How does Burna AI grade CTCAE adverse events?

A coordinator or clinician submits a clinical note, an audio recording, or a batch upload. The grading engine, twelve specialized agents in a cascading constraint pipeline, extracts adverse event phrases, standardises drug names, resolves terminology against the MedDRA hierarchy, matches against the 850 CTCAE v6.0 criteria, applies comorbidity differentiation, attaches per-drug attribution probability scores using WHO-UMC and Kramer algorithms, and assembles an evidence package containing the source sentence, the CTCAE criterion, the rationale, the attribution support, and the audit trail. The engine cannot produce a grade outside the valid CTCAE set and cannot skip citation. A clinician reviews and approves before the grade becomes a record.