Epic’s Approach to AI with Seth Hain

Why It Matters

Epic’s AI platform transforms massive de‑identified health data into real‑time clinical insights, accelerating personalized care and research while establishing trusted data‑sharing standards across U.S. hospitals.

Key Takeaways

• Epic leverages Cosmos to aggregate 300M de‑identified patient records
• Generative AI models trained on structured medical events improve with scale
• Trust and governance enable health systems to share data securely
• AI integration focuses on embedding insights directly into clinician workflows
• Epic’s R&D roadmap emphasizes cloud platforms and real‑time point‑of‑care analytics

Summary

In this NEJM AI Grand Rounds episode, Epic’s senior vice president of research and development, Seth Hain, outlines the company’s strategic approach to artificial intelligence. Central to the discussion is Cosmos, Epic’s de‑identified data repository that now contains over 300 million unique patient journeys, providing a foundation for large‑scale AI training and real‑world evidence at the point of care.

Hain explains how Epic has built generative medical‑event models that predict structured clinical sequences—diagnoses, orders, observations, and medications—rather than natural‑language text. The models demonstrate measurable performance gains as data volume and model size increase, underpinning applications such as Aura‑driven specialty‑diagnostic ordering and COVID‑test workflow automation. Governance is emphasized: a health‑system‑led council oversees data contribution, ensuring trust and prohibiting secondary uses beyond research and clinical decision support.

Illustrative examples include stitching patient records across multiple health systems to create comprehensive growth charts for rare diseases and enabling clinicians to instantly locate peers managing similar symptom constellations. Hain highlights the rapid rollout of generative AI in outpatient settings beginning in late 2022, aiming to augment clinician‑patient interactions and accelerate research initiatives.

The broader implication is that Epic’s AI infrastructure—built on cloud‑native platforms, scalable pre‑training pipelines, and tightly integrated workflow hooks—positions the company to drive personalized, data‑rich care across the U.S. health‑system landscape, while setting a benchmark for secure, collaborative data sharing.

Original Description

Clinical AI only helps patients if clinicians and health systems trust it. Seth Hain describes how Epic is building foundation models that respect institutional autonomy, minimize burden, and prioritize safety. He discusses scaling laws in structured medical data, cautious deployment for clinical interventions, and why understanding causality—not just correlation—is essential. This conversation reframes AI not as disruption, but as infrastructure for safer, more reliable care.

View the transcript and explore other episodes: https://www.podbean.com/eau/pb-atesa-1a43aa0

#podcast #artificialintelligence #aiinmedicine