Calibration
Whether a model's stated confidence matches reality. A well-calibrated model that says "70% risk" is right about 70% of the time. A model can discriminate well and still be poorly calibrated, so its numbers mislead.
A readmission model prints a percentage on every discharge summary, and case managers have started quoting it to patients. A local audit of six months of discharges finds that patients scored around 40% were readmitted about 20% of the time, roughly half the stated risk, although the model still put the riskiest patients near the top of the list. Ask for a calibration plot on your own patients, and recalibrate before the percentage is shown to staff or patients again.
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 →