HIV trends and disparities by gender and urban–rural residence among adolescents in sub-Saharan Africa

High levels of HIV infection among adolescents, especially girls, have been a major concern in sub-Saharan Africa, where about 80% of the world’s HIV-infected adolescents live [1, 2]. Adolescent girls and young women aged 15–24 years accounted for one in five new HIV infections, despite being just 10% of the population of sub-Saharan Africa [3]. UNAIDS estimated that, between 2010 and 2017, there was a 25% decline in new HIV infections among girls aged 10–19 years in Eastern and Southern Africa, but no decline in West and Central Africa. A recent analysis of pooled data from longitudinal community studies in Eastern and Southern Africa showed that adolescent girls had 5.9 and 3.2 times higher HIV incidence than adolescent boys during 2005–2015, respectively, with only limited evidence of a decline over time [4].

Several global and national initiatives have called for greater attention to adolescent girls and HIV, often with ambitious targets. The ‘All In to End Adolescent AIDS’ initiative, led by UNICEF and other global agencies, aims to reduce HIV infections among adolescents 10–19 years by 75% by 2020 and end the AIDS epidemic among adolescents by 2030 [5]. The DREAMS partnership, launched in 2014, in ten countries in sub-Saharan Africa aims to bring down adolescent and young women new infections by 40% in a 2-year period [6].

The HIV epidemic in sub-Saharan Africa is characterized by large differences between countries and subregions, and adolescents are no exception [7]. Understanding local epidemiology within each country is essential to effectively target HIV prevention efforts [8]. This includes consideration of distal socioeconomic and residential characteristics as well as biological and behavioural factors that affect HIV transmission [9, 10]. Within countries, HIV epidemiology tends to vary by socioeconomic and geographic characteristics [11]. National surveys often show higher HIV prevalence among adolescents and adults with more education, living in wealthier households or in urban areas compared to less educated, poorer and rural residents [12–15]. These inequality patterns differ from those observed for many other infectious diseases and for reproductive, maternal and child health but may change if urban, wealthier and more educated populations are quicker to adopt interventions or behavioural changes. An analysis of seven countries in sub-Saharan Africa found limited evidence such changes among educated adolescents prior to 2013 [16].

This study aims to assess sex-specific trends in HIV incidence and prevalence in adolescents by urban–rural residence, using national surveys with HIV testing and modules on sexual behaviour. This assessment will focus on the big picture from about 2005 to the most recent data points until 2018 for two subregions of sub-Saharan Africa, Eastern and Southern Africa and West and Central Africa.

The data were derived from national household surveys with HIV testing. We identified 77 surveys with HIV testing conducted in sub-Saharan Africa since 2003, including Demographic and Health Surveys (DHS), AIDS indicator Surveys (AIS) [17], Population HIV Indicator Assessment (PHIA) surveys [18] and specific national surveys (see Additional file 1). All surveys are nationally representative household surveys and included the collection of blood samples for HIV testing among female and male respondents 15–49 years (DHS and AIS) or 0–64 years (PHIA). Response rates in the surveys were high, as the median HIV testing and interview response rates for girls and boys 15–19 years in DHS and AIS surveys were about 90%, with only one survey with response rates below 70% (South Africa 2016).

Testing procedures for HIV prevalent infection may differ but all survey protocols involved confirmatory testing of positive cases [17, 18]. Recent advances have made it possible to directly estimate HIV incidence in cross-sectional surveys, using an algorithm based on limited antigen (Lag) avidity assay, viral load, and ARV use [19]. These statistics were available for respondents 15–24 years in eight PHIA survey reports that reported age-disaggregated data and one national survey in South Africa. The numbers were too small for disaggregation by urban–rural residence and no trend assessment could be done.

All surveys used a two-stage cluster-based sampling design to select households within selected census enumeration areas or equivalent unit. For the DHS and AIS surveys, conducted between 2003 and 2018, data sets are in the public domain and standardized indicators are available. The 11 PHIA surveys were conducted from 2015 but only four surveys data sets were available in the public domain at the time of our analyses (Eswatini, Malawi, Tanzania, and Zambia). We extracted information from the PHIA national reports for selected indicators if no primary data were available. Kenya, Nigeria, South Africa and Eswatini conducted other types of national HIV surveys for which only national reports could be accessed.

