Written informed consent was obtained from all participants. For participants aged under 18, written informed consent was obtained from a parent or guardian and assent was obtained from the child. The study was approved by the Medical Research Council of Zimbabwe on 27 ^th April 2018 (MRCZ/A/2243) and by the Imperial College Research Ethics Committee on 20 ^th June 2018 (17IC4160).

The data used in our study to measure social norms and structural factors and their influence on AGYW’s use of condoms for HIV prevention were collected in eight communities close to the Mozambique border in Manicaland, Zimbabwe’s eastern province, in a phased (one community at a time) general population survey conducted between mid-July 2018 and early December 2019. The survey started in the run-up to the first national elections to be held in the ‘New Dispensation’ period that followed the ousting of President Mugabe in November 2017 ²⁸ . The eight communities comprised two ‘high-density’ urban areas in Mutare, Manicaland’s provincial capital, two small towns in Nyanga and Makoni districts, two large tea and forestry plantations, and two areas of predominantly subsistence farming villages in Mutasa District. These communities were selected to reflect the main socio-economic population strata found in Zimbabwe. The survey provided baseline data for two community-randomised control trials of behavioural economics interventions to increase use of HIV prevention methods in AGYW (ClinicalTrials.gov registration number NCT03565575; registration date: June 20, 2018) and in young men aged 15–29 (ClinicalTrials.gov registration number NCT03565588; registration date: June 21, 2018).

The detailed survey procedures have been published previously ²⁹ . In brief, a census of all households in each community was carried out in one community at a time between July 2018 and November 2019. In each community, all women aged 15–24 and men aged 15–29 and random samples of two-thirds of women aged over 25 and men aged over 30 who had stayed in these households on the night before the census visit were invited to participate in detailed individual interviews. In the individual survey, participants were interviewed on a number of topics including their socio-demographic characteristics; views on AGYW’s sexual behaviour and use of HIV prevention methods; and HIV prevention cascade variables ³⁰ , and were tested for HIV infection. The survey questionnaire is available here.

To obtain measurements of actual risk of HIV acquisition in AGYW in Manicaland and to establish whether this risk differed by marital status and associated sexual risk behaviours, we used data from an earlier general population cohort survey (Gregson et al., 2017) that ran between 1998 and 2013 in eight communities in Manicaland – six of which were also included in the 2018/19 survey. Unmarried AGYW were considered to be potentially at higher risk of acquiring HIV infection if they reported having started sex and being sexually-active in the last 12 months. Married AGYW were considered to be at potentially higher risk if they reported multiple sexual partners or sex with a spouse who was already infected or who they believed to have multiple sexual partners. HIV incidence rates in AGYW were calculated by randomly allocating seroconversion dates within the inter-survey periods that individuals became infected using multiple imputation. When calculating incidence trends over time, incidence rates were directly age-standardised to allow for changes in the age-structure of the population in the study areas over the period of the cohort, using the age-structure of Manicaland province from the 2012 Zimbabwean Population Census ³¹ .

To quantify the pathways through which gender norms and other selected structural factors impact on AGYW’s use of condoms in our study population, we used data from the 2018/19 survey and the HIV prevention cascade framework proposed by Schaefer and colleagues ³⁰ . This framework builds from a core cascade comprising four main bars: 1) a ‘priority’ population at risk of HIV infection – and therefore in need of HIV prevention methods; 2) motivation to use an HIV prevention method amongst individuals in this priority population; 3) access to the method; and 4) effective use of the method. Each main bar in the cascade is measured conditional on inclusion in the previous bar. All survey respondents who reported effective use of the method were assumed to have been motivated and to have had access to it when they started to use it. It is considered that a gap between the last two bars in the core cascade would reflect limitations in capacity to use the method amongst individuals in the priority population who are motivated to use the method and are able to access it. The extended version of this HIV prevention cascade adds sub-bars for the gaps between each pair of consecutive main bars which represent the reasons or ‘explanatory factors’ for these gaps drawn from the theoretical and empirical literature ³² . Lack of social acceptability to use the prevention method (e.g. due to conflicting gender norms) is included explicitly in the extended cascade as an explanatory factor that can contribute to gaps in motivation to use an HIV prevention method. In addition, negative gender norms could limit access to a method (e.g. through lack of acceptable provision if service providers disapprove of AGYW having access to HIV prevention methods ³³ ) and also to capacity to use the method (e.g. by reinforcing possible partner objections to using the method). The core and extended versions of this HIV prevention cascade were measured using data collected specifically for this purpose in the population survey conducted in Manicaland in 2018/19. For further details of how the HIV prevention cascade variables were measured in this study, see Extended data (Tables S1–S4). Cascades were measured and compared for unmarried and married HIV-negative AGYW at higher risk of acquiring infection.

