Large language models extract standardized data from sarcoma reports
Agentic induction of common data elements and information extraction from sarcoma case report abstracts
Computation and Language
Summary
It can be hard to compare medical reports because doctors write them in many different ways. This paper shows how smart language AI agents can read thousands of sarcoma case reports and find common pieces of information automatically. The AI suggested many possible data elements, then people helped choose the best ones for typical use. The AI then extracted these data from most reports accurately, especially outside of anatomy and timing details. This method can help create standard data formats faster with AI and expert checks.
What this means in practice
- •For hospital data teams: Automatically create structured summaries from thousands of sarcoma case reports to support data consistency and research.
- •For biomedical text miners: Generate and refine data extraction schemas for disease-specific literature using AI agents before manual review.
Authors
Daniel Reichenpfader, Kwaku Appiagyei, Manovah Nadarajah, Ahmed Reda Bahr, Irina Filchenko, Fabio Dennstädt
Abstract
Common data elements (CDEs) support clinical data comparability but are usually defined manually. We investigated whether large language model agents could derive a CDE schema from sarcoma case report abstracts and apply it at scale. An agent proposed 72,909 candidate CDEs from 24,588 PubMed abstracts; agentic consolidation yielded 1,570 CDEs. We retained 53 supported by at least 50 abstracts and attempted evidence-linked extraction across the corpus. Processing succeeded for all retained CDEs in 79.2% of abstracts, including responses indicating non-mention. Two human reviewers approved 83.0% of retained CDEs for relevance, 67.9% for clarity, and 56.6% for correctness. In 100 sampled abstracts, exact agreement with the value extracted by the LLM was 94.9% among 1,116 pairs with a non-empty answer (abstract-level bootstrap 95% CI, 93.0-96.4%). Errors concentrated in anatomical and temporal fields. These findings support agent-assisted schema drafting followed by expert curation before large-scale extraction.