Filming Interactions to Nurture Development in Rural China

The intervention is the culturally adapted Filming Interactions to Nurture Development (FIND) program, a video feedback curriculum translated and localized for rural China. FIND uses a positive reinforcement model to highlight naturally occurring, developmentally supportive interactions between caregivers and young children. The program emphasizes caregivers' existing strengths and is grounded in the "Serve and Return" framework developed at the Harvard Center on the Developing Child. A "serve" occurs when a child initiates interaction through words, gestures, or attention; a "return" happens when the caregiver notices and responds. Within FIND, five specific elements of serve and return are introduced, one every other session.

The five FIND elements are:

Sharing the Child’s Focus: This occurs when the adult identifies or notices what the child is interested in and then puts his/her attention there as well.

Supporting and Encouraging: Having noticed the child’s focus of attention, the adult responds in a supportive and/or positive way, adding his or her own reaction by giving the child further information about or acknowledging what he/she is seeing, doing, or feeling.

Naming: An extension of the Supporting and Encouraging element, Naming occurs when the caregiver uses words to label what the child is seeing, doing, or feeling.

Back and Forth Interaction: After the child has “served” and the caregiver noticed and returned the serve by Supporting and Encouraging or Naming, the interaction continues. The interaction goes back and forth between child and adult, with the adult waiting for the child’s further initiations.

Endings and Beginnings: This occurs when a Back and Forth interaction between child and caregiver ends and a new serve and return interaction begins. The end of the back and forth interaction is signaled by the child or the episode naturally comes to its conclusion (e.g., the book is finished).

2026-05-31

2026-09-03

The intervention is built on the science of a strengths-based approach and serve and return interactions—the responsive back-and-forth exchanges between adults and children that build the foundation for healthy brain development. The key outcomes are caregiver-level parenting outcomes:

The first family is responsive caregiving / reciprocal interaction.
Self-reported responsive skills and the responsiveness subscale of the Responsive Caregiving Rating Scale for Children Aged 0 to 4 Years (RCRS).
Observed measures: Video-coded serve-and-return responsiveness, including cue-contingent caregiver responsiveness, coded from a 10–15-minute recorded caregiver-child interaction. LENA conversational turn counts and a measure of child-initiated conversational turn counts.

The second primary behavioral outcome family is parenting investment and developmental stimulation. Measures in this family include the Family Care Indicators (FCI). Additional measures include the Promoting Cognitive and Emotional Development subscale of the RCRS and prespecified caregiver-child interaction questions. This family captures the frequency and breadth of developmentally supportive activities and is conceptually distinct from reciprocal responsiveness.

We prespecify four intermediate outcome domains.
1. Parenting beliefs: Beliefs about responsive parenting investment; Beliefs about the returns to broader parenting investment; Scenario-based parenting-belief measures
2. Parenting self-efficacy will be measured using the short form of the Tool to Measure Parenting Self-Efficacy (TOPSE) and the Chinese Parenting Sense of Competence Scale (C-PSOC)
3. Caregiver-child relationship will be measured primarily using the Mothers’ Object Relations Scales (MORS), adapted as needed for mothers, grandmothers, and other primary caregivers.
4. Psychological capacity will be measured primarily using the Parental Stress Scale.

This change is to align with our pre-analysis plan before the endline survey was completed.

We will also examine secondary and exploratory caregiver-level outcomes: parenting discipline style, perceived social support, screen-related parenting and exposure, and broader caregiver well-being, including DASS-21 and happiness/life satisfaction. Downstream child-development outcomes include CDI language/vocabulary, CREDI development, and ASQ-SE socioemotional development. Any additional child assessments not explicitly listed in the outcome codebook will be labeled exploratory. Because these are distal outcomes at an immediate post-intervention endpoint, the FIND theory of change does not require large short-run child-development effects.

The study will take place in 100 rural villages across 4 counties in Hubei Province. Villages are eligible if they: (1) have at least eight resident children aged 12-30 months; (2) are located outside minority townships (where the majority population is non-Han); and (3) have at least one village resident willing to participate, as recommended by the village committee.

From the pool of eligible villages, we will randomly select 100. To minimize spillover effects, any neighboring villages will be replaced with another randomly selected eligible village, ensuring that all sampled villages are at least 5 km apart.

The target population includes children aged 12-30 months at baseline and their primary caregivers (primarily mothers and grandmothers). All eligible households in selected villages will be invited to participate, provided they do not plan to move during the study period.

We designed a 10-week, single-blind, clustered randomized controlled trial in 100 villages across Hubei Province. We created a stratified design at the county level, and randomly assigned villages to either the intervention group (n = 50) or the pure control group (n = 50) in a 1:1 allocation ratio within each stratum.

Cluster (village) level

Yes

100 villages

800 child-caregiver pairs (8 pairs per cluster)

50

Power was estimated based on the treatment effects of parenting interventions documented in the literature (Emmers et al. 2021) between study arms for the primary and secondary outcomes: 1) caregiver-level parenting outcomes, 2) child development outcomes. Our sample size estimates account for randomization at the cluster level at 0.05 significance level. For positive parenting behavior, the power of the design was calculated given a 0.39 standard deviation effect (Emmers et al. 2021), assuming an intra-cluster correlation of 0.01, and a cluster size of 8 babies per village. For the home learning environment, the power of the design was calculated, assuming a 0.31 standard deviation effect (Cunha et al., 2024), an intra-cluster correlation of 0.01, and a cluster size of 8 babies per village. For child development outcome, the power of the design was calculated given a 0.26 standard deviation effect (Emmers et al. 2021), assuming an intra-cluster correlation of 0.01 and a cluster size of 8 babies per village. Based on these reference effect sizes, our study design has a power of 0.99, 0.99, and 0.94 when different outcomes are used as the outcome variable, respectively.

analysis plan final.pdf