To help understand the possible influence of negative gender norms around AGYW’s sexual activity and use of HIV prevention methods, levels of support for AGYW’s pre-marital sex and access to condoms within the Manicaland population were measured in the 2018/19 survey (Q627) with results disaggregated by sex of survey respondent.

All survey participants were asked to say whether they agreed or disagreed with a series of statements ( Table 1) including the following:

and

In the current analysis, survey participants were considered to support condom use in marriage if they agreed that this was always acceptable.

The initial results showed widespread acknowledgement that many AGYW were starting sex before marriage these days and a substantial minority of participants said they would accept it if a teenage daughter had sex before marriage. Therefore we investigated patterns of divergence from these gender norms within the community by constructing multivariable logistic regression models disaggregated by sex to calculate fully-adjusted odds ratios for agreement with the above statements by a range of socio-demographic characteristics, including the participant’s own relationship to an AGYW, community roles and HIV infection status.

Finally, to explore the pathways through which negative gender norms and selected other structural factors influence condom use by AGYW in our study population, we used multivariable logistic regression models, disaggregated by sex and adjusted for single year of age and community location, to test for statistical associations between self-reports of perceived and/or experienced negative norms and other structural factors and each of the main bars in the core HIV prevention cascade for condoms. The other structural factors included in this analysis were HIV stigma and intimate partner violence – two factors that were measured in the 2018/19 survey and have been found or suggested in the literature to limit young women’s use of HIV prevention methods ^{34, 35} .

Sex differences were considered in the design of the study with sex of participants being defined based on self-report. All data analyses were done in STATA 17 (Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC). R is an open-access alternative with equivalent function to STATA 17.

HIV incidence in AGYW in our general population cohort in east Zimbabwe declined from 2.31% (95% CI, 1.69%–2.93%) between 1998 and 2003 to 0.98% (95% CI, 0.46%–1.49%) between 2009 and 2013, with a temporary increase to 2.02% (95% CI, 1.47%–2.57%) between 2003 and 2008 ( Figure 1A). Over the period of the cohort, HIV incidence was higher in sexually-active unmarried AGYW (3.77%; 95% CI, 2.63%–4.91%) and in married AGYW reporting sexual risk behaviours (2.51%; 95% CI, 1.32%–3.71%) than in other AGYW (1.11%; 95% CI, 0.85%–1.36%) ( Figure 1B).

Following these procedures, 96.4% (9803/10170) of the eligible women and men selected for the current survey (2018/19) agreed to participate; for which, 98.7% (9678/9803) had a known HIV status either through provider-initiated testing and counselling, laboratory testing on a dried blood spot collected in the survey or, for 339 participants, by self-report ( Figure 2). Forty-two percent of survey participants were male and 58% were female. The overall weighted prevalence of HIV was 11.3% (95% CI; 10.6-12.0). HIV prevalence was higher in females (12.3%; 95% CI, 11.4-13.2) than in males (9.81%; 95% CI, 8.85-10.9).

Sixty-one point five percent (95% CI, 53.5%–69.0%; N=161) of unmarried AGYW at risk and 4.3% (95% CI, 2.5%–6.8%; N=394) of married AGYW at risk used condoms at last sex. Twenty-seven point three percent (95% CI, 18.8%–37.1%; N=99) and 35.3% (95% CI, 14.2%–61.7%; N=17) of these unmarried and married AGYW who reported using condoms at last sex, respectively, also reported using condoms as a family planning method. The HIV prevention cascades for condoms show much larger gaps for married AGYW than for unmarried AGYW in motivation and in capacity to use condoms effectively (in women who were motivated and had easy access to condoms) ( Figure 3).

Eighty-six point three percent (139/161; 95% CI, 80.1%–91.2%) of unmarried AGYW at risk and 43.1% (170/394; 95% CI, 38.2%–48.2%) of married AGYW at risk were motivated to use condoms with their sexual partners. Amongst the 22 unmarried AGYW at risk who were not motivated to use condoms, lack of risk perception (95.5%), lack of social acceptability (100.0%), and poor knowledge about condoms (77.3%) were barriers to motivation ( Figure 3A). Amongst the 224 unmotivated married AGYW at risk, lack of risk perception (99.1%), lack of social acceptability (96.9%), and poor knowledge about condoms (69.2%) once again were the most common barriers to motivation ( Figure 3B).

