JMIR Publications

ABSTRACT

Background:

Generative artificial intelligence (GenAI) is increasingly used by health professional students during clinical placements. However, prevalence estimates and item-level descriptions do not indicate which dimensions of use and perception vary according to students’ perceived preparedness, which are shaped by professional training context, and which are broadly shared. Distinguishing these dimensions is necessary to design GenAI curricula combining common interprofessional foundations with readiness- and profession-sensitive components.

Objective:

To explore whether, and how, health professional students’ use and perceptions of GenAI during clinical placements are associated with self-perceived GenAI maturity, health profession, or both.

Methods:

We conducted a cross-sectional secondary analysis of a web-based survey of 388 medicine, pharmacy, nursing, midwifery, and physiotherapy students at Université Grenoble Alpes (July 17 to September 30, 2025). The main explanatory variables were self-perceived GenAI maturity, modeled as an ordinal 4-level index, and health profession. Outcomes were self-reported GenAI use during clinical placements, five composite perception indices, and five sentinel items. Logistic regression was used for reported use, and ordinary least squares regression with HC3 robust variance estimators was used for composite outcomes. Models were adjusted for health profession, gender, training cycle, and maturity as appropriate. Two pre-specified families of tests were Holm-corrected: 5 maturity effects and 20 profession contrasts.

Results:

Overall, 204 of 388 respondents (52.6%) reported GenAI use during clinical placements. Each 1-level increase in self-perceived maturity was associated with higher odds of use (adjusted OR 2.76, 95% CI 1.95-3.90; P<.001), with adjusted predicted probability rising from 17.7% for minimal maturity to 80.4% for high maturity. Across five composite perception outcomes, maturity remained associated after Holm correction with higher practical/informational benefit scores (β=0.165, 95% CI 0.052-0.278; Holm P=.017) and operational governance scores (β=0.122, 95% CI 0.043-0.201; Holm P=.013). The association with clinical benefits did not survive multiplicity correction. Health profession was not significantly associated with self-reported GenAI use after adjustment. However, nursing students reported lower practical/informational and clinical benefit scores than medical students (both Holm P=.016). Professional/cognitive risks were not differentiated by maturity, profession, or training cycle. Confidentiality concerns and the view that patients should be informed when AI is used in care were broadly endorsed, while a meaningful subgroup remained uncertain about the ethical legitimacy of GenAI use.

Conclusions:

GenAI use and related perceptions during clinical placements followed partly distinct patterns: self-reported use, practical/informational benefit perceptions, and operational governance perceptions were primarily maturity-sensitive; selected benefit perceptions were profession-sensitive; and perceived professional/cognitive risks and key ethical concerns were broadly shared across student profiles. GenAI curricula should therefore combine a common interprofessional foundation, readiness-sensitive progression, and profession-specific learning activities, supported by supervisor training and institutional governance.