What’s in a Name? A Mixed Method Study on How Young Women Who Sell Sex Characterize Male Partners and Their Use of Condoms

Introduction: Young women who sell sex (YWSS) are at disproportionate risk of HIV. Reducing YWSS' vulnerability requires engaging their male sexual partners. To achieve this, we need to understand the characteristics and dynamics of their sexual partnerships to inform effective interventions. Methods: We conducted a mixed-methods study to compare YWSS' qualitative descriptions of male partners with categories reported in a behavioral survey. Data were drawn from enrollment into an evaluation of the DREAMS initiative in Zimbabwe in 2017. As part of a respondent-driven sampling survey, we recruited 40 seed participants from 2 intervention and 4 comparison sites. We conducted semistructured interviews with 19 “seeds,” followed by a behavioral survey with 2387 YWSS. We interpreted quantitative and qualitative data together to understand how YWSS perceived male sexual partners, assess how well survey variables related to narrative descriptions, and describe patterns of risk behavior within partnerships. Results: Qualitative data suggest survey categories “husband” and “client” reflect YWSS' perceptions but “regular partner/boyfriend” and “casual partner” do not. In interviews, use of the term “boyfriend” was common, describing diverse relationships with mixed emotional and financial benefits. More than 85% of male partners provided money to YWSS, but women were less likely to report condomless sex with clients than regular partners (11% vs 37%) and more likely to report condomless sex with partners who ever forced them to have sex (37% vs 21%). Conclusions: Reducing HIV risk among YWSS requires prevention messages and tools that recognize diverse and changing vulnerability within and between sexual relationships with different male partners.


INTRODUCTION
HIV incidence in Southern Africa remains concentrated among adolescent girls and young women aged 15-24 years. 1 Young women who sell sex (YWSS) have a particularly high risk of acquiring HIV 2,3 due to the high number of partners, difficulties negotiating condom use, poor access to services, [4][5][6][7] and power imbalances within relationships. [8][9][10] Exposure to sexual and physical violence is a further driver of HIV among this group. 11 Increasingly, HIV prevention interventions for YWSS target "upstream" determinants of vulnerability, offering education subsidies, or cash transfers designed to lessen dependence on sexual relationships. 12,13 The DREAMS (Determined, Resilient, Empowered, AIDS-free, Mentored, and Safe) Partnership provided a combined package of skillsbuilding and entrepreneurial opportunities, social protection and sexual and reproductive health services in 10 sub-Saharan African countries. [14][15][16] DREAMS also acknowledged that reducing HIV risk among YWSS requires engaging their male sexual partners, and thus collected data on male sexual partners of high-risk adolescent girls and young women to better target them with HIV services.
Existing research on the male partners of adolescent girls and women focuses on their age, educational attainment, number of partners, and partner concurrency. 17 Evidence on HIV risk for women in age-disparate relationships is mixed, 18 but power differentials common to sexual partnerships with male partners 10-15 years older can exacerbate girls' and young women's susceptibility to HIV. 19,20 Qualitative studies exploring transactional sex find that young women identify multiple and distinct partner categories, from whom they receive a range of economic, material, social, and emotional support. 5,[21][22][23] Understanding how YWSS, including those self-identifying as sex workers, perceive and experience relationships with men has been less closely examined or used to inform programming.
We used mixed methods to characterize the male sexual partners of YWSS recruited to an evaluation of DREAMS in Zimbabwe. 24 Drawing on qualitative data, we examined how YWSS describe, understand, and navigate different kinds of sexual relationships. We used these qualitative insights to interpret quantitative data across predefined partner typologies, exploring associations between how YWSS characterize their partners, their behaviors with these partners, and likelihood of engaging in condomless sex. The aim of this analysis was to better understand YWSS' sexual relationship dynamics vis-à-vis risk to help inform targeted HIV prevention interventions.

