (27%) 348 (26%) 618 (47%)
HIV
Negative 2,269 (72%) 638 (28%) 327 (14%) 1,304 (58%) < .0001
Positive 904 (28%) 401 (39%) 147 (16%) 356 (45%)
Age (Mean (SD)) 38 (16.0) 29 (9.7) 30 (12.0) 47 (15.4) < .0001
Nationality
Elsewhere 1,002 (32%) 301 (30%) 146 (15%) 555 (55%) .049
South African 2,171 (68%) 738 (34%) 328 (15%) 1,105 (51%)
Evert tested for HIV
No 1,264 (40%) 367 (29%) 227 (18%) 670 (53%) < .0001
Yes 1,909 (60%) 672 (35%) 247 (13%) 990 (52%)
Education level (in years)
0 years 548 (17%) 75 (14%) 32 (6%) 441 (80%) < .0001
1–11 years 1,992 (63%) 702 (35%) 350 (18%) 940 (47%)
12+ years 633 (20%) 262 (41%) 92 (15%) 279 (44%)
Household assets
Lowest quintile 1,178 (37%) 407 (34%) 198 (17%) 573 (59%) .007
Middle quintile 1,010 (32%) 338 (33%) 137 (14%) 535 (53%)
Highest quintile 985 (31%) 294 (30%) 139 (14%) 552 (56%)
Alcohol consumption (per month)
None 2,470 (78%) 844 (34%) 254 (10%) 1,372 (56%) < .0001
1–3 days per month 341 (11%) 108 (32%) 100 (29%) 133 (39%)
Weekly 362 (11%) 87 (24%) 120 (33%) 155 (43%)
Open in a new tab

Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Table 2.

LCA multinomial logistic regression with direct effects (n = 3,173). Class 3 is the referent group.

Class 1 Class 2
OR 95% CI OR 95% CI
Gender
Female (base)
Male 1.83 [1.27 – 2.65] 7.59 [4.79 – 12.01]
HIV
Negative (base)
Positive 0.08 [0.01 – 0.91] 0.95 [0.29 – 3.10]
Age 0.84 [0.76 – 0.93] 0.88 [0.86 – 0.90]
Nationality
Elsewhere (base)
South African 1.76 [1.31 – 2.36] 1.68 [1.24 – 2.29]
Ever tested for HIV
No (base)
Yes 0.71 [0.54 – 0.93] 0.84 [0.62 – 1.13]
Education level (in years)
0 years (base)
1–11 years 0.47 [0.27 – 0.82] 0.59 [0.34 – 1.03]
12+ years 0.56 [0.32 – 0.96] 0.46 [0.24 – 0.88]
Household assets
Lowest quintile (base)
Middle quintile 1.16 [0.83 – 1.62] 1.16 [0.79 – 1.69]
Highest quintile 0.93 [0.69 – 1.25] 0.82 [0.58 – 1.16]
Alcohol consumption (per month)
None (base)
1–3 days per month 1.36 [0.87 – 2.11] 1.90 [1.19 – 3.05]
Weekly 1.33 [0.61 – 2.91] 2.54 [1.40 – 4.63]
Interaction
Age*HIV 1.11 [1.01 – 1.22] 1.02 [0.98 – 1.05]
Open in a new tab

Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Figure 2 shows predicted probabilities of class membership by gender, age, and HIV status. For HIV negative men, Class 1 had the highest probability until their late twenties, and Class 3 had the highest probability in their early-mid thirties. Class 2 had the second highest probability at all ages. For HIV positive men, Class 2 had the highest probability until their late twenties, and Class 3 had the highest probability in their early-mid forties. Class 1 had the second highest probability at all ages. For HIV negative women, the class with the highest probability changed from Class 1 to Class 3 in their mid-late twenties, with a low probability of being in Class 2 across all ages. For HIV positive women, the class with the highest probability of membership changed from Class 1 to Class 3 in their mid-thirties, with a relatively higher probability of being in Class 2 until their mid-late twenties compared to HIV negative women.

Figure 2.

Figure 2.

Open in a new tab

Predicted probabilities of latent class membership by gender, age, and HIV status, Agincourt, South Africa 2010 – 11 (n=3173). Results based on the final LCA multinomial logistic regression model with three classes. All other covariates were set at their mean levels for the estimation sample. Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Since the LCA model is restricted to those reporting any sexual partners in the past 2 years, Figure 3 presents the probabilities of being in one of the latent classes or reporting no sexual partners in the past two years, by gender, age, and HIV status. Women in older ages reported a higher probability of no sexual partners compared to men, regardless of HIV status. HIV positive women reported slightly higher probabilities of no sexual partners at ages 35–64 compared to HIV negative women.

