Parenting for Lifelong Health: a pragmatic cluster randomised controlled trial of a non-commercialised parenting programme for adolescents and their families in South Africa

Introduction

Investing in the future of the world’s 1.2 billion adolescents is now a pressing international agenda.1 Consistent parental supervision and positive involvement predict higher life expectancy, and lower risk behaviours, substance use and violence exposure.2–4 In contrast, abusive or neglectful caregiving increases risks of cancer, mental health problems, substance use, HIV infection and future violence victimisation and perpetration.5–9 Nine out of 10 youth—a billion adolescents—live in low-income or middle-income countries (LMIC).10 Challenges with raising adolescents are reported internationally, but disengagement and conflict increase when families face stressors such as extreme poverty, civil violence and illness. Child maltreatment and interpersonal violence outside of the home rise during adolescence,11 with the highest rates globally in the WHO African region.12 If we are to meet Sustainable Development Goals 3 (health), as well as 5 (gender equality) and 16 (violence prevention), and related goals, it will be essential to support the families in which adolescents grow up.

Consequently, international agencies and governments have identified an urgent need for evidence-based programmes to improve parenting and reduce violence against adolescents.13 Parenting programmes based on social learning theory and group problem-solving show promise,14 but existing research is almost entirely from high-income countries, and focused on younger children.15 16 Reviews show no published randomised trials of parenting programmes for adolescents in LMIC,17 18 although a recent trial in Thailand showed improved parent–child interactions and family cohesion.19 One unpublished trial of a combined voluntary savings and parenting programme in Burundi found reduced caregiver-reported harsh physical and verbal discipline, but no differences in positive discipline, family functioning, child well-being, mental health or problem behaviours.20 There are both cost and contextual barriers to implementing and testing parenting programmes in less developed settings. In high-income countries, several well-evidenced parenting programmes for younger children21 22 have been commercialised, with prohibitive costs for materials, training and accreditation. Other programmes require qualified health professionals for implementation, or use technological components (ie, video or internet) that may be inaccessible in low-resource contexts. Globally, those most in need of parenting support are missing out.

In response, the ‘Parenting for Lifelong Health’ initiative was launched in 2012, as a small, evidence-building collaboration between academics, students and colleagues at the WHO, Unicef and non-governmental organisations (NGOs). Its goal was to develop and test in randomised trials a suite of non-commercialised parenting programmes for low-resource settings. Programmes would not require professionals, videos, equipment or participant literacy, and would be freely available. Partners committed to ‘never-profit’ open-license the programmes if found effective. In this paper, we describe the pragmatic randomised controlled trial of the adolescent programme, ‘Sinovuyo Teen’, (‘we have joy’ in Xhosa). The programme was developed and adapted using a four-stage testing process from 2012 to 2016.23 Testing took place in South Africa—a country with high levels of violence against adolescents in both home and community settings.24 Study sites were in the Eastern Cape province, with the country’s lowest gross domestic product, high HIV prevalence, poor service access and infrastructure, and shortages of electricity and water. A partnership to plan the development and testing process was established between academics (Oxford University and the University of Cape Town), local NGOs (Clowns Without Borders South Africa, Unicef South Africa) and government (national and provincial Departments of Social Development and Basic Education). To reflect real-world conditions, programme implementation was by trained community members. Throughout testing, recruitment of families was community led, with no exclusion of concurrent conditions such as substance use, mental health problems, HIV infection and AIDS, and intimate partner violence.

In 2012, a first prototype manual was developed using systematic reviews of effective components of parenting programmes.8 9 This was revised after qualitative research,25 and input from 50 academic and programming experts. In 2013, version 2 was pilot-tested in a pre-post trial (n=60) and adapted.26 In 2014, version 3 was tested in a larger pre-post trial (n=230).27 These initial (non-controlled) tests showed no iatrogenic effects, significant reductions in harsh and abusive parenting, and improvements in parent and adolescent outcomes. Following participant input, the programme was extended to include sessions on (A) planning to protect adolescents from violence and exploitation in the community and (B) family financial management. In this trial, we tested the efficacy of the final version of Sinovuyo Teen. We hypothesised that the parenting programme would reduce abuse and improve parenting at the cluster level. We also hypothesised that it may have further effects on caregiver and adolescent well-being (eg, mental health, substance use, social support) and family outcomes (eg, economic welfare, risk avoidance planning).

