Computational Statistics, 2025 (SCI-Expanded)
The decision-making process in medicine often concludes with binary outcomes, determining whether a person has a medical condition and requires treatment or not. Various sources of information, such as patient complaints, symptoms, and diagnostic tests are utilized to rule-in or rule-out potential diseases. Despite certain diagnostic tests having limitations in perfectly discriminating between subjects, they are widely used in clinics due to their efficiency. Using a single cut-off value for classifying subjects in ordinal and quantitative diagnostic tests may lead to challenges when the distributions of diseased and healthy subjects overlap. This binary approach, particularly in the overlapped region, can result in false negatives and false positives. For this uncertainty, there exist various approaches known as a “gray zone” or “middle inconclusive area.” In this study, we intended to propose a novel solution for the boundaries of the gray zone based on information theory. Thus, we aimed to compare the performance of this proposed solution against existing methods. For equal variances, the proposed algorithm consistently achieves the smallest length of the gray zone in simulations. For unequal variances, it outperforms in certain cases. However, the performance of the proposed algorithm was found in second place in some scenarios if it did not yield with best performance. Therefore, the promising results of determining boundaries of gray zone with information criteria are obtained.