Negative predictive value
Of the patients a tool calls negative, the share who really do not have the condition. It is the number that matters for a rule-out tool, one whose job is to tell you who can safely skip the next step. Negative predictive value depends heavily on how common the condition is. When the condition is rare, it is high almost automatically, because nearly everyone is negative to begin with, so a high figure in a screening population says little about the tool. The useful questions are how many true cases it misses in absolute numbers and what happens to those patients. It is the counterpart of positive predictive value, and the two move in opposite directions as prevalence changes.
A tool that clears normal chest X-rays from the worklist, so no radiologist reads them, is a rule-out tool, and its negative predictive value is the safety case. In an outpatient population where most films are normal, a figure of 99% sounds reassuring and still means one abnormal film in every hundred cleared. Ask how many studies it will clear per week, do the multiplication, and decide whether that number of unread abnormal films is one your department can accept and explain.
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 →