指南与共识临床血液与体液质量管理与危急值

双变量z分数分析评估D-二聚体外部质评检测结果准确性

Assessment of accuracy of laboratory testing results, relative to peer group consensus values in external quality control, by bivariate z-score analysis: the example of D-Dimer.
指南 / 共识 · Clin Chem Lab Med · 2024-03-11

该研究基于ECAT基金会2022年国际D-二聚体外部质评计划四次调查结果,开发了一种可应用于外部质评计划调查的双变量z分数分析方法。方法先通过监督序列Hotelling T2控制图识别双变量离群值并剔除,对剩余数据稳健估计双变量正态分布参数,绘制95%和99.7%置信水平的嵌套同心椭圆。在2022-D4调查中对三种最常用方法进行分析,z分数对数量为830至857,双变量离群值为20至28。

要点

  • 双变量z分数图可同时提示系统偏差和随机偏差。
  • 椭圆形状有助于解读随机与系统偏差。
  • 同次调查多标本结果相关时适用双变量分析。

核心结论

  • 当同一调查中分发多个样本且结果可能存在依赖时,双变量z分数分析具有重要意义。
  • 椭圆形状可提供关于随机和系统偏差的重要信息,实验室可了解自身在最新分布中的位置及是否存在系统或随机偏差。

研究设计与样本

  • 方法学研究,使用ECAT基金会2022年国际D-二聚体外部质评计划四次调查结果。
  • 采用监督序列Hotelling T2控制图识别双变量离群值,剔除后稳健估计双变量正态分布参数并绘制95%和99.7%置信椭圆。
  • 在2022-D4调查中对三种最常用方法进行分析。

主要发现

  • 纳入的z分数对数量在830至857之间。
  • 双变量离群值数量在20至28之间。
  • z分数对之间的相关性在0.431至0.647之间。
  • 三种最常用方法的z分数对相关性在0.208至0.636之间。

临床怎么用

  • 实验室可利用双变量z分数图同时评估系统偏差和随机偏差。
  • 通过椭圆形状判断结果偏离零z分数的模式。
  • 适用于同一调查中多个样本结果可能存在依赖性的外部质评场景。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

OBJECTIVES: The aim of this study is to develop a practical method for bivariate z-score analysis which can be applied to the survey of an external quality assessment programme. METHODS: To develop the bivariate z-score analysis, the results of four surveys of the international D-Dimer external quality assessment programme of 2022 of the ECAT Foundation were used. The proposed methodology starts by identifying the bivariate outliers, using a Supervised Sequential Hotelling T2 control chart. The outlying data are removed, and all the remaining data are used to provide robust estimates of the parameters of the assumed underlying bivariate normal distribution. Based on these estimates two nested homocentric ellipses are drawn, corresponding to confidence levels of 95 and 99.7 %. The bivariate z-score plot described provides the laboratory with an indication of both systematic and random deviations from zero z-score values. The bivariate z-score analysis was examined within survey 2022-D4 across the three most frequently used methods. RESULTS: The number of z-score pairs included varied between 830 and 857 and the number of bivariate outliers varied between 20 and 28. The correlation between the z-score pairs varied between 0.431 and 0.647. The correlation between the z-score pairs for the three most frequently used varied between 0.208 and 0.636. CONCLUSIONS: The use of the bivariate z-score analysis is of major importance when multiple samples are distributed around in the same survey and dependency of the results is likely. Important lessons can be drawn from the shape of the ellipse with respect to random and systematic deviations, while individual laboratories have been informed about their position in the state-of-the-art distribution and whether they have to deal with systematic and/or random deviations.

来源:Clin Chem Lab Med,查看原文。版权归原作者所有。

引用

  1. Meijer P, Sobas F, Tsiamyrtzis P, et al. Assessment of accuracy of laboratory testing results, relative to peer group consensus values in external quality control, by bivariate z-score analysis: the example of D-Dimer. Clin Chem Lab Med. 2024. PMID: 38456711. doi:10.1515/cclm-2023-0835

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。