ATSCOT: Chain of Thought for Structuring Anesthesia Medical Records

This study proposes a method called ATSCOT (Anesthesia Text Structuring Chain of Thought), aimed at leveraging the powerful capabilities of Large Language Models (LLMs) to structure anesthesia medical records. The ATSCOT method enhances the semantic understanding abilities of Large Language Models through carefully designed prompts, breaking down complex anesthesia text processing tasks into several key steps. This Chain of Thought approach helps the model better understand and organize intricate medical records. The study constructed a specialized test set containing 100 anesthesia-related medical records and employed a combination of automatic evaluation, human assessment, and GPT-4 evaluation to comprehensively assess the model's performance. Evaluation criteria include the accuracy of information extraction, contextual understanding, and reasoning abilities. Experimental results indicate that the ATSCOT method performs excellently in structuring anesthesia texts, showing significant improvements compared to baseline methods. The findings suggest that the proposed approach has the potential to greatly assist in the efficient and accurate processing of anesthesia medical records, ultimately benefiting patient care and clinical decision-making.