Introduction

Child maltreatment, including corporal punishment, occurs at high rates around the globe and especially in low- and middle-income contexts (Hillis et al., 2016). This is also true of South Africa (Adam et al., 2016; Meinck et al., 2015; Ward et al., 2018). Both corporal punishment and harsher forms of child maltreatment are associated with a range of adverse outcomes for children, including internalising and externalising disorders (Afifi et al., 2017; Cuartas, 2021; Gershoff et al., 2018; Zielinski & Bradshaw, 2006). The importance of eradicating violence against children is recognized in the 2030 UN Sustainable Development Goals, most notably in Target 16.2, which calls for an end to ‘all forms of violence against…children’. This is a key issue in South Africa, which is recognized as one of the most unequal and violent countries in the world, creating situations of multiple disadvantage for children, especially amongst those in historically disadvantaged communities (Das-Munshi et al., 2016).

Parents tend to be amongst the chief perpetrators of violence against children (Meinck et al., 2015). Key family-level risk factors for parents to use corporal punishment or to perpetrate more severe forms of child maltreatment include not knowing alternatives (Cluver et al., 2018), the stress of parenting (Chiocca, 2017; Clément & Chamberland, 2008; Stith et al., 2008), intimate partner violence (Chiesa et al., 2018), poor parental mental health (Berlin et al., 2011; Stith et al., 2008), substance misuse (Staton-Tindall et al., 2013), and parents’ social isolation (Stith et al., 2008). Furthermore, abusive parenting and caregiver mental health have been shown to mediate between family disadvantage and adolescent outcomes (Meinck et al., 2015), suggesting that parenting interventions may help buffer the effects of family disadvantage. Caregiver praise and monitoring have also been identified as ‘development accelerators’, i.e., caregiver attributes that improve children’s lives across a number of Sustainable Development Goal targets simultaneously, including in contexts of deep adversity (Haag et al., 2022).

Fortunately, programs based on social learning theory and offering parents positive, non-violent alternatives to harsh discipline have been shown to be effective in reducing such violence (Barlow et al., 2006; Chen & Chan, 2016; Knerr et al., 2013; Vlahovicova et al., 2017). In particular, social learning theory-based programs have been shown to increase parent–child attachment (O’Connor et al., 2013) to reduce re-traumatising harsh parenting and increase positive parenting (which can buffer against adversity), improve self-regulation of both caregiver and child (Sanders & Mazzucchelli, 2013), and reduce parenting stress (Bennett et al., 2013), all of which are key aspects of trauma-informed work with children’s caregivers (Johnson et al., 2018). The importance of parenting programs as a key strategy, both to prevent violence against children and to support healthy relationships between caregivers and children, is recognized both by the World Health Organization (WHO; WHO, 2016; WHO, 2022) and the South African government (for instance, see the South African Children’s Act, No. 38 of 2005).

Few organisations, however, examine either trends in corporal punishment and positive parenting before implementing prevention programs, or whether parents are interested in the kind of support that programs offer. As a needs assessment, preparatory to beginning work in a small community in the rural Western Cape province of South Africa, we carried out three surveys to assess trends in parenting, risk factors for harsh parenting, and related child outcomes, and to ask parents if they would like to be involved in a parenting program. Ideally, to study community-wide interventions, one would recruit multiple communities and carry out a cluster randomized trial, but the intention with this work was to design and pilot test the intervention before carrying out a bigger study. In such situations, carrying out multiple pretests can allow for identification of trends and potential biases in estimating the effects of treatment (Shadish et al., 2002). The triple pretest was thus intended not only to explore family-level risk and protective factors playing a role in children’s mental health, and thus to drive the design of an appropriate intervention if indicated, but also to assess trends for the purposes of post-intervention causal inference. This paper thus reports on trends in children’s mental health and family-level risk factors in a particular community and the related intervention design.

Method

Study Design and Setting

The study took place in in Touwsranten, a small, historically disadvantaged, townshipFootnote 1 in the rural Western Cape province, South Africa. A door-to-door survey (June/July 2012) identified the number of households with children and was followed by three waves of a community-wide survey of caregivers (August 2012, March 2013, and January 2016), as well as focus group discussions in 2017. These three waves of data aimed to assess trends in parenting, child behaviour, and factors impacting parenting and child behaviour, in Touwsranten, with a view to developing an intervention to support parents and improve children’s mental health, if needed.