We focused on the assessment of trends in countries with surveys before and after 2010 (the most recent survey). We included Burkina Faso with its most recent survey in 2010 and Burundi with its first survey in 2010. We classified the countries into two geographic subregions, Eastern and Southern Africa, and West and Central Africa. We included Burundi and Rwanda in the West and Central Africa subregion because of their proximity to Central Africa and their greater HIV epidemiological similarity with this subregion (see Additional file). Trend data were available for 23 countries, with a median year of the first survey of 2005 and the second of 2014. 1

Subregional levels and trends were summarized with country medians and the annual average rate of change in HIV prevalence between the first (p1, referring to HIV prevalence at time t₁) and last survey (p2, HIV prevalence at time t₂), defined as the average relative percent change per year in HIV prevalence(ln(p₂/p₁)/(t₂ − t₁)). All analyses were conducted in Stata version 15 (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC) and MS Excel.

We tested for statistical significance of indicator trends over time by estimating the level of significance in the proportions’ differences using a formula developed by Altman and Bland [20]. We used 95% confidence intervals based on the Taylor linearization method used in DHS surveys and the jackknife repeated replication used for the PHIA surveys. The differences between the two methods are small in our case, as has been reviewed extensively elsewhere [21, 22].

HIV prevalence among adolescents has been used as a proxy measure for population incidence as the majority of infections represent recent sexually acquired infections and direct measurement of incidence was limited [23]. Even with HIV incidence measurement available, data are also needed to assess the distribution of prevalent infection in subpopulations. Because of the scale-up for antiretroviral therapy (ART) for children [24], prevalent adolescent HIV infection may include a rising number of pediatric infections who survive into adolescence, which affects the interpretation of adolescent prevalence statistics [25]. To obtain an idea of the size of this effect, we also examined HIV prevalence rates at ages 5–9 and 10–14 years, when sexual transmission is likely to be low, from recent PHIA surveys.

Even though our primary interest is in adolescents 15–19 years, we used data for the age group 15–24 years to disaggregate by urban–rural residence, to reduce sampling error. The correlation between HIV prevalence at ages 15–19 and 15–24 years was high for both females and males (Pearson’s correlation coefficient r > 0.90) (see Additional file). Urban–rural breakdowns were not available for the most recent surveys in Eswatini, Kenya and South Africa. 2

Urban–rural residence may affect HIV transmission through a set of proximate determinants which are behavioural in nature with a proven effect on the biological components of transmission such as risk of exposure and transmission efficiency [10, 26]. We examined trends in selected indicators of the proximate determinants, based on the availability of comparable time trend data from DHS, in urban–rural populations, disaggregated by gender:Sexual initiation: percentage of women/men 18–24 years who initiated sex before age 18Premarital sexual activity: percentage of young never married women/men age 15–24 who had sex in the last 12 months of all young single women/men surveyed among respondents 15–24 yearsMultiple partnerships: percentage of young women/men age 15–24 who had sex with more than one partner in the 12 months preceding the survey among women/men who had sexual intercourse in the 12 months preceding the surveyCondom use: percentage of young never married women/men age 15–24 who used a condom at last sex, of all young single sexually active women/men surveyedHIV testing: percent who received an HIV test in the last 12 months among sexually active respondents 15–24 years.

HIV prevalence among adolescent girls 15–19 years decreased in both subregions of sub-Saharan Africa between 2005 and 2015, by as much as 54% in Eastern and Southern Africa where prevalence is higher and by 43% in West and Central Africa where prevalence is generally 1% or lower. Among boys 15–19 years, HIV prevalence is generally less than half the levels of girls, the median prevalence declined in Eastern and Southern Africa by 43% but remained low in West and Central Africa (0.3%). These findings confirm the presence of a positive trend towards lower HIV transmission in many countries.

The prevalence trends among girls 15–19 years are likely to present an underestimate of the reductions in HIV transmission due to sexual behaviour. Prior to the introduction of antiretroviral therapy (ART) in children, only a small proportion of adolescents would have acquired HIV from their mothers, as most would have died before reaching age 15. Most adolescent infections would be sexually transmitted, almost exclusively to girls due to partnerships with infected older men, as their male peers have HIV prevalence close to zero in many countries. The comparison of HIV prevalence at 15–19 years with earlier age groups gives an indication of the size of the effect of increasing coverage with ART in children with vertically transmitted HIV. Prevalence was about 1% among children 5–9 and 10–14 years in the higher prevalence countries with little negligible girl-boy differences in both age groups and a much higher prevalence at 15–19 only among girls.

