Label Generation Process

Once you've uploaded data and defined variables for your project, it's time to generate labels.

Each generation takes time and resources. To make the best use of them, we recommend an iterative process:

1. Generate Labels for one patient to check data and variable setup.
   1. Review labels for that patient.
   2. Add or remove variables if you'd like.
   3. Change variable definitions and instructions, optimizing the variable if you haven't yet.
   4. Repeat single patient generation as necessary until you’re satisfied.
2. Generate Labels for a Batch of patients.
   1. Review that batch in Label Review, making data edits as necessary. These edits will automatically be included when optimizing variables.
   2. Optimize the variables to get better results on future batches.
3. Continue generating and reviewing patients in Batches until all of your data is abstracted and reviewed.
   1. You can optimize variables at any point to apply the learnings from your reviewed data points to future generations.

Once you’ve reviewed a data point, Brim will not overwrite that data point when you do a new generation unless you explicitly instruct it to overwrite human-reviewed labels.

Best Practice: There is a tradeoff between time spent refining variables and time spent reviewing abstracted labels. We recommend shifting your focus to reviewing when the labels pass 90% accuracy.

Generation History

As you generate labels, you can review a complete history of all your generations. Each generation shows the parameters used, resources consumed, and current status.

Generation Details

Each generation is grouped with the following information:

• Started: The timestamp when the generation began.
• Generation Parameters:
  • Patients: Whether you generated for "All" patients, a "Batch X (Y Patients)", or specific MRNs.
  • Filters: Any non-default filters applied, displayed in the order they appear in the form.
  • Variables: The variables included in this generation (comma-separated list).
  • Conditions: Any conditions applied, expressed as phrases.
  • Note IDs: The specific Note IDs included, or "last updated in
  • Overwrite: Shows only if not set to "Skip" — either "AI Values Only" or "AI and Human Values".
  • Scope: Shows only if not set to "All".

Tokens Used: The number of tokens consumed by the generation (formatted with thousands separators).

Time Elapsed: How long the generation took, displayed in human-readable format (e.g., 36 seconds; 1 minute, 27 seconds; 2 hours, 3 minutes, 56 seconds).