Blind grading, designated grader roles, and grade definitions beside the form

Week 7 of 2026 (February 9 to 15) made Burna AI blind grading a workflow a coordinator can run end to end: seed an organization with test cases, assign them in bulk to designated graders, and have each grader work a case without seeing anyone else's grade or the engine's intermediate reasoning. Grade definitions now open beside the grading form, so a boundary call can be checked against the CTCAE v6.0 wording without leaving the case. Quality leads running a rater program and the coordinators who route its cases are who this week was built for.

How blind grading works in Burna AI

Blind grading means two or more raters grade the same case independently, without seeing each other's work. The FAQ what is blind grading covers the method. From this week, the platform ran it in five steps.

1. Seed. One operation creates 50 test cases for an organization, drawn from 15 clinically realistic adverse event scenarios, among them cytokine release syndrome, colitis, oral mucositis, pancreatitis, epistaxis, atrial fibrillation and diarrhea.
2. Assign. A coordinator assigns graders to cases from a sidebar.
3. Grade blind. The grader sees the clinical note and the grading view only. The clinical evidence and drug attribution tabs are hidden, and no other rater's grade is shown.
4. Grade by hand. A grader can also identify an adverse event directly from the evidence, choose its severity grade and write the clinical justification.
5. Track. A case list shows totals for pending, in-progress and completed cases, with full-text search, a review-status filter and an expandable clinical note on each row.

Agreement between raters means something only when each rater graded alone. In a blind program, agreement is a measure of how consistently the team applies the criteria, a quality metric for the rating process rather than a verdict on anyone's accuracy. How agreement is measured explains how Burna AI treats agreement as evidence.

Designated grader roles and bulk assignment

Routing a case used to mean searching the full user list for someone qualified to grade it. This week:

• A user can be flagged as a grader, and the assignment sidebar lists only the graders in the active organization, filtered on the server.
• A coordinator can select between 2 and 10 cases and assign raters to all of them in one operation, with each case checked against the organization.
• Rater A, Rater B and the adjudicator must be three different people, because independent assessment depends on it.
• Case lists and case tabs show each rater's name.

Since then: in June 2026 the grader flag moved from the user profile to each organization membership, as separate rater and adjudicator capabilities, so a person can grade in one organization without grading in another.

CTCAE grade definitions beside the grading form

During an adverse event review, a Grade Definitions button on each term opens the CTCAE v6.0 descriptors for grades 1 through 5, with the currently assigned grade highlighted and a note of how the term was matched.

This matters most at boundaries such as grade 2 and grade 3. In CTCAE v6.0 (NCI, 2025), grade 2 is generally described as limiting instrumental activities of daily living, and grade 3 as limiting self-care activities of daily living. With the wording beside the case, the grader no longer leaves it to open the reference PDF. For how the descriptors moved between versions, see every CTCAE v6.0 change, term by term.

Patient assignment inside the grading workflow

• Web: a note cannot be submitted for grading until a patient is assigned. An assignment dialog appears when one is missing, and grading starts once the patient is set.
• Mobile: a new Patients tab with search, filtering, paging and in-place profile editing.
• Recording on the web: a five-second countdown before recording starts, which pauses while a dialog is open. Recording state now resets between sessions, so the countdown no longer skips on the next patient.

Smaller changes and fixes

• Encounter screen content now centers on wide displays.
• Fixed: the review status saved on adverse events.
• Mobile patient form: the return key moves to the next field.
• Mobile patient sheet: predictable snap points and keyboard behavior across device sizes.
• The keyboard now closes after a paste in the mobile recording view.
• Public website: dark mode across the marketing pages and a header that condenses on scroll. Since then: the public website was rebuilt in August 2026.

Since then

• Adjudication followed the next week: when two raters disagree, the case routes to an adjudicator. The FAQ on how disagreements between raters are handled describes it.
• The web app described here was retired in July 2026. Blind grading now runs in the platform at app.burna.ai, where multiple raters grade the same case without seeing each other, disagreements route to an adjudicator and inter-rater agreement is measured (what Burna AI does today).
• The mobile provider app is now called CTCAE by Burna AI.

To see how a blind grading program runs on your own cases, book a 15-minute conversation.