Large Language Models for Studying Amblyopia Treatment, Adherence, and Outcomes

PROJECT SUMMARY Amblyopia is a common pediatric eye problem impacting 2-4% of the population. This vision threatening disease can be inexpensively and effectively treated but proves quite challenging due to adherence problems. Studying treatment, adherence, and outcomes using real world electronic health record (EHR) data is limited due to inconsistent documentation and documentation of these concepts only in free-text clinical notes. Large language models (LLM) have the potential to extract these concepts from these notes automatically and at scale, enabling more research as well as the development of clinical decision support tools that use this data in clinical care. This project will develop, evaluate, and validate models using a single site’s data as well as multiple sites through the Sight Outcomes Research Collaborative (SOURCE) consortium.