RESUMO
A blind quality control system using a special type of commercial reference serum is described. As packaged, the product contains six pools of analyzed reference sera with vials identified by coded numbers. The test constituents in the pools cover the range of multichannel analyzers in a random manner, i.e., each pool contains constituents in the low, normal and elevated ranges. As specific pools are not identified, and the analyst is unaware of the expected values for each constituent, results collected for quality control purposes are unbiased, i.e., free of conscious or unconscious editing. Data analysis in the proposed system is by least-squares technics. During an eight-week period, performance was studied using the standard error of the estimate to evaluate precision and the slope of the regression line to evaluate accuracy on the SMA and Mark X instruments. As an empirical comparison, the data are expressed as relative errors and precision and accuracy of analyzer performance evaluated as mean and standard deviation. The proposed regression analysis-based technic has the advantage of simultaneously evaluating linearity, as well as checking accuracy and precision over the total dynamic range of the instrument by use of a series of related serum pools.