Lookup NU author(s): Dr Sara Graziadio,
Professor John Simpson,
Dr Joy Allen
This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
Quantifying diagnostic accuracy is an important first step in assessing whether a new diagnostic device is suitable for implementation into clinical practice. Without initial evidence as to whether a device is able to improve diagnostic performance, it is difficult to justify larger studies to assess the impact on patient outcomes.To many clinicians and researchers, statistical measures of diagnostic accuracy (which we refer to in this paper as ‘technical accuracy’) may appear counterintuitive and may not adequately reflect how a test result should influence decisions about the treatment of the patient.1 This difficulty arises because many test accuracy study results are expressed in terms of sensitivity and specificity rather than measures of ‘clinical accuracy’; that is, the probability that the patient has the disease or condition under consideration after receiving a positive or a negative test result.2 3There is also evidence that many clinicians find it difficult to extract usable probabilistic information from diagnostic test accuracy results in the way that they are typically reported.4 5 However, there are conflicting opinions on the extent to which this depends on the type of information provided.6The purpose of this article is twofold: to review the concepts of technical accuracy and clinical accuracy and highlight the measures of diagnostic performance that are particularly useful for statisticians, on the one hand, and patients and clinicians, on the other, and to demonstrate an interactive graphical interface to help medical educators and health professionals to teach, design and interpret the results of diagnostic accuracy studies.
Author(s): Fanshawe TR, Power M, Graziadio S, Ordóñez-Mena JM, Simpson AJ, Allen AJ
Publication type: Article
Publication status: Published
Journal: BMJ Evidence-Based Medicine
Online publication date: 24/01/2018
Acceptance date: 12/12/2017
Date deposited: 01/03/2018
ISSN (print): 1356-5524
ISSN (electronic): 1473-6810
Publisher: BMJ Group
Altmetrics provided by Altmetric