Unveiling Fall Risk Factors: Nurse-Driven Corpus Development for Natural Language Processing.

Area of Science:

• Clinical Informatics
• Nursing Research
• Natural Language Processing

Background:

• Hospital-acquired falls remain a significant clinical challenge.
• Nurse-generated clinical notes contain valuable data for fall prevention.
• Advanced analytical methods, like NLP, offer new avenues for data utilization.

Purpose of the Study:

• To develop and test a natural language processing (NLP) pipeline for extracting fall-related factors from registered nurse (RN) clinical notes.
• To leverage expert nursing knowledge in data annotation for improved NLP model training.
• To explore the potential utility of annotated clinical notes beyond fall prevention.

Main Methods:

• An iterative process of expert manual annotation of RN clinical notes was employed.
• The annotated data was used to train and test an NLP pipeline.
• Interrater reliability was assessed to ensure data quality.

Main Results:

• The annotated data corpus achieved moderately high interrater reliability (F-score=0.74).
• The NLP pipeline demonstrated capability in extracting fall-related factors.
• The extracted clinical concepts showed potential utility beyond patient falls.

Conclusions:

• Expert nurse annotation is a valuable resource for training NLP models to address clinical concerns like falls.
• The developed NLP pipeline shows promise for enhancing patient safety and extracting broader clinical insights.
• Further research is needed to optimize the efficiency of using nursing expertise in annotation tasks.