Engineering Cluster, Singapore Institute of Technology, Singapore.
Metabolic Laboratory, Genetics and Molecular Pathology Directorate, SA Pathology, Women's and Children's Hospital Site, South Australia, Australia.
Clin Biochem. 2020 Jun;80:42-47. doi: 10.1016/j.clinbiochem.2020.03.017. Epub 2020 Apr 2.
The performance of delta check rules has been considered to be dependent on the biological variation characteristics of the analyte of interest. The assumed relationships have not been formally studied. The mathematical relationship between biological variation and delta check rules is explored in this study.
From the mathematical model for absolute difference delta check, the threshold for specificity and sensitivity are observed to be normalized differently. For specificity, the threshold is normalized by the within-subject biological variation (expressed as a coefficient of variation, CV), whereas for sensitivity the threshold is normalized by the between-subject biological variation (expressed as a coefficient of variation, CV). This highlights the different roles the two biological variations play in affecting the absolute difference distribution for correct and switched patient samples. Analogous to absolute difference delta checks, for relative difference delta checks, the expressions for specificity and sensitivity are scaled by CV and CV, respectively. However, the expressions are independent of μ(the average of the population).
A comparison between the mathematical model and empirical/ historical laboratory data obtained from patients was conducted for both absolute and relative difference delta checks. In general it was found that the specificity obtained from the historical laboratory data was less than the model predicted values, while on the other hand, good correspondence was obtained between the experimental sensitivity and predicted sensitivity.
The difference in within-subject biological variation in different patients may contribute to the observed discrepancy in predicted and empirical delta check performance.
Delta 检验规则的性能被认为取决于所关注分析物的生物学变异特征。这些假定的关系尚未经过正式研究。本研究探讨了生物学变异与 Delta 检验规则之间的数学关系。
从绝对差值 Delta 检验的数学模型中，可以观察到特异性和敏感性的阈值以不同的方式进行归一化。对于特异性，阈值通过个体内生物学变异（表示为变异系数，CV）进行归一化，而对于敏感性，阈值通过个体间生物学变异（表示为变异系数，CV）进行归一化。这突出了两种生物学变异在影响正确和切换患者样本的绝对差值分布方面的不同作用。类似于绝对差值 Delta 检验，对于相对差值 Delta 检验，特异性和敏感性的表达式分别由 CV 和 CV 缩放。然而，这些表达式与 μ（总体平均值）无关。
对绝对和相对差值 Delta 检验进行了数学模型与患者的经验/历史实验室数据之间的比较。一般来说，发现从历史实验室数据获得的特异性小于模型预测值，而另一方面，实验敏感性与预测敏感性之间得到了很好的一致性。
不同患者个体内生物学变异的差异可能导致观察到预测和经验 Delta 检验性能之间的差异。
