Diagnostic efficiency in models for permissible measurement uncertainty

Abstract Limits for measurement uncertainty related to analytical imprecision and bias are most appropriately defined by the magnitude of tolerable diagnostic errors. A common mean to characterize the consequence of these errors is the diagnostic efficiency, which, in the case of data from a non-diseased population, is the rate of true-positive results (specificity). Three models have been identified by the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) for defining permissible uncertainty limits. Their model 1 is based on diagnostic requirements whereas models 2 and 3 do not primarily consider diagnostic errors. The present report links tolerable diagnostic error, empirical biological variation and the technical state of the art to derive the limits for measurement uncertainty. This approach combines the essential aspects of all three EFLM models and uses the diagnostic error, the clinically most relevant aspect, as the crucial criterion for the characterization of measurement uncertainty limits. The present approach is designed for the sole purpose of quality assurance.