The choice of Touwsranten as a location for this study was determined by several factors. Firstly, Touwsranten was a small, stable community, with defined geographic boundaries and was isolated from other communities. This meant that there were low rates of in- and out-migration, particularly amongst the majority Afrikaans-speaking population, which would make it possible to track caregiving and child behaviour over time without large changes in the sample. Secondly, the 2011 Statistics South Africa census data put the population of Touwsranten at 2,245 (personal communication, Statistics South Africa, 28 May 2021), thus providing a sufficiently large sample for effects to emerge. Thirdly, Touwsranten was home to a stable community-based organisation, the Seven Passes Initiative that had been providing afterschool care and educational support to children of the community since 2008. This provided a base for the research and likewise for any future parenting intervention.

Participants

A 2012 door-to-door survey identified 304 households with children between the ages of six and 18. In wave 1, one caregiver from each household where there were children of these ages was invited to participate in the study, and 223 caregivers agreed. The participant was asked to answer questions about one child, preferably the youngest child in the household. In wave 2, the same participants were invited to be re-interviewed 6 months later, along with other caregivers who were newly identified using the same inclusion criteria (n = 33). Fourteen of the wave 1 participants declined to be re-interviewed (a refusal rate of 6.3%), resulting in a sample size of 242 for wave 2.

These two waves were followed by a third in 2016. Waves 1 and 2 had indicated that parenting support would be valued and potentially helpful to parents and children. And in 2016, a grant from the World Childhood Foundation made it possible to begin planning the intervention. Given the gap between wave 2 and the award of the grant, a third wave of data collection was planned to establish whether there had been significant changes in the community. In preparation for wave 3, a second door-to-door survey was conducted, which identified 762 households, of which 481 households included children aged 0–18. The wider age range was included because if this wave identified that an intervention was still needed, we intended to offer it to parents of the full age range of children. In wave 3, the number of children per household in Touwsranten ranged from one to six (M = 1.74; SD = 1.12), and 838 children in total were identified in Touwsranten: 22 aged 4–7 months, 159 aged 12–30 months, 325 aged 31 months–9 years, and 332 aged 10 and older. In these 481 households, 506 caregivers met the criteria for an interview in wave 3 (i.e., they were residents of Touwsranten and the primary caregiver of a child under the age of 18), and 462 consented to be interviewed: 411 mothers and 62 fathers, a response rate of 91.3%. Thirty-three caregivers refused to be interviewed, a low refusal rate of 6.9%.

Each caregiver was asked to choose his/her youngest child under 18 years old as their ‘focus child’ and completed questionnaire responses about that child. The sample included 365 Afrikaans- and 108 isiXhosa-speaking caregivers.

Measures

Demographics

Caregivers were asked to report on their own and their child’s sex and ages, and their relation to the focus child, as well as their food security.

Caregivers’ Self-Report Measures of Their Parenting

The Alabama Parenting Questionnaire (APQ) Global Parent Report was used to assess the parenting behaviour of parents of children 6–18 years old in this study. The APQ is a 42-item questionnaire for parents (Essau et al., 2006) and has five different subscales, namely, (a) poor supervision and monitoring, (b) parental involvement, (c) positive parenting, (d) inconsistent discipline, and (e) corporal punishment (Shelton et al., 1996). The items are answered on a 5-point Likert scale ranging from 1 (never) to 5 (always), but wave 1 data revealed that parents struggled to differentiate between ‘never’ and ‘very little of the time’, and ‘often’ and ‘always’ (Ward et al., 2015). Answer categories were thus collapsed to ‘never/very little of the time,’ ‘sometimes,’ and ‘often/always.’

The APQ has shown adequate reliability in studies conducted in the US (Shelton et al., 1996), Australia, and Canada (Elgar et al., 2007), with Cronbach’s alphas greater than α = 0.70 for all subscales except poor monitoring and supervision (α = 0.67) and corporal punishment 0 (α = 0.55) in an Australian study (Dadds et al., 2003). In this study, the inconsistent discipline subscale had poor internal consistency (Cronbach’s alpha was 0.58), and it was therefore excluded; for the same reason, corporal punishment items were treated individually.

Two subscales of 7 items each (‘setting limits’ and ‘supporting positive behaviour’) from the Parenting Young Children Scale (PARYC; McEachern et al., 2012) assessed the parenting behaviours of parents of children 18 months to 5 years (e.g., ‘how many times in the past month did you teach your child new skills?’; ‘how many times in the last month did you stick to your rules and not change your mind?’). These subscales had response options of 1 (never) through 7 (almost daily in the past month). The original validation study demonstrated validity of the PARYC scale amongst high-risk caregivers from rural communities in Charlottesville and Pittsburgh (McEachern et al., 2012).

Caregiver Report Measures of Their Children’s Outcomes

The Child Behaviour Checklist (CBCL) for children aged 6–18 and the pre-school CBCL (for children aged 1½—5) were used to assess children’s emotional and behavioural problems (Achenbach & Rescorla, 2000, 2001; Ebesutani et al., 2010). The CBCL for children aged 6 to 18 is a 118-item self-completion scale for caregivers about the behaviour of their child (e.g. ‘drinks alcohol without parents’ approval,’ ‘argues a lot,’ and ‘overeating’). The preschool Child Behaviour Checklist is a 99-item self-completion scale that assesses child outcomes for children 18 months to 5 years (e.g., ‘can’t concentrate’, ‘can’t pay attention for long’). Item responses are on a 3-point Likert scale (not true, sometimes, often/very true). The CBCL is robust in a variety of cultural contexts, including South Africa (Achenbach & Rescorla, 2001; Calkins & Dedmon, 2000; Gross et al., 2006; Mesman, Bongers, & Koot, 2003; Nöthling, Martin, Laughton, Cotton, & Seedat, 2013).

Correlates of Parenting Behaviour and Child Outcomes

Parenting Stress

The Parenting Stress Index Short Form (PSI-SF; Abidin, 1990) was used to assess parenting stress. This is a 36-item self-completion scale that quickly screens for stress in the parent-child relationship (e.g., ‘my child is not able to do as much as I expected’). The PSI-SF has been used extensively in a variety of contexts and samples in the US (Haskett et al., 2006; Reitman et al., 2002; Smith et al., 2001), with good test-retest reliability (average score of 0.76) and high internal consistency (0.85) in the original validation study in rural and urban areas of Virginia (Abidin, 1990). Moreover, the PSI-SF has been found to have high test–retest reliability and validity in a sample of parents of South Africa children (Potterton et al., 2007) and sensitivity to change after parenting programs (Cowen & Reed, 2002; Wolfe & Hirsch, 2003).

Caregiver Mental Health

The General Health Questionnaire (GHQ; Goldberg, 1979) was used to assess caregiver mental health. The questionnaire consists of 28 items (e.g., ‘have you been getting scared or panicky for no good reason?’; ‘have you been getting edgy and bad tempered?’), with four possible responses scored using binary options, i.e., 0 (‘better than usual’ or ‘same as usual’) and 1 (‘worse than usual’ or ‘much worse than usual’). Using this scoring method, any total score greater than 3 indicated ‘psychiatric caseness’ (Goldberg, 1979). The instrument has been shown to be a reliable and valid measure of psychological well-being in over 38 different contexts (Jackson, 2007).

Caregiver Experience of Intimate Partner Violence

Sixteen items from the Conflict Tactics Scale—Revised (CTS-2; Straus et al., 1996) were used to assess levels of intimate partner violence (IPV). These items explored psychological and physical aggression. These ranged from mild (e.g. ‘my partner insulted or swore at me’) to severe (e.g. ‘my partner used a gun or a knife on me’) forms of violence. The CTS-2 is the most widely used measure of intimate partner violence (IPV) (Newton et al., 2001; Straus et al., 1996) and has been found to have good internal consistency and factor validity in diverse samples (Calvete et al., 2007; Lucente et al., 2001; Newton et al., 2001) including in South Africa (Swart et al., 2002).

