How to Export Data
Brim exports data as a "Comma-separated values" or CSV file.
To export your data, go to the “Data Export” section and select “Create Export File”.
Once the downloadable file is ready, a “Download Result” link will appear at the right of the Export History table.
Exported results include all LLM generated labels, the raw text associated with those labels, and the verdict from the human reviewer.
The export CSV has one line for each abstracted variable instance. The format is:
Name: This is the name of the variable
Type: The level of variable. Options:
- Document-level (Many per note or One per note)
- Patient-Level (One per patient)
- Patient-Level Dependent Variable
Patient_id: The patient ID imported originally.
Document_id: The document ID imported originally.
Value (Patient-Level only): The value if this is a patient-level variable. Otherwise, "—"
Status (Patient-Level and Dependent Variable only): The accept/edit/remove status of a reviewed variable instance.
Raw JSON Result (Document-Level and Dependent Variable only): The list of variable instances along with their raw text, confidence, and associated values in JSON format. Only for document-level variables and Patient-Level Dependent Variables.
Associated Text (Patient-Level and Dependent Variable only): the LLM reasoning for aggregated variables and dependent variables.
Approved: TRUE if a human reviewer selected Accept/Edit/Remove. Otherwise, FALSE.
Approving User: The username of the abstractor. Only applies if Approved is TRUE.
Create Time: The timestamp the variable instance was originally generated.
Edit Time: The timestamp the variable instance was edited.
Approve Time: The timestamp the variable instance was approved (only if Approved is TRUE).
Document-Level Variable Instances:
For Document_level variables, there is a group of columns for each instance of the variable in that document:
Value_0: The value for the first instance in that document.
Status_0: The status for the first instance in that document (accept/edit/remove)
Raw_text_0: The supporting raw text for the first instance in that document