RT Dissertation/Thesis
T1 Optimal cutoff points for classification in diagnostic studies: new contributions and software development
A1 López Ratón, Mónica
A2 Universidade de Santiago de Compostela. Facultade de Matemáticas. Departamento de Estatística e Investigación Operativa, 
K1 Biomarker
K1 Diagnostic test
K1 Empirical Likelihood
K1 Generalized Pivotal Quantity
K1 Logistic GAM regression model
K1 Misclassification costs
K1 Optimal threshold
K1 ROC curve
K1 R package
K1 Symmetry point
AB Continuous diagnostic tests (biomarkers or risk markers) are often used to discriminate between healthy and diseased populations. For the clinical application of such tests, the key aspect is how to select an appropriate cutpoint or discrimination value c that defines positive and negative test results. In general, individuals with a diagnostic test value smaller than c are classified as healthy and otherwise as diseased. In the literature, several methods have been proposed to select the threshold value c in terms of different specific criteria of optimality. Among others, one of the methods most used in clinical practice is the Symmetry point that maximizes simultaneously both types of correct classifications. From a graphical viewpoint, the Symmetry point is associated to the operating point on the Receiver Operating Characteristic (ROC) curve that intersects the diagonal line passing through the points (0,1) and (1,0). However, this cutpoint is actually valid only when the error of misclassifying a diseased patient has the same severity than the error of misclassifying a healthy patient. Since this may not be the case in practice, an important issue in order to assess the clinical effectiveness of a biomarker is to take into account the costs associated with the decisions taken when selecting the threshold value. Moreover, to facilitate the task of selecting the optimal cut-off point in clinical practice, it is essential to have software that implements the existing optimal criteria in an user-friendly environment. Another  interesting issue appears when the marker shows an irregular distribution, with a dominance of diseased subjects in noncontiguous regions. Using a single cutpoint, as common practice in traditional ROC analysis, would not be appropriate for these scenarios because it would lead to erroneous  conclusions, not taking full advantage of the intrinsic classificatory capacity of the marker.
YR 2016
FD 2016-04-15
LK http://hdl.handle.net/10347/14593
UL http://hdl.handle.net/10347/14593
LA eng
DS Minerva
RD 27 abr 2026