ENBIS-17 in Naples

9 – 14 September 2017; Naples (Italy) Abstract submission: 21 November 2016 – 10 May 2017

Quantifying Uncertainty in Medical Decision Making

11 September 2017, 16:00 – 16:20

Abstract

Submitted by
Hans Landsheer
Authors
Hans Landsheer (Utrecht University)
Abstract
A method for the quantification of uncertainty in Medical Decision Making, the bi-normal Uncertain Interval method, has as its goal to determine an interval of uncertain scores of a test or biomarker. A motivating example uses the Creatine Kinase biomarker for determination of possible carriers of Duchenne Muscular Dystrophy patients. Simulation results show the applicability of this method: the Uncertain Interval method successfully determines an interval of test scores that cannot be used for any useful ranking of patients with or without the targeted condition. Furthermore, these test scores proved to be unreliable, while the test scores outside the Uncertain Interval are more accurate and showed improved reliability. By dismissing the uncertain test scores, the method improves decision-making in comparison to the Youden cut-point method by indicating the patients whose test scores offer insufficient certainty for a decision. Results are also compared to that of the TG-ROC method, the most used trichotomization method which selects the best test scores. The properties of the middle interval are more stable for the Uncertain Interval. In addition, the middle Uncertain Interval is usually considerably smaller than TG-ROC’s middle interval, while the net accuracy for the test scores outside the uncertain interval is comparable or even larger.
View paper

Return to programme