Caregivers’ Self-Report Measures of Their Alcohol Use

The alcohol subscale from the ASSIST was used to assess the risk level of the respondents’ alcohol intake (Group, 2002; Humeniuk et al., 2008). The ASSIST has been validated in a number of diverse settings, including the USA, Spain, India, Zimbabwe, and South Africa (Humeniuk et al., 2008; Rubio Valladolid et al., 2014; van der Westhuizen et al., 2016). The ASSIST was found to have high internal consistency (α = 0.81–0.95 across different substances), as well as convergent and discriminant validity in a sample (n = 200) of South African emergency centre patients (van der Westhuizen et al., 2016).

Procedure

Fieldworkers for quantitative data collection were recruited by the Seven Passes Initiative and trained by two authors (CLW and CG) in ethics and interviewing skills. In waves 1 and 2 fieldworkers administered paper questionnaires. In wave three, questionnaires were administered using the Mobenzi (www.Mobenzi.com) digital platform. This made it possible to link each fieldworker to a specific phone and to manage the daily survey submissions and analytics on the Mobenzi web console, thereby facilitating cleaning and data analysis in real-time. While efforts were made in wave 3 to interview the same caregivers who had been interviewed in wave 2, this was not always possible, both because many of the children who were the focus of the interview in waves 1 and 2 were now older than 18, or because the children’s primary caregivers had changed.

Interviews took about 2 hours and were conducted in private. Refreshments were offered to each caregiver interviewed. Fieldworkers provided respondents with a list of local organisations that provide support for parenting, intimate partner violence, and substance misuse.

In addition to the quantitative data, five focus group discussions were held with twenty community members. Community members were recruited to participate in these through an announcement at a public meeting and flyers distributed throughout the community, inviting anyone who had been interviewed to attend. Only women volunteered to participate. To thank participants, they were given an R50 voucher for a local clothing store. Three themes were explored in these discussions: what it had been like to complete the questionnaires; what methods of discipline were primarily used in the community; and what stressors affected parenting in the township. Participants gave separate informed consent to participate in the focus group discussions.

Ethical clearance for the study was obtained from the Research Ethics Committee of the Faculty of Humanities at the University of Cape Town (reference no. PSY2015-049).

Data Analysis

Analyses of the quantitative data were conducted using R (R Core Team, 2022), with the aim of investigating the stability (or otherwise) in child and parental social risk factors within the Touwsranten cohort over 3 baseline waves. Both raw and standardised estimates (median and inter-quartile range) are reported for descriptive data. Since the majority of participants surveyed at wave 1 were followed up at wave 2 (89%), Wilcoxon signed-rank and McNemar’s tests were conducted to assess change across waves 1 and 2 for participants who were assessed at both waves. Given that the majority of participants in wave 3 were independent from those who were assessed at waves 1 and 2, Wilcoxon’s rank-sum and chi-squared tests were used to assess how participants at wave 3 differed from participants at wave 2. Qualitative data were analysed using thematic analysis (Braun & Clarke, 2006).

Results

The cohort was made up of a total of 462 unique caregivers, where all children were 6 years or older. Only 63 caregivers were interviewed across all 3 baseline waves: 155 caregivers completed any two of the three waves, while 244 only completed one wave. Most participants interviewed at wave 1 were followed up at wave 2 (216 participants, 89%), in comparison to 63 participants who completed all 3 baseline waves (13.6%), suggesting a heterogeneous sample at wave 3 relative to waves 1 and 2. Cronbach’s alphas for all scales ranged between 0.714 (total APQ score, excluding corporal punishment) and 0.961 (PSI-SF). The extent of poverty in Touwsranten was indicated by the finding that, in wave 1,183 families (60.4%) had run out of money to buy food more than four times in the preceding month.