Figure 3.

Figure 3.

Open in a new tab

Posterior probabilities of latent class membership, and proportions of individuals with no class membership due to missing covariate values and those reporting no sexual partners in the past two years, by gender, age, and HIV status, Agincourt, South Africa 2010 – 11 (n=4171). Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Finally, we tested the ability of a reduced set of sexual behavior indicators to accurately classify individuals into their most likely predicted class – selecting: union status and whether the respondent reported one or more than one sexual partner. This was based on assessing the most discriminating indicators in terms of high and low item-response probabilities across classes. Reporting multiple partners was 100% in Class 2 and very low in Classes 2 and 3. Conversely, being in a union was high in Class 3 and very low in Class 1. Assigning to Class 1 those reporting one partner and not being in union yielded sensitivity of .96 and specificity of .92. Assigning to Class 2 those reporting more than one sexual partner yielded sensitivity of 1.0 and specificity of .97. Assigning to Class 3 those reporting one sexual partner and being in union yielded a sensitivity of .86 and specificity of 1.0. Positive and negative predictive values are presented in Table 3.

Table 3.

Sensitivity, specificity and positive/negative predictive values (PPV/NPV) for simplified sexual behavior indicators on predicted latent class memberships (n = 3,173).

Sexual behaviour indicators Sensitivity 95% CI Specificity 95% CI PPV 95% CI NPV 95% CI
Not in a union & 1 sexual partner (Predicting class 1) 96.05 % 91.52 % 84.65 % 97.94 %
[94.68 – 97.15] [90.25 – 92.67] [82.46 – 86.66] [97.22 – 98.52]
2 or more sexual partners (Predicting class 2) 100 % 96.63 % 83.89 % 100 %
[99.22 – 100] [95.88 – 97.28] [80.60 – 86.83] [99.86 – 100]
In a union & 1 sexual partner (Predicting class 3) 86.08 % 100 % 100 % 86.75 %
[84.33 – 87.71] [99.76 – 100] [99.74 – 100] [85.07 – 88.31]
Open in a new tab

Class 1 – Single with consistent protective behaviors; Class 2 – Risky behaviors; Class 3 – In union with lack of protective behaviors.

Sensitivity: For those who belonged to a specific latent class, the proportion of participants to give a positive response to the set of sexual behavior indicators.

Specificity: For those who did not belong to a specific latent class, the proportion of participants to give a negative response to the set of sexual behavior indicators.

PPV: Given the positive response to the set of sexual behavior indicators, the probability of participants who were truly in a specific latent class.

NPV: Given the negative response to the set of sexual behavior indicators, the probability of participants who were not truly in a specific latent class.

DISCUSSION

This study is among the first to examine patterns of sexual behavior across gender and a wide age range in a high HIV prevalence setting. The vast majority of young individuals were characterized by being single with consistent protective behaviors, and the majority of older adults were characterized by being in a union which lacked protective behaviors. Patterns of sexual behaviors also varied significantly by age, gender, and HIV status. Shifts in the most common class, from single and consistent protective behaviors (Class 1) or risky behaviors (Class 2) to being in union with lack of protective behaviors (Class 3), occurred at later ages for HIV positive compared to HIV negative individuals. Being in the risky behaviors class was associated with alcohol use (M. Carrasco, Esser, Sparks, & Kaufman, 2016) and was more prevalent among young and middle-aged men compared to women.

Gender differences in the sexual behavior classes suggests that men and women vary in “who” they partner with but not “how”, underscoring the importance of an individual’s relationship status (Logan, Cole, & Leukefeld, 2002). For instance, the risky behaviors class represents a small group of almost entirely single women, but represents a larger group of some men, only some of whom are in a union. As evidence indicates that women tend to have older partners and men younger partners (Houle et al., 2018), this suggests that women in a union with lack of protective behaviors are at risk of HIV due to low rates of condom use with their male partners. Conversely, women aware of their husband’s behaviors (Houle et al., 2018) may be employing alternative protective strategies such as celibacy, as indicated by the increasing proportions of sexually inactive women in older ages (Mojola et al., 2015). However, for men in the risky behaviors class with multiple partners, their likely younger female partner(s) are also at risk, with potentially increased difficulty in negotiating safer sex with an older, male partner with higher SES (Bajos & Marquet, 2000; Kordoutis, Loumakou, & Sarafidou, 2000).