Seventy-three point nine percent (119/161; 95% CI, 66.4%–80.5%) of unmarried AGYW at risk and 30.7% (121/394; 95% CI, 26.2%–35.5%) of married AGYW at risk were motivated and felt able to access to condoms. For both unmarried AGYW and married AGYW at risk of HIV who were motivated to use condoms, reservations about the acceptability of service provision – privacy/confidentiality concerns and being embarrassed to ask – were the most common barriers to access (unmarried AGYW: 85.0%; 17/20; married AGYW: 81.6%; 40/49). Very few of these women reported practical difficulties – long distances or limited opening hours – or high costs as barriers to accessing condoms.

Amongst the small number of unmarried AGYW at risk of HIV who were motivated to use condoms and had access to them but did not use them at last sex (N=20), limitations in partner support, self-efficacy and/or social skills all contributed to lack of capacity to use condoms ( Figure 3a). In two-fifths of these cases, lack of practical skills – not having received instruction on condom use and/or not knowing what to do when condoms break – contributed to non-use. Amongst the married AGYW at risk of HIV who were motivated to use condoms and had access to them but did not use them at last sex (N=104), lack of partner support (80.8%) and not having the social skills to discuss condom use or to refuse sex with sexual partners (70.2%) were the main constraints on capacity to use condoms ( Figure 3b). For just over a half (51.9%) of these women, lack of practical skills contributed to non-use of condoms.

Social norms for unmarried AGYW. Almost everyone in our study population acknowledged that many young women now have sex before marriage ( Table 1). Nearly half of men (42.8%; 1744/4074) and a third of women (30.0%; 1719/5729) said that, if they had a teenage daughter who had sex before marriage, they would find this acceptable. Community members were split almost half-half on telling a teenage daughter – if they had or were to have one – about condoms and also pre-exposure prophylaxis (PrEP). Men were more likely than women to support this and also to support making condoms available in schools.

Opinions on teenage daughters’ having pre-marital sex varied substantially between and within the study communities ( Figure 4, Table S5). Surprisingly perhaps, for both sexes, tolerant views towards pre-marital sex were less common in ‘high-density’ city communities (31.1%; 498/1603) and town communities (29.9%; 780/2604) than in rural subsistence farming communities (37.0%; 1184/3198). However, men in agricultural communities were the most likely group to accept teenage daughters’ having pre-marital sex (52.2%; 651/1247); possibly because married men who work on these estates often live apart from their wives. Patterns of statistical association between individuals’ social roles and socio-demographic characteristics and acceptance of teenage daughters’ having pre-marital sex often differed between male and female community members. For example, approval levels increased with age for men ( Figure 4A) but decreased with age for women ( Figure 4B); never-married men were more likely to approve than currently- and formerly-married men whilst the reverse was true for women; and participation in local community groups was associated with reduced approval for men but increased approval for women. Male church leaders tended to be less likely to accept teenage daughters’ having pre-marital sex but sample sizes were too small to tell whether views varied amongst men and women performing other community roles. Men who had a teenage daughter themselves (36.9%; 110/298) or who were closely related to a teenage girl (35.0%; 411/1175) at the time of the survey were less likely than other men (47.4%; 1022/2155) to accept their daughters having sex before marriage; whilst mothers (31.7%; 282/889) and aunts (36.7%; 137/373) – but not grandmothers (23.2%; 80/345) – tended to be more tolerant than other women (28.7%; 794/2769).