Parenting and Child Outcomes Across the Three Waves

The three samples are described in Table 1. The median age of children across the three waves was 11, and child gender distributions were similar across waves. Most parents (92% wave 1, 94% wave 2, and 89% wave 3) were the biological parents of the children who were the focus of the survey. On the Alabama Parenting Questionnaire, median parenting scores suggested that parents often used positive parenting strategies, had slightly lower levels of parental involvement, and demonstrated poor monitoring of their children less than ‘sometimes’. Median scores suggested that spanking was prevalent, but that slapping and beating with an object were infrequent. Over the three waves, increasing numbers of parents reported poor mental health: 0 (0%) in the first wave, 0 (0%) in the second, and 43 (15%) in the third wave. Reports of alcohol use showed a similar pattern: in the first wave, 33 parents (14.6%) reported risky drinking patterns; in the second, 51 (21.8%) reported risky drinking; and in the third wave, 60 (22.5%) reported risky drinking. Reports of intimate partner violence amongst those parents who had a current partner were similar at the second (71; 29%) and third (73; 26%) waves but lower at the first wave (52; 23%). The qualitative data suggested that these apparent fluctuations may represent increasing trust in the confidentiality of the surveys. A majority of parents in the first (120; 54%) and second (159; 66%) waves reported being in the high range for parenting stress, but this fell considerably by the time of the third wave to 14 (5%).

Table 1 Longitudinal sample characteristics across baseline waves (1–3)

Children’s outcomes seemed to deteriorate slightly from the first to second waves and then hold steady (which may again represent greater trust in study confidentiality). Overall, in wave 1, 39 children (18%) were in either the clinical or borderline ranges for either internalising or externalising disorders; while at wave 2, this was 57 (28%); and at wave 3, it was 66 (26%). At the first wave, 32 children (14.4%) were in the borderline or clinical ranges on the internalising subscale of the Child Behaviour Checklist. This increased slightly to 36 (17.9%) at the second wave and 44 (17.8%) at the third wave. Fewer children appeared to demonstrate externalising behaviour problems: at the first wave, 19 (8.6%) were in the clinical or borderline range for externalising behaviour, whereas at the second (39; 19.3%) and third (51; 20.3%), a similar proportion of children were reported to have conduct problems.

Trends in Parent and Child Outcomes Over Time

Trends over time were examined by comparing results across waves one and two from those parents who completed both waves (dependent observations, Table 2) and then comparing the results from all parents who completed wave 2 with all who completed wave 3, excluding those who completed both waves 2 and 3 surveys (independent observations, Table 2).

Table 2 Dependent and independent observations (Wave 1 v Wave 2; and Wave 2 v Wave 3)

The comparison between reports of the same parents at waves 1 and 2 revealed that there were statistically significant but, in practical terms, very slight deteriorations in parenting behaviours, except for slapping children and hitting with the hand, which appeared to reduce slightly in frequency. There were marked increases in the proportion of parents reporting risky alcohol use (14.2 to 20.3%) and intimate partner violence (21 to 27%), and an increase in the proportion of parents reporting high levels of parenting stress (54 to 63%).

In terms of children’s outcomes reported by parents who participated in both wave 1 and wave 2, there were slight increases in the mean T-scores for children’s internalising and externalising, but with marked increases in the proportion of children reported to be in the clinical and borderline clinical ranges at wave 2: an increase in the proportion of children who might benefit from an intervention from 13.1 (n = 28) to 17.8% (n =38) for internalising and 7.5 (n = 16) to 19.1% (n = 41) for externalising.

When the changes from the independent samples at waves 2 and 3 were examined, there were no changes in parents’ use of spanking or slapping, or reports of intimate partner violence. There was a statistically significant increase in reports of risky alcohol use and a decrease in the proportion of parents reporting parenting stress. There was too little variation in reports of mental health symptoms to permit analysis. There was a small but statistically significant change in overall use of positive parenting strategies, apparently driven by a decrease in poor monitoring and a slight increase in involvement. There was no change in rates of children’s internalising or externalising disorders from wave 2 to wave 3.

Qualitative Data

Focus group discussions showed that respondents found some questions difficult to understand and others difficult to answer, especially if they were not the fulltime caregiver of the child in question (e.g. they were caring for a relative’s child). Most experienced the interview positively. A small number of participants were worried about what would happen to their data and were concerned that it might lead to a visit from a social worker investigating claims of child maltreatment. We had indicated in the consent form that if a child (or a caregiver) was in danger, we would have to report this to the mandated agency; however, this was not necessary during the surveys. Participants reported high levels of corporal punishment in the community, and some were uncomfortable answering questions about corporal punishment, suggesting that corporal punishment may have been under reported. Similarly, alcohol use in the community was reported as prevalent and has having a negative impact on caregiving. Intimate partner violence was also reported as taking place in many homes, and discomfort was expressed with answering questions about it. Finally, participants identified financial stress, partner’s infidelity, and difficult child behaviour as key sources of parenting stress.

