This analysis forms part of the Botsha Bophelo Adolescent Health Study (BBAHS), an interviewer-administered, cross-sectional survey of 830 adolescents aged 14-19 years living in Soweto.

The study was conducted at the Perinatal HIV Research Unit (PHRU) and at Kganya Motsha Adolescent Centre (KMAC) in Kliptown. Kliptown is the oldest district in Soweto, and it comprises a mixture of purpose-built housing and informal housing known as shacks [28]. Soweto is a large peri-urban township of about 1.3 million official inhabitants [29]. KMAC was founded in 2008 as an adolescent-friendly HIV management centre serving the HIV voluntary counseling and testing (VCT) and sexual and reproductive health needs of adolescents aged 14-19 years [30].

A targeted, stratified sampling approach was used based on geographic distribution across all areas of Soweto, as well as the sex and age of adolescents. Seven of the areas were informal areas (locally known as informal settlements), where adolescents lived in housing structures composed of metal structures (locally known as shacks), with shared access to water, toilets, and limited or no access to electricity. Eligibility criteria included age 14-19 years and living in Soweto. Participants were recruited from various locations within each area with the aim of ensuring representation from all areas of Soweto and recruiting hard-to-reach adolescents who may not have accessed health services. Research interviewers strategically targeted adolescents around malls, schools, and neighbourhood hangouts. Interviewers provided potential participants with recruitment flyers with contact information where interviewers could be reached. Interviewers were available on a dropin basis at KMAC and the PHRU or by appointment. To accommodate adolescent schedules, interview times were available after school and on the weekends. Some communities were more challenging to recruit from than others, particularly informal settlements where the visibility of adolescents was reduced due to the lack of community infrastructure for young people to congregate (ie, schools and malls). Participants learned about the study from research staff and were invited to follow-up if they were interested in participating. Participants were also invited to share the information about the study with their friends. Therefore, we have a convenience sample based on targeted recruitment strategies stratified by geographic location, age, and sex characteristics.

Surveys were conducted between June 2010 and June 2012. Participants received a face-to-face interviewer-administered survey assessing demographics, mobile phone access and use, sexual risk behavior, and alcohol use. Trained, multilingual interviewers administered the survey in English or IsiZulu (the most commonly used local languages in Soweto) via an online Internet platform, Survey Monkey, using iPads or computers [16,31]. Each survey took about 60 minutes.

The ethics committees of the University of the Witwatersrand, Johannesburg, South Africa and Simon Fraser University, Vancouver, Canada approved study procedures. Participants received 50 ZAR (~ $7) for travel reimbursement. Written informed consent was obtained for all participants. In addition, participants younger than 18 years required written parental consent together with their own written assent.

The primary outcome variable of interest in this analysis was social networking, defined as the use of social networking platforms such as Facebook, Twitter, and Mxit accessed through mobile phones. Social networking was measured by responses (yes vs no). Explanatory variables included demographics (including: gender, age, primary language spoken at home, education level [primary vs post primary]), socioeconomic (type of housing, source of spending money), sexual behavior (sexual orientation, sexually active, number of sexual partnerships, same sex partnerships, age at coital debut, condom use, intergenerational sex), alcohol and drug use in the last 6 months, depression, mobile phone use (eg, access to a mobile phone, access to the Internet, and source of Internet [computer vs mobile phone]), sexual partnerships ([none, one, and more than one], age at coital debut [less than 15 years, 15-16 years, and 17 or more years], intergenerational sex and condom use), and relational (eg, parental and adult presence in household) variables.

Participant characteristics of adolescents who accessed social networking sites are presented in frequencies and compared through bivariate analysis using chi-square and Fisher’s exact test. Variables with a significant association (P<.05) were considered for entry into the multivariate logistic regression model. Univariate and multivariate logistic regression were performed to determine crude and adjusted odds ratio (AOR), 95% confidence interval (CI), and their corresponding P values. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit statistic where the model was determined a good fit if the P value was not significant [32]. All statistical tests were two-sided and were conducted using STATA version 12.

In total, 956 interviews were completed between 2010 and 2012. Following extensive data cleaning, a final sample of 830 adolescent participants was reached. Due to the nature of our recruitment strategy, we were not able to determine absolute response or refusal rates. The median age for the sample (n=830) was 18 years (IQR 17-18) and 57% (n=475) were female. Of all participants, 60% (n=494) used social networking platforms. The most commonly used platforms were Mxit (46%) and Facebook (34%).

Table 1 shows the comparison of characteristics of social networking users (n=494) to nonsocial networking users (n=336). In bivariate analyses, social networking was significantly associated with sex (P=.041), age group (P=.002), type of housing (P<.001), educational level (P<.001), access to mobile phone (P<.001), access to Internet (P<.001), and source of Internet (P<.001). Other variables associated with social networking were alcohol use in the past 6 months (P<.001), daily mobile phone usage (P<.001), presence of an adult in the household (P=.001), sexually active (P=.032), sexual partnerships (P=.011), and age at coital debut (P=.012).

The unadjusted and adjusted predictors of social networking use are presented in Table 2. In the adjusted logistic regression model, adolescents who used mobile phones for 2-4 hours (AOR: 2.89, CI: 1.80-4.65) or more than 5 hours (AOR: 5.99, CI: 3.79-9.48) daily and those with one (AOR: 1.75, CI: 1.04-2.94) or more than one (AOR: 2.64, CI: 1.40-4.96) sexual partner had a higher odds of social networking.