Our results showed substantial age variation in HIV risk and sexual behaviors. For younger adults, our findings were similar to a study from Malawi amongst younger men that also identified low and high risk sexual behavior classes (M. A. Carrasco et al., 2018). Our results also suggest a “settling down” effect in older adults as the probability of being in the union with lack of protective behaviors class increased across age. Longitudinal evidence among U.S. men ages 15–26 showed an increasing transition of men from high to low risk classes (Dariotis et al., 2008). Our results suggest that the duration of membership in at-risk classes and the pace of transition to lower risk classes may be important in understanding life course HIV vulnerability, but longitudinal evidence is needed to clarify transitions in sexual behaviors at different life stages.

While our results demonstrate the patterning of multiple sexual behaviors across individuals, they also suggest that prior studies focusing on individual indicators may not adequately screen for sexual behavior risk. However, using only two indicators – whether the respondent was in union and whether they reported multiple partners – showed that these quickly obtainable measures divided respondents well into the three risk classes. These findings align with results from the 2016 South Africa Demographic and Health Survey (National Department of Health, 2019). For women, they found that having multiple partners was rare overall and particularly amongst married women. Having multiple partners was also uncommon amongst married men, but was higher than women (11% vs. 3%). The proportions of those with multiple partners reporting condom use at last sex were also similar to the reported levels for the risky behaviors class in this study. We conclude that these two indicators could be used effectively in screening and assessment activities to identify those in need of further risk and prevention counselling needs and match them to appropriate intervention approaches.

The multidimensional aspect underlying these measures can also inform HIV prevention, counselling, and intervention design and assessment. Our findings highlight the importance of recognizing the wider social contexts that influence individuals’ behavioral patterns at different life stages. For younger adults in the single with consistent protective behaviors class, prevention interventions are needed around changing social norms to support HIV status disclosure before sex. Evidence from Europe suggests that the social context where individuals first meet their partners may be an important determinant to discussions before first sex (Bajos & Marquet, 2000). The relatively lower levels of consistent condom use with one partner aligns with tailored messaging for specific relationship contexts, as adults are less likely to consistently use condoms with regular partners (Logan et al., 2002). Given high levels of intimate partner violence among young women in this setting (Pettifor et al., 2016), structural interventions that reduce poverty, unemployment and support changing gender norms (Dworkin, Treves-Kagan, & Lippman, 2013) may help reduce power differentials with older men in negotiating safer sex. For men, there is a higher probability of being in the risky behaviors class in age groups where circular labor migration is also common. Targeting migrant men at their workplace (Elwy, Hart, Hawkes, & Petticrew, 2002) and working towards supportive environments in workplace communities (Gilbert & Walker, 2002) and the women with whom they partner could benefit prevention program strategies. For older women, counselling activities are needed that include both partners around communication strategies supporting celibacy or condom use when men retire or return home (Lurie et al., 2003).

We note some limitations. First, study findings may have limited generalizability as they are based on a rural South African setting in 2010 – 2011 when ART availability was just becoming widespread (Houle, Clark, Gómez-Olivé, Kahn, & Tollman, 2014; Mee et al., 2014). Sexual behavior patterns in different population subgroups may have changed since then, as a result of greater treatment availability; however, more recent evidence on younger (Pettifor et al., 2016) and older adults (Rosenberg et al., 2017) suggests similarities with results from this study that focused on individual sexual behavior indicators (Houle et al., 2018). For instance, Pettifor et al. (2016) showed relatively low levels of condom use at last sex and older age partners for young women. Amongst adults ages 40 and older, Rosenberg et al. (2017) showed generally low condom use and multiple partnerships at moderate levels. The timing of ART availability at the time of this study is also important for understanding changes in sexual behavior. While studies of those on ART from generalized epidemic settings indicate a reduction in risky sexual behaviors (Berhan & Berhan, 2012; Doyle et al., 2014; Venkatesh, Flanigan, & Mayer, 2011), less is known in the general population that includes HIV-negative individuals (McGrath, Eaton, Bärnighausen, Tanser, & Newell, 2013). Our results on the multidimensionality of sexual behaviors provides an important comparison for how sexual behaviors might change, including the prevalence of different classes, and how the number and patterns within classes may change differentially by gender in light of the long-term impact of ART on general populations as well as other prevention programs. Our findings provide an important comparison for studies in other settings, as well as comparisons to other time periods when ART became more widespread. Second, the data are cross-sectional; longitudinal data are needed to understand how individuals transition between sexual behavior classes throughout their lives. Longitudinal data are particularly needed to understand how knowledge of one’s HIV status and treatment access may change sexual behavior patterns. Third, since individuals self-reported their sexual behaviors to local interviewers, our data may be subject to social desirability and recall biases (Bajos & Marquet, 2000; Houle et al., 2016), and in ways that vary by gender (Nnko, Boerma, Urassa, Mwaluko, & Zaba, 2004). Finally, while our LCA models were robust to inclusion/exclusion of other behavior indicators, the measurement model is inherently limited by the available measures in the survey. Despite our analysis of 11 indicators of recent and lifetime sexual behaviors, there may be other indicators or qualitative data (George, 1993) that may provide new explanatory information (Angotti, Houle, Schatz, & Mojola, 2018) or greater discrimination in selecting the number of sexual behavior classes. We especially point to our lack of information on partner behaviors, and believe our findings highlight the importance of including partners’ behaviors to contextualize sexual behavior patterns.