Study Location and Population
In Zimbabwe, DREAMS worked in partnership with the Centre for Sexual Health and HIV/AIDS Research (CeSHHAR) to reach YWSS within the national Sisters with a Voice program for female sex workers. YWSS were offered tailored HIV prevention and treatment services and referred into the DREAMS network of organizations providing the DREAMS "core package" of social, educational, and economic interventions. 24 Data were collected between April and July 2017 in 6 sites across Zimbabwe, 2 large cities where DREAMS was being implemented (anonymized as sites A and B), and 4 smaller towns without planned DREAMS activities (sites C, D, E, and F). 24 As described elsewhere, sociogeographical mapping was conducted to identify where and how young women sell sex and to recruit 44 "seeds," representative of the typology of YWSS, to initiate respondent-driven sampling. 25 Mapping identified different typologies of YWSS, including street-based YWSS, university students who transact sex during school terms, and rural migrants who sell sex to men with disposable income. 25 The 44 seed participants were given 2 coupons each to recruit women aged 18 to 24 years whom they knew and who sold sex to men, defined as "sex in exchange for money and/or material goods, and in the absence of the exchange, the sex would not happen." Each new recruit was assessed for eligibility, and after completion of survey procedures, given 2 coupons to recruit a further 2 YWSS. This process continued over 6 waves, with wave 1 women recruiting the second wave of women, who in turn recruited a third wave, until the target sample size of 2400 YWSS was reached by the sixth wave. 24

Qualitative Interviews
Qualitative data were collected from 19 seed participants. We intended to interview 20 women: 6 in each of the 2 DREAMS intervention cities (A and B) and 4 in 2 smaller comparison towns (C and F), selected for diversity in type and location of sexual exchange identified during mapping. 25 We completed all planned interviews except 1 in intervention site A. Semistructured interviews explored experiences of initiating selling sex, current involvement in sexual exchange, relationships with different male sexual partners, healthrelated risk perceptions, and engagement with services. The topic guides were developed for the initial mapping exercise to identify different YWSS typologies and guide recruitment into the respondent-driven sampling (RDS) survey and subsequent cohort, and thus specifically examined YWSS' perceptions of their sexual relationships, focusing on those for financial or material gain. A female researcher conducted interviews in a local language (Shona or Ndebele), which lasted roughly 45-60 minutes and were transcribed and translated into English by research assistants for entry into NVivo software.
Thematic content analysis was conducted using a 2stage process: first, each transcript was read and "case notes" written to summarize the respondent's relationship history and number/description of all current sexual partners. Based on the frequency of terms used to describe partners, we created 3 primary relationship nodes: "husband/permanent partner," "boyfriend," and "client" that we used to conduct "broad brush" coding of all interviews. Given the considerable overlap between these categories, particularly as women referred to the same individual using different terms, we next examined each of these 3 original nodes in detail, to identify patterns in characteristics, relationship dynamics, and behaviors for each partner type.

Behavioral Survey
Women enrolled into the DREAMS evaluation completed a questionnaire covering demographics, HIV service use, sexual behaviors, and history of selling sex, and whether they self-identified as a sex worker. YWSS were asked about their 3 most recent sexual partners, as follows: "How would you describe your relationship with [INITIALS] the last time you had sex?" Women could select: "husband," "regular/ steady partner/boyfriend," "casual partner known to you before having sex," "one-off partner not known to you before having sex," "sex work client," or could specify their own description. If women reported that last sex with the partner involved an exchange, they were asked whether they received money, school supplies, support with bills, groceries, or other items.
Using data on 3 most recent partners, we described total numbers, characteristics, and behaviors by 3 partner types: husband/regular partner, casual/one-off partner, or sex work client, as well as the number and percentage of partners with whom women reported any episode of condomless sex in the previous month. In regression analyses, the outcome of interest was condomless sex in the past month with a partner, and the unit of analysis was the partnership. Factors explored for their association with condomless sex were based on findings emerging from the qualitative analyses. As condomless sex in the previous month was ;10-40% across partner types in descriptive analyses, the log (probability of reporting condomless sex) was the outcome variable in our regression analysis; age-and site-adjusted and adjusted risk ratios were estimated using a generalized linear regression model, assuming that the outcome followed a normal distribution, with robust standard errors to allow for departures from this assumption. 26 Analyses were adjusted for women's age, level of education, marital status, self-identification as female sex worker (FSW), and site of recruitment. Data were weighted using the RDS-II estimator, 27 namely by the inverse degree of number of YWSS each woman reported knowing and normalized these by site. All seeds were excluded from analysis. Analysis was conducted using Stata 14.0. RDS diagnostics, described elsewhere, suggested our sites were broadly representative of age, HIV prevalence, and identification as FSW in 5 sites. 28 Findings from quantitative and qualitative data analysis were interpreted together to understand how YWSS perceived and categorized male sexual partners, identify whether and how well our prespecified measures related to narrative descriptions, and describe patterns of vulnerability and risk behavior within each type of partnership.

