Adolescents in low- and middle-income countries (LMIC) such as South Africa bear a disproportionate burden of abuse and violence [1, 2]. Globally, an estimated 95,000,000 children experience abuse annually, with the highest rates in the WHO Africa region [1]. In 2015, the first representative study in South Africa reported lifetime rates of 34 % physical abuse, 16 % emotional abuse and 20 % sexual abuse amongst 15–17 year olds [3]. Although the evidence-base is scattered, studies show that pathways of poverty, caregiver¹ mental health distress and HIV/AIDS contribute to harsh parenting and maltreatment [2]. There is clear evidence that exposure to abuse has severe and lasting adverse effects on adolescents’ health, education, employment, mortality and subsequent risk of HIV-infection [4, 5]. In the context of Africa’s population ‘youth bulge’ [6] and emerging evidence of continued brain development - neural, functional and emotional - during adolescence [7], it is increasingly clear that later childhood development is an important time for intervention.

In 2009, a systematic review of reviews identified parenting programmes as having a promising evidence-base for child abuse prevention [8], but also found a lack of research from LMIC. There is a small evidence-base in LMIC for parenting programmes targeted at younger children, with two emerging studies for children under 10 years-old in South Africa [9] and Liberia [10]. For adolescents, the evidence-base is severely lacking: to date there are no known published studies of parenting programmes for adolescent abuse prevention in LMIC [11].

A number of parenting programmes have shown effectiveness in child abuse prevention and reduction, and good transportability within high-income countries. These include The Incredible Years, Triple P, Parent–child Interaction Therapy (PCIT) and Parent Management Training Oregon (PMTO) (all for pre-teens) and the Nurse-Family Partnership (for pregnancy and infants) [12, 13]. All use social learning theory approaches, build parenting skills in behaviour management and promote family problem-solving. However, many existing evidence-based parenting programmes charge fees for training, manuals and equipment that make them prohibitive for low-resource contexts. Others are only able to be implemented by qualified nurses or psychologists - professions in extremely short supply in LMIC, and none are designed for adolescents.

In response to these challenges for LMIC, an international collaboration, ‘Parenting for Lifelong Health’ (http://tiny.cc/whoPLH) was initiated in 2012, with the World Health Organization (WHO), UNICEF, and academics from the global South and North. This collaboration is developing and rigorously testing a suite of child abuse prevention programmes for different child developmental stages [14], with support from many donor partners, LMIC governments and PEPFAR-USAID. Programmes are designed for low-resource contexts, are implemented by lay community workers in local NGOs or government services and require little or no equipment, nor availability of electricity. If found effective, programmes will become freely available under licensing that prohibits any profit-making or commercial interests.

This study concerns the adolescent programme. Development, testing and adaptation stages are iterative, using the Medical Research Council’s (MRC) model for intervention development [15]. First, systematic reviews were used to identify shared core components of effective parenting programmes [12], and the evidence for transportability of parenting programmes to low- and middle-income contexts [11, 13]. Second, draft manuals were co-developed by academics (Universities of Oxford and Cape Town) and NGO partners (Clowns Without Borders South Africa). Third, over fifty international academics, practitioners and programming experts provided comments and advice on initial drafts. Fourth, six months of in-depth qualitative research in South Africa identified adaptation needs such as including information on child rape, which is common in South Africa. Fifth, an initial pilot stage was conducted (n = 60 participants in rural South Africa), showing high levels of acceptability, pre-post test improvements in parent and adolescent outcomes, and no iatrogenic effects. Subsequently, the programme was adapted based on small-scale pre-post test results, observer-coding and qualitative feedback from participants and staff.

