CIRCE Data
Overview
The CIRCE dataset includes a complete record of all hospital-based encounters at Penn Medicine from July 1, 2017 to June 30, 2026, with encounters at Lancaster General Hospital from July 1, 2011 to June 30, 2026, and outpatient encounters at Penn Medicine from July 1, 2017 to June 30, 2026 (as of CIRCE version 20260630). The dataset contains structured data associated with each encounter such as laboratory results, diagnostic codes, and vital signs. It also contains unstructured data such as the text of clinical encounter notes.
The data are currently stored in three different locations that you can access depending on your computational workflows: Penn Medicine DataBricks, PMACS Shared Drive, and the CCEB LPC cluster.
CIRCE is organized by the level of data sensitivity:
- De-identified: direct identifiers (e.g., CSNs, MRNs) are disguised, date-times are zero-indexed at admission, and note text data is not present.
- Limited: direct identifiers are disguised, date-times are preserved, and note text data is not present.
- Identifiable: direct identifiers are preserved, date-times are preserved, and note text data is present.
Each level will vary in the available information (e.g., direct identifiers are only available in the identifiable level) and the pre-requisites necessary for access to CIRCE.
No CIRCE data can ever be copied to a new location (e.g. your laptop or another desktop machine). If you need to copy CIRCE data to a new location for your research (e.g. to a new computing environment, share data outside of Penn, or to merge with another dataset) please review your use case with the Penn IRB and Office of Research Services Privacy Office to ensure the necessary data use agreements and protections are in place, and notify the CIRCE study team
Documentation
You can read more details about CIRCE and data in our Quick Start Guide, which overviews CIRCE, and our Data Dictionary, which details the variables contained in each CIRCE data table -- find these on our Project Resources page!