Adverse event grade adjudication, live analytics, and a removal on purpose

In week 8 of 2026 (February 16 to 22), adverse event grade adjudication moved inside the platform: when two independent raters grade the same case differently, the disagreement is detected, routed to an adjudicator and resolved with a signed decision. Admins gained a managed adjudicator pool, and the analytics dashboard moved onto live data. In the same week, treatment recommendations came out of the multi-drug attribution workflow, by design.

How adverse event grade adjudication works when two raters disagree

Blind grading, shipped the week before, has two raters grade the same case without seeing each other's work. Disagreement is expected in any multi-rater program; what matters is what happens next. From this release, every disagreement has a route to a signed resolution.

The workflow now runs in six steps:

1. Each rater submits. Submission locks the rater's work: after it, they can no longer add or edit events on that case.
2. The platform compares. Once both raters have submitted, their grades are compared term by term. The same term graded differently is a conflict.
3. An adjudicator is assigned. The case goes to an adjudicator in the organization, choosing whoever holds the fewest adjudication assignments.
4. The adjudicator reviews side by side. Each disputed term shows both raters' grades and their rationales.
5. The adjudicator decides. They accept one rater's grade, with an optional note, and the choice is stored with the case under an electronic signature.
6. The case closes. It reads "consensus" when the raters agreed, or "adjudicated" once every disputed term is resolved.

Adjudicators are not sent one email per case. Once a day, each receives a single digest of the cases waiting on them. A send that fails is retried up to five times, and names, case identifiers and links are escaped before an email renders. For the short version, see how the adjudication workflow routes a disagreement.

Who can adjudicate: managing the adjudicator pool

Routing only works if a qualified pool exists before the disagreement does. Admins and owners now designate adjudicators for the whole organization, independent of any single case, from a sidebar that lists current adjudicators, filters by name or email, and removes the role after a confirmation step.

Only admins and owners can change the pool, and that check runs on the server, not only in the interface.

Since then (June 2026): rater and adjudicator became flags on each organization membership rather than on the user (June 11), and a disagreement now waits in an adjudication queue, where the first adjudicator to claim a case holds it (June 13).

Why treatment recommendations came out of multi-drug attribution

The multi-drug attribution workflow had a final step that suggested treatment modifications, monitoring requirements and contraindications alongside the causality analysis. Clinical review found that a useful suggestion needs context the workflow cannot reliably assess: comorbidities, concurrent medications, lab values and performance status. A suggestion made without that context could mislead, and in a trial the treatment decision belongs to the protocol and the site investigator.

So the step was removed on February 16. The workflow now ends at causality: temporal, pharmacokinetic, interaction and ranking analyses, with WHO-UMC and Kramer attribution per suspect drug. AI suggests, clinicians decide.

For the other capabilities left off the platform on purpose, see the features Burna AI removed by design.

Analytics on live platform data

The analytics dashboard replaced its placeholder values with live platform data across four tabs:

Tab	What it shows
Overview	Session volume, average documentation time, quality score
Productivity	Documentation time trend, daily session volume, template usage
Quality	Documentation quality scores, error count, revision rate
Practice-wide	Active users, usage trend, peak usage patterns

The figures come from the same event stream the clinical workflows write, so a research director reads current numbers without exporting anything.

Faster encounter capture and CTCAE term search

Capture Encounter now opens as a modal from anywhere in the dashboard, and from an encounter's detail page the patient is already filled in. A visit can also be started straight from a patient's profile.

In manual event entry, one searchable field covers the CTCAE term list, groups results by system organ class and fills in the SOC when a term is chosen. It replaces the two-step pick of SOC first, then term.

Smaller changes

• "Clinical Evidence" labels became "Drug Interaction" across web and mobile
• Loading states show a spinner with text in place of skeleton placeholders
• Workflow tabs show completion checkmarks, and causality details expand or collapse together with a Show Details toggle
• Grading case lists return only cases in the user's active organization
• Patient list item colors were corrected

Since then (July 2026): these screens lived in the web provider app, which was retired in favor of app.burna.ai, where blind grading, adjudication and agreement measurement run today. The current list is on what Burna AI does today.

To see a disagreement routed and resolved on cases like yours, book a 15-minute conversation.