Use of artificial intelligence large language models as a clinical tool in rehabilitation medicine: a comparative test case.

This is one textbook case, not a real patient or a series, so it cannot show how the tool performs in ordinary clinic work. The authors state the case was straightforward and real-world acute and chronic rehabilitation is more complex and uncertain. The authors describe the rehabilitation prescriptions as general. The model made a body-structure coding error and the authors report weakness in determining chronicity and selecting approaches based on prognosis prediction. Only two PMR clinicians reviewed the output, and the paper reports comments and comparison with a textbook standard. The authors state ChatGPT-4 may generate hallucinations, its knowledge stops at 2021, and LLMs are not suitable for replacing human decision-making.

Do not read this as evidence that ChatGPT-4 can write rehabilitation prescriptions or ICF codes without clinician review. It was one textbook case, the model made a body-structure coding error, and the authors state LLMs are not suitable for replacing human decision-making.