Confusion matrix
The table that counts a tool's results against the truth: true positives, false positives, false negatives and true negatives. Sensitivity, specificity, positive predictive value and negative predictive value are all ratios taken from these four cells, so the table is where those four metrics come from. It is also the plainest view of a tool, because it shows counts of patients, and counts are what your service has to staff. A tool with excellent specificity can still raise more false alarms than true finds when the condition is rare, and the table makes that visible at a glance. When a vendor quotes a metric, ask for the four counts at the threshold you will run, in a population with your prevalence.
A lung nodule tool is tested on 1,000 screening scans that contain 10 cancers. It finds 9 of them and flags 50 scans that turn out benign. Sensitivity is 90% and specificity about 95%, which is what the brochure reports. The person running the clinic reads the same table as 59 follow-up scans to schedule, 50 of them for patients without cancer, and one cancer missed.
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 →