False negative
When an AI tool misses something that is actually a problem. False negatives are the dangerous ones in healthcare, because the thing the model failed to flag still needs attention.
A head CT triage tool moves suspected bleeds to the top of the radiology worklist, and over time readers come to treat the unflagged scans as the quiet ones. If the department sees 40 bleeds a month and the tool's sensitivity is 92%, about 3 of them (8% of 40 is 3.2) arrive unflagged and wait in the routine queue, which on a busy night can mean an hour or more. Check how long an unflagged scan waits at the busiest time of day, and set the routine queue's turnaround target as if the tool did not exist.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →