Failure to report inconclusive test results can lead to misleading conclusions regarding the accuracy and clinical usefulness of a diagnostic tool. We show that these results are often overlooked in research on test accuracy and provide guidance on suitable approaches to reporting and analysing these problematic results. The results of studies on diagnostic test accuracy are often reported as a 2×2 classification matrix, in which test results are presented as a dichotomy and a reference standard is used to categorise individuals as with or without disease. This facilitates the calculation of many popular statistics used to summarise the discriminatory performance of tests (such as sensitivity and specificity, positive and negative likelihood ratios, and positive and negative predictive values). Restricting test results to be either positive or negative, however, fails to represent the reality of how they are used in clinical practice 1 ; in many cases, the results from a given diagnostic test do not exclusively fall into these "positive" and "negative" categories.
Bethany ShinkinsAlthough most test results provide useful information for diagnostic decision making, there is often a subset of results that are relatively uninformative and lead to an "inconclusive" diagnostic outcome. For example, the "normal" range on a standard biomarker is typically based on statistics that try to minimise the number of false classifications. As most biomarkers are far from perfect discriminators, however, results that fall close to the lower and upper limits of "normality" provide minimal information about the disease status of the patient and are therefore "inconclusive." In most cases inconclusive test results require extra attention from clinicians, such as repeating the test or using more costly (in terms of invasiveness, time, and expense) diagnostic tools. All of these outcomes, coupled with the inevitable delay in diagnosis and clinical decision making, directly affect patient care.In this paper we have focused on clinical scenarios in which there are two disease categories-disease present and disease absent-and discuss the reporting and analysis of inconclusive results produced by a single index test-that is, the diagnostic test under evaluation. Although this is a common scenario in research on diagnostic test accuracy, this is often a simplification of the actual clinical question. In practice, diagnostic tests are often used to distinguish between multiple diseases or different levels of severity, such as different staging in diagnosis of cancer. Additionally, the diagnostic process might involve a pathway of multiple tests rather than a single test.
Inconsistent reporting of inconclusive test results in research on diagnostic accuracyWe wanted to explore whether inconclusive results are consistently reported in diagnostic accuracy research. The STARD (STAndards for the Reporting of Diagnostic accuracy studies) statement-a reporting guideline encouraging high quality reporting of diagnostic accuracy studies-recommends...