Clinical stewardship for medical AI translation: A systematic review of technological, human, and organizational barriers
Clinical stewardship for medical AI translation: A systematic review of technological, human, and organizational barriers

原文摘要
出版方登记 · OpenAlexMedical artificial intelligence (AI) systems may perform well in controlled evaluations, yet challenges remain in their translation and implementation into routine clinical practice. This systematic review examines 247 original studies identified through searches limited to 2024-2025 publication dates and 31 clinical and translational journal titles. The most frequently coded translation-relevant evidence indicator is retrospective design (158 studies, 64.0%), followed by demographic or geographic data limitations (125, 50.6%), lack of qualifying independent external validation (71, 28.7%), small sample or data scarcity (51, 20.6%), and noisy or context-dependent labels (33, 13.4%). Categories are not mutually exclusive, and the estimates describe the sample defined by these filters, rather than all medical AI research. We integrate these findings with a "clinical stewardship" framework that distributes responsibility across clinical teams, developers, health systems, and patients while preserving the capacity of AI to reduce routine workload.
页面未能读取全文,以上为出版方登记的摘要;方法与图表见原文
来源:Cell Reports Medicine · doi.org