CONCLUSION

We provide one of the most detailed examinations of sexual behavior patterns between men and women across a wide age range in the context of a severe HIV epidemic. Greater understanding of the multidimensional nature of an individual’s sexual behaviors is critical to inform the targeting and prioritization of HIV prevention messaging programs and interventions. The aging of the HIV epidemic (Vollmer et al., 2016), coupled with demographic and social changes, also highlights the importance of understanding the contexts and life course stages in which behaviors occur, such as gender roles and socioeconomic disparities. Strategies are needed that address structural determinants and modify social norms to provide supportive environments promoting safer sexual practices. At a practical level, we find that with only two behavioral indicators (union status and multiple sexual partners), in this study’s specific context, individuals can be assigned to classes of sexual behavior risk that are relevant in considering intervention programs. Replication in other situations is needed to confirm these findings.

Supplementary Material

10508_2020_1663_MOESM1_ESM

Acknowledgments

We thank all the respondents who participated in this study and the Ha Nakekela field team. We are also grateful to members of the HIV Survey investigative team for their numerous contributions to the project. We also thank the people of the Agincourt sub-district for their long involvement with the MRC/Wits Rural Public Health and Health Transitions Research Unit.

Funding: We are grateful for funding support from: the U.S. National Institute on Aging – R01 AG049634 HIV after 40 in rural South Africa: Aging in the Context of an HIV Epidemic (PI Sanyu Mojola); the National Institutes of Health – R24 AG032112-05 Partnership for Social Science AIDS Research in South Africa’s Era of ART Rollout (PI Jane Menken); the University of Colorado, Innovative Seed Grant HIV after 40 in rural South Africa (PI Sanyu Mojola); the Eunice Kennedy Shriver National Institute of Child Health and Human Development – K01 HD057246 (PI Samuel Clark); and the William and Flora Hewlett Foundation 2009-4060 African Population Research and Training Program (PI Jane Menken). The MRC/Wits Rural Public Health and Health Transitions Research Unit and Agincourt Health and Socio-Demographic Surveillance System, a node of the South African Population Research Infrastructure Network (SAPRIN), is supported by the Department of Science and Innovation, the University of the Witwatersrand, and the Medical Research Council, South Africa, and previously the Wellcome Trust, UK (grants 058893/Z/99/A; 069683/Z/02/Z; 085477/Z/08/Z; 085477/B/08/Z). This work has also benefited from research, administrative, and computing support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development–funded University of Colorado Population Center (R24HD066613, P2C HD066613).

Footnotes

Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.

Conflict of interest: The authors declare that they have no conflict of interest.

Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

This study received ethical approvals from the University of the Witwatersrand Human Research Ethics Committee and the Mpumalanga Provincial Research and Ethics Committee.

Informed consent: Informed consent (assent for minors) was obtained from all individual participants included in the study.

