Effects of an LLM-based intelligent adaptive teaching program on clinical risk perception among newly employed nurses: a quasi-experimental study.

OBJECTIVE

To evaluate the effectiveness of a large language model (LLM)-based intelligent adaptive teaching program in improving adverse event risk perception among newly employed nurses.

METHODS

A prospective quasi-experimental study was conducted among 134 newly employed nurses recruited between July 2023 and July 2025 from a campus of a tertiary Grade A hospital in Zhejiang Province, China. Participants received training using an LLM-based intelligent adaptive teaching program. The Nurses' Risk Perception of Adverse Events Scale (NRPAES), the incidence of nursing adverse events, and training satisfaction were assessed before the intervention and 3 months after training. Differences in these outcomes before and after the intervention were analyzed. Statistical analyses were performed using SPSS version 26.0, with the chi-square test and Wilcoxon signed-rank test used as appropriate.

RESULTS

Following the LLM-based intelligent adaptive training, the adverse event risk perception score among newly employed nurses increased significantly from 14.95 (11.60, 17.29) before training to 16.22 (12.76, 19.00) after training (Z = -3.527, P < 0.001). The overall incidence of nursing adverse events decreased significantly from 36 events (1.01‰) before training to 28 events (0.58‰) after training (χ 2 = 5.060, P = 0.024). Visualization using a Sankey diagram further demonstrated an overall reduction in adverse events after the intervention. In addition, more than 90% of newly employed nurses reported satisfaction with each item of the satisfaction questionnaire for the LLM-based intelligent adaptive teaching program.

CONCLUSION

The LLM-based intelligent adaptive teaching program provides practical support for the individualized delivery of clinical risk perception training for newly employed nurses. The program may improve nurses' ability to perceive adverse event risks. These findings provide preliminary evidence supporting the integration of LLM-based tools into the design and delivery of nursing education programs.