Artificial intelligence (AI) has the potential to transform health care; however, successful integration of AI into health care requires overcoming obstacles, such as biases in data and AI models, and addressing challenges in generating sufficient clinical evidence for deployment. In this Viewpoint, we present a community-based, actionable framework for responsible and ethical development, deployment, and integration of AI-based solutions in health care, emphasising bias mitigation and clinical evidence generation. Our framework is intended for all members of the health-care team who interact with AI-based solutions, including software developers, data scientists, researchers, clinicians, hospital administrators, and institutional ethics and regulatory teams. We critically discuss the challenges associated with the use of such AI frameworks in health care. The framework, informed by multidisciplinary expertise, consists of four stages: (1) problem identification and study design, (2) model training and development, (3) silent deployment and clinical evaluation, and (4) operational deployment and lifecycle monitoring. This framework aligns with reporting standards such as SPIRIT-AI, CONSORT-AI, and TRIPOD+AI, offering practical steps for addressing biases, ensuring fairness, and validating clinical effectiveness. The framework provides action-oriented guidelines that can be used by institutions to support the ethical and efficient integration of AI into health care and equitable patient outcomes, either directly or by tailoring the guidelines with institution-specific resources.
Welch et al. (Sun,) studied this question.