1000 Charts in One Weekend: One Team's Success using Brim for Chart Review

April 25, 2025

“Brim significantly reduced the time required for chart reviews from an estimated 500 hours to about 25 hours; something my colleague completed over the weekend.”
- Senior Project Manager, Biomedical Informatics
Background
A research team needed to confirm the most recent patient characteristics recorded in approximately 1,000 patient charts in order to ensure the clinical alignment of therapy. This required reviewing post-visit summary notes and reconciling patient reports with structured data stored in the EHR. Manual review of each chart typically took up to 30 minutes, leading to an estimated 500-hour workload.
Challenge
Patient characteristics are often inconsistently recorded, buried in lengthy clinical notes, and subject to change over time. Reviewing charts manually was time-intensive, prone to inconsistency, and difficult to scale.
Solution: Brim’s AI-Guided Abstraction
The team used Brim to streamline the review process. Brim enabled them to:
- Define a custom variable that extracted the desired patient characteristics from provider notes.
- Specify synonymous terms and categorize responses for consistent labeling.
- Review results with traceability, including confidence intervals and text excerpts from the source notes.
- Manually adjudicate ambiguous cases, aided by seamless navigation to related text when needed.
Results
- Improved Efficiency: Total review time dropped from an estimated 500 hours to just 25 hours, completed by one team member over a weekend.
- Easy Quality Control: The team easily rejected or corrected 1–2% of proposed results using Brim’s point-and-click interface.
- Impact: Brim helped the team identify patients who should have been excluded based on updated patient records, improving the integrity of their dataset.
Conclusion
Brim delivered a 20x improvement in review efficiency, while increasing accuracy, consistency, and auditability. This project highlighted Brim’s value in handling evolving patient characteristics at scale—turning a burdensome manual task into a streamlined, AI-supported workflow.