The unadjusted and adjusted predictors of social networking use by sex are presented in Table 3. In the adjusted logistic regression by sex, males whose source of spending money was not provided by parents (AOR: 0.58, CI: 0.34-0.99) had lower odds of social networking. Those who used mobile phones for 2-4 hours (AOR: 3.06, CI: 1.69-5.51) or more than 4 hours daily (AOR: 6.16, CI: 3.46-10.9) had a higher odds of social networking. Adolescent males with one (AOR: 3.35, CI: 1.79-6.27) or more than one (AOR: 2.58, CI: 1.05-6.36) sexual partner had a higher odds of networking. Among females, a higher odds of social networking was associated with spending 2-4 hours (AOR: 3.53, CI: 2.08-5.97) or more than 5 hours (AOR: 6.49, CI: 3.92-10.73) daily on mobile phones and having more than one sexual partner (AOR: 1.99, CI: 1.08-3.67).

This is one of the first studies to report on social networking use via mobile phones among adolescents in South Africa. The data show that social networking use among a sample of adolescents in Soweto was associated with their source of spending money from either parents or others, longer duration of daily mobile phone use, and having more than one sexual partnership. In an analysis by sex, social networking was associated in both sexes with longer daily use of mobile phones, being sexually active and having multiple sexual partners. Interestingly, for females, there was a relationship between social networking use and alcohol use. For males, social networking use was associated with longer daily use of mobile phones and having more than one sexual partner.

Despite Soweto being a typically lower socioeconomic community in South Africa, social network access was prevalent among 60% of adolescents, indicating the opportunity to introduce HIV risk reduction interventions via social networking platforms to young people in this setting. Mobile phones were the main point of social networking access for the majority of the adolescents. In developed settings like the United States, social networking has been shown to be a feasible tool for health research among adolescents [33]. Studies (Yonker et al, 2015) have also shown that using social networking among adolescents and young adults concerning their health choices has proven to be an essential medium of communication [34]. Magidson et al (2015) demonstrated the effective use of social networking to engage HIV positive adult men who have sex with men in care [35]. In sub-Saharan Africa, published literature on social networking among adolescent samples is limited.

In our study, we found that access to social networking was lower among adolescents with lower socioeconomic markers including less educational attainment, living in informal housing arrangements and no adult presence in the household. We also found that those adolescents accessing social networking were more likely to be older, sexually active, and using alcohol. Taken together, these findings indicate the excellent potential to reach adolescents in Soweto with HIV prevention interventions via social networking, particularly those with higher risk sexual behaviors including being sexually active, having more than one sexual partner, and those who use alcohol. However, it may be harder to reach adolescents facing lower socioeconomic hardships via social networking.

Among participants under age 18 years, almost a third reported alcohol use, although we did not measure the extent of use. Findings from the second South African youth risk behavior survey (2008) showed that half of school-going adolescents aged 13–19 years had ever used alcohol and a third having consumed alcohol in the past month. Two-thirds of social networking users reported alcohol use in the past 6 months. Of those who reported social networking, almost half were underage. Furthermore, half of female adolescents who consumed alcohol in the past 6 months were more likely to use social networking platforms [36].

The link between alcohol advertising and use among adolescents has been documented before; however, research conducted in the United States has shown conflicting evidence about the effect of media campaigns to reduce problematic alcohol use. There is evidence that links alcohol advertising with alcohol initiation [37]. Furthermore, there is evidence from the United States that exposure to friends’ accounts of alcohol use on social networking platforms could have an indirect effect on willingness to consume alcohol [38]. By extension it may be reasonable to assume that if adolescents were exposed to normative antialcohol interventions on social networking platforms, they may in turn become less likely to initiate or continue alcohol use. In this way, social networking platforms could allow targeting of specific populations for novel antialcohol interventions. Male adolescents are more likely to use alcohol at an early age but there has been an increase in the number of younger female adolescents consuming alcohol [39,40], which is a concerning trend because females are physiologically more at risk from the effects of alcohol use than males [41,42,43]. Young females are an emerging target market for alcohol producers [40]. Adolescent alcohol use has been described as a factor in female sexual assault and engaging in risky sexual behaviors [44,45]. Among social networking users, almost a third reported multiple sexual partnerships. Adolescents with multiple sexual partners are a higher risk group for sexually transmitted infections [2]. However, multiple and concurrent sexual partnering is a risk factor for HIV only if condom use is inconsistent and if a partner is infected with HIV [46]. In South Africa, multiple and concurrent sexual partnerships has been associated with HIV infection among young people [2,47-50]. Of sexually active young people aged 15-24 years in South Africa, 22% had more than one sexual partner in the past 12 months [47]. Social networking applications can potentially be used to promote positive health messages through adolescent social networks.

This study did not document the type of content that adolescents were accessing on the social networking platforms; therefore, we could not determine whether adolescents demonstrated health-seeking behaviors on these platforms. Social desirability bias may have affected variables, possibly underestimating self-reported risk behaviors. Social networking based health applications may not be broadly applicable but would be effective for subsets of the adolescent population who are able to obtain data bundles to access social networking sites. There were limitations to the survey items because we did not assess the level of alcohol use and concurrency of sexual partnerships.

Our study suggests evidence about mobile phone–based social networking use among adolescents in Soweto, South Africa, and beyond. We recommend that future research addresses social networking platforms as a means of disseminating health interventions including positive alcohol and sexual behavior messaging among adolescent populations in South Africa.