Improving lives by accelerating progress towards the UN Sustainable Development Goals for adolescents living with HIV: a prospective cohort study

Background: Low-income and middle-income countries (LMICs) face major challenges in achieving the UN’s Sustainable Development Goals (SDGs) for vulnerable adolescents. We aimed to test the UN Development Programme’s proposed approach of development accelerators—provisions that lead to progress across multiple SDGs—and synergies between accelerators on achieving SDG-aligned targets in a highly vulnerable group of adolescents in South Africa. Methods: We did standardised interviews and extracted longitudinal data from clinical records at baseline (2014–15) and 18-month follow-up (2016–17) for adolescents aged 10–19 years living with HIV in the Eastern Cape province of South Africa. We used standardised tools to measure 11 SDG-aligned targets—antiretroviral therapy adherence, good mental health, no substance use, HIV care retention, school enrolment, school progression, no sexual abuse, no high-risk sex, no violence perpetration, no community violence, and no emotional or physical abuse. We also assessed receipt at both baseline and follow-up of six hypothesised development accelerators—government cash transfers to households, safe schools (ie, without teacher or student violence), free schools, parenting support, free school meals, and support groups. Associations of all provisions with SDG-aligned targets were assessed jointly in a multivariate path model, controlling for baseline outcomes and sociodemographic and HIV-related covariates, and adjusted for multiple outcome testing. Cumulative effects were tested by marginal effects modelling. Findings: 1063 (90%) of 1176 eligible adolescents were interviewed. Three provisions were shown to be development accelerators. Parenting support was associated with good mental health (odds ratio 2·13, 95% CI 1·43–3·15, p<0·0001), no high-risk sex (2·44, 1·45–5·03, p=0·005), no violence perpetration (2·59, 1·63–4·59, p<0·0001), no community violence (2·43, 1·65–3·86, p<0·0001), and no emotional or physical abuse (2·38, 1·65–3·76; p<0·0001). Cash transfers were associated with HIV care retention (1·87, 1·15–3·02, p=0·010), school progression (2·05, 1·33–3·24, p=0·003), and no emotional or physical abuse (1·76, 1·12–3·02, p=0·025). Safe schools were associated with good mental health (1·74, 1·30–2·34, p<0·0001), school progression (1·57, 1·17–2·13, p=0·004), no violence perpetration (2·02, 1·45–2·91, p<0·0001), no community violence (1·81, 1·30–2·55, p<0·0001), and no emotional or physical abuse (2·20, 1·58–3·17, p<0·0001). For five of 11 SDG-aligned targets, a combination of two or more accelerators showed cumulative positive associations, suggesting accelerator synergies of combination provisions. For example, the fitted probability of adolescents reporting no emotional or physical abuse (SDG 16.2) with no safe schools, cash transfers, or parenting support was 0·25 (0·16–0·34). With cash transfer alone it was 0·37 (0·33–0·42), with safe school alone 0·42 (0·30–0·55), and with parenting support alone 0·44 (0·30–0·59). With all three development accelerators combined, the probability of adolescents reporting no emotional or physical abuse was 0·76 (0·67–0·84). After correcting for multiple tests, four of the SDG-aligned targets (antiretroviral therapy adherence, no substance use, school enrolment, and no sexual abuse) were not associated with any hypothesised accelerators. Interpretation: The findings suggest the UN's accelerator approach for this high-risk adolescent population has policy and potential financing usefulness. Services that simultaneously promote several SDG targets, or combine to support particular targets, might be important to meet not only health-related targets, but also to ensure that adolescents in LMICs thrive within a new development framework. Funding: Nuffield Foundation, UK Research and Innovation Global Challenges Research Fund, UKAID, Janssen Pharmaceutica, International AIDS Society, John Fell Fund, European Research Council, Economic and Social Research Council, Philip Leverhulme Trust, and UNICEF. We did standardised interviews and extracted prospective data from clinical records at baseline (2014–15) and 18-month follow-up (2016–17) from adolescents aged 10–19 years living with HIV in the Eastern Cape province of South Africa. The study was established as the region's first large-scale community-traced cohort of this group, tracking not only clinical outcomes, but also social, educational, familial, sexual health, and community experiences. The province is characterised by high morbidity, low human development, and poor infrastructure. In a district comprising rural, urban, and periurban settlements, we identified all 52 primary clinics, hospitals, and community health centres providing HIV treatment to adolescents. In each facility, all files (paper and computer) were reviewed to identify all individuals aged 10–19 years who had ever initiated HIV treatment. To include those lost to follow-up as well as those retained in care, adolescents were traced to 180 communities and interviewed at home or a location of their choice. At 18 months post baseline, all adolescents who had given consent to be re-approached were asked for consent for follow-up. Because of migration, participants lived in six provinces at follow-up: Eastern Cape, Free State, Gauteng, KwaZulu-Natal, North-West, and Western Cape. Ethical approval was given by the University of Cape Town (Cape Town, South Africa; CSSR 2013/4), Oxford University (Oxford, UK; CUREC2/12-21), Provincial Departments of Health and Education, and all participating health-care facilities. All adolescents and their primary caregivers gave written informed consent at both timepoints in their language of choice (Xhosa or English), which was also read aloud in cases of low literacy. There were no financial incentives, but the study's adolescent advisory group recommended a