Methods

Study design and participants

In this pragmatic cluster randomised controlled trial, we selected 40 communities (located in 34 rural villages and 3 large periurban townships) within a 2-hour drive of a rural town (the research team’s base) in South Africa’s Eastern Cape. Local traditional and political leaders agreed to participation for all communities, and the programme was presented as support for families in raising adolescent children. To reflect recruitment in real-world community-based services, families with an adolescent aged 10–18 years were identified by a range of sources, including self-referrals, local chieftains, community-selected representatives, schools, social services and local NGOs, as well as door-to-door visits to find families with adolescents. Between April and August 2015, around 1910 households were assessed for eligibility and of these 960 had resident adolescents. Families completed two screening questions asking, ‘do you and your teen argue and shout a lot?’ and ‘do you sometimes end up hitting your teen when things are really stressful?’ Six hundred and twenty families responded positively to at least one question. These were visited by the trained local community programme team and asked whether they could attend daytime programme workshops (after dusk there are high crime levels in these communities). Five hundred fifty-two families were included in the trial and completed baseline assessments by both adolescents and caregivers. Due to high levels of orphaning and fostering in South Africa, there were no requirements for a biological relationship between adolescent and primary caregiver but they had to reside in the same dwelling at least four nights per week. Adolescents who had learning difficulties so severe that they were unable to provide informed consent were not included. No other exclusion criteria applied (figure 1).

Request permission

Figure 1

Trial profile. CG, caregiver; ITT, intention to treat.

We used a sample size calculation to determine the number of clusters required. The calculation was based on the primary outcome of child abuse. We estimated intracluster correlations (ICC) of 0–0.08 based on pilot-testing, and note that there are large numbers of zeros in reporting of child abuse in any pragmatically recruited sample. Using Optimal Design software, 40 equal clusters with 12 families per cluster were required for a minimum detectable effect size of 0.36 for desired power of 0.80 with two-tailed P<0.05 and ICC of 0.08. We note that this study required a substantially higher effect size than found in a recent meta-analysis of parenting programmes for prevention of child abuse, which showed average programme effects of 0.20,28 thus potentially underestimating our programme effects.

There were no monetary incentives for participation, although families received snacks at pretest and small food parcels at post-test as thanks. An independent trial steering committee oversaw trial conduct. The protocol was published29 and the trial was registered on the Pan-African Clinical Trials Registry PACTR201507001119966 on 27 April 2015. Written informed consent was obtained from all adolescent participants and their primary caregivers. Confidentiality was maintained, unless participants were at risk of significant harm or requested assistance. If participants reported severe abuse, rape, recent suicide attempts or other significant harm, immediate referrals and follow-up support were made to child protection and health services (21 referrals, 10 in intervention arm and 11 in control arm). The Consolidated Standards of Reporting Trials (CONSORT) 2010 checklist for cluster randomised trials and the CONSORT 2008 extension for pragmatic trials were used for reporting.

Results

The trial was completed on 18 August 2016, when the 5–9 month outcome assessment ended (figure 1). Baseline characteristics are shown in table 2, baseline characteristics for secondary outcomes are shown in online supplementary table 1. Caregivers were 95% female, with a mean age of 49 (SD 14.7). Forty-two per cent were biological parents and 28% were grandparents. Adolescents were 44% female, with a mean age of 13.8 years (SD 2.39). Two-thirds of families reported no household employment, and 77% lacked food in the home for at least 1 day during the past week (caregiver report).

Caregivers in the intervention arm attended an average of 50% and adolescents 64% of all sessions. Nine per cent of caregivers and 5% of adolescents attended no sessions, but all, except four families, received brief home catch-ups. Retention rates at endpoint in the intervention group were 263 (97%) of caregivers and 260 (96%) of adolescents, and in the control group 277 (98%) of caregivers and 270 (96%) of adolescents. There were no differences in attrition between treatment arms. This population has high rates of migration, and 53 caregiver–adolescent dyads were no longer living together at endpoint, although all were included in ITT analyses. Caregivers lost to follow-up were more likely to be HIV positive and reported less involved parenting compared with those not lost to follow-up. There were no differences in adolescents lost and retained at follow-up (online supplementary table 2).

The primary outcomes are summarised in table 3. Abuse, positive parenting, parental supervision and involved parenting were measured at 1 month and 5–9 months postintervention. Neglect, corporal punishment and inconsistent discipline were only measured at 5–9 months.

