Transfer Learning for FHIR Questionnaire Terminology Binding
A new research paper explores methods for binding Fast Healthcare Interoperability Resources (FHIR) Questionnaire items with Logical Observation Identifiers Names and Codes (LOINC) to improve electronic prior authorization workflows. The study compares six different techniques, including TF-IDF, MiniLM, BioBERT, and a novel BioLORD model, on a dataset of 54 items. BioLORD demonstrated the best top-rank accuracy, while a contrastively fine-tuned MiniLM model achieved the highest recall at ranks 5 and 10. The research also identified specific error types, such as wrong specificity and ambiguous text, as key challenges. AI
IMPACT Enhances accuracy in healthcare data binding, potentially streamlining prior authorization processes and improving clinical data consistency.