Use of robust Z-score to assess creatinine proficiency testing data
10.3760/cma.j.issn.1009-9158.2011.12.023
- VernacularTitle:利用稳健Z比分数评价肌酐能力验证的数据
- Author:
Qi ZHOU
;
Wei XIE
;
Jianping XU
;
Shaonan LI
;
Xiaopeng LI
- Publication Type:Journal Article
- Keywords:
Creatinine;
Algorithms;
Normal distribution;
Statistics,nonparametric
- From:
Chinese Journal of Laboratory Medicine
2011;34(12):1144-1148
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo evaluate creatinine proficiency testing data by robust Z-score analysis.MethodsThe data were collected from three proficiency surveys of routine biochemical test in 2009,to which 1 179,1 169 and 1 168 laboratories participated respectively.Creatinine data were divided into Jaffe group and enzymatic group based on the analytical method used.The results tested by both methods were compared using Mann-Whitney test.The outliers were deleted using the TUKEY fence established by quartile values.The normality of raw data and trimmed data was tested using one-sample Kolmogorov-Smirnov test.The performance of the laboratories was assessed using robust Z-score,whose values were considered satisfactory when ｜ Z-score｜ ≤2,questionable when 2 ＜ ｜ Z-score ｜ ＜ 3 and unsatisfactory when ｜ Z-score ｜ ≥3.Results86.7％ results tested using Jaffe and enzymatic methods were not comparable.The raw data in all research groups were not normally distributed.After deletion of outliers,73.3％ trimmed data in most research groups were normally distributed.For the three proficiency tests in 2009,in Jaffe group,the satisfactory rates were 89.8％( 495/551 ),87.2％( 468/537 ) and 89.5％( 476/532 ) respectively,unsatisfactory rates were 3.3％ ( 18/551 ),6.5％ (35/537) and 4.5％ (24/532) respectively; while in enzymatic group,the satisfactory rates were 88.8％ (558/628),89.3％ (564/632) and 88.1％ (560/636) respectively,unsatisfactory rates were 5.6％( 35/628 ),5.2％ (33/632) and 6.6％(42/636) respectively.Conclusion It is reasonable to choose robust Z-score as a proficiency testing assessment index,because it avoids the influence of the outliers on evaluation results.
