Intelligence without intuition: a mixed-methods pilot study on reasoning models in musculoskeletal physiotherapy for low-back pain
-
By
-
March 18, 2026
Objective:
To evaluate the validity and reliability of specific reasoning models for clinical reasoning in musculoskeletal physiotherapy for low-back pain.
Key Findings:
- Reasoning models demonstrated sufficient test-retest reliability and were competent in clinical reasoning dimensions, but weaknesses in logical coherence, patient-centeredness, empathy, and intuition were identified, particularly in the domain of intuition.
Interpretation:
The findings highlight the potential of reasoning models in clinical reasoning but also underscore the need for improvements in areas such as empathy and intuition to better align with expert reasoning, suggesting a pathway for future model development.
Limitations:
- The study's scope was limited to specific case vignettes and reasoning tasks, which may not fully represent the diversity of clinical scenarios.
- Potential biases and inaccuracies in large language models due to web-scale pretraining could affect the reliability of the findings.
Conclusion:
A multidimensional framework is essential for evaluating language model outputs, and guidance for model selection and prompting strategies can enhance clinical reasoning performance, ultimately improving patient care.