Discussion

This is the first known randomised controlled trial of a parenting programme for adolescents in Africa. It found that a low-resource programme, implemented by trained community members, has a range of positive outcomes. On both caregiver and adolescent reports, those receiving the parenting programme had reduced abuse (at least in the short term), improved involved parenting and parental supervision, improved household economic welfare and financial management, improved family planning to avoid adolescent violence victimisation in the community and reduced substance use among both caregivers and adolescents. Caregivers also reported reduced depression and stress, less attitudes condoning corporal punishment and improved social support. The programme did not improve all aspects of parenting, nor did it reduce adolescent depression or behaviour problems in adolescent report. However, the study showed positive intervention impacts of the programme on a range of parenting, family, caregiver and adolescent outcomes 5–9 months after the end of the intervention, and no harmful effects. These findings demonstrate the promise of this parenting programme in an LMIC setting.

It is important to note a number of limitations in this study. First, the trial was conducted by the programme developers, and future studies should be conducted independently. Second, as is usual in trials of parenting programmes, blinding of participants and data collectors was limited by participants talking about the programme. Third, as is standard in the parenting programme evidence for older children and adolescents, self-report data were used. It is generally recognised that observations would be equally prone to measurement error, and this study took a number of measures to improve reliability of reporting. In addition to using tablet-based ACASI for sensitive items, we collected both caregiver and adolescent report for shared outcomes of parenting, family processes and economic welfare. All interviews were conducted separately and in private, and with different interviewers for caregivers and adolescents. Six measures were only reported by caregivers (eg, caregiver depression, caregiver substance use) and five measures were only reported by adolescents (eg, adolescent depression, adolescent substance use). Fourteen measures were reported by both caregivers and adolescents (eg, all parenting measures, shortages of essential goods in the household, corporal punishment). Of these 14 shared measures, 10 concurred between caregivers and adolescents and 4 showed differences in significance of effects (abuse at 5–9 months, positive parenting at 5–9 months, corporal punishment at 5–9 months and positive parenting at 1 month and 5–9 months). It is notable that these differences in caregiver-adolescent report were driven by differential reporting in the control group. In the intervention group, both adolescents and caregivers reported reductions in abuse, corporal punishment and improved positive parenting. In the control group, caregivers reported no or lesser improvements, while the control group adolescents reported improvements equal to those in the intervention group. Response bias must always be considered in any self-report measure, but it is unclear why the control group reports differed on these measures (and not others). We note that the International Rescue Committee (IRC) reported the same patterns in reporting of harsh punishment in the parenting trial in Thailand.19

Fourth, qualitative data and quantitative findings suggest that there may have been a ‘Hawthorne effect’ of the research process. Control group participants commented in focus group discussions that the pretest and post-test interviews were helpful to them, and in particular that a sympathetic researcher asking them about family relationships had prompted them to reflect on and improve these. Similar reports have been noted in trials of other adolescent and family interventions, especially in settings with very low service access where an interview about parenting may be the only ‘intervention’ ever received.45

Fifth, the study was not able to conduct follow-up beyond 9 months. A recent review of parenting in conflict settings highlights that impacts on child outcomes may be delayed when mediated through improved parenting practices.46 Further follow-ups could valuably assess parent and adolescent outcomes over time and into young adulthood. Sixth, the trial did not measure impacts of the programme on other adults or children within the households or within the wider communities, and future research should test for such potential effects. Finally, during the pilot-testing stages, there were unexpected high rates of community-level dissemination. For example, villages established additional Sinovuyo groups, and programme messages were disseminated widely by local pastors and school principals. The trial stage had been originally planned as an individually randomised trial, but high risk of contamination due to the expansion of the programme within communities consequently required a change to cluster randomisation at community level. Although the community-level dissemination demonstrated high programme acceptability, the subsequent clustered trial was only powered to detect effects that were substantially larger than the average for parenting programmes, thus potentially underestimating programme effectiveness.

The trial also has a number of strengths. It used standardised outcome measures with a 5–9 month follow-up period. In contrast, recent reviews identify that the majority of parenting programme trials use only immediate postintervention or 1-month follow-ups. We used robust cluster randomisation methods and ITT analyses. We measured actual acts of abusive behaviour, while many programmes instead measure proxies such as parental depression and stress or attitudes towards corporal punishment.14 Another key strength of the study is its external validity. In order to reflect real-world service delivery in LMIC, we explicitly used pragmatic randomised trial methods. These included recruitment methods typical of NGO and government services, and an intervention implemented by a local NGO in community settings, using non-professional staff and with no participant exclusion criteria (apart from learning difficulties too severe to allow consent). These pragmatic methods increase the generalisability of the findings and their applicability to programming. Indeed, the intervention and trial were conducted during sustained political and civil violence in research sites. During the study period, most research sites were without electricity or water on multiple days of the week. This suggests that the programme—and robust methods to test it—may be feasible even in very constrained contexts.