Partner Typology
During qualitative interviews, women referred to 3 partner categories, of which 2 corresponded to predefined variables used in our survey. Approximately half the interview respondents (9/19) referred to having a spouse or "permanent" partner at some time, defined by a history of setting up a shared home, having a child/children together, and/or traditional or legal marriage. At the other end of the spectrum, "clients" paid cash in direct exchange for sex, at the time of sex, and the relationship did not involve personal attachment.
The largest category, however, was "boyfriend," covering numerous, diverse relationships that did not match the survey's use of "regular" or "casual" partner. For some YWSS, "boyfriend" implied emotional attachment and/or hopes for marriage. Others described how clients could become "boyfriends" through increasing frequency or amount of financial contributions. YWSS who did not self-identify as sex workers referred to clients as "boyfriends," perhaps reluctant to adopt the language of sex work. Having 2-5 "boyfriends" was a common means of maximizing financial security. Often, one boyfriend was considered the most important emotionally and might provide regular support such as food and rent, instead of cash. YWSS were more likely to establish informal arrangements with boyfriends, who were expected to pay regular household expenses instead of paying money at the time of sex.
A further distinction was based on time, ie, husbands were referred to solely in the past, with initiation of selling sex after the end of the marital relationship. In the present, YWSS called partners "boyfriends." "Permanent partner" referred to previous spouses or current relationships that they defined as "serious." Age differences did not feature prominently in interviews. Table 1 provides illustrative excerpts from interviews for the 3 partner categories.

Exposure to Violence
YWSS experienced sexual and physical violence across relationships, feeling most vulnerable when it occurred within a romantic relationship by a husband/permanent partner or "boyfriend" for whom they felt personal attachment. Three young women described how their spousal relationships started with sexual assault or rape.
A particularly violent case was a YWSS who was just 12 or 13 at the time of the rape. After the episode below, she stayed with her assailant until her second pregnancy with him at age 15. He then abandoned her, leading her to sell sex to support herself and her baby:  Others did not hide their involvement in sex work from boyfriends but tried to avoid confronting them directly with its reality to avoid violence.

Condomless Sex
YWSS reported that negotiating condoms with clients was possible, but depended on immediate economic needs.
But was there a time when you had to sleep with a client when they refused to use condoms?
Yahhh. . It was not often. It was when I saw that I was desperate and needed money I had to go to school. So if that client had money and did not want to use condoms, I had to risk because l knew l had to go to school. (Age 24, completed school, no children, DREAMS site A) A few YWSS reported that they themselves disliked condoms. One described preferring the female condom, and another explained low condom use as her preference for condomless sex. Although women reported unplanned pregnancies and STIs, usually within established relationships, these were not considered as serious as HIV. As illustrated in the quote above, there was little motivation to avoid these outcomes through condom use with partners described as "boyfriends."

Analyses of the Behavioral Survey
Through recruitment chains, 2387 women were recruited to the study; 20.9% (n = 448) were aged 18 years, 44.4% (n = 1060) had completed some secondary education, and 67.3% (n = 1637) self-identified as FSW. Most women were confident in discussing HIV testing and condom use with regular and/or new sexual partners ( Table 2).

