[A model for evaluating internal accuracy].

Abstract

OBJECTIVE: To provide an evaluation model of internal accuracy illustrating its use in a quality control program of clinical chemistry. METHODS: The model uses data transformed to percentage of assigned value (%AV) that allows the pooling of different controls. Precision is evaluated by the mean of the intracontrol coefficients of variation, and accuracy by the mean of means of the %VAs and its standard deviation. The model was used in 17,280 measurements of 23 analytes assayed in seven pairs of controls (medium and high concentration) during four years in two automated clinical chemistry analyzers. RESULTS: The model established inaccuracy of 5% to 25% in seven analytes, five of them enzymes. It also detected interanalyzer differences in accuracy of five analytes (22%) and an erroneous assigned value in 19 of 304 controls (6%). CONCLUSIONS: The model showed its capacity to dissect the sources of variation of accuracy. There was inaccuracy in 30% of the analytes but further studies are needed to validate these inaccuracies with controls of other sources. The model can be useful for control programs of any quantitative assay system. A summary description of the model's operation is given.