Exploring the ability of natural language processing to extract data from nursing narratives

Area of Science:

• Informatics
• Nursing Informatics
• Clinical Informatics

Background:

• Nursing narratives contain rich patient data not always captured by structured electronic health records.
• Natural Language Processing (NLP) offers a method to process unstructured clinical text.
• Existing NLP tools require domain-specific adaptation for optimal performance in nursing.

Purpose of the Study:

• To evaluate the effectiveness of the Medical Language Extraction and Encoding (MedLEE) NLP system in processing oncology nursing narratives.
• To identify key nursing concepts related to patient care, symptoms, and interventions within free-text nursing notes.

Main Methods:

• The Medical Language Extraction and Encoding (MedLEE) NLP system was applied to 553 oncology nursing process notes.
• The system extracted concepts related to patient signs, symptoms, and nursing interventions.
• Extracted data was analyzed to determine frequently recorded concepts.

Main Results:

• MedLEE successfully extracted 490 concepts from the nursing narratives.
• Commonly extracted signs and symptoms included adverse chemotherapy reactions, shortness of breath, nausea, pain, and bleeding.
• Frequently recorded nursing interventions involved chemotherapy administration, blood cultures, medications, and blood transfusions.

Conclusions:

• NLP, specifically MedLEE, shows feasibility for extracting nursing data from free-text notes, potentially enhancing patient safety and quality measurement.
• The study highlights the need to expand NLP lexicons with nursing-specific terms and abbreviations to improve performance in this domain.