Community members views on telling teenage daughters about condoms also varied between and within the study communities ( Figure 5, Table S6). Comparing the different types of study community, men in agricultural estates (60.5%; 754/1247) tended to be more likely to agree that this was a good idea than those in rural communities (56.3%; 722/1282) (p=.10) ¹ ( Figure 5A); possibly reflecting their more tolerant views on pre-marital sex. However, there was no evidence for this link amongst women ( Figure 5B): fewer women living in town communities (39.1%; 646/1654) supported telling teenage daughters about condoms than was the case in rural communities (40.1%; 769/1916) (p=.018) but more women in city communities (54.5%; 549/1008) than in rural communities supported the idea (p<.001) despite their lower tolerance of pre-marital sex. Teenagers and older people (ages 55 years and above) of both sexes were the least likely community members to support teenage daughters being told about condoms. For both men and women, those living with HIV were more likely to approve than their uninfected counterparts. Men’s agreement with telling teenage daughters about condoms varied with marital status (lowest agreement in currently married men), socio-economic status (progressively lower agreement with increasing poverty), religion (lowest agreement in Apostolic churches), and community engagement (lower agreement among men in community groups). Male teachers (78.0%; 39/50) had twice the odds of agreement with telling teenage daughters about condoms as men with no particular community leadership role (58.7%; 2254/3840) but had similar views about making condoms available in schools (Figure S1, Table S7). Women’s agreement with telling teenage daughters about condoms varied with marital status (highest agreement in divorced women), education (rising steadily with increasing education), religion (lowest agreement in Apostolic churches), and community engagement (higher agreement among women in community groups). Amongst community leaders, female village health workers had four times greater odds of agreement with telling teenage daughters about condoms compared to regular community members (72.0%; 18/25 versus 41.9%; 2204/5257); and church leaders were also more likely to be supportive (48.8%; 163/334). Men who currently had a teenage daughter themselves had similar views to other men (62.8%; 187/298 versus 61.1%; 1316/2155; p=0.24) on discussing condoms with daughters but brothers (55.3%; 492/890), uncles (48.0%; 98/204) and grandfathers (37.0%; 30/81) were all less in favour ( Figure 5a). Mothers (45.3%; 403/889), sisters (43.1%; 583/1353) and aunts (47.5%; 177/373) all had similar views on this question to other women in the community (43.0%; 1190/2769) ( Figure 5b). Grandmothers of teenage girls (25.2%; 87/345) were less likely to support telling teenage girls about condoms than other women in the community (43.0%; 1190/2769) but this difference was not statistically significant in the multivariable model (p=.21).

Social norms for married AGYW. Twenty-one point three percent of men and 32.5% of women expressed the view that condoms were acceptable with a marital partner under all circumstances ( Table 2). Before accounting for other characteristics, married and unmarried men and women were equally likely to agree with this. For both sexes, condom use was most acceptable when one spouse was HIV-positive (men: 92.9%; women: 95.8%), followed by when a spouse had another sexually transmitted infection (90.2%; 94.9%), when a spouse had other sexual partners (84.4%; 91.2%), and when condoms were being used to prevent unintended pregnancies (83.0%; 79.1%).

Condom acceptability under all circumstances within marriage was lowest for men in city communities (6.7%; 40/595), intermediate in small town (18.6%; 177/950) and rural subsistence farming (21.7%; 278/1282) communities, and highest in agricultural estate communities (28.8%; 359/1247) ( Figure 6A); differences that were statistically significant after adjusting for other characteristics (Table S8). For women, there was a consistent urban-rural gradient in condom acceptability within marriage: lowest in city communities (20.7%; 209/1008), higher in small town communities (25.8%; 426/1654), higher again in agricultural estate communities (34.3%; 395/1151), and highest in rural subsistence farming communities (40.4%; 775/1916) ( Figure 6B, Table S8). Men and women aged 30–54 were more likely than younger people to support condom use in marriage. As with telling teenage daughters about condoms, for both men and women, those living with HIV were more likely to approve of condom use in marriage than uninfected people. After adjusting for other characteristics, men’s agreement with condom use in marriage varied with marital status (lowest agreement in currently married men) but not by education, socio-economic status, religion, community engagement or being in a community leadership role. Women’s agreement with condom use in marriage varied with marital status (highest agreement in never-married women and lowest agreement in divorced, separated and widowed women) and education (greater agreement at higher levels of education) but not by socio-economic status, religion, community engagement or having a community leadership role. Similar proportions of men with teenage daughters themselves and men not related to a teenage girl approved of condom use in marriage (26.5%; 79/298 versus 23.3%; 503/2155) but brothers (16.0%; 142/890), uncles (17.6%; 36/204) and grandfathers (13.6%; 11/81) were all less likely to be in favour ( Figure 6A). More mothers (41.3%; 367/889) than other women (31.3%; 865/2765) supported condom use within marriage but this difference was not statistically significant after also adjusting for community location and other individual-level characteristics (p=.13) ( Figure 6B). No evidence for differences was found for other female relatives of teenage girls.

In this section, we look at how AGYW whose circumstances put them at risk of HIV infection (AGYW ^@Risk) perceive other people's opinions (i.e. the community norms described above) to be and whether these perceptions influence their own views and use of condoms.

AGYW ^@Risk’s perceptions of the influence of gender norms and other structural factors on their decisions to use condoms by marital status. Similar proportions of unmarried and married HIV-negative AGYW ^@Risk reported perceiving negative community views (31.7% versus 29.2%), negative religious leaders views (27.4% versus 27.1%), and negative family views (27.7% versus 27.8%) that influenced their decisions on whether to use condoms with sexual partners ( Table 3). No differences were found in the proportions of HIV-negative AGYW ^@Risk reporting negative gender norms or family member views as influencing their decisions on condom use between those who were unmarried by single, divorced and widowed status or between those who were married by form of HIV risk but sample sizes were small (data not shown). Currently married HIV-negative AGYW ^@Risk were more likely than unmarried HIV-negative AGYW ^@Risk to believe that their peers were not using condoms (92.6% versus 69.3%).