DISCUSSION
In this mixed-methods analysis, we found consistencies and divergence in how YWSS characterized male partners in semistructured interviews and a behavioral survey. Our qualitative data suggest that although survey categories of "husband" and "client" reflected YWSS' definitions fairly well, the prespecified labels "regular partner/boyfriend" and "casual partner" did not. These categories seemed subsumed within wider use of the term "boyfriend," referring to relationships along a continuum rather than a specific "type." "Boyfriend" could signify close emotional attachment, a former client transitioning from direct exchange to longer-term financial support, or a short-term client when used by YWSS who did not consider themselves sex workers. Survey respondents might allocate "boyfriends" across categories in unpredictable ways, making understanding risk across relationships challenging to determine or usefully apply to an intervention design.   Nonetheless, survey and interview findings reinforce that condomless sex is more common with longer term, more "regular" partners, as found elsewhere. [29][30][31] YWSS reported the highest condomless sex at last sex and in the past month with "regular" partners. "Regular" partners were younger than casual partners and clients, being more similar in age to the young women themselves. Almost all these relationships involved material exchange, including money and assistance with rent, groceries, and other household expenses likely to be longer-term support. YWSS might value these contributions more highly than cash if they are more reliable or signify greater personal involvement in daily life, thus catalyzing a partner's transition from "client" to "boyfriend." Increasing financial reliance on a boyfriend might be one reason YWSS report lowest perceived condom negotiation confidence with "regular" partners. In South Africa, a nuanced account of men's HIV risk profiles revealed 2 groups of moderate-to-high risk younger men who engaged in transactional sex but had limited access to available HIV services. 32 These male partners may be subsumed within the "regular" partners described by the women in our study. Critical to the HIV response is developing strategies to reach these "regular" partners with HIV prevention and care services.
Women were more likely to report experiences of IPV for "regular" partners, which was nearly double that reported for "casual" or "client" partnerships. As highlighted in our qualitative data, selling sex could be a consequence and a determinant of IPV. Experience of violence led to some YWSS initiating sex work as an alternative to dependence on the violent partner, yet selling sex could also exacerbate violence because of partners' jealousy or feeling disrespected. Exposure to IPV is known to be a risk factor for HIV, independently and because of its association with alcohol use, [33][34][35][36] and DREAMS' core package targeted IPV as a structural driver of HIV. 16 To minimize risk of IPV among YWSS, prevention programs need to understand the drivers of IPV, such as poverty and interpersonal communication, with a focus on partners classified by women as "regular," to deliver effective intervention strategies. 37 We found women's own dislike of condoms negatively affected use with nonclient partners, suggesting factors other than unequal power dynamics determine HIV prevention practices. YWSS reported their own reluctance as the second most common reason for not using a condom in the past month with a "regular" partner, suggesting in interviews that they associated condoms and HIV risk with clients more than with boyfriends, reducing motivation to use condoms with the latter. This highlights the need for greater attention to YWSS' risk of unwanted pregnancy and other STIs, both of which were discussed in interviews but did not seem to motivate YWSS to use condoms. The sole focus on HIV and neglect of other sexual and reproductive health outcomes for female sex workers has attracted previous criticism. 38,39 It also suggests that PrEP could be better promoted as a means to increase pleasure in sex with regular partners in addition to offering protection where condoms are difficult to negotiate. 40,41 Overall, this study suggests that the partner labels commonly used on behavioral surveys remain a useful indicator of partnerships that likely place women at higher HIV risk, but remain blunt tools. Although YWSS had mostly clear delineations for "husband/spouse" and "client" on either side of the emotional continuum, they applied the term "boyfriend" to a very diverse range of relationships that are unlikely to be captured through "regular" or "casual" categories. This suggests the need for an extensive and indepth qualitative inquiry to understand local perceptions and behaviors, and how these map on to the risk of HIV and other outcomes before selection of targeted behavioral change messages.
Our analysis is subject to limitations. Our quantitative analysis excluded seed participants, yet our qualitative analysis focused solely on seed participants. Seed participants, however, represented the typology of YWSS in study sites so we thus consider our qualitative findings likely to reflect YWSS relationship dynamics in this context. Selfreported data on condomless sex and violence are subject to bias and likely to be underreported. Our finding that confidence in condom negotiation was associated with fewer occurrences of condomless sex may be due to reverse causality. The women in our study reported a high number of partners in the past month, yet our quantitative analysis is limited to characteristics of and behaviors with their 3 most recent partners, making our findings potentially less generalizable to all partners.

CONCLUSION
Among adolescent girls and young women, YWSS are at a disproportionately high risk of HIV. Our mixed-methods analysis found that partners defined as "regular" are diverse but often characterized by stronger emotional ties and an increased risk of violence and condomless sex than other partner types. For YWSS in Zimbabwe, the most salient category of male partner was "boyfriend," which subsumed a wide range of experiences, including sex work clients for those YWSS who did not self-identify as FSW. This complexity adds to the challenges of appropriately targeting messaging and programs to YWSS. To reduce HIV risk among YWSS, prevention programs need to move beyond relying on the limitations of partner labels and focus on improving women's access to multiple HIV prevention options, including integrating IPV services within broader sexual and reproductive health services. Programs need to recognize that young women's needs change over time, within relationships and between partners, and provide services that are flexible to these changing needs.
J.B. and B.H. jointly conceived and drafted the paper and led qualitative and quantitative data analysis, respectively. I.B., J.R.H., and S.F. led study design for the DREAMS evaluation. S.F. and S.T.C. contributed to quantitative data analysis and commented on previous drafts. T.C. and P.M. led qualitative data collection. F.M.C. was the PI for the DREAMS evaluation in Zimbabwe and helped revise the paper. All authors have read and approved the manuscript.