In wave 1, we also asked parents if they would be interested in parenting support, and if so, how would they like it delivered. The majority said they wanted home visits (53.8%), followed by a course (38.6%), a pamphlet to read (27.1%), or a once-off workshop (16.2%). Less attractive options were a parenting hotline (11.6%), a cellphone app (4.6%), or information on the internet (1.6%).

Discussion

Our earlier work examining relationships between variables at wave 1 reveals that parenting stress, parental mental health, and corporal punishment were associated with children’s internalising and externalising behaviour, and that intimate partner violence was also associated with children’s externalising (Ward et al., 2015). Analyses of these three waves of survey data suggest that although there were changes that were statistically significant, overall there was little or no substantive change. Some apparent changes (for instance, in the sample who completed both waves 1 and 2, increased reports of intimate partner violence and risky alcohol use) may reflect increased trust in the confidentiality of the research team over the repeated waves of the survey, rather than actual changes in prevalence.

It is also notable that parents were reporting high proportions of children to be in either the borderline or clinical ranges of the Child Behaviour Checklist; approximately 1 in 5 children fell into one or both of these ranges for either internalising or externalising behaviours. This prevalence is similar to CBCL parent-reported rates of disorders in US military families (Kelley et al., 2003), considerably higher than those reported in an Iranian sample of children aged 6–12 (Tehrani-Doost et al., 2011), and similar to those of children in a sample in Sao Paulo, Brazil (Rocha et al., 2013). Regardless of similarity to other samples, however, this is a very high proportion of children in a community to need care. Furthermore, the clinic in this community provided no mental health services, and the nearest services were located in the town of George, a 30-min drive away, making them inaccessible for parents without cars (i.e. all parents in the community) in an area with very poor and expensive public transport.

Likewise, parent reports of their mental health symptoms and the stress they experienced in parenting were also high. Reported rates of intimate partner violence and substance misuse were low compared with anecdotal reports in the community and were no doubt underreported in the survey, at least in the first wave. Corporal punishment was also revealed as a common parenting strategy. Analyses of the wave 1 data showed that all these problems (mental health problems, parenting stress, intimate partner violence, and substance misuse) were related to child behaviour problems (Ward et al., 2015), and these three waves of data found that they changed little over time.

Given this lack of change over time, it was clear that spontaneous reduction in harsh parenting and improvements in child mental health were unlikely, and that an intervention would be necessary, if children’s mental health were to be improved in the community. The question then arose: how best to protect and support children in a community with few formal services? Based on these results, the Seven Passes Initiative decided to offer parent skills training programs, on the basis that parenting programs have evidence for reducing violence against children and increasing positive parenting skills (Coore Desai et al., 2017; Knerr et al., 2013; Mikton & Butchart, 2009) and improving children’s mental and behavioural health (Medlow et al., 2016; Pedersen et al., 2019). Although inconclusive, there is some evidence that suggests that parenting interventions may improve children’s outcomes even in contexts of intimate partner violence (Coore Desai et al., 2017; Latzman et al., 2019), substance misuse (Bröning et al., 2012; Reupert et al., 2017), and poor parent mental health (Suarez et al., 2016). Furthermore, while the evidence is again inconclusive, some studies have found that parenting interventions improve parent mental health, at least in the short-term (Barlow & Coren, 2018). Reductions in parenting stress have also been found in several parenting programs (Barlow & Coren, 2018). In recognition of the broader context of parenting, the Seven Passes Initiative also undertook to support parents who needed help with intimate partner violence, substance misuse, mental health problems, and food insecurity, and built up a network of providers to whom they could refer parents with these needs.

