Pragmatic Approaches to the Evaluation and Monitoring of Artificial Intelligence in Health Care: A Science Advisory From the American Heart Association

Abstract

The rapid development and integration of artificial intelligence (AI), including predictive, generative, and emerging agentic tools, into cardiovascular and stroke care is outpacing traditional evaluation frameworks and the generation of robust clinical evidence. This science advisory addresses the urgent need for pragmatic, risk-proportionate approaches to the evaluation and monitoring of health care AI. AI implementation practices often rely on real-world or anecdotal evidence, with considerable variability in local validation, bias assessment, and postdeployment monitoring. Several evaluation frameworks exist, but they can be difficult to operationalize, especially outside of well-resourced health systems. We propose and discuss evaluation across 3 phases: predeployment, implementation, and postdeployment. We also provide 4 pragmatic guiding principles for health systems that are beginning to set up AI governance processes, including strategic alignment, ethical evaluation, usefulness and effectiveness evaluation, and financial performance, to inform health system selection, validation, deployment, and actionable monitoring of AI tools. The American Heart Association's extensive hospital and volunteer network and commitment to evidence-based practice position it as a trusted leader in advancing responsible AI governance. By grounding evaluation and monitoring in these principles, this science advisory aims to ensure that AI adoption in health care is safe, effective, equitable, and sustainable, ultimately improving patient outcomes and supporting high-quality AI-enabled care.

Ähnliche Arbeiten