JMIR Med Educ. 2026 Feb 5;12:e75534. doi: 10.2196/75534.
ABSTRACT
BACKGROUND: Artificial intelligence (AI) shows promise in clinical diagnosis, treatment support, and health care efficiency. However, its adoption in real-world practice remains limited due to insufficient clinical validation and an unclear impact on practitioners’ competence. Addressing these gaps is essential for effective, confident, and ethical integration of AI into modern health care settings.
OBJECTIVE: This study aimed to evaluate the effectiveness of informed AI use, following a tailored AI training course, on the performance of general practitioners (GPs) and internists in test-based clinical competence assessments and their attitudes toward clinical AI applications.
METHODS: A pre-post intervention study was conducted with 326 physicians from 39 countries. Participants completed a baseline test of clinical decision-making skills, covering diagnosis, treatment planning, and patient counseling; attended a 1.5-hour online training on effective AI use; and then took a similar postcourse test with AI assistance permitted (GPT-4.0). Test performance and time per question were compared before and after the training. Participants also rated AI accuracy, efficiency, perceived need for structured AI training, and their willingness to use AI in clinical practice before and after the course.
RESULTS: The average test scores improved from 56.9% (SD 15.7%) to 77.6% (SD 12.7%; P<.001), and the pass rate increased from 6.4% (21/326) to 58.6% (191/326), with larger gains observed among GPs and younger physicians. All skill domains (diagnosis, treatment planning, and patient counseling) improved significantly (all P<.001), while time taken to complete the test increased slightly from before to after the course (mean 40.25, SD 16.14 min vs 42.29, SD 14.02 min; P=.03). By the end of the intervention, physicians viewed AI more favorably, reporting increased confidence in its accuracy and time efficiency, greater appreciation for the need for structured AI training, and increased confidence and willingness to integrate AI into patient care.
CONCLUSIONS: Informed use of AI, based on tailored training, was associated with higher performance in test-based clinical decision-making assessments and greater confidence in using AI among GPs and internists. Building on previous research that often lacked structured training, focused primarily on model performance, or was limited in clinical scope, this study provides empirical evidence of both competence and perceptual improvement following informed AI use in a large, multinational cohort, enhancing the generalizability. These findings support the integration of structured AI training into medical education and continuing professional development to improve clinical performance and promote competent use of AI in clinical practice.
PMID:41643188 | DOI:10.2196/75534