HIV incidence results in recent surveys are consistent with this finding. Annual incidence among females 15–24 years was at least 0.5 per 100 per year in most countries with higher prevalence, while male levels are close to zero. Unfortunately, incidence measurement in surveys is still affected by sample size limitations for this relatively rare event, and no estimates for 15–19 years or time trends are available.

These results confirm the importance of sexual transmission among girls and not among boys, given the difference between 10–14 years and 15–19 years prevalence among girls but not boys, and the large sex differences in HIV prevalence at 15–19 years which are not expected as a result of only vertical transmission. Elsewhere, maternal transmission rather than sexual transmission was also considered a likely explanation for the higher prevalence among adolescent orphans 15–17 years old compared to non-orphaned peers [27].

In virtually all countries, urban and rural HIV prevalence among adolescents and young people 15–24 years declined at a similar pace, leaving the gap unchanged. Urban residents however are however still disadvantaged with higher HIV with urban rural ratios among females of 1.6 and 2.1 in Eastern and Southern Africa and West and Central Africa, respectively.

Self-reported indicators of behavioural risk did not help explain this urban disadvantage. Sexual risk behaviours as measured by sexual initiation before age 18, premarital sexual activity and multiple partnership among sexually active adolescents and young people, were remarkably similar in the 23 countries with two surveys conducted around 2005/2006 and 2015/2016. The urban rural gaps remained similar. There was however an increase in reported condom use at premarital sex in both higher and lower prevalence countries, and a major increase in HIV testing rates among adolescents and young adults of both sexes in urban and rural respondents. The urban–rural gap reduced, as rural residents experienced a faster increase. The increase in condom use in both urban and rural adolescents is encouraging, even though there are concerns about the shift in emphasis on long term methods in family planning programs leading to less emphasis on condoms which is the only method that protects against pregnancies and sexually transmitted infections [28]. Even though there are substantial knowledge gaps about what works best, the provision of sexuality education and sexual and reproductive health services. increasing awareness, acceptance, and support for youth-friendly sexual and reproductive health education and services and addressing gender inequalities have been singled out as priority interventions [29].

Our study has several limitations. Our focus on subregional inequalities and trends ignored the substantial variation between countries with the subregions. The most prominent outlier was Ethiopia within Eastern and Southern Africa with much lower HIV prevalence than all other countries in this subregion, but also differences between Eastern and Southern African countries were large in terms of levels and trends. Also, within West and Central Africa there were important differences between countries, but it is beyond the scope of this paper to discuss these individual trends. The urban–rural analyses were based on ages 15–24 years for disparity analysis to reduce sampling errors. We however showed that HIV prevalence at 15–24 years was highly correlated with 15–19 years and is likely to be indicative of urban–rural differences at ages 15–19 years. All other indicators focused on adolescent HIV and related risk behaviours through a combination of current status data and recall by survey respondents 15–24 years.

Regarding behavioural indicators, we considered only multi-country and subregional trends for a limited number of indicators with comparable data. The urban–rural differences in HIV and behaviours appeared weakly correlated. In fact, several behavioural indicators in this study would suggest higher HIV risks for rural than urban adolescents, which suggests that other proximate factors are likely to play a role. These may include risk related to age mixing (younger girls having sex with older men), concurrent partnerships and sexual networks, other sexually transmitted infections, and male circumcision [30–34]. There is some evidence from surveys of increases in male circumcision rates in multiple countries, and more so in urban areas [35]. In addition, sexual behaviour reporting is subject to multiple biases, of which underreporting related to stigma or desirability bias is the most problematic, especially by girls and young women [36–39]. The consistency of the results between consecutive surveys in the same country was however remarkable, suggesting no major change in bias over time. In addition, our analysis did not consider within urban inequalities, especially the HIV risk among the urban poor [40]. This issue needs more attention in future research.

Our main objective was to assess the general sex-specific HIV trends in urban and rural adolescents in two major subregions of sub-Saharan Africa. HIV prevalence data suggest that HIV transmission among adolescents 15–19 years have reduced considerably in the past decade in both Eastern and Southern Africa and West and Central Africa, in spite of increasing survival of children who acquired HIV from their mothers into adolescence. This may be due to increased testing and ART and increased condom use. However, incidence and prevalence issue indicate that sexual transmission is still a major issue for adolescent girls. Adolescent and young adult HIV prevalence was still 1.5 to two times higher in urban populations compared to rural population in both subregions, with little change over time. Urban adolescent girls remain at greatest risk of HIV especially in higher prevalence countries and need to be targeted in prevention programs.