Large Language Models Versus Human Examiners for Grading Physiotherapy Clinical Cases

Condition(s) studied

Investigational drug(s) / intervention(s)

LLM-based assessment: Assessment of each anonymized examination by three large language models (for example, Claude, ChatGPT, and Gemini, in the versions available during data collection). Each model receives an identical standardized prompt embedding the study rubric and returns a score per criterion, a global score, and structured qualitative feedback. Each model is queried in duplicate in independent sessions under fixed generation parameters to estimate intra-model (test-retest) reliability, and outputs are compared across models to estimate inter-model agreement.

Faculty assessment (reference standard): Assessment of the same anonymized examinations by faculty with expertise in the course, applying the identical rubric, serving as the reference standard. In the preferred scenario, two faculty members score each examination independently (paired human correction); if faculty workload precludes this, a single expert faculty rating, or the official course grade already assigned, is used as the reference. Faculty and LLM raters are blinded to one another's scores.

Study summary

This study evaluates whether large language models (LLMs) can reliably assess written clinical-reasoning case examinations completed by undergraduate physiotherapy students, compared with faculty assessment. In the course "Specific Methods in Physiotherapy" (third year of the Physiotherapy Degree), students solve complex clinical cases that require clinical reasoning, technical knowledge, and therapeutic decision-making. These cases are traditionally graded by faculty, a time-consuming process that may show inter-rater variability.

A set of de-identified student case examinations will be assessed using the rubric currently applied in the course, which covers clarity and structure of clinical reasoning, integration of the biopsychosocial model (ICF and APTA frameworks), accuracy in identifying pain mechanisms, coherence between diagnosis, hypotheses, and treatment, originality and depth of analysis, and professional writing. Each examination will be scored independently by three LLMs (for example, Claude, ChatGPT, and Gemini), each receiving an identical standardized prompt that embeds the same rubric, and by faculty serving as the reference standard.

To avoid overloading faculty, full double human grading may not be feasible; the human reference will therefore consist of expert faculty grading by one independent rater or, when resources allow, two independent raters. In contrast, paired assessment is fully implemented across the AI models: each examination is scored by several LLMs, and each model is queried in duplicate, allowing the study to estimate agreement between models and the test-retest stability of each model.

The primary aim is to quantify agreement between LLM-generated scores and the faculty reference score. Secondary aims include agreement among the LLMs, test-retest reliability of each model, criterion-level agreement, the quality and usefulness of the qualitative feedback generated, the time and cost associated with each approach, and students' perceptions of the usefulness of human versus AI feedback.

The findings will clarify the strengths and limitations of LLMs as supportive tools for formative assessment in health-professions education and will inform criteria for their responsible and effective use. No LLM output will affect students' official grades, which remain the sole responsibility of faculty.

Eligibility

Primary outcome measure(s)

• Agreement between LLM global scores and the faculty reference global score — Single cross-sectional assessment during the data-collection period (approximately 2 months)
  Agreement between the global examination score generated by each large language model (LLM) and the faculty reference global score, computed for the same anonymized examinations. Agreement is quantified with the intraclass correlation coefficient (ICC), two-way random-effects model, absolute-agreement definition, single- and average-measures forms \[ICC(2,1) and ICC(2,k)\], with 95% confidence intervals. Systematic bias is examined with Bland-Altman analysis (mean difference and 95% limits of agreement). ICC is interpreted as poor (\<0.50), moderate (0.50-0.75), good (0.75-0.90), or excellent (\>0.90). Pre-specified target: ICC \>= 0.75.

Trial sites (1)

More Neuron, Spain trials in Spain