Associations of social media and health content use with sexual risk behaviours among adolescents in South Africa

Figures & data

Figure 1. Mzantsi Wakho cohort study flow chart. Abbreviations: T1, wave 1; T2, wave 2; T3, wave 3

Table 1. Descriptive characteristics of the study population

	Sample size	All participants	Boys	Girls	Adolescents not living with HIV	Adolescents living with HIV
Wave 1, 2014–2015
Female	1410	56.7%	..	..	60.3%	55.1%
ALHIV	1410	69.4%	72.0%	67.5%	..	..
Age	1410	13.8 (0.08)	13.2 (0.11)	14.3 (0.11)	14.3 (0.10)	13.6 (0.6)
Rural residence	1410	27.4%	25.5%	28.9%	29.0%	26.8%
Informal housing	1409	18.0%	15.7%	19.8%	16.5%	18.7%
Poverty	1409	67.2%	64.0%	69.7%	65.6%	67.9%
In a relationship	1410	29.3%	25.4%	32.3%	40.6%	24.3%
School enrolment	1410	94.3%	98.4%	91.2%	94.7%	94.2%
Wave 2, 2015–2017
Use	1410
No access		49.6%	56.6%	44.2%	42.9%	52.5%
Social media alone		33.3%	23.6%	40.7%	49.2%	36.3%
Health content		17.1%	19.8%	15.1%	7.9%	21.2%
Sex after substance use	1221	7.2%	8.4%	6.3%	11.5%	5.3%
Unprotected sex	1410	14.7%	9.0%	19.0%	20.6%	12.1%
Multiple sexual partnerships	1410	15.1%	16.7%	13.9%	18.3%	13.7%
Inequitable sexual partnership	1352	10.7%	9.7%	11.4%	10.4%	10.8%
Wave 3, 2017–2018
Use	1353
No access		42.7%	53.2%	34.7%	37.4%	45.1%
Social media alone		34.1%	24.7%	41.3%	55.2%	44.5%
Health content		23.2%	22.1%	24.0%	7.4%	30.3%
Sex after substance use	1265	7.5%	9.3%	6.2%	11.1%	5.9%
Unprotected sex	1353	21.7%	12.3%	28.9%	31.9%	17.0%
Multiple sexual partnerships	1353	16.0%	19.6%	13.3%	20.2%	14.1%
Inequitable sexual partnership	1237	9.9%	6.2%	12.7%	12.9%	8.6%

Note. Data are % or mean (SD) unless otherwise specified. Abbreviations: SD, standard deviation.

Figure 2. Sexual risk behaviours by participant's sex across waves. Abbreviations: Cl, confidence interval

Figure 3. Sexual risk behaviours by participant's HIV status across waves. Abbreviations: Cl, confidence interval. HIV, human immunodeficiency virus . HIV-, adolescents not living with HIV. HIV+, adolescents living with HIV

Table 2. Multivariable association between mobile phone access and use and self-reported sexual risk behaviours. Results from mixed-effects logistic regression models on ten imputed datasets (n = 1353)

	Sex after substance use		Unprotected sex		Multiple sexual partnerships		Inequitable sexual partnerships
	Adjusted average marginal effects (95% CI)	p-value	Adjusted average marginal effects (95% CI)	p-value	Adjusted average marginal effects (95% CI)	p-value	Adjusted average marginal effects (95% CI)	p-value
Model 1 (Mobile phone access vs. no access)
No mobile phone access (Reference)	0		0		0		0
Mobile phone access	–0.0056 (–0.0269 to 0.0157)	0.605	0.0081 (–0.0187 to 0.0350)	0.552	0.0146 (–0.0110 to 0.0404)	0.264	0.0003 (–0.0245 to 0.0251)	0.981
Model 2 (Social media alone and health content vs. no mobile phone access)
No mobile phone access (Reference)	0		0		0		0
Social media alone	0.0235 (–0.0016 to 0.0487)	0.066	0.0468 (0.0156 to 0.0780)	0.003	0.0177 (–0.0108 to 0.0461)	0.223	0.0015 (–0.0258 to 0.0287)	0.916
Health content use	–0.0532 (–0.0739 to –0.0324)	<0.001	–0.0750 (–0.1063 to –0.0436)	<0.001	0.0095 (–0.0230 to 0.0420)	0.569	–0.0016 (–0.0323 to 0.0290)	0.917