This trial also highlights the potential impacts of collaboration between science and policy. International agencies, researchers, local NGOs and local leaders worked in close partnership throughout development and testing. This engagement increased the relevance of the research to policymakers and programmers. There is a strongly recognised need for evidence-based, non-commercialised parenting programmes for LMIC, and the Sinovuyo Teen programme is currently being adapted and taken to scale by a number of national governments, international, regional and local NGOs within Africa. By 2020, an estimated 200 000 families will receive the programme within Democratic Republic of Congo, Lesotho, South Africa, South Sudan, Uganda, Tanzania and Zimbabwe. Further countries planning to scale up the programme include Afghanistan, Haiti, Israel, Lithuania and the Philippines.

This raises further research questions. Although effectiveness on many outcomes was shown in this South African study, we do not know the extent to which these findings are generalisable to other countries and regions. In two of the countries undertaking scale-up, randomised trials are planned by implementing agencies. Each country has adapted the programme for local languages and cultures, and some have added components such as menstrual hygiene, child labour information or HIV prevention education. Furthermore, versions of the programme are being implemented with diverse groups, such as deinstitutionalised children and adolescent children of sex workers. It will be important to test whether and how such adaptations and different recipients may affect programme impacts in differing cultural and country contexts.47 It will also be important to understand whether there are differential effects of the programme on highest risk families such as those experiencing HIV/AIDS or intimate partner violence. Future moderator analyses are required in order to examine subgroup differences across country settings. The International Committee of Medical Journal Editors recently called for individual participant data sharing to be normed for clinical trials.48 Sharing of data across trials of parenting programmes in LMIC could provide substantive value.

Findings of this trial can also inform our understanding of processes of family strengthening in low-resource contexts. This programme shares many common elements with other rigorously evaluated programmes, such as the IRC’s interventions for children in postconflict settings, Families Matter!49 and other Parenting for Lifelong Health interventions for infants, toddlers and young children.50 51 Qualitative feedback suggests that collaborative learning and non-blaming approaches may be key. Trying out skills at home and having opportunities to problem-solve challenges within a supportive group may enhance caregivers’ sense of agency.52 Importantly, programmes aim to capitalise on caregivers’ already-held aspirations of how they would like to parent, and families identify their own goals. There may also be important practical elements—for example, in areas with high burden of HIV illness or other disease, home visit catch-up sessions may be necessary in order to ensure programme access for affected families. Further research is required in order to understand the mechanisms of change by which a parenting programme can work in LMIC settings, and future mediation analyses of possible pathways of change would be of value—for example, the potential roles of reduced economic hardship, reduced substance use and caregiver depression in improving family relationships.

In conclusion, this pragmatic cluster randomised trial demonstrates that a parenting programme showed improvements across a range of parenting, family and violence prevention outcomes. It showed reduced emotional and physical abuse at immediate post-test, with possible longer term effects. There were no impacts on positive parenting, neglect or inconsistent discipline but the programme showed improved positive involved parenting and parental supervision, which evidence suggests may be particularly important in reducing HIV risk behaviour among adolescents. The programme showed no impact on adolescent depression, behavioural problems or exposure to community-level violence at 5–9 months postintervention. However, the trial showed increased family communication about reducing risks for adolescents in community settings, caregivers had reduced depression and parenting stress—both of which are strongly associated with child outcomes in the parenting literature—as well as improved social support and reduced attitudes condoning corporal punishment. Both caregivers and adolescents reported reduced alcohol and other substance use. These suggest a strengthening of caregiving capacities and lowering of family risks 5–9 months after the programme ends. In addition, families showed improvements in financial self-efficacy and planning, and reported direct impacts of improved budgeting, namely reductions in month-end shortages of basic essentials. It is possible that there may be particular value to including budgeting elements within family interventions.

The trial also demonstrates that positive effects are possible even in a high-deprivation area and during a period of sustained and violent civil unrest. All manuals and programme tools are freely available online, and a number of regional NGOs are establishing skills in training and supervision. The Global Partnership to End Violence against Children provides further guidance on Parenting for Lifelong Health programmes in their INSPIRE (Seven strategies for Ending Violence against Children package) package. Further research is essential, but this study is a step towards closing the global gap in evidence-based parenting support for adolescents.