Notes
A Note on Uncertainty of Diagnostic Measures
Uncertainty represents imperfect or incomplete information1. When quantifiable, we can express it with probability.
A. Measurement Uncertainty
Given the intrinsic variability of measurements, measurement uncertainty is defined as a parameter associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand. This measurement uncertainty concept supplants the traditional notion of total analytical error for diagnostic assays.
B. Sampling Uncertainty
Diagnostic measures are derived from screening or diagnostic tests applied to population samples. The variability within these samples contributes to their overall uncertainty. This intrinsic heterogeneity is present even when simple random sampling techniques are used.
C. Combined Uncertainty
Combined uncertainty is computed via uncertainty propagation rules, employing a Taylor series approximation.
Estimating, evaluating, and mitigating the combined uncertainty of diagnostic measures is critical in medical diagnosis.
We have already developed software tools for exploring the uncertainty of diagnostic accuracy measures and Bayes' theorem-derived posterior probability for disease, which can significantly impact their clinical usefulness2, 3.
Theodora Chatzimichail, M.R.C.S.,
tc@hcsl.com
Related Publications
1. Kallner A, Boyd JC, Duewer DL, Giroud C, Hatjimihail AT, Klee GG, et al. Expression of Measurement Uncertainty in Laboratory Medicine; Approved Guideline. CLSI document EP29-A. Clinical and Laboratory Standards Institute; 2012:1-56.
Abstract
2. Chatzimichail T, Hatjimihail AT. A Software Tool for Calculating the Uncertainty of Diagnostic Accuracy Measures. Diagnostics. 2021;11(3):406.
Abstract
3. Chatzimichail T, Hatjimihail AT. A Software Tool for Estimating Uncertainty of Bayesian Posterior Probability for Disease. Diagnostics. 2024;14(4):402.
Abstract