Design of method comparison and bias estimation of multi tests on multi instruments / 中华检验医学杂志

ABSTRACT

Objective To design an experiment for method-comparison and bias estimation of multi tests on multi instruments.Methods According to the procedure described by the NCCLS approved guideline EP9-A and improving the method of sample collection,we took 8 patient mixing samples each day to analyze all comparison tests on 11 auto-chemistry analyzer within following 5 days.The duplicates were assessed within the same run.The coefficient of correlation and average bias% were calculated,and the system errors at medical decided levels were assessed.Results Taking ALT as an example,the coefficients of correlation were between 0.994-1.000,and the average bias% were between-0.460%-4.927%,SE at 40U/L was-1.510-1.834 and SE at 300 U/L was-3.101-9.188.Conclusion In all tests that joined the comparison among the different instruments,28 tests were acceptable,2 tests were acceptable after modifying the coefficients,and AMY and LIP were not acceptable.