Artificial Intelligence Tool Use and Perceptions Among Australian General Practitioner Trainees: A National Survey (Preprint)

Abstract

<sec> <title>BACKGROUND</title> Artificial intelligence (AI) is increasingly prevalent in healthcare. However, little is known about how frequently Australian general practitioner (GP) trainees (registrars) use AI or how they perceive its potential impact on training. </sec> <sec> <title>OBJECTIVE</title> This study aimed to explore how frequently Australian GP registrars use AI tools during training, what types of AI technologies they use, and their perceptions of AI’s benefits, risks, and training needs. </sec> <sec> <title>METHODS</title> A cross-sectional national online survey was conducted from May 7 to June 4, 2024. GP registrars (3743) enrolled in the Australian General Practice Training and the Fellowship Support programs run by the Royal Australian College of General Practitioners were invited to participate; 727 (19.5%) responded. The survey included 13 questions exploring self-reported frequency of AI use in clinical, educational, and personal contexts, practice AI governance, context and types of AI tool use, perceived impacts of AI on training, extent of AI-related education received, and ethico-legal concerns. Results are reported as percentages. </sec> <sec> <title>RESULTS</title> Most registrars reported infrequent or no use of AI. For clinical tasks, 7.3% (53/727) used AI at least weekly. Similarly, only 4.9% (36/727) used AI weekly or more for educational tasks. Practice-level adoption was limited, with 10.8% (78/725) of registrars reporting their practice subscribed to any AI tools. General large language models, remote patient monitoring devices and note-taking voice-to-text tools were the most common AI tools used in the clinic. For education, general large language models dominated, used by 63.5% of those using AI (141/222), far more than any other tool. Most registrars anticipated benefits from AI, particularly for learning efficiency and reducing administrative workload. However, 25% (181/724) believed AI could increase medicolegal risk. Formal training was rare; most had received minimal or none and thus relied on self-directed learning. Most registrars rated formal training in AI competencies as at least moderately important. </sec> <sec> <title>CONCLUSIONS</title> At the time of the survey, AI use among registrars was limited, though most recognized its potential. Enthusiasm was tempered by concerns about safety, ethics, and legal risks. Structured AI training and clear guidelines are needed to support safe, effective adoption in general practice and education. </sec>

Ähnliche Arbeiten