Adverse event data export in one click, and workflows that resume where they failed

Week 5 of 2026 (January 26 to February 1) added adverse event data export to the web app: one click packages a site's CTCAE grading history, clinical session notes or decision logs into a downloadable archive. The same week, the mobile grading workflow learned to retry from the step that failed instead of starting over, attribution and decision-support results began to be saved for the audit trail, and each visit gained its own review status. Coordinators preparing for a monitoring visit and investigators working through graded encounters get the most from these changes.

Adverse event data export in one click

Settings in the web app gained a data export section with three buttons. Each one gathers the matching records, packages them as a ZIP archive, stores the archive and returns a download link along with the number of records it contains.

Export	What it contains	Format
CTCAE grading history	Each grading run: date, run ID, status, the input the engine read and the output it produced	JSON in a ZIP archive
Clinical session notes	Date, title, clinical note, patient ID and visit type for each visit that has a note	CSV in a ZIP archive
Clinical decision logs	Each run of the drug interaction analysis workflow, with the same fields as the grading history	JSON in a ZIP archive

The two JSON exports also record when the export was made, by whom and for which organization.

When a monitor or an auditor asks for a site's adverse event grading history, the file handed over is the record the platform holds, not values copied between screens.

Since then: the web app described here was retired in July 2026, and the platform now runs at app.burna.ai, where the graded record exports for a sponsor or a monitor and every access to patient data is recorded (what Burna AI does today). A patient-level export, assembled as FHIR R4 bundles in a signed archive, shipped on September 12, 2026.

Workflow resumption: retry from the step that failed

Before this week, the CTCAE grading workflow in the mobile app ran as one block. If the audio upload or a later step failed, the coordinator started again from the beginning.

The workflow now runs as discrete steps:

1. Create the workflow run.
2. Attach the recording or the pasted note.
3. Update the visit.
4. Start grading.

The app records the last step that completed. When a step fails, the progress panel shows an error state with a Retry button, and a retry runs from the failed step without repeating the ones that finished. Recorded notes and pasted notes each have their own error handling, so a failed upload does not strand a typed note.

Networks in clinics drop, and a phone gets pulled away for the next patient. Work that survives those interruptions is work coordinators can trust with a real encounter. The FAQ on how Burna AI handles workflow failures describes the current behavior.

Since then: in March 2026 the web app gained the same pattern for attribution, with a Retry Workflow button that reruns a failed attribution without regrading the encounter.

Attribution and decision-support results saved for the audit trail

Attribution and decision-support results are now written to the database as each run finishes, with their confidence scores and drug attributions, so they sit in the record beside the grading runs. A reviewer can trace an AI suggestion back to the run that produced it and the input that run read. The write happens in the background, so it does not delay the result a clinician is waiting for.

How access to those records is controlled and logged is covered in the architecture a security review asks about.

A review status on every visit

Each visit now carries a review status, calculated from its grading run and the clinician's review edits:

• Pending review when no review exists, or when only some suggested events have been reviewed.
• Accepted when every suggested event has been accepted.

A database trigger recalculates the status whenever a review is created, updated or deleted. The mobile case list now filters by that status on the server rather than in the app, and the web list moved to the same server-side filter the following week.

Since then: the first version also had a Rejected state. It was removed on June 9, 2026, and a visit's review status is now pending review or accepted.

Mobile app: simpler navigation and clearer feedback

• Routing in the mobile provider app was flattened, so the four tabs (Search, Record, Cases, Settings) sit at one level, and sign-in no longer loops between onboarding and the main screens.
• More than 15 screens now use one shared search input, and onboarding uses one shared text input, with consistent keyboard handling in sheets and modals.
• Profile photo uploads, adverse event acceptance, note saves and file uploads show a loading state while they run.
• Sign-in codes: an error clears as soon as the user types again, and a resent code is confirmed on screen.

Since then: the mobile provider app is now called CTCAE by Burna AI.

Also in week 5

• A new encounter detail page with the clinical evidence and multi-drug attribution in their own panels, redesigned patient and encounter lists, and in-place patient editing.
• A redesigned settings window on the web, with a sidebar that scrolls on its own and sign-out in the sidebar.
• A redesigned CTCAE search screen on the web, with clickable example searches for terms such as anemia and nausea.
• An ICD-10 codes panel beside CTCAE search results. Since then: removed on June 9, 2026, because ICD-10 codes beside CTCAE terms confused providers.

To see how a graded encounter becomes an adverse event export a monitor can read, book a 15-minute conversation.