Truace
Subscribe

Effectiveness of Generative AI-Enabled Teaching for Clinical Reasoning in Health Care Education: Protocol for a Systematic Review and Meta-Analysis

Background: Clinical reasoning skills are key competencies required by health care professionals. Generative AI (GenAI)-enabled teaching may be able to offer personalized learning; however, its effectiveness compared with traditional teaching methods for developing clinical reasoning capabilities remains underexplored. Objective: This systematic review and meta-analysis will evaluate, synthesize, and compare the evidence on effectiveness in clinical reasoning capabilities between GenAI-enabled teaching and tradi…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1554· cite & verify
Effectiveness of Generative AI-Enabled Teaching for Clinical Reasoning in Health Care Education: Protocol for a Systematic Review and Meta-Analysis
Reading Wikipedia in the Classroom for Secondary School Students 07 by James Rhoda. CC BY-SA 4.0 · https://creativecommons.org/licenses/by-sa/4.0

In brief

This protocol sets out a planned systematic review and meta-analysis comparing generative AI-enabled teaching with traditional teaching for developing clinical reasoning among health care professionals. It specifies the databases to be searched, eligible study designs, independent screening by two reviewers, and risk-of-bias tools.

The topic matters because clinical reasoning is a core competency and institutions are adopting generative AI teaching before the evidence is settled. No pooled effect is reported yet, so whether these methods outperform traditional teaching remains unknown until the review is completed.

Main points

  1. The primary outcome is objective clinical reasoning performance, with attention to validity for the construct, population, and intended purpose.
  2. Eligible designs include randomized trials, non-randomized trials, quasiexperimental studies, and cohort studies with suitable controls.
  3. A random-effects model with the Hartung-Knapp-Sidik-Jonkman method will pool effects, and heterogeneity will be assessed with I2 and τ2.
  4. Searches retrieved 4831 studies, and 42 were included in the final review after screening.
  5. Subgroup analysis may consider the type and functionality of generative AI and the professional background of learners.

The problem

Health care education is adopting generative AI teaching methods faster than evidence shows they build clinical reasoning better than traditional teaching.

The rundown

The review was funded in September 2025, a trial MEDLINE search was completed in December 2025, and final database searches finished in February 2026. Searches covered Embase, MEDLINE, CINAHL, Scopus, and Cochrane CENTRAL, supplemented by Google Scholar, reference lists, and citation searches.

Of 4831 retrieved records, 2083 duplicates were removed and 2558 were excluded at title and abstract screening. A total of 190 underwent full-text screening and 42 were included.

Risk of bias will be assessed with ROB 2 for randomized trials and ROBINS-I for other designs. Where studies are too diverse, results will be presented narratively, and prediction intervals may show how true effects vary. Completion is expected in autumn or winter 2026.

Sources

  1. Peer-reviewedJMIR Research Protocols2026-10-15

The debate