certificate, snack, and small gift pack including soap and pencils. These were provided to all adolescents, regardless of interview completion. To avoid unintended disclosure of HIV status or HIV-associated stigma, we presented the research focus as general adolescent health and social needs, and 456 neighbouring adolescents were additionally interviewed (not included in analyses). Confidentiality was maintained except in cases of risk of harm. For individuals who had been abused or raped, or showed suicidal planning or behaviour or untreated severe illness (eg, symptomatic tuberculosis), researchers made immediate health and social service referrals with follow-up support. Adolescent clinical records were reviewed, and research staff supported participants to complete tablet-based questionnaires in the language of their choice, lasting 60–90 min. Input to questionnaire design was provided by an adolescent advisory group; the South African National Departments of Health, Social Development, Basic Education and National AIDS Council; UNICEF, PEPFAR-USAID, and local non-governmental organisations. Pre-piloting was done locally with 25 adolescents living with HIV. Given that the study was initiated before 2015, we retrospectively identified 11 adolescent outcomes that were aligned to (or proxies for) specific targets within four SDGs: antiretroviral therapy adherence in the past week, good mental health, no substance abuse, HIV care retention, school enrolment, school progression, no sexual abuse, no high-risk sex, no violence perpetration, no community violence, and no emotional or physical abuse (table 1). Identification of alignment was undertaken in consultation with UNDP, UNICEF, and UNAIDS, followed by literature review to check for possible effects across development domains. All were measured at baseline and 18-month follow-up. SDG targets, definitions, and scales used in this analysis SDG=Sustainable Development Goal. PTSD=post-traumatic stress disorder. α=internal reliability of each scale in this sample. Following evidence that provisions need to be sustained and predictable to benefit adolescents, we assessed receipt of real-world government and community services with high external validity as access at both baseline and follow-up for six hypothesised accelerators (ie, provisions that are positively associated with adolescent outcomes across three or more SDGs). (1) Free school, measured as either a no-fees school or an individual school fees exemption. (2) Government cash transfer, measured using items based on South Africa's census as household receipt of any of child support grant, foster child grant, pension, disability, or care dependency grant. (3) Safe school, measured as no past-year physical violence by teachers or students, using the Social and Health Assessment.26 (4) Access to a support group, measured as regular monthly attendance of a youth-focused or general support group (either HIV-specific or non-specific) in the community or clinic. (5) Free school meals, measured using items drafted with the National Department of Basic Education as provision of either lunch or breakfast at school every day. (6) A high level of parenting support, measured as consistently high parental monitoring using this subscale of the Alabama Parenting Questionnaire.27 Covariates included nine sociodemographic and HIV-related cofactors, all measured at baseline: age (dichotomised as 10–14 or 15–19 years); sex; urban or rural residence; housing (formal or informal shack housing); maternal orphanhood and paternal orphanhood, both measured using items based on South Africa's Census; mode of infection (vertical or horizontal) assessed using clinical files, date of antiretroviral therapy initiation, and parental HIV status;28 general health, self-reported using the WHO International Classification of Functioning, Disability and Health checklist (past 6 months);29 and poverty, measured as access to eight highest socially perceived necessities for children in the nationally representative South African Social Attitudes Survey (enough food, money for school fees, to see a doctor when needed, school uniform, basic clothing, soap, school books, and shoes).30 Analysis took place in six steps in SPSS, STATA, and MPlus. The first three steps were to assess frequencies, sample representativeness, and reliability of outcome measures. In step 1, eligible participants included in the study were compared with those excluded for sociodemographic characteristics that were available for both groups (age, sex, and urban or rural location) using χ2 tests (reported in text). Next, participants who were retained and not retained at follow-up were compared on all variables used in this study. In step 2, frequencies for all SDG-aligned targets, hypothesised accelerator provisions, and covariates were reported. In step 3, self-reported antiretroviral therapy adherence was validated in a multivariable logistic regression against undetectable viral load, controlling for all nine covariates. In step 4, we tested for associations of hypothesised development accelerators with SDG-aligned targets by doing a multivariate path analysis in MPlus that applied the remaining five provisions and nine covariates to all the outcomes simultaneously; two of the 45 paths had to be removed because a sparse and an empty cell prevented model fit. In step 5 we controlled for test multiplicity using the Benjamini-Hochberg method, checking for associations between predictors and outcomes for a false-positive rate of 0·1. We also checked for collinearity. In step 6, we tested for possible cumulative effects between the development accelerators that emerged, using marginal effects models with each combination of the accelerators, holding significant covariates at mean values, and providing probabilities and 95% CIs. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. LDC had full access to all the data in the study and had final responsibility for the decision to submit for publication.