Randomized Controlled Trials Evaluating AI in Clinical Practice: A Scoping Evaluation

Abstract

ABSTRACT Background Artificial intelligence (AI) has emerged as a promising tool in healthcare, with numerous studies indicating its potential to perform as well or better than clinicians. However, a considerable portion of these AI models have only been tested retrospectively, raising concerns about their true effectiveness and potential risks in real-world clinical settings. Methods We conducted a systematic search for randomized controlled trials (RCTs) involving AI algorithms used in various clinical practice fields and locations, published between January 1, 2018, and August 18, 2023. Our study included 84 trials and focused specifically on evaluating intervention characteristics, study endpoints, and trial outcomes, including the potential of AI to improve care management, patient behavior and symptoms, and clinical decision-making. Results Our analysis revealed that 82·1% (69/84) of trials reported positive results for their primary endpoint, highlighting AI’s potential to enhance various aspects of healthcare. Trials predominantly evaluated deep learning systems for medical imaging and were conducted in single-center settings. The US and China had the most trials, with gastroenterology being the most common field of study. However, we also identified areas requiring further research, such as multi-center trials and diverse outcome measures, to better understand AI’s true impact and limitations in healthcare. Conclusion The existing landscape of RCTs on AI in clinical practice demonstrates an expanding interest in applying AI across a range of fields and locations. While most trials report positive outcomes, more comprehensive research, including multi-center trials and diverse outcome measures, is essential to fully understand AI’s impact and limitations in healthcare.

Ähnliche Arbeiten