Study design and sample
In this pragmatic cluster randomized controlled trial, the total sample compromised 552 dyads of adolescents and their parents\caregivers (parents\caregivers M = 49.37; SD = 14.69 and adolescents M = 13.84; SD = 2.38) who were recruited from 40 communities (located in 34 rural villages and three large peri-urban townships) in South Africa’s Eastern Cape. 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 for at least four nights per week. Further information about the study design and sample and inclusion, exclusion criteria is available in Cluver et al. [10].
Randomization was stratified by urban location and conducted after baseline using a random number generator by an independent, blinded statistician (CL). Complete randomization within strata used a ratio of 1:1 intervention: control. The sample included 270 families in the intervention arm and 282 families in the control arm (M = 14 families per cluster, SD = 1.9). Blinding of participants and program providers was not feasible for parenting programs.
Ethical approval was given by the University of Oxford (SSD/CUREC2/11–40), University of Cape Town (PSY2014-001), and government Departments of Social Development and Education.
Procedure and data collection
Parents\caregivers and adolescents completed a structured self-report questionnaire at three points in time: pre-test (Baseline), 1 month post-intervention (with a limited sub-set of items) and 5–9 months post-intervention. The analyses of the current study were conducted based on the data at baseline and follow-up tests.
Intervention group
Dyads (parent\caregiver and adolescent) in the intervention group received a 14-session parenting programme called “Parenting for Lifelong Health/Sinovuyo Teen”. Each session lasted for 1–1:30 hours a week. All sessions took place in public and community places such as churches, community halls, schools and under trees.
Based on Social Learning Theory [28], the programme involves parenting principles, such as praising each other, managing anger and stress, joint problem-solving, non-violent discipline, rules and routines, keeping adolescents safe in the community, and responding to crises. In addition, the programme includes economic strengthening components of family budgeting and saving sessions. Sessions included songs, collaborative problem-solving techniques (not didactic methods) and traditional stories, role-play, modelling and stress reduction activities. The programme was designed for low-resource settings with no technology (such as video) or literacy requirements.
Participants were encouraged to engage in home practice in the week following each session. For participants unable to attend sessions due to illness or disability, catch-up meetings were arranged to give brief session content at home or in the hospital. A simple lunch was included at the beginning of each session as many participants found difficulty in concentrating due to hunger. The programme was delivered by local community members, who were trained by a local NGO, Clowns Without Borders South Africa, and supported through weekly supervision.
Control group
Dyads in the control group received a one session (five hours) of hygiene programme called “SinoSoap”. This programme was implemented by the NGO “Clown without Borders” in South Africa, and involved drama-based skills-building on safe water conservation and hand washing for children. The session was delivered through performance and activities. All children received a soap which – when used – had a small toy inside.
Measurements
Parents\caregivers and adolescents completed self-report questionnaires, using tablets at baseline, 1 month post-intervention and 5–9 months following the intervention. All questionnaires were pre-piloted with local adolescents and parents\caregivers. All measurements were translated into isiXhosa, one of the 11 official languages spoken in South Africa, and back-translated.
Alcohol and substance use among parents\caregivers was assessed by using the adapted version of the WHO Alcohol Use Disorders Identification Test (AUDIT) [29] and the WHO Global School-based Health Survey. This variable was reported by parents\caregivers (4 items; α = .529; e.g., “In the past month, have you had a drink?”; “Did you take any drugs to help you relax?”). Responses were: 0 = No and 1 = Yes. One overall score was derived by computing the sum of the items.
Alcohol and substance use among adolescents was measured by using three items from the Child Behavior Checklist Scale [30]. This variable was reported by adolescents (3 items; α = .547; e.g., “During the past month, I drank alcohol without the permission of my caregivers’ approval”; “I smoke cigarettes”; “I use drugs like dagga (marijuana) or other drugs”). Responses ranged from 0 = Not true to 2 = Very true.
Parenting stress was measured using 18 items (α = .770; e.g., “I am happy in my role as a parent”; “Caring for my children sometimes takes more time and energy than I have to give”) from the Parental Stress Scale [31]. Items were measured on a five-point Likert type scale, ranging from 0 (Strongly disagree) to 4 (Strongly agree). One overall score was derived by computing the sum of the items.
Parental depression was assessed by using 20 items (α = .876; e.g., “I felt very sad even with help from my family and friends”; “I didn’t feel like eating”; “My appetite was poor”) from the Centre for Epidemiological Studies Depression Scale [32]. Responses ranged from 0 (Not at all) to 4 (Less than even day). One overall score was derived by computing the sum of the items. Items were measured on a five-point Likert type scale, ranging from 0 (Strongly disagree) to 4 (Strongly agree).
Family poverty was measured as monthly consistent access to necessities including food, electricity, communication, and transport [33]. This variable was assessed by using 9 items (α = .683; e.g., “Afford 3 meals a day”; “Afford the costs of the school”; “Afford enough warm clothes”). Responses were: 0 = No and 1 = Yes. One overall score was derived by computing the sum of the items.
All variables were measured at baseline and 5–9 months follow-up after the intervention was completed.
Data analyses
Analyses used intention-to-treat (ITT) for all clusters and families irrespective of intervention uptake and included families who were no longer living together at follow-up (n = 53). Independent sample t-tests were conducted to compare means of outcomes and mediator differences at baseline and follow-up between intervention and control groups.
A linear Structural Equation Modeling (SEM) was used with AMOS21 statistics program. The SEM procedure combined measurement modeling (Confirmatory Factor Analyses – CFA) and structural equation modeling. Items that were theoretically and empirically perceived as describing the variable were used in the measurement model.
Goodness of fit for the final model was assessed using the Comparative Fit Index (CFI) and the Root Mean Standard Error of Approximation (RMSEA). We also report χ2 fit statistics but acknowledge that the test is inflated by sample size of the study.