How does Burna AI tell a treatment-related adverse event from a pre-existing condition?

By reading the patient's known comorbidities alongside the note. When the engine grades an event, the patient's recorded comorbidities and current medications are part of what it weighs, and where a comorbidity may explain the symptom it attaches a note saying so: peripheral neuropathy in a patient with long-standing type 2 diabetes, for example, may predate the regimen. The proposed grade and attribution still come with their source sentence, and the investigator decides whether the event is new and whether it is related to treatment. Attribution also weighs alternative causes, such as the disease itself or another medication, for every drug in the regimen.