Implementation Study With Embedded RCT of a Digital System of Parenting Support Within Government Services in South Africa

Uptake, engagement, and completion are measured using process data embedded within the chatbot, including programme uptake, frequency of logging in, duration of engagement, and completion rates of assigned content modules.

Sexual violence vulnerability is assessed using caregiver-reported items from the Global Parenting Initiative, measuring children's and adolescents' exposure to unsafe situations in their communities. The measure includes five items for adolescents and four for children. Caregivers report how often in the past week their child or teen: spent time with untrusted friends; was unsupervised at home or outside (children) or went to taverns or drank with friends (teens); took a lift with someone unknown; or stayed away overnight without the caregiver knowing their whereabouts or returned late. Caregivers of adolescents also report their adolescents receiving sexual comments from older individuals. Responses are coded as (0) Never, (1) Sometimes (1-2 times), or (2) Often (3 or more times), with "I don't know" coded as 777 and "prefer not to say" coded as 888. Scores are summed for a total vulnerability score.

Positive involvement refers to the amount of caregiver participation in games and activities with their child or teen, and is measured using the one item from the positive involvement subscale of the Alabama Parenting Questionnaire (APQ; Essau et al., 2006), namely "How often in the past week did you play games or do other fun activities with your child/teen?" Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888.

Positive parenting is measured using the one item from the positive parenting subscale of the Alabama Parenting Questionnaire (APQ; Clerkin et al., 2007), namely "How often in the past week did you praise your child/teen when they behaved well?" Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888.

Parent mental health is measured using the Patient Health Questionnaire (PHQ-2; Löwe et al., 2005). The PHQ-2 measures depressed mood and anhedonia, and consists of the last two items of the PHQ-4 (Kroenke et al., 2009). The two items include, "How often in the past week have you felt down, depressed, or hopeless?" and "How often in the past week did you have little interest or pleasure in doing things?" The response options for this study range from 0 (Never) to 2 (Often - 3 or more times), with an option for "prefer not to say" coded as 888, to fit the responses of the other measures. The total score is calculated by summing the two responses. The validity of the PHQ-2 has been recognised in South African studies among patients attending primary health facilities (Bhana et al., 2015; Levis et al., 2020). Previous studies indicate that the PHQ-2 has sound reliability, with Cronbach alpha scores ranging from 0.78 to 0.92 (Löwe et al., 2005; Staples et al., 2019; Lee et al., 2026).

Safety planning communication is assessed using locally derived items from the Global Parenting Initiative by asking caregivers of a child: "Have you talked to your child in the past week about keeping safe?" Caregivers of teens are presented with two questions: "Have you talked to your teen in the past week about keeping safe in any of these situations?" and "Have you talked to your teen in the past week about keeping safe from situations such as sexual violence?" The safety planning communication questions are presented immediately after the items on sexual violence vulnerability; thus, "any of these situations" refers to the risky situations mentioned on sexual violence vulnerability items. Response items include (0) No, (1) Yes, or "prefer not to say" coded as 888. The sum of the item scores is used to obtain a total score.

Sexual violence exposure is assessed in the survey, within the chatbot, using four items for caregivers of children and teens. Caregivers are asked how often their child/teen has been exposed to four instances of sexual violence in the past month. These instances include whether their child/teen displayed inappropriate sexual behaviour, was touched sexually, was shown something sexual, or was made to do anything sexual. These occurrences provided are a combination of sexual violence items retrieved from the Sexual Abuse subscale of the International Society for the Prevention of Child Abuse and Neglect (ISPCAN) Child Abuse Screening Tool-Trial Version (ICAST-T; Meinck et al., 2018). Response items include (0) Never, (1) Sometimes (1-2 times), or (2) Often (3+ times), with "I don't know" coded as 777 and "prefer not to say" coded as 888. The sum of the item scores is used to obtain a total score.

Parent support of education is measured using caregiver-reported items from prior research (Ceballo et al., 2014) and applied in Parenting for Lifelong Health (PLH) studies. The scale includes two items each for caregivers of children and adolescents, assessing engagement in education-related activities. Caregivers report how often in the past week they praised their child or teen for working hard at school. Caregivers of teens also report how often they supported their teen's schoolwork, while caregivers of children indicate how often they named, counted, or drew with their child, based on the Family Care Indicators from the UNICEF Multiple Indicator Cluster Survey (Kariger et al., 2012). Responses are coded as (0) Never, (1) Sometimes (1-2 times), or (2) Often (3 or more times), with "prefer not to say" coded as 888. The sum of the item scores is used to obtain a total score.

This outcome assesses changes in programme users' perceived openness in communication with their children. Caregiver-Child/Adolescent Communication is measured using one item for caregivers of children and two items for caregivers of teens. Caregivers of children and teens are asked, "How often in the past week did your child/teen let you know what is bothering them?", while caregivers of teens are also asked "How often in the past week could you talk about personal things like boyfriends, girlfriends, or sexual behaviour, with Your Teen?". These items are adapted from a subscale of the Parent-Child Communication (PCC) scale, which was used in the Fast Track Intervention Study (McCarty & Doyle, 2001). Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888.

