Number of patient samples affected before error detection: Strategic implications for internal quality control and patient-based quality control practices.

Abstract

Introduction: There is a general impression that patient-based quality control (PBQC) requires a high volume of laboratory results to detect errors effectively. However, internal quality control (IQC) performed infrequently may be associated with increased risk of missed error (i.e. low power of error detection).

Methods: Using simulations and routine sodium and aspartate aminotransferase (AST) as examples, this study examined how the "average number of patient samples affected before error detection' (ANPed) metrics can provide linkage to compare relative performance of QC practices in various IQC and PBQC settings.

Results: Smaller numbers of IQC samples tested per IQC run or larger average number of patient samples measured between IQC runs are associated with higher ANPed. The ANPed for sodium and AST PBQC models were smaller than IQC performed once every 100 patient samples, except when the systematic error was small for AST.

Discussion: Use of ANPed clearly illustrate the relative impact of different IQC frequencies and number of IQC levels tested. Patient-based quality control can outperform IQC even for laboratories with small testing volume. Laboratory practitioners can use this metric to design a QC strategy that suit their desired risk profile.