Perceived stigma around pre-marital sex (38.5%) and about use of condoms in marital relationships (73.6%) were reported by substantial proportions of HIV-negative unmarried and married AGYW ^@Risk, respectively ( Table 3). Unmarried and married HIV-negative AGYW ^@Risk reported similar levels of HIV stigma (35.6% versus 34.3%) and intimate partner violence (16.6% versus 21.8%) ( Table 3).

Associations between gender norms and other structural barriers and AGYW ^@Risk’s motivation, access and use of condoms. In the generic HIV prevention cascade explanatory framework, lack of social acceptability is hypothesised to be a barrier to use of HIV prevention methods by contributing to gaps in motivation ³⁰ . However, gender norms and associated structural factors could also contribute to gaps in access and capacity to use prevention methods. For example, where healthcare workers hold similar opinions on restricting young women’s sexual activity and access to prevention methods to those in the wider community or feel obliged to uphold these views.

In Figure 7, we show the results on associations between gender norms and potentially related structural barriers and gaps at each step in the HIV prevention cascades for condom use in unmarried and married AGYW ^@Risk.

For unmarried AGYW ^@Risk, no association was found between perceived overall community disapproval of condom use and reported condom use at last sex (p=.69) ( Figure 7A). However, two-thirds (69.3%) of these AGYW reported that their friends were not using condoms ( Table 3) and fewer of these women reported using condoms themselves than was the case for those who believed their friends were using condoms (39.4% versus 68.3%; p<.001). Looking closely at the gaps in the HIV prevention cascades for unmarried AGYW ^@Risk, more of the young women who perceived community disapproval of their using condoms lacked motivation to use condoms (21.2%) than was the case for those who didn’t perceive community disapproval (10.3%) (p=.053); but no differences were found for lack of access (p=.68) or lack of capacity to use condoms effectively (p=.82). Twenty-one percent (22/105) of the AGYW in this group who thought their friends did not use condoms lacked motivation to use condoms themselves compared to none of those who said their friends did use condoms (N=56; Fishers exact test, p<.5). No differences were found between AGYW ^@Risk who did and did not think their friends were using condoms and lack of access (p=.38) or lack of capacity to use condoms effectively (p=.72). No differences were found according to whether unmarried AGYW ^@Risk felt there was pre-marital sex stigma (p=.67) or HIV stigma (p=.13) and condom use. However, reports of HIV stigma were positively associated with lack of motivation to use condoms in unmarried AGYW ^@Risk (24.1% versus 7.9%; p=.004); and, pre-marital sex stigma was positively associated with difficulties in accessing condoms in unmarried AGYW ^@Risk who wanted to use condoms (27.3% versus 6.0%; p=.001). Unmarried AGYW ^@Risk who experienced IPV were less likely to use condoms at last sex than those not experiencing IPV (29.4% versus 52.0%; p=.023). These women had similar levels of motivation to use condoms to their peers but tended to be more likely to report lack of access (26.3% versus 12.5% in motivated unmarried AGYW; p=.08) and lack of capacity to use (28.6% versus 15.2% in motivated unmarried AGYW with access; p=.13) condoms with their sexual partners.

For married AGYW ^@Risk, no associations were found between perceived overall community disapproval of condom use (p=.73) or believing friends were not using condoms (p=.48) and reported condom use at last sex ( Figure 7B). However, married AGYW ^@Risk who thought their friends did not use condoms were more likely to lack motivation to use condoms (58.4% versus 37.9%; p=.033) and, for those who were motivated, there was a trend towards more limited access (30.9% versus 11.1%; p=.07). Condom use in marriage stigma showed a weak association with non-use of condoms (p=.13) which appears to reflect greater lack of motivation to use condoms (63.1% versus 39.4%; p<.001) and, in motivated married AGYW ^@Risk, greater lack of access (34.6% versus 19.0%; p=.048). No differences were found for married AGYW ^@Risk in condom use or in any of the gaps in the condom HIV prevention cascade between those who did and did not report community HIV stigma as a factor in their decisions to use condoms ( Figure 7B). A fifth of married AGYW ^@Risk reported recent IPV. These women were more likely than other married AGYW ^@Risk to be motivated to use condoms (55.8% versus 39.6%; p=.013); but, for those who were motivated and could access condoms, there was a trend towards lower capacity to use condoms effectively (8.8% versus 16.1%; p=.17).

