Background: A study performed in 2007 comparing the evaluation procedures used in European external quality assessment schemes (EQAS) for hemoglobin and leukocyte concentrations showed that acceptance criteria vary widely. For this reason, the Hematology working group from the European Organisation for External Quality Assurance Providers in Laboratory Medicine (EQALM) decided to perform a statistical study with the aim of establishing appropriate acceptance limits (AL) allowing harmonization between the evaluation procedures of European EQAS organizers.
Methods: Eight EQAS organizers from seven European countries provided their hematology survey results from 2010 to 2012 for red blood cells (RBC), hemoglobin, hematocrit, mean corpuscular volume (MCV), white blood cells (WBC), platelets and reticulocytes. More than 440,000 data were collected. The relation between the absolute value of the relative differences between reported EQA results and their corresponding assigned value (U-scores) was modeled by means of an adaptation of Thompson's "characteristic function". Quantile regression was used to investigate the percentiles of the U-scores for each target concentration range. For deriving AL, focus was mainly on the upper percentiles (90th, 95th and 99th).
Results: For RBC, hemoglobin, hematocrit and MCV, no relation was found between the U-scores and the target concentrations for any of the percentiles. For WBC, platelets and reticulocytes, a relation with the target concentrations was found and concentration-dependent ALs were determined.
Conclusions: The approach enabled to determine state of the art-based ALs, that were concentration-dependent when necessary and usable by various EQA providers. It could also easily be applied to other domains.
Keywords: acceptance limits; characteristic function; external quality assessment (EQA); hematology; quantile regression.