Clinical Text Summarization: Adapting Large Language Models Can Outperform Human Experts

Area of Science:

• Artificial Intelligence
• Clinical Informatics
• Natural Language Processing

Background:

• Clinicians face significant time burdens summarizing electronic health records (EHR).
• Large language models (LLMs) show promise for NLP but require validation in diverse clinical summarization tasks.

Conclusions:

• This study provides the first evidence of LLMs outperforming human experts in multi-task clinical text summarization.
• Integrating adapted LLMs into clinical workflows can potentially alleviate documentation burdens.
• LLM adoption may enable clinicians to dedicate more time to patient care and the human aspects of medicine.