Abstract:

Background and objectives:
The transition from theoretical knowledge to autonomous clinical practice remains a long-standing challenge in physiotherapy education, particularly for the development of clinical reasoning and the therapeutic alliance. Within the course Specific Methods in Neuromusculoskeletal Physiotherapy, we designed a structured methodology aimed at transforming clinical observation into an active learning process anchored on the International Classification of Functioning, Disability and Health (ICF) framework. The objective of this study was to evaluate the educational efficacy and predictive validity of complementing this framework with an AI-based patient simulator powered by a large language model (LLM).

Methods:
A quasi-experimental pilot study was conducted with 37 final-year physiotherapy students. The intervention combined a dual-component assignment, which integrated ICF-based clinical assessment with self-directed functional anatomy monographs, with an LLM-based simulator configured to portray three patients with complex chronic pain. The intervention group (n = 28) completed six interactions with the simulator before the final clinical examination, while the control group (n = 9) followed the standard curriculum. Clinical performance was assessed by tutors using the Lasater Clinical Judgment Rubric (LCJR).

Results:
The intervention group obtained significantly higher LCJR scores (M = 30.9, SD = 7.3) than the control group (M = 24.2, SD = 4.6), with a large effect size (Cohen's d = 0.99). Performance within the AI simulator showed a moderate-to-strong positive correlation with real-patient clinical scores (rho = 0.618, p < 0.001), supporting its predictive validity. Notably, no association was found between student satisfaction and objective clinical performance (rho = -0.07).

Conclusions:
Embedding LLM-based simulation within a structured ICF-based curriculum appears to enhance clinical reasoning and therapeutic alliance, shifting the student from passive observer to active clinical analyst. The tool offers a scalable, evidence-informed bridge between theoretical learning and professional practice. Further randomised controlled trials are warranted to confirm and extend these findings.

Keywords:

Artificial intelligence, clinical simulation, large language model, physiotherapy education, teaching innovation, therapeutic alliance.