Comparing individual and peer-supported ChatGPT use in nursing case analysis: Effects on nursing diagnostic skills and cognitive load in a randomized controlled trial.
Level 2 - randomized trial
Individual randomized controlled trial with three parallel groups.
PubMed 42636698 · doi:10.1016/j.nepr.2026.104949
What was done
A pretest-posttest, parallel-group randomized controlled trial evaluated 108 eligible nursing students (91 completed analyses) randomly assigned to three conditions during a 3-week case-based surgical learning curriculum: Individual work, Individual + ChatGPT, or Peer + ChatGPT. Primary outcomes included nursing diagnostic skills (Nursing Diagnosis Evaluation Form), cognitive load (Cognitive Load Measurement Scale), and agent perceptions (Agent Persona Instrument).
What was found
Diagnostic skills improved across all groups over time, but there was no statistically significant interaction effect among the study arms (p = .936). Total cognitive load decreased in all three groups. Germane cognitive load showed a descriptive increase in the Peer + ChatGPT group, but this difference did not reach statistical significance.
Why it matters
Despite interest in generative AI for clinical training, integrating ChatGPT—either individually or in peer pairs—did not provide measurable diagnostic skill gains compared with standard individual case analysis.
Limits
The sample was modest (n = 91 analyzed, with 17 dropouts from the initial 108), limited to a single educational context and surgical case scenarios over just three weeks. Long-term skill retention, clinical performance, and specific score values were not reported in the abstract.