Time series modeling for quality control in clinical chemistry

Abstract

Autocorrelation of clinical chemistry quality-control (Q/C) measurements causes one of the basic assumptions underlying the use of Levey-Jennings control charts to be violated and performance to be degraded. This is the requirement that the observations be statistically independent. We present a proposal for a new approach to statistical quality control that removes this difficulty. We propose to replace the current single control chart of raw Q/C data with two charts: (a) a common cause chart, representing a Box-Jenkins ARIMA time-series model of any underlying persisting nonrandomness in the process, and (b) a special cause chart of the residuals from the above model, which being free of such persisting nonrandomness, fulfills the criteria for use of the standard Levey-Jennings plotting format and standard control rules. We provide a comparison of the performance of our proposed approach with that of current practice.