However, even when widely offered in a community, it is unlikely that all parents will attend a parenting program (Finan et al., 2018). Furthermore, the effect of parenting programs may not always sustain overtime (Barlow & Coren, 2018). For this reason, a social activation process was to be offered alongside the parenting programs (Parker & Becker-Benton, 2016). The process involved an action research activity following the Action Media method (Parker & Becker-Benton, 2016) to identify community perspectives on parenting, identify community-led actions to support parenting and parent–child engagement, and develop a manifesto, slogans, songs, and branding to support change processes. A steering committee was established to lead the process, and activities over the period of the study included community mural painting, street clean-ups, playground repair, family-level endorsement of the parenting manifesto, display of support through branded stickers on front doors, and a series of community-events including fund-raising activities, parent–child games evenings, including parenting activities to national day celebrations such as Youth and Women’s Day (both national holidays in South Africa), amongst other opportunities for parents and children to spend time together.

The four Parenting for Lifelong Health programs (see https://www.who.int/teams/social-determinants-of-health/parenting-for-lifelong-health) were chosen as the parenting programs, on the basis that they covered ages of children from pregnancy to 18, had been tested in randomized controlled trials (Cluver et al., 2018; Cooper et al., 2009; Vally et al., 2015; Ward et al., 2020), and were accessible: costs lie in training and coaching of facilitators, not materials; and training was available in South Africa. These programs include:

  1. (i)

    A home-visiting program for pregnant women until the baby is 6 months old, with the goal of improving maternal attachment (Cooper et al., 2009), which has theoretical links to later child mental health (Barlow et al., 2016; Fearon et al., 2010; Sroufe, 2005);

  2. (ii)

    A group-based dialogic book-sharing program with evidence for improving infant vocabulary and theory of mind (Vally et al., 2015), as well as improving parent-child attachment, all of which have theoretical links to violence reduction and child mental health (Dodge et al., 2008; Ogilvie et al., 2014);

  3. (iii)

    A group- and social learning theory-based intervention for parents of children aged 2–9, with evidence for reducing violence against children (Ward et al., 2020);

  4. (iv)

    A group- and social learning theory-based intervention for parents and children aged 10–17, with evidence for reducing violence against children and a range of child conduct problems (Cluver et al., 2018).

The theory of change for this intervention as a whole (see Fig. 1) was thus in summary as follows: that the delivery of parenting programs (which cover the full age range of children, i.e. zero to 18), combined with a social activation process, would lead to a community-wide shift towards positive parenting. This would be achieved through two pathways. The delivery of the parenting programs (three of which are group-based) would increase the use of positive parenting strategies, reduce corporal punishment, and increase parent social support, which could lead to improved parent mental health, reduced parenting stress, and improved communication and relationships between caregivers and children. In parallel, a social activation process would identify and amplify existing community values about positive parenting, undertake activities to support those values, and disseminate messages of positive parenting and care widely across the community, thus both increasing uptake of the parenting programs and enabling community-wide change towards positive parenting. The theory of change was informed by several assumptions: (1) that the PLH programs could be adapted to be contextually and culturally relevant, (2) that parents would participate in the parenting programs, and (3) that community members would take ownership of and lead the social activation process.

Fig. 1
figure 1

Intervention theory of change

Pregnant women were to be enrolled with the assistance of the local clinic nurses, who would inform them about the home visiting program, and if they agreed, refer them to the Seven Passes Initiative (SPI). Other parents were to be recruited through community outreach of the parenting facilitators, for instance by going door-to-door to hand out fliers, or through the social activation program. Group programs were to be offered in the SPI offices or a local church, which had rooms of sufficient size to accommodate groups. Monitoring and evaluation were to be provided by the University of Cape Town and the Institute for Security Studies, who would conduct follow-up waves of surveys 18 months after the start of the intervention, and again another 18 months later.

This paper reports on a detailed assessment of the need and appetite for parenting interventions in a rural South African community, which was relatively isolated from services. The data reported here are limited in that it proved impossible to follow the identical group of parents across all three waves of the study, and of course, the data are also limited to this particular community. However, the study did establish that a high proportion of children were experiencing mental health symptoms, that parents were frequently using corporal punishment, and that a number of parents were experiencing mental health problems, substance misuse, intimate partner violence, and high levels of parenting stress; and that these issues were all related. Despite small fluctuations, rates of these problems remained fairly stable over time; suggesting that without intervention, they would continue to do so. Results of this study led to the design of a community intervention to support parenting, which was to be tested and, if successful, offered to other communities.