The current study comprises the next major step in programme testing: a larger pre-post trial in order to test indicative effects, programme acceptability and to identify further adaptation needs in real-world conditions. Programme implementation was delivered by childcare workers within King William’s Town Child and Youth Care Centre, part of Isibindi, a community-based programme founded by the National Association of Child and Youth Care Workers. These local staff were trained and supervised by another NGO, Clowns Without Borders South Africa, itself funded by the Provincial government’s Department of Social Development and UNICEF South Africa. The study took place in six rural and peri-urban communities in one of South Africa’s poorest provinces: the Eastern Cape. The area was selected by the Department of Social Development and UNICEF South Africa, as representative of the country’s most deprived and service-challenged contexts – it is characterized by high levels of HIV, unemployment, crime and by overburdened social and health services [16]. Low levels of infrastructure in rural areas include very limited roads, no electricity and frequent lack of water supplies: such a context is potentially more generalizable to other parts of Southern Africa than the major cities in which programmes are more frequently first tested. Programming and research took place in close partnership with the Provincial and National Departments of Social Development and Education and UNICEF. Aims of the study were: 1) to identify indicative effects of the programme on child abuse and related outcomes; 2) to investigate programme safety for testing in a future randomised controlled trial, and 3) to identify potential programme adaptations for improvement.

This is the first known quantitative study of a child abuse-prevention programme for adolescents in a LMIC, albeit an early stage test of the programme, using a pre-post design without comparison group. Following the intervention, adolescents and caregivers independently reported reductions in child abuse, adolescent problem behaviour, and negative parenting, and improvements in positive and involved parenting, compared to their levels at baseline. Analyses of secondary outcomes showed additional reductions in adolescent and caregiver depression, parenting stress and caregiver substance use as well as improvements in perceived receipt of social support among both adolescents and caregivers. Overall, the study showed no evidence of harmful effects and positive changes on all primary outcomes (8 out of 8) and most secondary outcomes (7 out of 11). Of course it is important to be cautious about pre-post results without a control group, but these findings suggest possible programme benefits in a range of outcomes, and viability of the programme for further testing in a randomised controlled trial.

The study also provides evidence and raises questions about how a parenting support programme may interact with high levels of deprivation in a Southern African context. Challenges to implementation included flooding (making dirt roads in rural areas impassable), outbreaks of xenophobic violence, and tensions surrounding concurrent national elections that resulted in safety concerns for participants, NGO and research staff. It was essential to establish clearly that the programme was not associated with a political party or electioneering, whilst gaining the support and consent of local politicians and leaders – a difficult balance. Participants also reported high levels of poverty, intimate partner violence and HIV/AIDS, and low levels of education and literacy. For example, some children were severely ill throughout much of the programme, and sessions were delivered to them in hospital with visiting families. For other families, very high levels of alcohol or other substance use made engagement difficult and home visits dangerous for staff. Finally, the research was committed to building local capacity rather than bringing expertise from major cities, and as a result training of research and implementation teams was intensive, reflecting the systematic deprivation of educational capacity in ‘homelands’ during apartheid rule. Given these constraints, successful programme completion, high attendance rates and positive outcomes among both caregivers and adolescents were encouraging. These findings can be compared to evidence from high-income countries. For example the Nurse-Family Partnership found that maternal exposure to intimate partner violence diminished overall high programme effectiveness [52], whilst in some trials of Incredible Years parenting intervention in Europe, parents with the highest levels of depression improved most [53].

This pre-post test also provides important lessons for forthcoming stages of a randomised trial and potential scalability. Firstly, participants were included in the programme through a range of community and service referrals and door-to-door visiting – reflecting real-world approaches to identifying families in need in Southern African contexts. This probably resulted in inclusion of families who were less in need of child abuse prevention services, but also certainly resulted in inclusion of families who would often be excluded from initial programme testing, for example those with severe alcohol and drug problems, mental health problems, domestic violence and terminal illness. Second, the findings, measure development and attrition rates of this study have allowed sample size calculations to be made for a forthcoming cluster randomised trial – previously impossible due to lack of any prior studies of prevention of abuse of adolescents in a LMIC. Third, the unanimous support of traditional leaders, provincial and local government, and school principals was essential to implementation feasibility. Although some families initially expressed suspicion, programme acceptability was far higher than expected and staff were approached daily in study sites with requests to join the programme. Indeed, entirely unanticipated, extensive programme diffusion took place within communities, although nothing is known about the quality or fidelity of diffusion activities. Such diffusion has not been recorded in previous literature on parenting programmes. It seems that South African cultural approaches were adapting what had been conceptualized as a small-group intervention into a shared, community-level resource.

