Diagnostic tests’ decision-making rules based upon analysis of ROC-curves

Abstract

In this paper we propose the model which substantiates diagnostics decision making based on the analysis of Receiver Operating Characteristic curves (ROC-curves) and predicts optimal values of diagnostic indicators of biomedical information. To assess the quality of the test result prediction the standard criteria of the sensitivity and specificity of the model were used. Values of these criteria were calculated for the cases when the sensitivity of the test was greater than specificity by several times, when the number of correct diagnoses was maximal, when the sensitivity of the test was equal to its specificity and the sensitivity of the test was several times greater than the specificity of the test. To assess the significance of the factor characteristics and to compare the prognostic characteristics of models we used mathematical modeling and plotting the ROC-curves. The optimal value of the diagnostic indicator was found to be achieved when the sensitivity of the test is equal to its specificity. The model was adapted to solve the case when the sensitivity of the test is greater than specificity of the test.