Statistical approach for optimization of external quality assurance (EQA) studies of molecular and serological viral diagnostics.

Abstract

Management of viral diagnostic quality is based on external quality assurance (EQA), where laboratories involved in diagnostics of a targeted virus are offered to analyze a panel of blinded samples. The utility of EQAs is compromised because of the absence of an approach to EQA design which upfront defines acceptance criteria and associated statistical analysis ensuring fair and consistent interpretation. We offer a rigorous statistically based approach for EQA planning. Instead of a conventional performance characteristic (the score) which is calculated as the sum of the points for correctly identified samples in a blinded test panel, Youden index is used as the performance measure. Unlike the score, Youden index requires an estimate of sensitivity and specificity and incorporates the relationship of these performance parameters. Based on the assumption that the coordinator is a reputable expert of viral diagnostics, the performance of the coordinator's laboratory is defined as a proficiency standard for performance evaluation. The immediate goal of EQA is defined as to obtain a statistically reliable estimation for every laboratory whether its performance meets the proficiency standard, while the overall goal is to match every laboratory to its specific performance level. Dependence of informational capacities of test panel from the panel size and content is quantitatively analyzed and the optimal design and informational capacities of both idealized panels (whose size is not restricted by financial factors) and currently feasible panels are considered. Our approach provides the basis both for rational design of currently feasible EQA test panels and for an increased panel size.