Study finds F score ranks clinical diagnostic tests inconsistently with patient benefit

Written by

in

A methodological study published in Diagnostic and Prognostic Research compared how the F score and net benefit rank clinical diagnostic tests and prediction models. In a cancer detection simulation, F1 ranked a test with higher specificity — but one that may miss diagnoses — first, whereas net benefit ranked a test with 99% sensitivity and 47% specificity first. The study notes that F-beta lacks an interpretable clinical unit, relies on an arbitrary dichotomization, and cannot represent a no-test strategy, leading the authors to recommend that medical evaluations use discrimination, calibration, and decision curve net benefit instead.

Source:BIOENGINEER.ORG · Read original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *