Nonlinear comparison study on 4 hemotology analyzers using curve estimation program
Zhijiang Mo · Chinese Journal of New Clinical Medicine · 2009
Objective To compare the measured results from 4 hemotology analyzers using nonlinear procedure. Methods Daily 5 blood samples with high, normal and low measured values from patients were collected and analyzed respectively using 4 hemotology analyzers, Bayer120, CD3500, BC5500 and BC3000. The measured results of WBC, RBC, HGB and PLT from CD3500, BC5500 and BC3000 were respectively related to Bayer120 through 11 regression models (including linear model) from SPSS curve estimation program. Results The coefficient of determination (R2) of all best regression models found out was more than 0.981, and then method comparison and bias estimation between different hemotology analyzers could be made with more exact bias estimation and wider comparative range than only using linear model. Conclusion Nonlinear comparison is suitable for method comparison and bias estimation for clinic laboratory.