ParentText for Crisis: A Randomised Controlled Trial of Chatbot-Delivered Mental Health and Parenting Content For Palestinian Caregivers in Jordan

This individually randomised controlled trial (RCT) is an evaluation of Palestinian ParentText for Crisis in Jordan, a rule-based chatbot to provide tips to strengthen parental mental health, improve parenting practices, and help parents support children in crisis settings. This randomised controlled trial is conducted through a collaboration with the University of Oxford, Parenting for Lifelong Health (PLH), and Sanadak. ParentText for Crisis content is adapted from PLH and "Hope Groups", which is a mental health and parenting programme originally developed from PLH for war and crisis settings. Hope Groups, delivered either in-person, have demonstrated effectiveness in improving caregiver mental health and parenting practices in crisis settings among Palestinians in Jerash Camp, Jordan. This study aims to evaluate the effectiveness of delivering this mental health and parenting content via 9 chatbot-delivered lessons over WhatsApp to Palestinian families within the same camp (Jerash Camp, Jordan). The RCT includes an intervention arm consisting of individuals who are randomised to join CrisisText immediately, and a waitlist control arm consisting of individuals who are randomised to join CrisisText after the intervention arm. This is an individually randomised study, with each individual participant as the unit of randomisation, and randomisation deployed using the treatment_assign function in the RCT package in R. The first cohort aimed to include N=240 Palestinian participants (N=120 randomised to intervention and N=120 randomised to control). As data collection for the first cohort completed, additional funding was secured to include a second cohort of another N=240 Palestinian participants (N=120 randomised to intervention and N=120 randomised to control). Thus, this RCT will roll out over two cohorts, conducted as two back-to-back RCTs, following procedures from staircase trials, cumulative trials, and adaptive trials. A Bayesian interim analysis will be conducted after the first cohort of N=240 participants to inform the trial's second cohort. The interim analysis will inform whether a) if harmful effects are identified, all activity will be paused; b) if the intervention demonstrates promise for statistical and clinical significance, the trial will continue into a second cohort; c) if the intervention does not demonstrate promise for statistical and clinical significance, the intervention will be enhanced prior to the second cohort. All details for the interim analysis are included in the posted statistical analysis plan.

Overall, the primary aim of this RCT is to test the effectiveness of ParentText for Crisis 30-days after enrolment, using an individually randomised controlled trial design. The final analysis will be conducted after the completion of the second cohort. Investigators hypothesise that ParentText for Crisis will lead to statistically significant improvements in caregiver mental health and positive parenting practices, and reductions in violence against children.

• Timeline for participants across arms (within both the first and second cohorts) will be as follows:
• Individuals who are randomised to the intervention arm will be invited to join an in-person onboarding session. At this session, participants will be given a trigger word to message to the chatbot's number on WhatsApp. Once participants send this message, ParentText for Crisis will guide them through informed consent, followed by a short (8 question) baseline survey. After this, the chatbot will begin delivering the first intervention session. Participants will be encouraged to complete this first session at the in-person event, then complete one addition session every day for the next 8 days.
• After the intervention arm concludes the ParentText for Crisis intervention, survey questions will appear again. (This data is for exploratory pre-post comparisons.)
• At +30 days after onboarding, the chatbot will message participants to complete a follow-up survey including the same survey questions. (This +30 days follow-up is considered trial endline.)
• During the week of the +30-day follow-up for the intervention arm, the wait-list control arm will be gathered for an in-person meeting to begin the intervention. At this in-person meeting, participants will complete informed consent and a survey. This data from the control arm will be compared with the intervention arm at their +30 day follow up for the primary trial analysis.
• Therefore, this trial's primary analysis will involve a single timepoint comparison: Estimating differences between the intervention arm at 30-day follow-up and the untreated control arm at this same calendar timepoint. To assess baseline balance in this randomised trial, investigators will compare non-time-varying participant demographic variables (e.g., age, sex, number of children) between arms. If there is imbalance, investigators will control for demographic variables in the trial analysis.