Abstract
Diagnostic tests are approaches used in clinical practice to identify with high accuracy the disease of a particular patient and thus to provide early and proper treatment. Reporting high-quality results of diagnostic tests, for both basic and advanced methods, is solely the responsibility of the authors. Despite the existence of recommendation and standards regarding the content or format of statistical aspects, the quality of what and how the statistic is reported when a diagnostic test is assessed varied from excellent to very poor. This article briefly reviews the steps in the evaluation of a diagnostic test from the anatomy, to the role in clinical practice, and to the statistical methods used to show their performances. The statistical approaches are linked with the phase, clinical question, and objective and are accompanied by examples. More details are provided for phase I and II studies while the statistical treatment of phase III and IV is just briefly presented. Several free online resources useful in the calculation of some statistics are also given.
Highlights
An accurate and timely diagnostic with the smallest probability of misdiagnosis, missed diagnosis, or delayed diagnosis is crucial in the management of any disease [1, 2].e diagnostic is an evolving process since both disease and diagnostic approaches evolve [3]
E diagnostic tests are frequently reported in the scientific literature, and the clinicians must know how a good report looks like to apply just the higher-quality information collected from the scientific literature to decision related to a particular patient. is review aimed to present the most frequent statistical methods used in the evaluation of a diagnostic test by linking the statistical treatment of data with the phase of the evaluation and clinical questions
(i) Nonparametric: empirical method (estimated Area under the ROC curve (AUC) is biased if only a few (i) AUC 1 ⟶ perfect diagnostic test and smoothed-curve accuracy) methods such as kernel density method (not (ii) AUC ∼ 0.5 ⟶ random classification reliable near the extremes of the receiver operating characteristic (ROC) curve) (iii) 0.9 < AUC ≤ 1 ⟶ excellent accuracy (ii) Parametric: binomial method (iv) 0.8 < AUC ≤ 0.9 ⟶ good accuracy (tighter asymptotic confidence bounds for (v) 0.7 < AUC ≤ 0.8 ⟶ worthless samples less than 100)
Summary
An accurate and timely diagnostic with the smallest probability of misdiagnosis, missed diagnosis, or delayed diagnosis is crucial in the management of any disease [1, 2]. Each guideline is accompanied by a short checklist describing the information needed to be present in each section and include some requirements on the presentation of statistical results (information about what, e.g., mean (SD) where SD is the standard deviation, and how to report, e.g., the number of decimals). Different designs of experiments received more attention, and several statistical guidelines, especially for clinical trials, were developed to standardize the content of the statistical analysis plan [9], for phase III clinical trials in myeloid leukemia [10], pharmaceutical industry-sponsored clinical trials [11], subgroup analysis [12], or graphics and statistics for cardiology [13].
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Computational and mathematical methods in medicine
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