Almost a quarter (24.4%) of unmarried AGYW interviewed in this study reported having started sex with the great majority of those currently sexually active and at risk of HIV infection being motivated to use condoms. We found a substantial degree of community acceptance of this pattern of behaviour: that most young women had sex before marriage nowadays and, to a lesser but still considerable extent, that this was acceptable and that teenage girls should be told about condoms and given access to them in schools. These results indicate that the traditional gender norm against female pre-marital sex has probably been weakening in Manicaland; a finding broadly in line with findings from studies in other parts of sub-Saharan Africa ^{20, 38} . Nevertheless, those AGYW ^@Risk who reported community disapproval (31.7%) or friends not using condoms (69.3%) were less likely to be motivated to use condoms than their peers which suggests that traditional social norms and peer pressure remain important influences on many unmarried AGYW’s sexual activity and condom use. Interestingly, the widespread perception amongst unmarried AGYW ^@RISK that most young women in their situation are not using condoms appears to be a misconception as 61.5% of these women reported that they did use condoms when they last had sex. Therefore, this barrier to wider condom use could be rectified in interventions. HIV stigma perceived by unmarried AGYW ^@Risk was also associated with their lacking motivation to use condoms ³⁹ and may be reinforcing these effects. A sizeable minority (16.6%) of unmarried AGYW ^@Risk reported having experienced IPV which – as was the case in an earlier study in Johannesburg and Harare ⁴⁰ – was associated with less condom use. In Manicaland, we also found that this was linked to greater difficulties accessing condoms and to lower capacity to use them effectively.

Some community members’ or groups’ views may have particularly strong or more limited influence on AGYW’s sexual behaviour and use of condoms. Generally, we found that female community members were less likely than their male counterparts to find it acceptable for their daughters to have sex before marriage and to have access to condoms. Women who were mothers of teenage daughters at the time of the survey had similar views to women as a whole; however, aunts of teenage girls, who – in the Shona culture which predominates in Manicaland – play an important role in helping young women to navigate marital arrangements, expressed more liberal views on pre-marital sex. Fathers, and male relatives generally, were more likely than other men to be against female pre-marital sex. In in-depth interviews and focus group discussions, held in the same study areas in Manicaland, parents reported conflicting attitudes towards pre-exposure prophylaxis (PrEP)’s being made available for AGYW ⁴¹ . We found that while ‘… parents wanted to see girl-children protected from HIV … they struggle to reconcile this with traditional ‘good girl’ notions that stigmatise pre-marital sex’. As a consequence, parents often ‘co-produce public gender norms that prevent AGYW from accessing PrEP’ and ‘AGYW require permission and experience limited autonomy to use PrEP due to intersecting sexuality- and PrEP-related stigmas’ ⁴² . It seems likely that parents experience similar mixed feelings towards their daughters having access to condoms.

AGYW ^@Risk – in common with most other members of their communities – reported that ‘many young women have sex before marriage these days’. In these circumstances, young women might be expected to experience strong peer pressure to become sexually active especially given the opportunities to obtain material benefits that early sexual activity can bring ^{10, 43, 44} . However, younger and older AGYW ^@Risk differed in their views on this: whilst relatively few teenage AGYW ^@Risk supported pre-marital sex and AGYW’s having access to condoms, those aged 20–24 had similar views on pre-marital sex to older women and were more likely to approve of their having access to condoms. Therefore peer pressure to engage in pre-marital sex in Manicaland may become stronger as young women get older. Two-thirds of unmarried AGYW ^@Risk said that friends and community members’ views either encouraged them to use condoms or were unimportant. In a systematic review of studies investigating peers influence on young people’s sexual behaviour in sub-Saharan Africa, Fearon and colleagues identified 11 studies that tested 37 possible associations ⁴⁵ . They found evidence for an association between peers and sexual behaviour for 51% (19/37) of the associations tested but noted that longitudinal designs with biomarker outcomes were needed to provide conclusive evidence on the role of peers in adolescent behaviour.

A quarter (27.4%) of unmarried AGYW ^@Risk said that church leaders had a strong influence on their use of condoms. More male church leaders than other men in the community were against the idea of young women having pre-marital sex but they didn’t differ in their levels of support for making condoms available for young women. Female church leaders were actually more in favour of this than other women. Church leaders views on young women’s early sexual activity could vary between religious denominations. Sample sizes were too small for this to be investigated directly using the survey data but there was no evidence for differences in the population as a whole: members of all churches were less likely to support female pre-marital sex than people with no religion. Members of Apostolic churches were less likely to support making condoms available to unmarried young women than people subscribing to other religions ⁴⁶ .

