Analyzing Nursing Records in Wound Care Using a Large Language Model.

Area of Science:

• Oncology Nursing
• Health Informatics
• Artificial Intelligence in Healthcare

Background:

• Unstructured nursing records pose challenges in extracting critical patient information.
• Efficient summarization of cancer wound management documentation is essential for quality care.
• Large Language Models (LLMs) offer potential solutions for processing clinical text.

Purpose of the Study:

• To evaluate the efficacy of a large language model (LLM) in summarizing unstructured nursing records for cancer wound management.
• To assess the quality and accuracy of LLM-generated summaries against established clinical guidelines.
• To identify types of errors in low-quality summaries for future model improvement.

Main Methods:

• Retrospective analysis of 80 unstructured nursing records from specialized cancer wound care nurses.
• Extraction of 21 key variables based on British Columbia Cancer Agency guidelines.
• Application of an LLM for record summarization, followed by quantitative and qualitative quality assessments.

Main Results:

• The LLM demonstrated complete accuracy for nursing intervention variables and high accuracy (approx. 80%) for nursing assessment variables.
• Factual consistency received the highest scores in both quantitative and qualitative evaluations.
• Approximately 50% of low-quality summaries were attributed to reasoning errors.

Conclusions:

• LLMs can effectively summarize unstructured nursing records for cancer wound management.
• The technology shows significant potential to support clinical decision-making and improve patient care.
• Further refinement of LLMs may address reasoning errors to enhance summary quality.