[Evaluating an internal quality control procedure: application to multidimensional control]

Abstract

Internal quality control keeps in constant evolution in the industrial world. Introducing in clinical chemistry new QC methods derived from the industrial practice raises the point of the means for their evaluation. The main evaluation criteria are discussed in this paper. The importance of ARL (average run length) as a key-criterion of the efficiency of a quality control procedure is stressed. These principles were applied to the multivariate approach of multilevel control with the Hotelling's T2. This method led to a better detection of random errors than the independently managed conventional Shewhart (Levey- Jennings) charts. Applied to systematic errors, both methods gave similar results with a certain lack of sensitivity. However the multivariate method can be sensitised using EWMA (exponentially weighted moving average), a method specific for improved bias detection. EWMA efficiency outperforms that of the older systems of rules. Moreover, in any kind of error, multivariate approach secures a well-defined false rejection rate, whereas this rate is dependant on unknown inter-level correlation coefficients in conventional QC.