Data are average marginal effects (and 95% CI) adjusted for participant’s sex, rural residence, informal housing, HIV status, household poverty and marital status. Data were obtained from mixed-effects logistic regression models on ten imputed data (n = 1353). Model 1 presents data from regression of individual sexual risk behaviours on mobile phone access (access vs. no access), adjusting for covariates. Model 2 presents data from regression of individual sexual risk behaviours on mobile phone use (social media alone and health content vs. no access), adjusting for covariates. Abbreviations: CI, confidence interval.

Table 3. Comparisons of the multivariable association between mobile phone access and use and self-reported sexual risk behaviours for boys, girls, adolescents living with and without HIV. Results from mixed-effects logistic regression models on ten imputed datasets (n = 1353 including 587 boys, 766 girls, 933 living with HIV, and 420 not living with HIV)

	Sex after substance use		Unprotected sex		Multiple sexual partnerships		Inequitable sexual partnerships
	Adjusted average marginal effects (95% CI) p-value		Adjusted average marginal effects (95% CI)	p-value	Adjusted average marginal effects (95% CI)	p-value	Adjusted average marginal effects (95% CI)	p-value
Model 3 (Mobile phone access vs. no access)
Among boys
No mobile phone access (Reference)	0		0		0		0
Mobile phone access	–0.0060 (–0.0447 to 0.0327)	0.760	–0.0019 (–0.0396 to 0.0359)	0.923	0.0535 (0.0111 to 0.0959)	0.013	–0.0068 (–0.0462 to 0.0326)	0.735
Among girls
No mobile phone access (Reference)	0		0		0		0
Mobile phone access	–0.0055 (–0.0300 to 0.0190)	0.660	0.0135 (–0.0237 to 0.0508)	0.477	–0.0144 (–0.0454 to 0.0166)	0.362	0.0053 (–0.0257 to 0.0363)	0.738
Among adolescents not living with HIV
No mobile phone access (Reference)	0		0		0		0
Mobile phone access	0.0021 (–0.0372 to 0.0413)	0.917	–0.0313 (–0.0810 to 0.0185)	0.218	0.0226 (–0.0227 to 0.0679)	0.328	0.0008 (–0.0401 to 0.0418)	0.968
Among adolescents living with HIV
No mobile phone access (Reference)	0		0		0		0
Mobile phone access	–0.0102 (–0.0355 to 0.0151)	0.428	0.0292 (–0.0024 to 0.0608)	0.071	0.0059 (–0.0262 to 0.0380)	0.719	0.0015 (–0.0292 to 0.0322)	0.923
Model 4 (Social media use and health content use vs. no mobile phone access)
Among boys
No mobile phone access (Reference)	0		0		0		0
Social media use	0.0659 (0.0117 to 0.1201)	0.017	0.0431 (–0.0049 to 0.0912)	0.078	0.0441 (–0.0057 to 0.0940)	0.082	0.0035 (–0.0450 to 0.0520)	0.888
Health content use	–0.0683 (–0.1150 to –0.0216)	0.004	–0.0683 (–0.1143 to –0.0222)	0.004	0.0759 (0.0145 to 0.1373)	0.015	0.0106 (–0.0492 to 0.0703)	0.729
Among girls
No mobile phone access (Reference)	0		0		0		0
Social media use	0.0069 (–0.0197 to 0.0334)	0.611	0.0617 (0.0182 to 0.1051)	0.005	–0.0061 (–0.0390 to 0.0268)	0.716	0.0096 (–0.0239 to 0.0430)	0.575
Health content use	–0.0340 (–0.0621 to –0.0059)	0.018	–0.0786 (–0.1250 to –0.0322)	0.001	–0.0284 (–0.0689 to 0.0121)	0.169	0.0243 (–0.0199 to 0.0684)	0.281
Among adolescents not living with HIV
No mobile phone access (Reference)	0		0		0		0
Social media use	0.0059 (–0.0349 to 0.0467)	0.775	–0.0208 (–0.0729 to 0.0314)	0.435	0.0179 (–0.0283 to 0.0642)	0.447	–0.0132 (–0.0543 to 0.0280)	0.531
Health content use	–0.0180 (–0.0714 to 0.0355)	0.510	–0.1002 (–0.1699 to –0.0305)	0.005	0.0503 (–0.0288 to 0.1295)	0.213	0.0874 (0.0087 to 0.1662)	0.029
Among adolescents living with HIV
No mobile phone access (Reference)	0		0		0		0
Social media use	0.0429 (0.0074 to 0.0784)	0.018	0.0939 (0.0533 to 0.1344)	<0.001	0.0114 (–0.0257 to 0.0486)	0.546	0.0200 (–0.0168 to 0.0568)	0.286
Health content use	–0.0641 (–0.0873 to –0.0408)	<0.001	–0.0600 (–0.0944 to –0.0256)	0.001	–0.0100 (–0.0476 to 0.0277)	0.606	–0.0202 (–0.0551 to 0.0146)	0.256

