Remote Intervention for Children with Tuberous Sclerosis Complex: Lessons on Adaptability and Engagement from an AI-Assisted, CFIR-Guided Analysis

Introduction: Children with Tuberous Sclerosis Complex (TSC), a rare genetic condition frequently involving epilepsy and developmental delay, experience high medical burden and contextual instability. The TRAIN study represents the first fully remote randomized trial of JASPER, a caregiver-mediated social communication intervention. This study leveraged treatment-group data from this completed trial to examine implementation factors shaping engagement and continuity in medically complex families. Methods: We conducted a qualitative-dominant, multi-source secondary analysis of 22 treatment-group families with children aged 12-34 months (82% epilepsy). Longitudinal data included interventionist coaching notes, caregiver diaries, and structured surveys across 24 weeks. Using the Consolidated Framework for Implementation Research (CFIR), we coded data to 20 constructs and conducted cross-case pattern analysis. A custom GPT model (ChatGPT Enterprise) supported multiple stages of analysis, including structured item-level coding, construct mapping, participant-level summaries, and cross-case synthesis. AI outputs were systematically reviewed, refined, and validated by researchers, preserving transparency, auditability, and alignment with standard CFIR procedures. This demonstrates an AI-supported CFIR workflow for rigorous secondary translational analysis. Results: Four cross-construct themes emerged. Telehealth enabled real-time adaptation within household constraints. Engagement was shaped by fluctuating child and caregiver capacity, with three observed patterns: flexible engagement during instability, temporary disengagement until stability returned, and continued disengagement. Families who established adaptable routines early were more likely to sustain participation. Disruption was common, and re-entry after pauses required explicit scaffolds. Findings informed concrete intervention design principles, including stability-timed starts, explicit adaptability rules, restart pathways, and sustainability supports. Discussion: This study demonstrates how existing clinical trial data can be systematically leveraged to generate implementation-relevant insights beyond efficacy outcomes. Findings inform the design of remote behavioral interventions for children with developmental disability and epilepsy by emphasizing adaptability, routinization, and structured re-entry supports. The AI-supported CFIR workflow enhanced analytic exhaustiveness, traceability, and feasibility after initial trial funding, offering a scalable model for translational evaluation across CTSA contexts.