Selection of an optimal combination of decision criteria for the internal quality control in an automatic analyzer

After some years using an automatic analyser model Hitachi 705, we faced the task of selecting a reasonable combination of statistical procedures for internal quality control in order to use it in a computer program, by means of which, one could apply all the decision criteria in a rapid and automatic manner. As a result of the first 3 years of work, an algorithm was adopted that includes the multirule procedure originally proposed by Westgard, but with three main differences: (1) All the control rules are always applied; (2) the 4(1S) and 10x control rules are used as criteria for warning or caution but not for run rejection; and (3) the control limits are recalculated daily using all the accumulated control results (just the accepted values). The 2(2S), 4(1S) and 10x rules are applied first across materials (considering consecutive observations in different control materials), and then within materials (for consecutive observations on the same control material). The program not only has permitted us to improve our capacity for error detection, but it has also permitted the reduction of run rejections making an important decrease in costs possible and a significant improvement in quality.