Data are average marginal effects (and 95% CI) from models with interaction between participant’s sex, HIV status and mobile phone access (Model 3) or mobile phone use (Model 4), adjusting for rural residence, informal housing, household poverty and marital status. Data were obtained from mixed-effects logistic regression models on ten imputed data (n = 1353 including 587 boys, 766 girls, 933 living with HIV and 420 not living with HIV).

Appendix 1: STROBE Statement—Checklist of items that should be included in reports of cohort studies

	Item No	Recommendation	Page No
Title and abstract	1	(a) Indicate the study’s design with a commonly used term in the title or the abstract	1
		(b) Provide in the abstract an informative and balanced summary of what was done and what was found	2
Introduction
Background/rationale	2	Explain the scientific background and rationale for the investigation being reported	3
Objectives	3	State specific objectives, including any prespecified hypotheses	3
Methods
Study design	4	Present key elements of study design early in the paper	4
Setting	5	Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection	4/5
Participants	6	(a) Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up	4/5
		(b) For matched studies, give matching criteria and number of exposed and unexposed	N/A
Variables	7	Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable	4/6
Data sources/ measurement	8*	For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group	4/6
Bias	9	Describe any efforts to address potential sources of bias	6
Study size	10	Explain how the study size was arrived at	5
Quantitative variables	11	Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why	6/7
Statistical methods	12	(a) Describe all statistical methods, including those used to control for confounding	6/7
		(b) Describe any methods used to examine subgroups and interactions	6/7
		(c) Explain how missing data were addressed	6
		(d) If applicable, explain how loss to follow-up was addressed	6
		(e) Describe any sensitivity analyses	6/7
Results
Participants	13*	(a) Report numbers of individuals at each stage of study—eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed	7/8
		(b) Give reasons for non-participation at each stage	5/6
		(c) Consider use of a flow diagram	5
Descriptive data	14*	(a) Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders	7/8
		(b) Indicate number of participants with missing data for each variable of interest	N/A
		(c) Summarise follow-up time (e.g., average and total amount)	N/A
Outcome data	15*	Report numbers of outcome events or summary measures over time	8/9
Main results	16	(a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included	10/11
		(b) Report category boundaries when continuous variables were categorized	10/11
		(c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period	N/A
Other analyses	17	Report other analyses done—e.g., analyses of subgroups and interactions, and sensitivity analyses	10/13
Discussion
Key results	18	Summarise key results with reference to study objectives	13/15
Limitations	19	Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias	13/15
Interpretation	20	Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence	13/15
Generalisability	21	Discuss the generalisability (external validity) of the study results	14/15
Other information
Funding	22	Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based	19

*Give information separately for exposed and unexposed groups.

Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is available at http://www.strobe-statement.org.

Appendix 2: Comparison between participants interviewed in both waves (2 and 3) and those who dropped out of the study (between wave 2 and 3)

	Total (N=1,410)	Included (N=1,353)	Excluded (N=57)	p-value
Rural residence	27.4%	27.5%	26.3%	0.85
Informal housing	18.0%	17.8%	22.8%	0.34
Poverty	32.8%	32.9%	29.8%	0.63
Age (in years)	13.8 (2.9)	13.8 (3.0)	14.5 (2.7)	0.062
Sex				0.85
Male	43.3%	43.4%	42.1%
Female	56.7%	56.6%	57.9%
ALHIV	69.4%	69.0%	80.7%	0.059
In a relationship	29.3%	29.3%	29.8%	0.93
School enrolment	94.3%	94.4%	93.0%	0.65
Mobile phone use				0.015
No access	49.6%	49.8%	43.9%
Social media only	33.3%	32.6%	49.1%
Health content use	17.1%	17.6%	7.0%
Sex after substance use	7.2%	7.1%	9.8%	0.46
Multiple sexual partnerships	15.1%	15.2%	12.3%	0.54
Unprotected sex	14.7%	14.5%	19.3%	0.31
Inequitable sexual partnerships	10.7%	10.8%	5.9%	0.26

Appendix 3: Missing data imputations

	Missing – N (%)
Wave 1
Female	0 (0%)
ALHIV	0 (0%)
Age	0 (0%)
Rural residence	0 (0%)
Informal housing	1 (0.07%)
Poverty	1 (0.07%)
In a relationship	0 (0%)
Wave 2, 2015–2017
Mobile phone use	0 (0%)
Sex after substance use	183 (13.53%)
Unprotected sex	0 (0%)
Multiple sexual partnerships	0 (0%)
Inequitable sexual partnership (defined as self-report of sexual intercourse with partners 5+ years older than participants or in exchange of material gifts.)	52 (3.84%)
Wave 3, 2017–2018
Mobile phone use	0 (0%)
Sex after substance use	88 (6.50%)
Unprotected sex	0 (0%)
Multiple sexual partnerships	0 (0%)
Inequitable sexual partnership	116 (8.57%)

Determining whether the data on outcomes are MCAR

Little's χ² test indicated that data were not missing completely at random Chi-square = 33.97 (with degrees of freedom 11) in wave 2 and 58.02 in wave 3 (with degrees of freedom 11)

Appendix 4: Spearman’s bivariate correlation coefficients (95% CI; p-value) between outcome variables. Below the diagonal (value 1) are between-participants correlations and above the diagonal are within-participants correlations

	Sex after substance use	Unprotected sex	Multiple sexual partnerships	Inequitable sexual partnerships
Sex after substance use	1	0.06 (0.02 to 0.10); p=0.003	0.16 (0.12 to 0.20); p<0.001	0.04 (0.00 to 0.08); p=0.043
Unprotected sex	0.25 (0.20 to 0.30); p<0.001	1	0.02 (-0.02 to 0.06); p=0.288	0.01 (-0.03 to 0.05); p=0.680
Multiple sexual partnerships	0.44 (0.40 to 0.48); p<0.001	0.38 (0.34 to 0.43); p<0.001	1	0.17 (0.13 to 0.21); p<0.001
Inequitable sexual partnerships	0.27 (0.22 to 0.32); p<0.001	0.39 (0.34 to 0.43); p<0.001	0.47 (0.42 to 0.51); p<0.001	1

