Written and maintained by CASRAI Editorial Board
Last updated
This guide outlines research data systems and regulatory frameworks; it does not constitute clinical or medical advice.
Epic Cosmos is a research and evidence-generation data network built from electronic health record (EHR) data contributed by healthcare organizations that use Epic software. Its published scale is about 310 million patients, and it is used for real-world evidence, observational research and cross-institution analysis.
What Is Epic Cosmos?
Cosmos is described in the peer-reviewed literature as a cross-institution data aggregation tool for customers of a single EHR vendor, Epic. The first description of the platform reported that data move to a central Cosmos host as encrypted HL7 C-CDA documents over Epic’s Care Everywhere network. A later paper in ACI Open (2026) reports that matches made through Care Everywhere are sent as cryptographically hashed IDs, which allows chart deduplication without introducing re-identification risk.
Researchers use Cosmos instead of querying a single institution’s data warehouse, for example to study rare conditions or to compare outcomes across many organizations. For background on the health data exchange standard that Epic and other vendors support, see Fast Healthcare Interoperability Resources (FHIR).
Who Runs It and What It Covers
Epic Systems operates the platform. Governance sits with a Cosmos Governing Council, which the 2026 ACI Open paper describes as a body of clinicians, researchers and informaticists from participating healthcare organizations. Its members are nominated for election by the participating community.
According to the official Cosmos site (checked November 7, 2026) and the published literature:
- Patients: The Cosmos site lists 310 million patients in the Cosmos community.
- Facilities: The site lists 2,334 hospitals and 51.9K clinics.
- Organizations: The 2026 paper reports about 300 million deduplicated patient records and 337 organizations as of April 2025, with participants in Canada, Lebanon, Saudi Arabia and the United States.
- Public research arm: Epic Research publishes studies, and most of them use the Cosmos database, which it describes as “a universe of electronic health record data that drives evidence-based research and individualized patient care.”
Patient counts change often, so check the official site for the current figure.
How Many Patients Are in Epic Cosmos?
The official Cosmos site lists 310 million patients (checked November 7, 2026). The 2026 ACI Open paper reports about 300 million deduplicated patient records as of April 2025, and notes that 117 million patients (39.03%) have five or more years of data. Deduplication means a person who received care at several participating organizations is intended to be counted once.
How Do I Access Epic Cosmos?
The Cosmos site has a “Request Access” link, and the platform is aimed at users at participating organizations. The published description of the researcher tools is:
- SlicerDicer: an Epic-developed no-code visualization and analysis application that works on a limited dataset. It reports small counts as “10 or fewer.”
- Data Science Virtual Machines (DSVMs): users “with required paid training” can use DSVMs to work with de-identified line-level data.
Which tools you can use depends on your organization’s participation and on its own approval process, so start with your institution’s informatics or research office. Researchers planning clinical studies on secondary data should also review Good Clinical Practice, clinical trial protocol design and clinical trial registration and reporting compliance.
Does Epic Cosmos Cost Money?
The 2026 ACI Open paper states that DSVM use requires paid training, and that “fees may be charged when funding is received or additional resources are needed.” It also notes that most work in Cosmos is unfunded. Specific prices are not published in the sources checked, so ask your organization or Epic for current costs.
Does Epic Cosmos Use AI?
The Cosmos site describes Look-Alikes, which connects providers across the Cosmos community who have treated patients with similar symptoms or conditions so they can collaborate on rare conditions or medical mysteries. The 2026 ACI Open paper adds that “early research indicates Cosmos may be capable of producing novel insights with generative AI technology.” See data masking for how de-identification techniques fit into protected data work.
Strengths and Limits
Strengths
- Scale: A very large patient base across many organizations supports studies of uncommon conditions.
- Longitudinal depth: 117 million patients have five or more years of data, according to the 2026 paper.
- De-identification: The expert-determination dataset meets HIPAA’s expert determination standard, using Gaussian noise on some values and a consistent random date shift per patient.
Limits
- Single vendor: The authors note bias inherent to a population drawn from organizations using one EHR vendor.
- Structured data: Unstructured and non-discrete data elements are not available.
- Contribution model and control: The first description lists an “all or none” contribution model and limited control over queries run on an institution’s data.
- Method risks: The 2026 paper also notes documentation variation, workflow differences and the risk of statistically significant results that are not clinically meaningful.
Epic Cosmos vs Alternative Research Networks
Other networks used in translational research differ in who operates them and which data model they use.
| Dimension | Epic Cosmos | TriNetX | PCORnet | N3C |
|---|---|---|---|---|
| Operator | Epic Systems with the Cosmos Governing Council | TriNetX, LLC (commercial) | PCORI-funded network of partner organizations | NIH NCATS-led |
| Data model | Epic-based, single EHR vendor | Proprietary harmonization | PCORnet Common Data Model | OMOP Common Data Model |
Frequently Asked Questions
What is Epic Cosmos?
Epic Cosmos is a research data network that aggregates EHR data from healthcare organizations using Epic software, used for evidence generation and cross-institution research.
How many patients are in Epic Cosmos?
The official site lists 310 million patients (checked November 7, 2026). A 2026 paper reports about 300 million deduplicated patient records and 337 organizations as of April 2025.
How do I access Epic Cosmos?
Use the “Request Access” link on the official Cosmos site and work through your organization’s research or informatics office. SlicerDicer provides aggregate views, and Data Science Virtual Machines require paid training.
Does Epic Cosmos cost money?
The published literature says DSVM use requires paid training and that fees may be charged for funded work or extra resources. No price list was found in the sources checked.
Does Epic Cosmos use AI?
The Cosmos site describes Look-Alikes for connecting providers who treated similar patients, and a 2026 paper notes early research on generative AI using Cosmos data.
Sources
All sources checked November 7, 2026.
- Epic Systems. Epic Cosmos. https://cosmos.epic.com
- Howat M, Martell S, Tarabichi Y, Kaelber DC, Chen SS, Freedman H. “Advancing Real-World Evidence Generation: Growth and Lessons Learned from the Early Years of Cosmos.” ACI Open. 2026;10(1):e10-e23. https://pmc.ncbi.nlm.nih.gov/articles/PMC13404524/
- Tarabichi Y, Frees A, Honeywell S, et al. “The Cosmos Collaborative: A Vendor-Facilitated Electronic Health Record Data Aggregation Platform.” ACI Open. 2021;5(1):e36-e46. https://pmc.ncbi.nlm.nih.gov/articles/PMC8775787/
- Epic Research. About Our Data. https://www.epicresearch.org/about-our-data
