Linking Free Text Documentation of Functioning and Disability to the ICF With Natural Language Processing.

Area of Science:

• Medical Informatics
• Computational Linguistics
• Rehabilitation Medicine

Background:

• Electronic Health Records (EHRs) contain unstructured free text with valuable patient functioning information.
• Extracting and organizing this data is crucial for clinical decision-making and research.
• Natural Language Processing (NLP) is key to unlocking insights from clinical documentation.

Purpose of the Study:

• To apply NLP methods to analyze patient functioning information within clinical documents.
• To utilize the International Classification of Functioning, Disability, and Health (ICF) framework for classifying functioning data.
• To develop and evaluate machine learning-based NLP models for automated ICF coding of functional status.

Main Methods:

• Analysis of clinical documents from U.S. Social Security Administration disability benefits claims.
• Annotation of functional status information by expert clinicians, focusing on mobility, self-care, and domestic life.
• Training machine learning NLP models to automatically assign ICF categories to documented functional activities.

Main Results:

• Extensive and varied patient functioning data identified in free-text records.
• Over 2,400 Mobility and 3,900 Self-Care/Domestic Life activity mentions were annotated.
• NLP models achieved over 80% macro-averaged F-measure, demonstrating strong performance in automated ICF coding.

Conclusions:

• NLP facilitates the use of expressive clinical documentation for standardized, comparable data analysis.
• The ICF provides a valuable, though practically limited, framework for organizing patient functioning information.
• Developed ICF-based NLP technologies have significant implications for disability management, clinical care, and research.