Many young women in Manicaland are members of self-assessed ‘well-functioning’ local community groups ⁴⁷ . In previous studies, we have seen that participation in these groups can contribute to social changes (e.g. early adoption of safer sexual behaviours ⁴⁸ and uptake of new HIV services including HIV testing and prevention of mother-to-child transmission ⁴⁹ ). In the current study, we found that women who were members of community groups were more likely to accept unmarried AGYW starting sex and having access to condoms. Therefore community group participation may be one pathway through which traditional social norms on young women’s sexual activity are being eroded.

The views of local healthcare workers may also be an important influence on unmarried AGYW ^@Risk‘s access to HIV prevention methods. If they disapprove of female pre-marital sex, this will probably contribute to AGYW ^@Risks’ sense that they can’t access condoms due to lack of acceptable provision. In the survey, nurses and community health workers (who are both largely female) were more likely to express supportive views towards AGYW having access to condoms than other women in the community. However, in in-depth interviews, some were self-critical of how they, as a group, interacted with AGYW. Those in rural areas, especially, acknowledged how their community-embeddedness affected their perceptions of AGYWs’ sexuality ³³ . This mix of community-embeddedness and self-awareness may help to explain the patterns we found in the condom cascade for unmarried AGYW ^@Risk – i.e. a relatively small gap between the motivation and access bars in the cascade but lack of acceptable provision being the largest (>75%) barrier contributing to this gap.

Only a small minority of married AGYW ^@Risk (4.3%; 17/394) reported using condoms when they last had sex. Six of these women reported using condoms as a family planning method and two had multiple sexual partners themselves and may have had their last sex with a non-regular partner. Therefore, underlying levels of condom use for HIV prevention within marriage are probably even lower than 4%. If so, this would be consistent with the limited community support we found for condom use generally within marriage (34.1%), the high condom in marriage stigma reported by married AGYW ^@Risk (73.6%), and the small minority of these young women who think their friends use condoms (7.4%). Reports of condom in marriage stigma and believing that friends are not using condoms were both associated with less motivation to use condoms themselves ³ which could be indicative of peer pressure. However, only a minority (29.2%) of married AGYW ^@Risk felt that community disapproval was an obstacle to their using condoms and no difference was found in motivation to use condoms between those who did and did not report community disapproval as being an obstacle. Therefore, other factors, such as low risk perception and perceived negative consequences of using condoms (e.g. loss of sexual pleasure and fertility desires) may be stronger barriers to these women’s being motivated to use condoms.

Community views on condom use within marriage were less favourable in cities than in rural areas; perhaps surprisingly as rural communities might have been expected to be more conservative and HIV prevalence, typically, has reached higher levels in urban and peri-urban areas ^{50, 51} . Within the study communities, as elsewhere in sub-Saharan Africa, younger people (15–24 years) were less likely than older people (≥25 years) ³⁸ and currently married people were less likely than never-married people to support condom use in marriage ⁵² . This may be, in part, because this is a demographic who are keen to have children or feel that their risk of HIV infection is low. As a consequence, any influence from married AGYW ^@Risks peer groups would tend to affirm and reinforce their adversity to condom use. Parents of teenage daughters and other potentially influential community members expressed similar views to the wider communities in which they lived.

Married AGYW ^@Risk who reported having experienced IPV were equally likely to report condom use as those who did not report IPV. Whilst more of these women wanted to use condoms, those who were motivated to do so showed a trend towards lower capacity to use condoms. In an earlier analysis of a wider age-range (ages 15–54) of married women in Manicaland, we found a positive association between experiencing IPV and being at risk of HIV infection (72.3% versus 58.5%; p<.001); but, as here with married AGYW ^@Risk, no difference in condom use amongst those at risk ⁵³ . The HIV prevention cascades for women reporting and not reporting IPV were similar; both showing large gaps in motivation and capacity to use condoms. Women reporting motivation and access were more likely to report their partner being a barrier to condom use if they experienced IPV (84.8% versus 75.5%; p=.015).

Comins and colleagues conducted a survey of AGYW recruited from selected venues using time-location sampling and used latent class analysis to identify intersecting social- and structural-level determinants of HIV/STI acquisition in Ethiopia ⁵⁴ . Previous qualitative studies have described community views on aspects of AGYW’s sexual behaviour ⁴ with some finding evidence for changing community views on AGYW sexuality and use of HIV prevention methods ²⁰ . However, to the best of our knowledge, this is the first study to measure contemporary support for traditional gender norms surrounding young women’s sexuality in a representative population survey, in a high HIV prevalence sub-Saharan Africa setting, and to investigate their influence on condom use for HIV prevention.