REFERENCES

  1. Akers AY, Cohen ED, Marshal MP, Roebuck G, Yu L, & Hipwell AE (2016). Objective and Perceived Weight: Associations with Risky Adolescent Sexual Behavior. Perspectives on Sexual and Reproductive Health, 48(3), 129–137. doi: 10.1363/48e11416 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Akobeng Anthony K. (2006). Understanding diagnostic tests 1: sensitivity, specificity and predictive values. Acta Paediatrica, 96, 338–341. [DOI] [PubMed] [Google Scholar]
  3. Angotti N, Houle B, Schatz E, & Mojola SA (2018). Classifying and contextualizing sexual behavior practices across the life course: implications in later life. Paper presented at the Population Association of America, Denver, CO, USA. [Google Scholar]
  4. Bajos N, & Marquet J (2000). Research on HIV sexual risk: Social relations-based approach in a cross-cultural perspective. Social Science & Medicine, 50(11), 1533–1546. [DOI] [PubMed] [Google Scholar]
  5. Berhan A, & Berhan Y (2012). Is the Sexual Behaviour of HIV Patients on Antiretroviral therapy safe or risky in Sub-Saharan Africa? Meta-Analysis and Meta-Regression. AIDS Research and Therapy, 9(1), 14. doi: 10.1186/1742-6405-9-14 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Carrasco M, Esser M, Sparks A, & Kaufman M (2016). HIV-Alcohol Risk Reduction Interventions in Sub-Saharan Africa: A Systematic Review of the Literature and Recommendations for a Way Forward. AIDS Behav, 20(3), 484–503. doi: 10.1007/s10461-015-1233-5 [DOI] [PubMed] [Google Scholar]
  7. Carrasco MA, Nguyen TQ, & Kaufman MR (2018). Low Uptake of Voluntary Medical Male Circumcision Among High Risk Men in Malawi. AIDS Behav, 22(2), 447–453. doi: 10.1007/s10461-016-1633-1 [DOI] [PubMed] [Google Scholar]
  8. Cederbaum JA, Gilreath TD, & Barman-Adhikari A (2014). Perceived risk and condom use among adolescents in sub-Saharan Africa: a latent class analysis. Afr J Reprod Health, 18(4), 26–33. [PubMed] [Google Scholar]
  9. Clark SJ, Collinson MA, Kahn K, Drullinger K, & Tollman SM (2007). Returning home to die: circular labour migration and mortality in South Africa. Scand J Public Health, 69, 35–44. doi: 10.1080/14034950701355619 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cochran SD, de Leeuw J, & Mays VM (1995). Optimal scaling of HIV-related sexual risk behaviors in ethnically diverse homosexually active men. J Consult Clin Psychol, 63(2), 270–279. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Cockerham WC (2005). Health lifestyle theory and the convergence of agency and structure. Journal of Health and Social Behavior, 46(1), 51–67. [DOI] [PubMed] [Google Scholar]
  12. Danielson CK, Walsh K, McCauley J, Ruggiero KJ, Brown JL, Sales JM, … Diclemente RJ (2014). HIV-related sexual risk behavior among African American adolescent girls. J Womens Health (Larchmt), 23(5), 413–419. doi: 10.1089/jwh.2013.4599 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Dariotis JK, Sonenstein FL, Gates GJ, Capps R, Astone NM, Pleck JH, … Zeger S (2008). Changes in sexual risk behavior as young men transition to adulthood. Perspect Sex Reprod Health, 40(4), 218–225. doi: 10.1363/4021808 [DOI] [PubMed] [Google Scholar]
  14. Davies SL, Cheong J, Lewis TH, Simpson CA, Chandler SD, & Tucker JA (2014). Sexual risk typologies and their relationship with early parenthood and STI outcomes among urban African-American emerging adults: a cross-sectional latent profile analysis. Sex Transm Infect, 90(6), 475–477. doi: 10.1136/sextrans-2013-051334 [DOI] [PubMed] [Google Scholar]
  15. Doyle JS, Degenhardt L, Pedrana AE, McBryde ES, Guy RJ, Stoové MA, … Hellard ME (2014). Effects of HIV Antiretroviral Therapy on Sexual and Injecting Risk-Taking Behavior: A Systematic Review and Meta-analysis. Clinical Infectious Diseases, 59(10), 1483–1494. doi: 10.1093/cid/ciu602 [DOI] [PubMed] [Google Scholar]
  16. Dworkin S, Treves-Kagan S, & Lippman S (2013). Gender-transformative interventions to reduce HIV risks and violence with heterosexually-active men: a review of the global evidence. AIDS Behav, 17(9), 2845–2863. doi: 10.1007/s10461-013-0565-2 [DOI] [PubMed] [Google Scholar]