Online safety assesses programme user changes in their engagement with their children regarding their children's use of the internet and/or mobile devices. It is assessed for caregivers of children and teens using two items, namely "How often in the past week did you know what your child was doing when they were using the internet, cellphone, or other device?" from the the Global Kids Online survey (GKO; 2020) and "How often in the past week did you talk to your child about their use of the internet or device?" from the EU Kids Online Survey (Šmahel et al., 2020). Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888. Sores on the items are summed to obtain a total score for online safety.

Caregiver involvement is assessed with three items from the Family Care Indicators (FCI) subscale of the UNICEF Multiple Indicator Cluster Survey (MICS; Kariger et al., 2012). Caregiver involvement items assess caregivers' perceptions of their involvement with their children, including active participation in activities and decision-making processes. The three items ask how often in the past week a caregiver actively participated in activities with their child/teen, such as playing, sharing books, or helping with homework; were involved in making decisions regarding their child/teen's education, health, and overall well-being; and showed their child/teen affection or love. Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888. The three items will be summed to create a total score.

If applicable, the caregiver completing the programme also responds to similar items to those they completed for themselves, reflecting their perspective on another caregiver's involvement in their child's life. Other caregiver involvement is assessed with three items from the Family Care Indicators (FCI) subscale of the UNICEF Multiple Indicator Cluster Survey (MICS; Kariger et al., 2012). The three items include how often the other caregiver actively participated in activities with the child/teen, such as playing, sharing books, or helping with homework; was involved in making decisions regarding the child/teen's education, health, and overall well-being; and showed the child/teen affection or love. Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888. The three items will be summed to create a total score.

To assess programme user changes in harsh parenting, items adapted from the physical abuse (1 item) and emotional abuse (1 item) subscales of the International Society for the Prevention of Child Abuse and Neglect (ISPCAN) Child Abuse Screening Tool-Trial Version (ICAST-T; Meinck et al., 2018) are used. These items measure caregivers' behaviours towards their children, indicative of maltreatment. The two items include: "How often in the past week did you discipline your child/teen physically, such as hitting with your hand, belt, or another object?" (physical abuse) and "How often in the past week did you shout or scream at your child/teen?" (emotional abuse). Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888. The sum of the item scores is used to obtain an overall score for harsh parenting.

This outcome assesses whether programme users report changes in their child or adolescent's behavioural problems (externalising symptoms). Child or Adolescent Behavioural Problems are measured using one item, namely "How often in the past week did your child/teen not obey, or it was difficult to make them obey?" from the Child and Adolescent Behaviour Inventory (CABI; Cianchetti et al., 2013), which assesses caregivers' perceptions of externalising behaviours in their children (Cianchetti et al., 2013). Response items include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say" coded as 888.

The accessibility of the ParentText programme will be measured using a single question embedded in the chatbot: "What did you mainly use to connect to ParentText?" with response options (0) "Mobile data", (1) "WiFi at home", (2) "WiFi at work", (3) "Hotspot in the community", (4) "Other", or "prefer not to say" coded as 888.

The acceptability of the ParentText programme will be measured using the following question embedded in the chatbot: "How helpful was ParentText to you and your family?" with response options (0) "Not helpful", (1) "A bit helpful", (2) "Very helpful", or "prefer not to say" coded as 888.

This outcome aims to assess adolescent-reported sexual violence vulnerability via a brief individual telephonic survey. Adolescents' sexual violence vulnerability is measured with the questions: In the past week (7 days), how often: Did you spend time with friends that your parent doesn't know or doesn't trust? Did you go to a tavern or drink alcohol with your friends? Were you out of the house at night alone, or your parent didn't know where you were, or you came home later than your parent expected? Did you get a lift in a car with someone you don't know? Did you receive sexual comments from men or women who are older than you? Did your parent know what you were doing when you were using the internet or using the cell phone? Item responses include (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say/skip" coded as 888. Item responses are summed to compute a total score.

Questions include: In the past week (7 days), did your parent talk to you about: Spending time with friends that they don't know or trust? Going to a tavern or drinking alcohol with friends? Being out of the house at night alone, or somewhere where they didn't know where you were, or about coming home later than your parent expected? Getting a lift in a car with someone you don't know? Receiving sexual comments from men or women who are older than you? Keeping safe from situations such as sexual violence? Responses to these items are coded as (0) No, (1) Yes, or "prefer not to say/skip" coded as 888. In the past week (7 days): How often could you talk to your parent about personal things like boyfriends, girlfriends, or sexual behaviour? How often did your parent talk to you about using your cell phone or being on the internet? These two items are coded as (0) Never, (1) Sometimes (1-2 times), (2) Often (3 or more times), or "prefer not to say/skip" coded as 888.

Data embedded in the chatbot will track the number of referral requests made and the types of referral services selected throughout the programme.

The total cost of the intervention, including set-up and operating costs, as well as the average cost to engage with the intervention per participant (unit cost), will be calculated. These calculations will be based on detailed project budgets including all set-up, ongoing, and expected expenses associated with the rollout of the intervention.