AI in Medicine—JAMA’s Focus on Clinical Outcomes, Patient-Centered Care, Quality, and Equity

Abstract

The transformative role of artificial intelligence (AI) in health care has been forecast for decades, 1 but only recently have technological advances appeared to capture some of the complexity of health and disease and how health care is delivered. 2ecent emergence of large language models (LLMs) in highly visible and interactive applications 3 has ignited interest in how new AI technologies can improve medicine and health for patients, the public, clinicians, health systems, and more.The rapidity of these developments, their potential impact on health care, and JAMA's mission to publish the best science that advances medicine and public health compel the journal to renew its commitment to facilitating the rigorous scientific development, evaluation, and implementation of AI in health care.JAMA editors are committed to promoting discoveries in AI science, rigorously evaluating new advances for their impact on the health of patients and populations, assessing the value such advances bring to health systems and society nationally and globally, and examining progress toward equity, fairness, and the reduction of historical medical bias.Moreover, JAMA's mission is to ensure that these scientific advances are clearly communicated in a manner that enhances the collective understanding of the domain for all stakeholders in medicine and public health. 4For scientific development of AI to be most effective for improving medicine and public health requires a platform that recognizes and supports the vision of rapid cycle innovation and is also fundamentally grounded in the principles of reliable and reproducible clinical research that is ethically sound, respectful of rights to privacy, and representative of diverse populations. 2,3,5he scientific development in AI can be viewed through the framework used to describe other health-related sciences. 6n these domains, scientific discoveries begin with identifying biological mechanisms of disease.Then inventions that target these mechanisms are tested in progressively larger groups of people with and without diseases to assess the effectiveness and safety of these interventions.These are then scaled to large studies evaluating outcomes for individuals and populations with the disease.This well-established scientific development framework can work for research in AI as well, with reportable stages as inventions and findings move from one stage to the next.The editors seek original science that focuses on developing, testing, and deploying AI in studies that improve under-