What You Should Know:
– Today, Epic announced the launch of a new form of healthcare intelligence, named Comet, developed to help clinicians, health systems, and patients make more informed decisions.
– Built on Epic Cosmos, a secure data platform, Comet is designed to predict what is likely to happen next in a patient’s journey by estimating disease risk, length of stay, and treatment outcomes. The platform has learned from more than 100 billion patient medical events to calculate these likely futures.
– The development of Comet, detailed in a paper on arxiv, was a collaboration with researchers from Yale and Microsoft. The platform’s goal is to address the uncertainty that clinicians face daily when making decisions about patient care, such as whether a patient should be admitted or discharged.
How Comet Works
Comet models are trained on sequences of time-ordered medical events, including diagnoses, labs, medications, and encounters, from millions of de-identified patient records and more than 100 billion data points. Using the same core technology as today’s large language models, Comet learns how clinical patterns evolve over time. When given a patient’s current state, it can generate a number of plausible future timelines that reflect real-world complexity, such as diagnoses resolving or emerging, complications arising, and care needs shifting.
These simulations will be summarized into insights and presented to clinicians in their workflows, providing a quantitative, data-driven view of what is likely to happen next. In a majority of the 78 cases evaluated, Comet was found to outperform individual models that were each specifically designed for one of the cases.
Designed for Security and Shared Learning
Comet is built entirely within Epic Cosmos, a secure, collaborative data platform for large-scale health research and innovation. Cosmos uses de-identified patient data and operates under rigorous privacy, security, and compliance standards. Because the platform is updated as care is delivered, the data powering Comet reflects evolving clinical practices and emerging trends across diverse health systems.
Operating within this governed environment, Comet can surface insights to participating health systems without compromising patient privacy. The platform’s architecture is designed to enable shared learning across the network while preserving privacy and security.
Availability
Starting in February 2026, researchers from participating Cosmos organizations will be able to explore Comet in a virtual lab to test new use cases and help advance its clinical relevance.