Appendix 5: Summary of statistical models

Model	Main association	Adjustment variables
Model 1	Average marginal effects from the multivariable mixed-effects logistic regression of individual sexual risk behaviours outcomes on mobile phone access (access vs. no access).	Participant’s sex, mean-centred age, HIV status, relationship status, and household poverty, rural residence, informal housing, school enrolment, and survey wave.
Model 2	Average marginal effects from the multivariable mixed-effects logistic regression of individual sexual risk behaviours outcomes on mobile phone use (social media use alone and health content use vs. no access).	Participant’s sex, mean-centred age, HIV status, relationship status, and household poverty, rural residence, informal housing, school enrolment, and survey wave.
Model 3	Average marginal effects from the multivariable mixed-effects logistic regression of individual sexual risk behaviours outcomes on mobile phone access (access vs. no access), participant’s sex, interaction between ownership, participant’s sex, and HIV status. Model 3 presents the average marginal effects from the association between mobile phone access and individual sexual risk behaviours outcomes for boys, girls, adolescents living with HIV, and not living with HIV.	Participant’s mean-centred age, relationship status, and household poverty, rural residence, informal housing, school enrolment, and survey wave.
Model 4	Average marginal effects from the multivariable mixed-effects logistic regression of individual sexual risk behaviours outcomes on mobile phone use (social media use alone and health content use vs. no access), participant’s sex, interaction between ownership, participant’s sex, and HIV status. Model 4 presents the average marginal effects from the association between mobile phone use and individual sexual risk behaviours outcomes for boys, girls, adolescents living with HIV, and not living with HIV.	Participant’s mean-centred age, relationship status, and household poverty, rural residence, informal housing, school enrolment, and survey wave.

Appendix 6a: Sensitivity analyses. Multivariable association between mobile phone access and use and self-reported sexual risk behaviours. Results from mixed-effects logistic regression models on complete cases

	Sex after substance use		Unprotected sex		Multiple sexual partnerships		Inequitable sexual partnerships
	Adjusted average marginal effects (95% CI)	p–value	Adjusted average marginal effects (95% CI)	p–value	Adjusted average marginal effects (95% CI)	p–value	Adjusted average marginal effects (95% CI)	p–value
Model 1 (Mobile phone access vs. no access)
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Mobile phone access	–0.0065 (–0.0271 to 0.0140)	0.551	0.0062 (–0.0201 to 0.0335)	0.611	0.0153 (–0.0109 to 0.0411)	0.242	0.0005 (–0.0247 to 0.0250)	0.961
Model 2 (Social media alone and health content vs. no mobile phone access)
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Social media alone	0.0252 (–0.0021 to 0.0482)	0.077	0.0456 (0.0142 to 0.0767)	0.005	0.0184 (–0.0105 to 0.0465)	0.209	0.0017 (–0.0251 to 0.0285)	0.905
Health content use	–0.0538 (–0.0746 to –0.0331)	<0.001	–0.0758 (–0.1072 to –0.0445)	<0.001	0.0107 (–0.0220 to 0.0433)	0.534	–0.0016 (–0.0322 to 0.0291)	0.919
N	1332		1351		1351		1333

Appendix 6b: Sensitivity analyses. Comparisons of the multivariable association between mobile phone access and use and self-reported sexual risk behaviours for boys, girls, adolescents living with and without HIV. Results from mixed-effects logistic regression models on complete cases

