Comparison of different statistical approaches used to evaluate the performance of participants in a proficiency testing program

Abstract

Five statistical approaches were applied for assessing the performance of participants in 19 rounds (2007-2011) of the Proficiency Testing Program for lead in blood determination (PEP-Pbs), conducted by Instituto Adolfo Lutz. The performance evaluation was provided by using the z-score. The following statistical approaches were tested: 1 – mean and standard deviation, after rejecting outliers; 2 – median and normalized inter-quartile range; 3 – robust mean and robust standard deviation; 4 – robust mean and standard deviation for proficiency assessment of 3 μg/100 mL (for concentrations up to 40 μg/100 mL and a standard deviation for proficiency assessment of 7.5 % of the assigned value (for concentrations above 40 μg/100 mL); 5 – robust mean and standard deviation for proficiency assessment of 2 μg/100 mL (for concentration up to 40 μg/100 mL) or 5 % of the assigned value (for concentrations above 40 μg/100 mL). The approach 4 showed to be the most adequate statistical methodology to assess the performance of participating laboratories in the PEP-Pbs.