Epic announced today the launch of Comet, a healthcare intelligence platform designed to predict patient health journeys by analyzing disease risk, length of stay, and treatment outcomes. Developed in collaboration with researchers from Yale and Microsoft, the platform represents a significant advancement in clinical decision support technology, leveraging Epic's vast Cosmos data platform to address the daily uncertainties clinicians face when determining patient care pathways.
The platform's foundation rests on an unprecedented scale of medical data, having learned from more than 100 billion patient medical events across millions of de-identified patient records. Comet employs the same core technology underlying today's large language models, but specifically adapted for healthcare applications. The system processes sequences of time-ordered medical events including diagnoses, laboratory results, medications, and clinical encounters to understand how clinical patterns evolve over time.
When presented with a patient's current medical state, Comet generates multiple plausible future timelines that reflect real-world clinical complexity. These simulations account for various scenarios such as diagnoses resolving or emerging, potential complications arising, and shifting care needs. The platform then synthesizes these projections into actionable insights, presenting them directly within clinicians' existing workflows to provide a quantitative, data-driven view of likely patient trajectories.
In validation studies encompassing 78 different clinical cases, Comet demonstrated superior performance compared to individual models that were each specifically designed for single use cases. This broad applicability across diverse clinical scenarios underscores the platform's potential to enhance decision-making across multiple specialties and care settings.
Security and privacy considerations are central to Comet's architecture. The platform operates entirely within Epic Cosmos, which maintains rigorous privacy, security, and compliance standards while using de-identified patient data. This design enables shared learning across participating health systems without compromising patient privacy. As the platform receives continuous updates from ongoing care delivery, it reflects evolving clinical practices and emerging healthcare trends across diverse health systems, ensuring its predictions remain current and relevant.
Beginning in February 2026, researchers from participating Cosmos organizations will gain access to Comet through a virtual laboratory environment, where they can explore new use cases and contribute to advancing the platform's clinical relevance and applications.
Click here for the original news story.