Making nursing visible: AI-assisted standardization of electronic health record interventions using generative...

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Nursing Data Standards

Background:

• Nursing care significantly impacts patient outcomes but is under-examined due to a lack of standardized nurse-generated data.
• Standardizing nursing intervention terms in electronic health records (EHRs) is crucial for large-scale data analysis.

Purpose of the Study:

• To evaluate the performance of retrieval-augmented generation (RAG) and generative pre-trained transformer (GPT) models in standardizing nursing intervention terms within EHRs.
• To assess the utility of AI models in creating a "gold standard" dataset for nursing terminology.

Main Methods:

• Manual mapping of local nursing terms to standardized terms to create a gold standard dataset.
• Implementation and assessment of a RAG pipeline utilizing GPT models for term standardization.
• Integration of AI model outputs into an AI-assisted manual mapping workflow.

Main Results:

• Generative pre-trained transformer (GPT) models demonstrated performance comparable to human mappers in standardizing nursing terms.
• The accuracy of GPT models was insufficient for fully automated standardization of nursing intervention terms.

Conclusions:

• AI models, particularly GPT, can significantly reduce human workload and streamline the standardization of nursing intervention data.
• Advancements in AI-assisted standardization are vital for making nursing contributions visible and enabling data-driven evaluation of nursing practices.