Closing the Loop on Radiology Follow-Up: Inside the Brim and Primordial Partnership
November 12, 2026

Radiology reports are full of findings that need follow-up, whether it’s a lung nodule, a renal lesion, or an aortic aneurysm. Health systems have invested in flagging them, but then a navigator has to open the chart and figure out whether the flag is real. There are errors in both directions: reports that needed followup don’t get flagged, or a navigator has to spend time on a chart that doesn’t need followup.
That is why Brim and Primordial Design are working together. Primordial's Follow-up Manager tracks imaging follow-up in more than 400 facilities. Brim can now read incoming radiology reports and tag what the radiologist actually found and recommended, so the worklists navigators see are built from accurate, structured data. In existing use, that has cut the number of reports wrongly sent to navigators for chart review by more than 50%.
Chart Abstraction for Follow-up
To demonstrate some of the challenges in chart abstraction for followup, here are a few examples of common patterns. They are fake patients, but the patterns show up in real radiology text.
The word is there, but the finding isn't. "The aorta is ectatic however no discernable aneurysm." This one would trigger on a keyword search on "aneurysm", but a person or interpretive tool can understand the negative phrasing.
The template recommends follow-up, but the radiologist didn't. Many aortic and thyroid reports end with a standard guideline table full of "follow-up" and "referral" language. The table appears on every study, including ones that say, for example "Negative for abdominal aortic aneurysm."
The recommendation is buried in the body. The findings section says a 5 mm nodule needs a CT in 3 to 6 months. The impression says "No acute abnormality... other incidental findings, as above." This is easy to miss for both a busy reader and a rules-based tool.
Brim leverages large language models (LLMs) hosted by Primordial, and combines them with structured guardrails to achieve interpretation and predictable results. It reads the whole report, follows the radiologist's own words rather than making its own clinical call, and records the finding, its size, and any follow-up recommendation as separate fields. Then Primordial builds worklists using the combination of these datapoints, such as "pulmonary nodule larger than 6 mm with an imaging follow-up recommendation." Brim records what the radiologist said, and Primordial's rules apply the clinical thresholds. Every answer links back to the sentence in the note it came from, so a navigator can check it quickly.
What changes for Primordial customers
- A cleaner worklist. Fewer reports that turn out not to need follow-up, so navigators spend their time on patients who do.
- More findings tracked. Brim has let Primordial expand the types of findings Follow-up Manager can track.
- Verifiable answers. Every result shows the supporting text from the report, for review and audit.
- Same familiar workflow. Navigators keep working in Follow-up Manager.
What it was like to build with Brim
“What stood out about Brim was how much control it gave our team,” said Andy Stahl, General Manager at Primordial Design. “Our staff built and validated the finding definitions, and when we want to track something new, we can build it internally. That lets us improve Follow-up Manager and add other use cases quickly.”
Brim is a configurable chart abstraction tool deployed in an institution’s cloud environment, with no PHI leaving. Because of Brim’s flexibility, Primordial didn't have to wait on us to define their use case. Instead, Primordial's staff built the finding definitions and validated them using the Brim tool. When they wanted to track something new or tighten a definition, they could handle it themselves.
Brim also offers flexible integration tools. To fit Brim into their existing workflow, Primordial integrated with Brim's REST API. Brim processes each radiology result and returns structured tags for the finding, its size, and any follow-up recommendation. Primordial queries those tags through the API and runs its own rules to decide which patients go on a navigator worklist, all inside their controlled environment.
The combination creates a powerful partnership: Brim’s flexible AI chart abstraction engine combined with Primordial’s radiology expertise and dedicated workflow tools.
Building something similar?
Whether you run a follow-up program or build software for teams that do, Brim can turn unstructured clinical notes into structured, traceable data inside your own environment. Get in touch to talk about your use case.