  17. Elwy AR, Hart GJ, Hawkes S, & Petticrew M (2002). Effectiveness of interventions to prevent sexually transmitted infections and human immunodeficiency virus in heterosexual men: a systematic review. Archives of Internal Medicine, 162(16), 1818–1830. [DOI] [PubMed] [Google Scholar]
  18. George LK (1993). Sociological perspectives on life transitions. Annual Review of Sociology, 19(1), 353–373. [Google Scholar]
  19. Gilbert L, & Walker L (2002). Treading the path of least resistance: HIV/AIDS and social inequalities—a South African case study. Social Science & Medicine, 54(7), 1093–1110. [DOI] [PubMed] [Google Scholar]
  20. Gómez-Olivé FX, Angotti N, Houle B, Klipstein-Grobusch K, Kabudula C, Menken J, … Clark SJ (2013). Prevalence of HIV among those 15 and older in rural South Africa. AIDS Care, 25(9), 1122–1128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gorbach PM, Drumright LN, & Holmes KK (2005). Discord, Discordance, and Concurrency: Comparing Individual and Partnership-Level Analyses of New Partnerships of Young Adults at Risk of Sexually Transmitted Infections. Sexually Transmitted Diseases, 32(1), 7–12. doi: 10.1097/01.olq.0000148302.81575.fc [DOI] [PubMed] [Google Scholar]
  22. Hallfors DD, Iritani BJ, Miller WC, & Bauer DJ (2007). Sexual and drug behavior patterns and HIV and STD racial disparities: the need for new directions. Am J Public Health, 97(1), 125–132. doi: 10.2105/ajph.2005.075747 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Halpern CT, Hallfors D, Bauer DJ, Iritani B, Waller MW, & Cho H (2004). Implications of racial and gender differences in patterns of adolescent risk behavior for HIV and other sexually transmitted diseases. Perspect Sex Reprod Health, 36(6), 239–247. doi: 10.1363/psrh.36.239.04 [DOI] [PubMed] [Google Scholar]
  24. Haydon AA, Herring AH, & Halpern CT (2012). Associations between patterns of emerging sexual behavior and young adult reproductive health. Perspect Sex Reprod Health, 44(4), 218–227. doi: 10.1363/4421812 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Houle B, Angotti N, Clark SJ, Williams J, Gómez-Olivé FX, Menken J, … Tollman SM (2016). Let’s talk about sex, maybe: interviewers, respondents, and sexual behavior reporting in rural South Africa. Field Methods, 28(2), 112–132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Houle B, Clark SJ, Gómez-Olivé FX, Kahn K, & Tollman SM (2014). The unfolding counter-transition in rural South Africa: mortality and cause of death, 1994–2009. PLoS One, 9(6), e100420. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Houle B, Mojola SA, Angotti N, Schatz E, Gomez-Olive FX, Clark SJ, … Menken J (2018). Sexual behavior and HIV risk across the life course in rural South Africa: trends and comparisons. AIDS Care. doi: 10.1080/09540121.2018.1468008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hunter M (2007). The changing political economy of sex in South Africa: the significance of unemployment and inequalities to the scale of the AIDS pandemic. Social Science & Medicine, 64(3), 689–700. doi: 10.1016/j.socscimed.2006.09.015 [DOI] [PubMed] [Google Scholar]
  29. Joint United Nations Programme on HIV/AIDS (UNAIDS). (2014). The Gap Report. Retrieved from http://files.unaids.org/en/media/unaids/contentassets/documents/unaidspublication/2014/UNAIDS_Gap_report_en.pdf [PubMed] [Google Scholar]
  30. Kabudula C, Houle B, Collinson M, Kahn K, Gómez-Olivé F, Clark S, & Tollman S (2017). Progression of the epidemiological transition in a rural South African setting: findings from population surveillance in Agincourt, 1993–2013. BMC Public Health, 17, 424. doi: 10.1186/s12889-017-4312-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kabudula C, Houle B, Collinson MA, Kahn K, Tollman S, & Clark S (2016). Assessing changes in household socioeconomic status in rural South Africa, 2001–2013: a distributional analysis using household asset indicators. Social Indicators Research, 1–27. doi: 10.1007/s11205-016-1397-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Kahn K, Collinson MA, Gómez-Olivé FX, Mokoena O, Twine R, Mee P, … Khosa A (2012). Profile: Agincourt health and socio-demographic surveillance system. International Journal of Epidemiology, 41(4), 988–1001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Kordoutis P, Loumakou M, & Sarafidou J (2000). Heterosexual relationship characteristics, condom use and safe sex practices. AIDS Care, 12(6), 767–782. doi: 10.1080/09540120020014318a [DOI] [PubMed] [Google Scholar]
