Evaluating the Real-World Clinical Performance of AI

As artificial intelligence (AI) continues to transform healthcare, its shift from controlled research environments to real-world clinical settings presents both immense promise and critical challenges. This collection in npj Digital Medicine showcases cutting-edge research that rigorously examines how AI systems perform in diverse, dynamic, and often unpredictable clinical contexts.

We invite contributions that explore the following areas:

• Clinical utility: Demonstrating how AI tools enhance diagnostic accuracy, inform treatment decisions, and improve patient outcomes in prospective or retrospective evaluations
• Safety and reliability: Assessing risks, unintended consequences, and robustness of AI systems in real-world deployment
• Equity and fairness: Evaluating performance across diverse populations to uncover and mitigate algorithmic bias
• Scalability and generalizability: Understanding how AI systems adapt across institutions, specialities, and care settings
• Workflow integration: Investigating how AI fits into clinical processes, team dynamics, and decision-making, and understanding barriers and facilitators to AI adoption in clinical workflows.
• Post-deployment monitoring and evaluation: Including model updating, performance decay, feedback loops, audits, prospective trials, and longitudinal studies of AI in practice over time
• Transparency and reproducibility: Promoting open methods, interpretable models, and replicable results to foster trust and accountability
• Patient-centered outcomes: Measuring impact of AI on care quality, experience, and health equity
• Human factors and clinician-AI interaction: including trust, interpretability, and decision support dynamics.

By spotlighting real-world evidence, this collection aims to bridge the gap between algorithmic innovation and clinical impact—ensuring that AI technologies not only work in theory but truly deliver value in clinical practice.