Semantic Convergence with LLMs for Head and Neck Cancer Quality Indicators

We developed a novel method for leveraging large language models (LLM) to systematically filter and categorize large numbers of clinical quality indicators (CQI) for head and neck cancer. This was used to transform a tedious, human-resource intensive review process into a more efficient, knowledge-driven approach. Although we have successfully demonstrated the successful application of this approach to reduce manual effort overall, it is not possible to rely entirely on language models for such a task, and human oversight remains essential. We have delivered a generalizable approach that offers a promising pathway for more efficient and systematic clinical quality indicator management in other settings.