600 Section - Using artificial intelligence to enhance resource allocation in pediatric rehabilitation: study protocol for a pilot study

Children with developmental disabilities require services tailored to functioning-related needs rather than diagnostic categories. Ontario’s SmartStart Hubs provide a central entry point to assessments and services to families who have concerns about their children's development. SmartStart Hubs use the “About My Child” (AMC) questionnaire to collect semi-structured and free-text data from caregivers. This study aims to validate a semantic network-based artificial intelligence algorithm (text-to-ICF algorithm developed by Headai) that performs automated linking of free-text responses to International Classification of Functioning, Disability, and Health (ICF) categories using established linking rules. The goal is to enable ICF-coded data to support service planning and resource allocation.

Design:

Three Children’s Treatment Centres in Ontario will provide 300 anonymized AMC questionnaires. Manual ICF coding by trained raters will serve as the gold standard. The text-to-ICF algorithm, which operates offline and does not learn from input data, will be adapted to English-language AMC responses. Agreement between manual and AI coding will be assessed using Cohen’s Kappa, with a target score of >0.9. Stakeholder feedback will be collected via surveys and townhall meetings.

Results:

Initial manual coding has shown high inter-rater reliability. Algorithm adaptation is underway. The study will evaluate the algorithm’s ability to generalize across languages and cultural contexts, supporting its scalability in international settings.

Conclusion: This project introduces an algorithm validated for automated ICF coding from clinical free-text data. While piloted in pediatric rehabilitation, the algorithm’s prior validation in adult populations supports broader applicability across age groups, clinical domains, and healthcare systems globally. By embedding functioning as a third health indicator alongside morbidity and mortality, this approach contributes to a more holistic understanding of health and disability.