Top Things to Know: Pragmatic Approaches to the Evaluation and Monitoring of Artificial Intelligence in Health Care
Published: November 10, 2025
Prepared by Paul St. Laurent, RN, DNP
- Artificial intelligence (AI) is rapidly transforming cardiovascular and stroke care, but its adoption is outpacing traditional evaluation frameworks, creating an urgent need for pragmatic approaches to evaluation and monitoring.
- AI tools are associated with unique risks, such as mechanistic opacity, performance changes over time, speed to scale of impact, and human-AI interaction effects that demand monitoring over time rather than one-time validation.
- Current evaluation practices for assessing AI tools are inconsistent and often insufficient, leaving significant gaps in ensuing safety, fairness, and effectiveness of AI-enabled care.
- While frameworks exist for evaluating and monitoring AI tools, they are often operationally burdensome and difficult to deploy, particularly for less well-resourced health care systems.
- Risk-tiering can help determine the intensity of evaluation and frequency of monitoring, balancing appropriate rigor with sustainability.
- There are three phases of evaluation: pre-deployment, implementation, and post-deployment, with rigor calibrated to the level of risk and context of use.
- AI governance should be anchored in four guiding principles: strategic alignment, ethical integrity, demonstrable usefulness and clinical effectiveness, and financial sustainability.
- Responsible, high-value AI use requires innovation not only in the development of novel AI tools, but also in our capacity to act on new information within existing or new clinical workflows.
- The American Heart Association is uniquely positioned to leverage its national network to advance real world evidence generation, pragmatic evaluation methods, and dissemination of best practices.
- Ultimately, responsible health care AI integration requires a risk-proportionate, principle-driven approach that advances clinical and patient outcomes, promotes equity, fosters trust, and ensures sustainable value.
Citation
Jain SS, Goto S, Hall JL, Khan SS, MacRae CA, Ofori C, Pegus C, Pencina M, Peterson ED, Schwamm LH; on behalf of the Council on Clinical Cardiology; Stroke Council; Council on Quality of Care and Outcomes Research; Council on Cardiovascular and Stroke Nursing; and Council on Lifelong Congenital Heart Disease and Heart Health in the Young. Pragmatic approaches to the evaluation and monitoring of artificial intelligence in health care: a science advisory from the American Heart Association. Circulation. Published online November 10, 2025. doi:10.1161/CIR.0000000000001400