The diffusion also had implications for research. It indicated that a planned individually randomised trial would be at high risk for contamination between intervention and control families within the same villages. Consequently, the research design was adjusted to a cluster trial with randomization at a village-level (currently underway, Pan-African Clinical Trials Registry registration number: PACTR201507001119966). It also highlighted the need for qualitative research alongside the trial to better understand the dynamics of a parenting programme in South Africa.

This study raises further questions about the possible mechanisms of change produced by the programme. Skills-building was based on collaborative problem-solving, home practice, and vignettes, following systematic reviews from high-income countries. However, in a complex intervention, we do not know whether these were the vital factors in programme effectiveness. For instance, caregivers reported regular and supportive contact with their ‘Sinovuyo Buddies’ during and beyond the programme, and this ‘unsupervised’ component may have contributed to positive outcomes. Moreover, participants adopted programme songs as reminders of session principles and when staff arrived in villages, children would surround them singing the past week’s song. Thus again, local cultures seemed to have modified, and potentially improved, the programme for their own context.

This study has several limitations. Findings from a pre-post, non-randomised test cannot determine causality and are only indicative of potential programme results. Post-test interviews occurred shortly after intervention and therefore we do not know whether reported changes endured over time, nor the extent to which participants continued to meet with their programme groups or ‘buddies’. It may well be that formal or informal follow-up sessions boost programme effects, but the need for or value of this remains an empirical question in the current context. It would also be valuable in future trials to assess potential impacts on other children and caregivers within study households. Although the study area was chosen for infrastructural challenges and very limited services that may be representative of conditions across Southern Africa, generalizability to other LMIC settings remains unknown and further testing is clearly required. Additionally, there remain questions about how gender, adolescent age and biological/non-biological caregiving relationships may moderate programme effects. Some findings remain unexplained – for example why caregivers reported reductions in substance abuse, but adolescents did not, despite reported reductions in related problems of aggression and delinquency.

Finally, a number of outcomes were analysed, although all were complementary and conceptualised as part of the intervention’s theory of change. Statistical adjustment for multiple testing was consequently not conducted due to concerns of overcorrection and misinterpretation of treatment effects [54], and because tests will be replicated in the forthcoming randomised-controlled trial. Nevertheless, checking with a conservative Bonferroni correction (α = 0.05/19), all effects remained statistically significant except adolescent-reported positive parenting.

Despite these limitations, this study also has notable strengths. External validity is high: local NGO childcare workers (trained by a NGO) ran the programme in low-resource locations without electricity or running water. This pragmatic approach meant that the research team were not able to assure quality or programme fidelity, but reflected real-world conditions of delivery in Southern African contexts. There were no exclusion criteria for participants, reflecting the reality of service provision rather than ideal research conditions. These conditions boost the external validity of findings by reflecting real-world programming in low-resource contexts in Southern Africa. The study also used standardized outcome measures and elicited independent reporting from adolescents and caregivers within each household, bolstering findings that suggest that in low-resource contexts an abuse-prevention programme is feasible and scalable.

This pre-post test has three key findings. First, it demonstrates feasibility and initial effectiveness of a parenting programme to reduce abuse of adolescents in a low-resource setting. Second, it shows that rigorous parenting research is possible in high deprivation contexts and in close partnership with NGOs and government. Third, it adds to the parenting programme literature as the first study of its kind in a LMIC, taking place in a region with some of the world’s highest rates of child abuse [3]. While this study by no means provides definitive answers, it suggests strong possibilities in a global initiative towards evidence-based solutions.