Variance Reduction in Analytical Processes in the Clinical Laboratory by Digital Filtering

Abstract

A model is postulated describing the fluctuations in analytical chemical processes in the clinical laboratory. In this model the process variations are described by a non-stationary stochastic process with a significant time-varying mean value. Experiments demonstrate a short-term variance within a run and a long-term variance between runs determined by the time-varying mean value. For four different analytical systems used for determining six serum analytes between-run variance was demonstrated to be significantly greater than within-run variance. Based on the model a digital filtering procedure is presented which in each run estimates the process mean and subsequently corrects serum samples for its deviation. Thus significant variance reductions are obtained. The filtering procedure was tested for the determination in inorganic phosphate with a continuous-flow system in an experimental environment.