Using Structured Data in Brim

Brim is designed to work with both unstructured notes (like clinical narratives), and convert it to structured data.

However sometimes you want to compare structured data (like lab values or medication lists) with the outputs from your notes. This article shows you how to bring in the data you already have in a structured format and evaluate it side-by-side with abstracted data from notes.

Uploading Structured Data

• Upload your structured data file the same way you would upload an unstructured note.
  • Brim can read and understand values in text tables. See an example below.
• If you have structured data from multiple dates (for example, labs from different encounters), you can enable longitudinal reasoning by uploading each date’s data as a separate note.
• Once uploaded, Brim treats the structured data like any other note: it becomes part of the project’s dataset.

Below is an example of a text table inside of the NOTE_TEXT column of a Brim upload data file. Exporting this as a CSV should put the NOTE_TEXT inside of quotes so that it imports correctly into Brim.

Using Variables with Structured Data

After upload, you can create variables in Brim to pull the information out of your structured "notes", just as you would with unstructured text.

For example:

• You can define a variable for “Hemoglobin value” and have it extract the field from the structured data.
• You can set up a variable for “Medication: Prednisone” and confirm its presence across structured entries.

This is a good opportunity to limit the document types for the variable; you can read more about that in Advanced Settings for Brim Variables.

Combining Structured and Unstructured Data

One of Brim’s strengths is the ability to combine variable values intelligently to achieve hierarchies:

• Create variables for structured fields and unstructured abstracted fields.
• Use these variables as input to variables if you need to read from the note, or create dependent variables that use values from structured & unstructured sources to make a decision.

In Summary, In Brim, you can think of structured data as just another kind of “note.” Once it’s in Brim, you can abstract, validate, and compare it seamlessly alongside your unstructured documents.