JMIR Publications

ABSTRACT

Background:

Digital health transformation has accelerated AI-enabled training technologies—LLM-based tutoring, AI-embedded immersive simulation, and intelligent tutoring systems—in nursing education. However, evidence regarding their effectiveness in clinical nurse training beyond undergraduate settings remains fragmented. Key gaps include inconsistent outcome reporting, technology variability, and limited understanding of adoption determinants.

Objective:

This scoping review aimed to characterize AI applications in clinical nurse training, synthesize evidence regarding educational outcomes, and identify gaps related to technology adoption and implementation. The evidence base comprised clinical nurses and nursing learners training in clinically transferable competencies, per prespecified eligibility criteria.

Methods:

Following the Arksey and O’Malley framework and PRISMA-ScR guidelines, eight databases (PubMed, CINAHL, Wanfang, VIP, CNKI, CBM, Web of Science, and Europe PMC) were systematically searched from January 2020 to May 2026. A supplementary EBSCOhost search covering CINAHL Ultimate, MEDLINE Complete, and Education Source was conducted in September 2026 using model-specific terms. Three reviewers independently screened titles, abstracts, and full texts, with disagreements resolved through consultation with a fourth reviewer. Evidence synthesis was guided by the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) framework. Methodological quality was assessed using design-specific Joanna Briggs Institute (JBI) critical appraisal tools, AMSTAR-2 (A MeaSurement Tool to Assess Systematic Reviews), and a 6-item JBI Text and Opinion checklist for bibliometric and system-development reports.

Results:

Of 22,595 identified records, 127 studies were included (91 English-language studies, 71.7%; 36 Chinese-language studies, 28.3%), comprising 8 RCTs, 18 quasi-experimental studies, 10 systematic reviews, 13 scoping reviews, and 78 studies using other designs. AI applications spanned four domains: LLM-based tutoring, AI-embedded simulation, adaptive learning, and competency assessment. In two small nursing-student RCTs, a study-reported Cohen d=0.96 (empathy/communication outcomes) and an author-calculated Hedges g=0.70 (perceived clinical competency) were identified. Methodological quality varied: RCTs scored 5–10/13 on the JBI checklist, and all ten systematic reviews had AMSTAR-2 limitations. UTAUT2 mapping showed entirely indirect (proxy) support for performance expectancy and no direct or indirect/contextual evidence for facilitating conditions, price value, or habit.

Conclusions:

AI-enabled training technologies show promising educational benefits in clinical nurse training; however, current evidence remains largely derived from non-specialist populations. Before broader clinical implementation, three major gaps require further investigation: population representativeness, implementation determinants (the implementation gap), and methodological rigor. Seven research priorities are proposed.