Background: When a systematic error is detected in the analytical process, ideally, one seeks to retest only patient samples between the onset of the error and the time the error was detected. In practice, the onset of error is often unknown, and patient samples are retrospectively retested back to the last acceptable QC sample. This can be wasteful of reagents and operator time.
Methods: An alternative approach that is based on the expected number of spurious results is described to determine when retrospective retesting should terminate. Assuming each patient sample was independently measured by an analytical process with an underlying Gaussian distribution, a Bayesian model that takes into account the difference between the original and retest result of each patient sample was developed.
Results: We are able to significantly reduce the number of samples retested, while ensuring that the average number of spurious results observed under the proposed retesting procedure was similar to or only marginally higher than the baseline number of spurious results when the assay was in control.
Conclusion: Patient samples measured after the systematic error have high probabilities of being retested under the proposed retesting procedure.
Keywords: Bayesian; Bias; Evidence-based; Quality; Retesting; Systematic error.
Copyright © 2014 Elsevier B.V. All rights reserved.