Burna AI's first study of its own grading is the concordance study it is running as a Mayo Clinic Platform_Accelerate participant: 1,600 cases from a de-identified dataset of 3 million patients, across three campuses, compared against an adjudicated two-clinician reference panel. Results read out in December 2026, with peer-reviewed publication as the aim. Until then, the approach rests on published literature: a kappa of 0.59 to 0.68 between two reviewers identifying CTCAE toxicities in the same notes (Fairchild et al., 2020), and attribution that changed at least once in 31 to 36% of cases when the same adverse event recurred in the same patient (Hillman et al., JCO 2010). The study design is at burna.ai/validation.
Interactive demo
Try Burna right now, no sign up required.