JMIR Publications

ABSTRACT

Background:

Generative artificial intelligence (GenAI) is increasingly used in medical education, including clerkship settings where learning is closely tied to patient safety. Although GenAI may assist students with reasoning, documentation, and task preparation, its educational value in safety-critical clinical environments remains unclear.

Objective:

This study aimed to examine how GenAI-related factors are related to patient safety learning and ward-based infection-prevention practice among medical clerkship students, with particular attention to the mediating roles of self-efficacy in AI-generated usage and perceived AI self-control balance.

Methods:

A sequential exploratory mixed-methods design was used. In the qualitative phase, interview findings were used to identify and contextualize key GenAI-related constructs relevant to medical clerkship learning. These constructs then informed the quantitative phase, in which data from 214 active medical clerkship students from private universities in the Jakarta Metropolitan Area were analyzed using partial least squares structural equation modelling. Participants were recruited through purposive sampling. The model examined the relationships of perceived algorithmic fairness, trust calibration literacy, explainability quality exposure, and privacy concern toward AI monitoring on safety-oriented learning through human-centered mediators. Promotion of patient safety was tested as a moderator.

Results:

Among 214 medical clerkship students, self-efficacy in AI-generated usage (β=.477, P<.001) and perceived AI self-control balance (β=.407, P<.001) showed the strongest positive associations with ward-based infection-prevention practice. Explainability quality exposure was positively associated with self-efficacy in AI-generated usage (β=.279, P<.001) and perceived AI self-control balance (β=.401, P<.001). Perceived algorithmic fairness was also positively associated with self-efficacy in AI-generated usage (β=.185, P=.004) and perceived AI self-control balance (β=.236, P=.008). Promotion of patient safety strengthened the association between self-efficacy in AI-generated usage and ward-based infection-prevention practice (β=.122, P=.044). By contrast, the paths involving trust calibration literacy and privacy concern toward AI monitoring were not supported.

Conclusions:

GenAI may support patient safety learning in medical clerkship, but its value appears to depend less on the technology itself than on students’ confidence in using it and their ability to retain human oversight. These findings support integrating GenAI into clinical training through structured supervision, verification, and explicit attention to infection-prevention practice.