Using graph machine learning to identify functioning in patients with low back pain in terms of ICF

Area of Science:

• Health Informatics
• Machine Learning Applications
• Rehabilitation Medicine

Background:

• The World Health Organization's International Classification of Functioning, Disability, and Health (ICF) offers a standardized framework for assessing patient functioning.
• The complexity of the ICF classification hinders its widespread adoption in clinical practice.
• Automating ICF code extraction from electronic health records (EHRs) is crucial for efficient implementation.

Purpose of the Study:

• To evaluate the efficacy of the Headai Graphmind graph machine learning engine in identifying ICF codes within Finnish EHRs.
• To assess the performance of the machine learning model against a domain expert's manual coding.
• To determine the potential of automated ICF coding for clinical practice and retrospective data analysis.

Main Methods:

• A dataset of 93 adult patients (18-65 years) with chronic low back pain was compiled.
• The Headai Graphmind engine was applied to a sample of 20 patient records to extract ICF codes.
• Performance metrics (precision, recall, F1 score) were calculated by comparing machine-extracted codes against those identified by a human expert.

Main Results:

• Headai Graphmind demonstrated high performance with a precision of 0.95, recall of 0.83, and an F1 score of 0.89.
• The engine identified 112 distinct ICF codes, closely aligning with the 119 codes identified by the domain expert.
• The study confirmed the capability of Headai Graphmind to recognize ICF codes in EHRs for chronic low back pain.

Conclusions:

• The Headai Graphmind engine shows significant promise for automating ICF code recognition in clinical settings.
• This technology can facilitate the practical implementation of the ICF classification system.
• Automated ICF coding can enhance the retrospective analysis of medical data, supporting research and quality improvement.