  34. Logan T, Cole J, & Leukefeld C (2002). Women, sex, and HIV: Social and contextual factors, meta-analysis of published interventions, and implications for practice and research. Psychological Bulletin, 128(6), 851–885. doi: 10.1037//0033-2909.128.6.851 [DOI] [PubMed] [Google Scholar]
  35. Lurie MN, Williams BG, Zuma K, Mkaya-Mwamburi D, Garnett GP, Sweat MD, … Karim SS (2003). Who infects whom? HIV-1 concordance and discordance among migrant and non-migrant couples in South Africa. AIDS, 17(15), 2245–2252. [DOI] [PubMed] [Google Scholar]
  36. Mackesy-Amiti ME, Ouellet LJ, Finnegan L, Hagan H, Golub E, Latka M, … Garfein RS (2014). Transitions in latent classes of sexual risk behavior among young injection drug users following HIV prevention intervention. AIDS Behav, 18(3), 464–472. doi: 10.1007/s10461-013-0601-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Masters NT, Beadnell B, Morrison DM, Hoppe MJ, & Wells EA (2013). Multidimensional characterization of sexual minority adolescents’ sexual safety strategies. J Adolesc, 36(5), 953–961. doi: 10.1016/j.adolescence.2013.07.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Masters NT, Casey E, Beadnell B, Morrison DM, Hoppe MJ, & Wells EA (2015). Condoms and Contexts: Profiles of Sexual Risk and Safety Among Young Heterosexually Active Men. J Sex Res, 52(7), 781–794. doi: 10.1080/00224499.2014.953023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. McGrath N, Eaton JW, Bärnighausen TW, Tanser F, & Newell M-L (2013). Sexual behaviour in a rural high HIV prevalence South African community. AIDS, 27(15), 2461–2470. doi: 10.1097/01.aids.0000432473.69250.19 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. McMahon RC, Stanforth ET, Devieux JG, & Jean-Gilles M (2016). HIV risk behavior and internalizing/externalizing psychopathology among adolescents in court-ordered treatment. Am J Drug Alcohol Abuse, 42(2), 187–195. doi: 10.3109/00952990.2015.1132719 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Mee P, Collinson MA, Madhavan S, Root ED, Tollman SM, Byass P, & Kahn K (2014). Evidence for localised HIV related micro-epidemics associated with the decentralised provision of antiretroviral treatment in rural South Africa: a spatio-temporal analysis of changing mortality patterns (2007–2010). Journal of Global Health, 4(1), 010403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mojola SA, Williams J, Angotti N, & Gómez-Olivé FX (2015). HIV after 40 in rural South Africa: A life course approach to HIV vulnerability among middle aged and older adults. Social Science & Medicine, 143, 204–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Mustanski B, Byck GR, Dymnicki A, Sterrett E, Henry D, & Bolland J (2013). Trajectories of multiple adolescent health risk behaviors in a low-income African American population. Dev Psychopathol, 25(4 Pt 1), 1155–1169. doi: 10.1017/s0954579413000436 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. National Department of Health, Statistics South Africa, South African Medical Research Council, and ICF. (2019). South Africa Demographic and Health Survey 2016. Preotria, South Africa and Rockville, Maryland, USA: NDoH, Stats SA, SAMRC, and ICF. [Google Scholar]
  45. Nnko S, Boerma J, Urassa M, Mwaluko G, & Zaba B (2004). Secretive females or swaggering males? An assessment of the quality of sexual partnership reporting in rural Tanzania. Social Science & Medicine, 59(2), 299–310. doi: 10.1016/j.socscimed.2003.10.031 [DOI] [PubMed] [Google Scholar]
  46. Noor SW, Ross MW, Lai D, & Risser JM (2014). Use of latent class analysis approach to describe drug and sexual HIV risk patterns among injection drug users in Houston, Texas. AIDS Behav, 18 Suppl 3, 276–283. doi: 10.1007/s10461-014-0713-3 [DOI] [PubMed] [Google Scholar]
  47. Nylund KL, Asparouhov T, & Muthén BO (2007). Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569. doi: 10.1080/10705510701575396 [DOI] [Google Scholar]
