Are large language models capable of generating safe and guideline-concordant rehabilitation recommendations in orthopaedics and sports medicine? A review
Journal of Pre-Clinical and Clinical Research · 27.05.2026
DOI: 10.26444/jpccr/222263 ↗Kurz gefasst
Leitlinie in der Zeitschrift Journal of Pre-Clinical and Clinical Research, online erschienen am 27.05.2026. Verfasst von Bartosz Gołembiewski und 10 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Introduction and Objective.Large language models (LLMs) are being increasingly used in medicine, including orthopedics and sports medicine, where they may support the development of rehabilitation recommendations and making decisions on return to physical activity.However, concerns remain regarding safety and adherence to clinical guidelines.The aim of the review is to summarize the available evidence on the quality of LLM-generated recommendations, their concordance with current guidelines, and their potential clinical implications.Review Methods.A narrative review of the literature was conducted using the MEDLINE (PubMed) and Scopus databases, covering the period from November 2022 (after the widespread introduction of models such as ChatGPT) to 1 March 2026.English-language studies evaluating LLM-generated rehabilitation recommendations, their concordance with clinical guidelines, and safety aspects were included. Brief description of the state of knowledge.LLMs can generate coherent and often useful rehabilitation recommendations, although their quality is variable.Many studies report only partial concordance with clinical guidelines, with key elements -such as exercise parameters or progression criteria -frequently omitted.Another important limitation is the sensitivity of responses to prompt formulation.While the recommendations are generally reasonable, they often require specialist verification before their use in clinical practice.Summary.LLMs may serve as a valuable tool to support rehabilitation, particularly in patient education and treatment planning; however, they should not be considered a standalone source of clinical recommendations.Their use requires specialist oversight and further validation in clinical studies.
Zitieren
APA
Gołembiewski, B., Mulski, G., Manicka, M., Baranowicz, M., Mazurkiewicz, W., Borecka, Z., Sobczak, A., Szymczak, A., Jaworowicz, A., Piasecka, J., & Łapaj, Ł. (2026). Are large language models capable of generating safe and guideline-concordant rehabilitation recommendations in orthopaedics and sports medicine? A review. Journal of Pre-Clinical and Clinical Research. https://doi.org/10.26444/jpccr/222263
Vancouver
Gołembiewski B, Mulski G, Manicka M, Baranowicz M, Mazurkiewicz W, Borecka Z, et al. Are large language models capable of generating safe and guideline-concordant rehabilitation recommendations in orthopaedics and sports medicine? A review. Journal of Pre-Clinical and Clinical Research. 2026. doi:10.26444/jpccr/222263
BibTeX
@article{Goembiewski2026,
author = {Gołembiewski, Bartosz and Mulski, Grzegorz and Manicka, Martyna and Baranowicz, Michał and Mazurkiewicz, Weronika and Borecka, Zuzanna and Sobczak, Agnieszka and Szymczak, Alicja and Jaworowicz, Anna and Piasecka, Joanna and Łapaj, Łukasz},
title = {Are large language models capable of generating safe and guideline-concordant rehabilitation recommendations in orthopaedics and sports medicine? A review},
journal = {Journal of Pre-Clinical and Clinical Research},
year = {2026},
doi = {10.26444/jpccr/222263}
}
EUVIMED