	Sex after substance use		Unprotected sex		Multiple sexualpartnerships		Inequitable sexualpartnerships
	Adjustedaveragemarginaleffects(95% CI)	p–value	Adjustedaveragemarginaleffects(95% CI)	p–value	Adjustedaveragemarginaleffects(95% CI)	p–value	Adjustedaveragemarginaleffects(95% CI)	p–value
Model 3 (Mobile phone access vs. no access)
Among boys
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Mobile phone access	–0.0064 (−0.0453 to 0.0317)	0.731	–0.0021 (−0.0399 to 0.0352)	0.899	0.0546 (0.0125 to 0.0972)	0.011	–0.0066 (−0.0460 to 0.0328)	0.742
Among girls
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Mobile phone access	–0.0062 (−0.0313 to 0.0187)	0.618	0.0125 (−0.0250 to 0.0487)	0.537	–0.0140 (−0.0451 to 0.0169)	0.359	0.0051 (−0.0257 to 0.0361)	0.659
Among adolescents not living with HIV
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Mobile phone access	0.0004 (−0.0390 to 0.0399)	0.909	–0.0355 (−0.0853 to 0.0141)	0.151	0.0243 (−0.0212 to 0.0699)	0.296	0.0007 (−0.0401 to 0.0417)	0.919
Among adolescents living with HIV
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Mobile phone access	–0.0105 (−0.0356 to 0.0150)	0.422	0.0293 (−0.0024 to 0.0609)	0.071	0.0063 (−0.0258 to 0.0380)	0.711	0.0017 (−0.0293 to 0.0321)	0.923
Model 4 (Social media use and health content use vs. no mobile phone access)
Among boys
No mobile phone access (Reference)	0	.	0	.	0	.	0	.
Social media alone	0.0655 (0.0113 to 0.1197)	0.015	0.0432 (−0.0052 to 0.0914)	0.077	0.0451 (−0.0045 to 0.0953)	0.073	0.0040 (−0.0447 to 0.0526)	0.873
Health content use	–0.0681 (−0.1145 to −0.0217)	0.004	−0.0684 (−0.1143 to −0.0225)	0.002	0.0775 (0.0160 to 0.1391)	0.014	0.0107 (−0.0491 to 0.0704)	0.723
Among girls
No mobile phone access (Reference)	0	.	0	.	0	.	0
Social media alone	0.0066 (−0.0203 to 0.0335)	0.630	0.0598 (0.0162 to 0.1034)	0.007	−0.0062 (−0.0389 to 0.0265)	0.710	0.0094 (−0.0241 to 0.0429)	0.582
Health content use	−0.0349 (−0.0631 to −0.0065)	0.016	−0.0803 (−0.1267 to −0.0340)	<0.001	−0.0273 (−0.0677 to 0.0131)	0.186	0.0242 (−0.0199 to 0.0683)	0.282
Among adolescents not living with HIV
No mobile phone access (Reference)	0	.	0	.	0	.	0
Social media alone	0.0046 (−0.0362 to 0.0455)	0.825	−0.0248 (−0.0771 to 0.0276)	0.354	0.0196 (−0.0269 to 0.0661)	0.409	−0.0133 (−0.0545 to 0.0280)	0.528
Health content use	−0.0196 (−0.0728 to 0.0335)	0.470	−0.1035 (−0.1731 to −0.0339)	0.004	0.0534 (−0.0266 to 0.1334)	0.191	0.0876 (0.0089 to 0.1663)	0.029
Among adolescents living with HIV
No mobile phone access (Reference)	0	.	0	.	0	.	0
Social media alone	0.0406 (0.0009 to 0.0700)	0.018	0.0943 (0.0533 to 0.1349)	0.001	0.0113 (−0.0258 to 0.0485)	0.551	0.0201 (−0.0167 to 0.0569)	0.284
Health content use	−0.0646 (−0.0876 to −0.0410)	0.001	–0.0601 (−0.0945 to –0.0257)	<0.001	–0.0092 (−0.0467 to 0.0284)	0.633	–0.0202 (−0.0550 to 0.0147)	0.256
N	1332		1351		1351		1333

Data sharing and data availability statement

Prospective users, policymakers/government agencies/researchers (internal/external) will be required to contact the study team to discuss and plan the use of data. Research data will be available on request subject to participant consent and having completed all necessary documentation. All data requests should be sent to the Elona Toska ([email protected]) or William Rudgard ([email protected]).