A double-blind, 12-month, parallel-group cluster randomized controlled trial with 1:1 allocation ratio will be implemented in Changsha, Hunan Province, China. This study will strictly adhere to the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) guidelines (Checklist 1)[31,32]. This protocol version is 1.0 and was enacted on August 1, 2024.

A 2-stage recruitment strategy will be conducted. First, 12 streets and towns were selected from 5 administrative districts in Changsha, including 3 urban districts (Kaifu, Yuhua, and Yuelu) and 2 rural areas (Ningxiang and Liuyang). Each district consisted of at least 2 streets and towns that were eligible for study recruitment (Figure 1). Eligible streets and towns had a primary health care institution present that (1) provides basic public health services for a population with over 2000 children younger than 5 years of age and (2) reports they are not involved in any active child injury prevention programs.

One health care provider per institution will facilitate participant recruitment through (1) face-to-face communication, leaflet distribution, and poster display during routine services; (2) digital advertisements and instructional videos shared through established communication platforms, such as WeChat groups; and (3) collaboration with local schools serving children aged 3 to 5 years, where teachers will distribute flyers to caregivers.

Families will be eligible if children (1) are aged ≤5 years at the time of baseline survey; (2) live with a primary caregiver who provides daily support; (3) reside in the selected study site; (4) have a caregiver aged ≥18 years who is literate, owns a smartphone, and provides informed consent; and (5) are not enrolled in other structured injury prevention programs. Children with severe illnesses requiring specialized care or hospitalization will be excluded.

Previous research reported a 12-month incidence of unintentional injury of 15.5% among children younger than 6 years of age in Changsha [33]. We hypothesize that the proposed mHealth intervention will reduce the injury incidence rate by 25% [34], with a design effect of 1.5 for the cluster sampling [35]. We also very conservatively assume 30% loss to follow-up [36]. Given these assumptions, a minimum sample size of 5178 children is needed to achieve 80% statistical power at a .05 significance level with a 1:1 allocation ratio between intervention and control groups (G*Power 3.1.9.7). Given that the actual values of certain parameters (eg, incidence rate and effect size) may be somewhat lower than assumed [28,34], we empirically expanded the sample size to recruit approximately 5600 children, with (2800 children from urban areas and 2800 from rural areas, including at least 350 children from each selected urban street and 700 from each rural town) within the constraints of available research resources. The 1:2 recruitment ratio of urban streets and rural towns is based on prior field experience, which suggests greater recruitment challenges in urban areas compared to rural ones.

The iCURE program is designed as an expert system, defined as a computer-based system that simulates the decision-making ability of a human expert, integrating the latest evidence-based knowledge on child injury prevention and applying rules of logic to provide context-specific guidance and tailored education to caregivers [37]. Its development is guided by expert system methodologies and the intervention mapping framework [38]. The design incorporates core principles from injury prevention and behavior change theories, along with empirical evidence and key stakeholders’ perspectives [39,40]. These components were integrated through a systematic development process involving literature review, needs assessment, and expert consultation [38]. The program applies algorithms such as keyword matching, text similarity analysis, decision tree models, and rule-based logic to deliver personalized recommendations and support [37,40-42].

A pilot study was conducted to assess the feasibility and operational performance of the iCURE program. The findings from the pilot were used to optimize its design and implementation, including improvements in system stability, registration flow, reward mechanisms, and data quality control.

The iCURE program consists of 2 components, a WeChat service account for caregivers and a web-based platform for health care providers (Figure 2 and Multimedia Appendices 1 and 2). This dual-platform approach facilitates information delivery and, crucially, supports caregiver-provider interaction.

The WeChat service account serves as the main channel for delivering intervention content to caregivers. The web-based platform is only accessible to health care providers, who support the system by managing and customizing educational content, monitoring caregiver engagement, and tracking self-reported injury events. All data are saved securely once collected and can be visualized on the platform. Providers and researchers can also export relevant data, including risk assessment outcomes, injury events, and intervention engagement, as needed (Multimedia Appendix 1).

The WeChat service account for caregivers comprises 5 key modules:

During routine consultations, primary health care providers will assist caregivers in registering for the iCURE WeChat service account. Upon registration, all caregivers will complete a questionnaire providing demographic information, their child’s injury history, and an injury risk assessment survey.

