Many diagnostic tests report only positive or negative diagnoses, as if these results were definitive. Should more tests allow a ‘neutral zone’, corresponding to an uncertain diagnosis?

Diagnostic tests often use single continuous biomarker measurements to return a positive or a negative result. However, in clinical practice, a diagnosis might often be reached based on multiple signs, symptoms and test results rather than using only a single measurement. These multiple individual characteristics are sometimes incorporated into a statistical model that aims to estimate the probability of disease, which helps in making a diagnosis. Guidelines have been published for optimal development, validation and report of such tools.1 2

The receiver operating characteristic (ROC) curve remains a popular technique for showing diagnostic performance in terms of sensitivity and specificity, that is, how accurately the test can correctly classify people with and without disease. Traditionally, statistical methods that follow this approach assume that individuals can be classified into just two categories: either a positive or a negative diagnosis.

The recent paper by Jeske et al 3 considers methods for identifying what has been termed a ‘neutral …