The study had a good sample size with a good participation rate, was conducted in a mix of different socio-economic settings (making the findings generalisable to much of Zimbabwe), used biomarker data to identify uninfected individuals, and utilised bespoke validated HPC data collected in the survey that covered multi-level demand and supply side barriers to condom use ³⁰ . The data analysis was stratified by marital status which highlighted important differences in the gaps and barriers to condom use between these two groups of AGYW at higher risk of HIV infection. The survey data used in the main analyses for the study were cross-sectional so we have not been able to establish the direction or causality of the associations we observed. Longitudinal studies are needed to quantify changes in support for and adherence to traditional and emerging new social norms over time. The population cohort survey data used to compare rates of new infections in unmarried versus married AGYW were only available for an earlier period (1998–2013) and did not include data for the two urban sites included in the 2018/19 survey so the observed relative levels of HIV incidence may not represent the situation in 2018/19. Informal confidential voting interview methods ⁵⁵ , adapted for use with questionnaires administered using tablets in the 2018/19 survey, were used to reduce social desirability bias in the data on sexual behaviour. However, some residual under-estimation of the priority populations for HIV prevention methods is likely and the measurements of social norms and other structural barriers also will be subject to this bias. When populating the HPCs, we assumed that the AGYW who reported using condoms were doing so – at least in part – for HIV prevention and were motivated and had access to condoms for this purpose. Where AGYW ^@Risk reported using condoms for family planning, this assumption may not have been correct; in which case, our estimates for all bars in the HPCs will have been overstated.

PrEP is an additional HIV prevention option available now for AGYW ^@Risk in the general population in Zimbabwe and elsewhere in sub-Saharan Africa ⁵⁶ . The current study was limited to condoms; but, despite PrEP’s fundamental advantage in being a method over which women can, in principle, exercise direct control, there are early signs that the norms that we have examined here can also act as obstacles to AGYW’s use of PrEP ^{42, 57} .

In an analysis of Zimbabwe Demographic and Health Survey data, Sambisa and colleagues ⁵⁸ described differences in AGYW’s sexual risk behaviours for HIV infection (SRBs) between the Shona and Ndebele tribes and concluded that condom promotion approaches could be strengthened by recognising and responding to cultural forces that reproduce and perpetuate risky sexual behaviours. In a community-randomised control trial in South Africa, Pettifor and colleagues found that a community mobilization intervention reduced negative gender norms in men but did not reduce IPV or increase condom use ³ . The authors concluded that more time might be needed to change behaviour or that interventions may need to address behaviours more directly. Following an analysis of survey data collected from sexually-experienced students in Nyanza, Kenya – and consistent with our HIV prevention cascade framework – Maticka-Tinsdale and Eric Tenkorang ¹⁷ concluded that key factors influencing reported condom use existed at the school/community levels as well as at the individual level. They recommended that HIV prevention interventions should take account of barriers at each of these levels.

Based on the findings from these and other studies, the PEPFAR-supported multi-country DREAMS programme, developed to reduce SRB and increase AGYW’s use of HIV prevention methods, incorporated activities to address barriers at community, sexual partner and individual levels ⁵⁹ . Community-strengthening programmes included community conversations with male opinion leaders to promote positive gender norms and IPV reduction activities. Early findings from the comprehensive set of planned DREAMS evaluations show mixed evidence for an impact of the programme in reducing new infections – with faster reductions in new diagnoses in pregnant women in settings where more DREAMS activities have been implemented ⁶⁰ but no evidence for an acceleration in HIV incidence declines in the period immediately following their introduction ⁷ . This may be because early uptake of community activities was low in some intervention settings ⁶¹ . However, whilst our findings support the overall layered approach adopted in the DREAMS programme, they suggest a need to establish social spaces where helpful norms can be cemented, unhelpful norms (e.g. moralistic attitudes and conservative gender ideologies) renegotiated, and popular misconceptions (e.g. on levels of condom use amongst unmarried AGYW ^@Risk) challenged. This can be achieved through community conversations, an approach that provides community members with a platform to come together to share perspectives and collectively address community challenges (like those discussed above) ^{62, 63} . In other health contexts, where gender inequalities have constituted a problem, community conversations have proved productive in transforming gender norms and practices ⁶⁴ .