Caregivers randomly assigned to the intervention group will then receive the full version of the iCURE program, including real-time feedback, tailored educational content, and access to the Q&A module. The control group will access a simplified version involving completion of assessment surveys that are used only for data collection; they will receive no feedback, education, or supplemental materials. Both groups will continue to receive routine safety education provided by preschool teachers and primary health providers. Group allocation is detailed in Table 1. Details regarding the timeline for recruitment and intervention period are shown in Figure 3.

The primary outcome is unintentional injury incidence, calculated as:

Injury events will be identified based on established criteria from previous studies [48], defined as any of the following: (1) receiving medical attention at a health care facility and being diagnosed with a specific injury; (2) receiving first aid from a nonmedical responder (eg, family member and teacher), such as wound dressing or bandaging; and (3) missing school for more than half a day due to an injury. We will also compare injury rates at 3-month intervals to assess the effect over time.

In addition, several secondary outcomes will be assessed:

(1) Change in child injury risk level: A 46-item questionnaire will assess children’s risk of unintentional injury, classifying children as low-to-moderate or high risk. The questionnaire was adapted from existing validated instruments and covers 8 common types of injuries: drowning (5 items), road traffic injuries (6 items), choking (6 items), falls (6 items), burns and scalds (6 items), poisoning (5 items), animal bites (5 items), and sharp and blunt object or foreign body injuries (7 items) [43-47].

(2) Change in caregiver supervision behavior score: Caregivers’ caregiver supervision behavior will be measured using a simplified version of a validated 5-item questionnaire, with responses rated on a 5-point Likert scale [49]. Items overlapping with the child injury risk assessment questionnaire were excluded to avoid redundancy.

Randomization will be conducted at the street or town level to minimize potential contamination. Each street or town will be randomly assigned to either the intervention or control group, stratified by administrative division, using a random number sequence generated by R 4.4.1 software (R Core Team). The group allocation will remain concealed from both participants and the researcher conducting data analysis.

To maximize caregiver engagement with iCURE, we will implement 4 strategies: (1) during registration, caregivers will be instructed to disable the “Do Not Disturb” mode in their WeChat service account; (2) caregivers in the intervention group will earn lottery entries for completing questionnaires and digital points for viewing educational materials, with both entries and points available to be redeemed in ranking competitions, as supported by prior research [28,50] and findings from a needs assessment; (3) safety-themed activities will be organized on designated special days, such as World Drowning Prevention Day, to reinforce injury prevention knowledge; and (4) iCURE materials will be shared through existing WeChat groups by teachers and health care providers in intervention areas to leverage community networks and expand outreach.

Primary outcome data will be collected every 3 months to minimize recall bias. Secondary outcomes, which are less sensitive to recall bias, will be assessed annually. The baseline survey includes demographic information, child injury history over the past 12 months, caregiver supervision behaviors, and injury risk assessment, selected based on established associations with child injury and potential confounders [51]. All data will be self-reported, collected via the internet-based iCURE system, and securely stored on encrypted databases and hard drives. Caregivers will receive automated reminders followed by telephone follow-up if necessary to improve response rates.

The iCURE system will automatically record user engagement indicators, including (1) whether educational content was viewed, (2) frequency and duration of viewing, (3) messages to iCURE research team, and (4) online subscription to the iCURE WeChat service account.

Data will be analyzed following intention-to-treat principles and including all participants as originally assigned, regardless of adherence. Descriptive analysis (mean, SD, median, IQR, and proportion) will be used to describe baseline characteristics by groups. Differences in unintentional injury incidence across 5 timepoints (baseline, 3, 6, 9, and 12 mo) and other outcomes at 2 timepoints (baseline and 12 mo) will be examined using rank-sum tests for ordinal outcomes, chi-square tests for categorical outcomes, and independent 2-sample t tests for continuous outcomes. Within-group changes over time will also be evaluated.

Generalized linear mixed models with a Poisson link function will be used to compare the 12-month injury incidence between groups, adjusting for baseline covariates and accounting for clustering at the street or town level. Risk ratios and 95% CIs will be reported. Generalized estimating equations will also be used to assess repeated measures of injury incidence across follow-up timepoints, providing population-averaged estimates while accounting for within-subject and within-cluster correlations. Secondary outcomes, including changes in injury risk levels and caregiver supervision behaviors, will be analyzed using regression models appropriate to the type and distribution of the variable. User engagement indicators will be included as covariates of the effectiveness analysis model.

