[Evaluation of diagnostic tests performance: the ROC curve]

Rev Med Brux. 2002 Apr;23(2):92-5.
[Article in French]

Abstract

Most of the diagnostic tests applied in clinic measure continuous or discrete ordinal variables and do not provide an "yes or no" response. Therefore, we can not assess tests nor compare them on the basis of a single pair of values of sensitivity and specificity. The ROC curve plots values of sensitivity against 1-specificity and allows evaluation of the discriminatory power of a test for all ranges of sensitivity and specificity when applied in two populations, one presenting the target disease and the other not. Evaluation of the discriminatory power or comparison of different tests are further deducted from the measure of the area under the curve. The discriminatory power of the test demonstrated, one has to determine the cutoff point on the ROC curve, that will provide the best information for clinical application.

Publication types

  • Evaluation Study

MeSH terms

  • Diagnostic Tests, Routine / statistics & numerical data*
  • Humans
  • ROC Curve
  • Sensitivity and Specificity