A Novel Approach to Nursing Clinical Intelligent Decision-Making: Integration of Large Language Models and Local Knowledge Bases

Abstract

Abstract Objective: The purpose of this study is to develop a novel method for nursing clinical intelligent decision-making that integrates Large Language Models (LLMs) with local knowledge bases, aiming to enhance the accuracy and reliability of clinical decisions in nursing. Methods: Initially, we established a multi-level classified nursing knowledge base by collecting textual knowledge from public knowledge platforms and integrating selected contents from peer-reviewed nursing journals, academic papers, textbooks, and nursing standards. Subsequently, data knowledge was collected from clinical records and normalized to form a data knowledge base. Additionally, we proposed a nursing clinical decision-making system paradigm based on prompt learning in “LLMs + professional knowledge bases”, addressing the issue of catastrophic forgetting common in domain-specific question-answering systems due to the “data + fine-tuning” paradigm. Results: Utilizing the aforementioned methodology, we successfully constructed a nursing knowledge base and developed a decision-making system. The evaluation results demonstrate that this system possesses high accuracy, logical coherence, completeness, and readability in clinical nursing decisions. It enhances the convenience and efficiency of medical staff in clinical decision-making and effectively improves the applicability of LLMs in the field of nursing. Conclusion: This study validates the effectiveness of the approach that combines LLMs with local knowledge bases in nursing clinical decision-making. This method not only enhances the accuracy of decisions but also provides efficient decision support in resource-limited scenarios. In the future, this approach is expected to be applied in a broader range of nursing settings, offering new perspectives and tools for clinical nursing practice and research.