Epic Cosmos
Epic Cosmos is the largest repository of patient data in the world. Organizations choose to participate in Cosmos and contribute a HIPAA‐defined limited data set for all patients to the centralized Cosmos database. Every Cosmos portal user gets access to the community wide data set that will continue to grow over time. Users explore the multi-center data in Cosmos using SlicerDicer, Epic's data exploration tool. With additional training, researchers can extract de-identified row-level data into a Cosmos Data Science Virtual Machine, with access to programs such as R and SQL.
To access Cosmos, UVA users must:
• Have a uvahealth.org email
• Be a UVA Health Epic user with access to SlicerDicer in UVA EPIC Hyperspace ("Clinical View Only" is recommended for users without other Epic responsibilities)
• Have an EPIC UserWeb account
• Submit UVA Cosmos Access Request
• There must be an available UVA Cosmos "seat" (note that we are limited to 10 seats per UVA super user, so consider becoming a super user!)
Patients are included in Epic Cosmos if there were seen at participating institution and if they meet base criteria (at least two visits at least two years apart). The earliest available patient data starts in January 2017. Actual data start dates, quality, missingness, and recording practices may vary by site. Patients seen at multiple sites have records linked when possible (deduplication). Patient data is refreshed every two weeks.
Exploratory queries can be run on limited dataset data using the Cosmos SlicerDicer interface. Only "green lit" data sources that have passed Epic Cosmos quality checks are used in these queries and results are presented in aggregate.
All research on patient row-level (source data) has to be done in an approved workspace inside of a Cosmos Data Science Virtual Machine. The data science environment presents de-identified data as tables using a Cosmos DS version of Caboodle. Note that access to the Data Science environment requires additional training that can be found here in the Epic Course Catalog. The researcher must cover this cost. Cosmos may also charge for the use of the data, depending on the funding associated with the research project.
When your study is done, you'll want to share your findings with the medical and scientific communities. Before doing so, please review the Cosmos Publication Checklist, accessible via Galaxy and the Cosmos Portal. After you've checked your study against the checklist, you're ready to reach out to transfer your files. Results of your analysis must be aggregated before export so that no individual patient data is retained. Ensure that the files meet the following guidelines, then contact your Epic Cosmos representative to complete the file transfer process. All files will be vetted by the Epic Cosmos security team. Turnaround time is usually within one week. Requirements:
a) Contents of the files relate to your project
b) Files must not be overly large (there's no cap, but a 50MB file is subject to more scrutiny than a file that is 10s of KB)
c) Only consist of summarized data like counts or percentages. Exporting line‐level data is prohibited.
d) Low counts (<10) must be translated to rates or obfuscated when demographics are included.
e) Low counts may be included if they are required for the type of calculation to be useful and don't pose additional re‐identification risk.
f) Demographic data must be dispersed enough not to be organization‐identifiable.
Post-Export Analyses
• After the AGGREGATE data are reviewed and approved by EPIC Cosmos security team and extracted from Cosmos, they can be freely share and used as aggregate data for other analyses.
For guidance related to this resource and assistance with EPIC Cosmos Slicer Dicer feasibility queries visit PHS @ the Library.
For users conducting analyses in a Cosmos Data Science Virtual Machine, the Research and Clinical Trial Analytics (RCTA) team can help build your project dataset. This support is provided at their regular hourly rate.