Linking Free Text Documentation of Functioning and Disability to the ICF With Natural Language Processing.
Denis Newman-Griffis, Jonathan Camacho Maldonado, Pei-Shu Ho and 4 others
PMID 35694445WHAT IT FOUND
Natural language processing models mapped free-text disability claim notes to ICF Mobility, Self-Care, and Domestic Life codes for many common activities, but the ICF category for looking after one's health was too broad.
Key findings
01For automatic ICF coding, classification models outperformed candidate selection models, and giving the model the exact location of the activity word improved performance.
02The best models achieved high automated coding performance for several common activities, including walking, maintaining body position, driving, dressing, preparing meals, and assisting others.
03Looking after one's health dominated the Self-Care/Domestic Life coding, accounting for 43.6% of actions, and the authors found it too broad for practical use.
STILL TO COME
How it was doneWhat they foundWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The records were disability benefit claim documents, not ordinary electronic health record notes in most health systems. The people described were claimants for federal disability benefits, not a general rehabilitation population. Many documents had noise from converting scanned pages to text. The ICF categories themselves caused problems: some activities did not fit cleanly, and looking after one's health was too broad. The study evaluated text coding performance, not whether the systems help patient care.
Declared interests
Funded by the NIH Intramural Research Program and the U.S. Social Security Administration. The authors declared no commercial or financial relationships that could be construed as a conflict of interest.
The easy way to misread this
Do not conclude that these models are ready for routine clinical use or that ICF coding of free text is clinically validated. They were tested on OCR-processed disability claim documents, not on ordinary clinical notes or patient care outcomes.