Concentration-dependent profiles for describing the scatter of results of interlaboratory surveys

In an interlaboratory survey for the quantitative determination of a clinical chemical quantity, samples of the same specimen are analysed in different laboratories. If the number of participating laboratories is sufficiently large, then the differences between the 50th percentile (median) and e. g. the 25th and 75th percentiles of the results give a very reliable impression of the range of interlaboratory scatter for the particular analytical technique. Results from a relatively large number of interlaboratory surveys, in which specimens containing different concentrations of the analyte are investigated, can be handled in the same way. If the resulting differences between the chosen percentiles are plotted against the median, and the corresponding two regression lines (upper and lower) are drawn, the results are asymmetric scatter profiles covering the concentration range of the specimen collective. Numerous options are available. Thus, a profile's power of characterizing the scatter correctly can be improved by weighing of the results. Moreover, scatter profiles can be based on different variables of the survey, such as the analytical method, or the observation period, etc. They may be based on the total collective of all results for a given quantity, or they can be constructed for subcollectives of results obtained with a single analytical method. Further, it is possible to present the results of all subcollectives in a single pair of scatter profiles. This latter type of analysis provides profiles of the average scatter for a collective of different analytical methods, which are unaffected by any systematic differences that may exist between the methods.