  48. Pettifor A, The, H. p. t., MacPhail C, Selin A, Gómez-Olivé FX, Rosenberg M, … Kahn K (2016). HPTN 068: A Randomized Control Trial of a Conditional Cash Transfer to Reduce HIV Infection in Young Women in South Africa—Study Design and Baseline Results. AIDS and Behavior, 20(9), 1863–1882. doi: 10.1007/s10461-015-1270-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Pflieger JC, Cook EC, Niccolai LM, & Connell CM (2013). Racial/ethnic differences in patterns of sexual risk behavior and rates of sexually transmitted infections among female young adults. Am J Public Health, 103(5), 903–909. doi: 10.2105/ajph.2012.301005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Rosenberg MS, Gómez-Olivé FX, Rohr JK, Houle BC, Kabudula CW, Wagner RG, … Tollman SM (2017). Sexual Behaviors and HIV Status: A Population-Based Study Among Older Adults in Rural South Africa. Journal of Acquired Immune Deficiency Syndromes (1999), 74(1), e9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Satorra A, & Bentler PM (2010). Ensuring Positiveness of the Scaled Difference Chi-square Test Statistic. Psychometrika, 75(2), 243–248. doi: 10.1007/s11336-009-9135-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Tibbits MK, Caldwell LL, Smith EA, Vergnani T, & Wegner L (2016). Longitudinal Patterns of Active Leisure among South African Youth: Gender Differences and Associations with Health Risk Behaviours. World Leis J, 58(1), 60–68. doi: 10.1080/16078055.2015.1089317 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. UNAIDS. (2016). Country overview: South Africa. Retrieved from http://www.unaids.org/en/regionscountries/countries/southafrica
  54. Varga CA (2003). How gender roles influence sexual and reproductive health among South African adolescents. Studies in Family Planning, 34(3), 160–172. doi: 10.1111/j.1728-4465.2003.00160.x [DOI] [PubMed] [Google Scholar]
  55. Vasilenko SA, Kugler KC, Butera NM, & Lanza ST (2015). Patterns of adolescent sexual behavior predicting young adult sexually transmitted infections: a latent class analysis approach. Arch Sex Behav, 44(3), 705–715. doi: 10.1007/s10508-014-0258-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Venkatesh K, Flanigan T, & Mayer K (2011). Is expanded HIV treatment preventing new infections? Impact of antiretroviral therapy on sexual risk behaviors in the developing world. AIDS, 25(16), 1939–1949. doi: 10.1097/QAD.0b013e32834b4ced [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Vollmer S, Harttgen K, Alfven T, Padayachy J, Ghys P, & Bärnighausen T (2016). The HIV epidemic in sub-Saharan Africa is aging: Evidence from the Demographic and Health Surveys in sub-Saharan Africa. AIDS Behav. doi: 10.1007/s10461-016-1591-7 [DOI] [PubMed] [Google Scholar]
  58. Waller MW, Hallfors DD, Halpern CT, Iritani BJ, Ford CA, & Guo G (2006). Gender differences in associations between depressive symptoms and patterns of substance use and risky sexual behavior among a nationally representative sample of U.S. adolescents. Arch Womens Ment Health, 9(3), 139–150. doi: 10.1007/s00737-006-0121-4 [DOI] [PubMed] [Google Scholar]
  59. Wechsberg WM, Myers B, Kline TL, Carney T, Browne FA, & Novak SP (2012). The Relationship of Alcohol and Other Drug Use Typologies to Sex Risk Behaviors among Vulnerable Women in Cape Town, South Africa. J AIDS Clin Res, S1(15). doi: 10.4172/2155-6113.S1-015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Wechsberg WM, Peasant C, Kline T, Zule WA, Ndirangu J, Browne FA, … van der Horst C (2017). HIV Prevention Among Women Who Use Substances And Report Sex Work: Risk Groups Identified Among South African Women. AIDS Behav, 21(Suppl 2), 155–166. doi: 10.1007/s10461-017-1889-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Wellings K, Collumbien M, Slaymaker E, Singh S, Hodges Z, Patel D, & Bajos N (2006). Sexual behaviour in context: a global perspective. Lancet, 368, 1706–1728. [DOI] [PubMed] [Google Scholar]
  62. Wu J, Witkiewitz K, McMahon RJ, & Dodge KA (2010). A parallel process growth mixture model of conduct problems and substance use with risky sexual behavior. Drug Alcohol Depend, 111(3), 207–214. doi: 10.1016/j.drugalcdep.2010.04.013 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

10508_2020_1663_MOESM1_ESM
NIHMS1580689-supplement-10508_2020_1663_MOESM1_ESM.docx (140.2KB, docx)

RESOURCES