Two sets of sensitivity analyses will be performed to assess the robustness of findings: (1) per-protocol analyses and (2) models excluding participants with missing primary outcome data. Missing data will be documented and addressed using appropriate statistical methods as needed.

This study will adhere to the CONSORT (Consolidated Standards of Reporting Trials) 2025 [52] statement to perform data analysis and report results.

This study was reviewed and approved by the Ethics Committee of Xiangya School of Public Health, Central South University (XYGW-2024-71). All participants will provide electronic informed consent via the iCURE WeChat service account prior to participation, in accordance with national and institutional ethical guidelines. To ensure privacy and confidentiality, all data will be securely stored on encrypted databases and password-protected hard drives. Data will be de-identified before analysis, and no personally identifiable information will be retained in the analytic dataset. Participants will receive a modest digital monetary incentive ranging from 5 to 50 Chinese Yuan through a lottery-based mechanism via WeChat for their time and participation. No identifiable information will be included in the manuscript or any supplementary materials.

This protocol outlines a 12-month cluster randomized controlled trial designed to evaluate the effectiveness of iCURE, an intelligent, WeChat-based mHealth intervention for child unintentional injury prevention. The intervention is structured to align with the current primary health care service model in China. If proven effective, iCURE can be generalized and implemented nationwide as a low-cost and easy-to-use unintentional injury prevention intervention.

Childhood injuries result in high morbidity and mortality rates worldwide, including in China. Although injury prevention has been integrated into China’s National Basic Public Health Service system, implementation has been limited due to workforce shortages, high workloads, inadequate training, and the absence of operational guidelines. This study addresses these gaps through the development and evaluation of iCURE. Similar mHealth initiatives have demonstrated effectiveness in child unintentional injury prevention [18-30]. However, unlike traditional app-based programs, iCURE operates via a widely used platform (WeChat) without requiring app installation and incorporates real-time, caregiver-reported injury monitoring. These features are particularly suited to resource-constrained settings and may support scalable implementation in LMICs.

The iCURE program adopts a structured and theory-informed design that is rooted in behavioral science and an injury prevention framework. By integrating an AI-assisted dissemination strategy that is guided by personalized risk assessment reports, it sends individualized and tailored education materials to each caregiver. Leveraging mHealth technologies, iCURE offers cost-effective, easily implementable, and real-time interactive solutions. In addition, the evaluation uses a rigorous cluster-randomized controlled trial design, with blinded outcome assessment and control for potential confounders. It includes clearly defined outcome measures and a standardized data collection process, enhancing the validity of findings.

Potential challenges to the study include inadequate caregiver enrollment, loss to follow-up, and reporting bias. To address these concerns, the study includes a multifaceted engagement strategy that includes incentives, safety-themed campaigns, and active follow-up via automated reminders and telephone outreach. The involvement of health care providers in participant recruitment and follow-up is crucial to maintaining engagement throughout the study. Data will be collected every 3 months to minimize recall bias.

To minimize potential recall bias from caregivers’ reports, we purposively adopted follow-up assessments every 3 months and strictly implemented standardized definitions of unintentional injury and prompts for minor events (eg, bandaging and disinfection). Nevertheless, such strategies cannot fully overcome recall biases as caregivers may interpret and treat injury events differently, and some might not report minor injuries that were not specified by the operational definition or were not perceived as injuries [53].

If proven effective, iCURE could enhance the capacity of China’s primary health care system to deliver standardized, accessible child injury prevention services. It also offers a practical model for integrating digital tools into national child health programs in LMICs, where injury prevention resources are typically scarce.

This study responds to the urgent need for practical, evidence-based strategies to reduce childhood injury in China and other LMICs. Through a structured development process and rigorous evaluation, it aims to generate high-quality evidence on the effectiveness of an intelligent, system-integrated injury prevention program. If successful, iCURE could serve as a cost-effective, scalable solution for child injury prevention and surveillance, with potential for adaptation and dissemination through national public health platforms globally.