两种测量方法定量测量结果的一致性评价

[目的]比较两种测量方法定量测量结果的一致性评价方法的优劣。[方法]应用配对t检验、简单相关分析、组内相关系数以及Bland-Altman法对某一致性较好的实例以及其衍生出的系统误差较大、随机误差较大以及测量范围局限3种情况进行一致性评价。[结果]对实例一致性评价中,简单相关分析、组内相关系数与Bland-Altman法显示一致性较好,配对t检验显示一致性较差;系统误差较大情况下,简单相关分析显示较好一致性,而配对t检验、组内相关系数与Bland-Altman法显示一致性较差;随机误差较大情况下,配对t检验显示一致性较好,而简单相关分析、组内相关系数与Bland-Altman法显示一致性较差;在测量范围局限而随机误差和系统误差均小情况下,配对t检验与Bland-Altman法显示较好一致性,而简单相关分析与组内相关系数显示一致性较差。[结论]配对t检验与简单相关分析作为一致性评价方法有明显缺陷,而组内相关系数与Bland-Altman法虽有一定局限但可作为一致性评价的优选方法。

EVALUATION ON DIFFERENT ASSESSMENT METHODS OF CONSISTENCY OF QUANTITATIVE MEASUREMENTS

[ Objective] To evaluate different assessment methods of consistency of quantitative measurements. [ Methods] The paired t test, simple correlation analysis, intra-class correlation analysis and Bland-Altman method were adopted to evaluate the consistency ofa true example with good consistency in fact and three hypothetical situations developed from the true one. [Results] As to consistency assessment of the true example, simple correlation analysis, intra-class correlation analysis, and Bland-Altman method showed good consistency, except for paired t test. In the situation of obvious systematic bias, intraclass correlation analysis, paired t test and Bland-Altman method showed that the data were not reliable, whereas the Pearson correlation suggested that the data were highly reliable. In the situation of comparatively large random error where low consistency was expected, intra,class correlation analysis, simple correlation analysis and Bland-Altman method accurately illustrated this, while paired t re.at suggested good consistency. In the situation of low range of measurements where both systematic bias and random error were small, paired t test and Bland-Altman method suggested good consistency, but intra-class correlation analysis and simple correlation analysis failed to demonstrate that. [Conclusions] Both paired t test and simple correlation analysis have defects in assessing consistency, but intra-clasa correlation analysis and Bland-Altman method are the preferable techniques to evaluate the consistency of quantitative measurements although there are some limitations associated with the use of these techniques.