Detection of outliers and establishment of targets in external quality assessment programs

BackgroundThe establishment of target values is important in external quality assessment (EQA) programs since most of the programs use the target as the assessment criterion. Results submitted to EQA programs usually are not Gaussian-distributed due to the contamination with outliers. The traditional or robust statistical method can be chosen for the truncation of outliers. We investigated the two methods when setting targets for glucose in an EQA program.MethodsThe results of glucose were analyzed as an all-methods group and divided into four subgroups according to the analytical methods prior to testing of each data distribution. Based on the distribution, the traditional or robust statistical method was used to detect outliers. After removal of outliers, the data distributions were retested and if Gaussian distributions were obtained, the mean values were used as the target.ResultsOriginal data sets were not Gaussian-distributed for all tested groups. Therefore, the robust statistical method was employed to detect outliers. After truncation of outliers, the data sets showed Gaussian distributions and the means were used to set target values.ConclusionsThe targets of glucose were determined for all-methods and individual methods from the mean values following removal of outliers using the robust statistical method. This led to comparable targets among the tested groups.