Researchers have explored the use of large language model agents to automate the creation and application of common data elements (CDEs) from medical abstracts. An agent proposed over 72,000 candidate CDEs from nearly 25,000 PubMed abstracts, which were then consolidated into 1,570 CDEs. After retaining CDEs supported by at least 50 abstracts, the system successfully processed 79.2% of abstracts for evidence-linked extraction. Human reviewers found 83% of the retained CDEs to be relevant, though clarity and correctness scores were lower, suggesting a need for expert curation. AI
IMPACT Demonstrates potential for LLM agents to accelerate scientific literature analysis and data standardization.
RANK_REASON Research paper detailing a novel application of LLM agents for scientific data extraction. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →