evidoria

← Back to browse

Good practice Imported

AI-Assisted Problem-Based Learning Lifts Knowledge Retention in Physiotherapy Education — Istanbul–Toledo RCT

Türkiye · Istanbul · See the Türkiye profile · See the Istanbul profile

Evidence: Randomised controlled trial Top 89% 27/100 · Ask Evidence Copilot about this practice

A 40-student RCT at Istanbul Medipol University and Universidad de Castilla-La Mancha found ChatGPT-assisted problem-based learning gave physiotherapy students far larger 2-week knowledge gains than traditional PBL (d=3.14), though clinical competence barely differed.

40
Students randomized
35
Trial completers
3.14 Cohen's d
Effect size, knowledge retention (Cohen's d)
1.70 Cohen's d
Effect size, AI self-efficacy (Cohen's d)
0.42 Cohen's d
Effect size, clinical competence (Cohen's d, not significant)

Details

Maturity
Pilot
Promoter
Istanbul Medipol University (with Universidad de Castilla-La Mancha, Spain)
Period
June 2025 (published October 2025)
Keywords
higher education, physiotherapy education, problem-based learning, clinical reasoning

Context

Developing realistic clinical scenarios for problem-based learning (PBL) is a core but time-intensive part of physiotherapy education, and instructors have limited tools to verify whether AI-assisted versions of this method actually improve what students retain.

Activities

In June 2025, Istanbul Medipol University (Turkey) and Universidad de Castilla-La Mancha (Spain) ran a prospective, two-arm randomized controlled trial: 40 physiotherapy students were randomized to either AI-assisted PBL, using ChatGPT-4.0 in small groups to build simulated patient scenarios, evaluation plans and management strategies for chronic low back pain under instructor guidance, or traditional PBL without AI. Of the 40 randomized, 35 completed the trial (19 AI-PBL, 16 traditional PBL).

Results

At 2-week follow-up, the AI-PBL group showed very large gains in knowledge retention (Cohen's d = 3.14) and AI self-efficacy (d = 1.70) relative to traditional PBL, and higher reading motivation (d = 1.31). Clinical competence, measured with the Mini Clinical Evaluation Exercise, was only marginally higher in the AI-PBL group and the difference was not statistically significant (d = 0.42).

Conclusions

The trial's strength is its randomized, peer-reviewed design; its limitation is scale -- 35 completers at two universities, a single clinical topic, and a two-week follow-up window -- so the very large effect sizes should be read as an early, promising signal rather than a settled result.

Implementation
Indicative cost
Low (< €50k) — Academic research trial using ChatGPT-4.0; no infrastructure cost beyond existing coursework.
Time to results
Short (< 1 year) — Trial conducted June 2025 with a 2-week follow-up.
Staffing & skills
Istanbul Medipol University researchers, Universidad de Castilla-La Mancha researchers

Conditions for success

  • Instructor guidance retained alongside AI tool use rather than unsupervised AI use

Common failure modes

  • Small completer sample (35) at two universities, single clinical topic, two-week follow-up window limits generalisation
  • No significant improvement in clinical competence despite gains in knowledge retention

Where it fits

Governance type
academic research collaboration
Scale
two universities (Turkey, Spain)
Income level
upper-middle-income / high-income

Replication kit

Reusable artefacts from this practice — as published by their sources.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful