Three ways to quantify uncertainty in individually applied "minimally important change" values
Three ways to quantify uncertainty in individually applied "minimally important change" values
复制标题
DOI：
10.1016/j.jclinepi.2009.03.011
复制
发表时间：
2010-01-01
影响因子：
7.2
通讯作者：
Terwee, Caroline B.
中科院分区：
文献类型：
--
作者：
de Vet, Henrica C. W.;Terluin, Berend;Terwee, Caroline B.
目的：确定“最小重要变化”（MIC）有助于解释多项目工具上的变化评分。本文重点介绍MIC值应用于个体患者时应如何解释。研究设计和设置：在400例改善和100例未改善患者的样本中，使用受试者工作特征（ROC）方法和三种方法来量化不确定性，确定假设问卷“Q”的MIC值。问卷Q上的MIC值为10.5。首先，MIC值的95%置信区间（CI）（问卷Q：5.6-14.2）量化了MIC值估计的不确定性。其次，“我们对每个患者的MIC值的确定程度”通过灵敏度（74%）和特异性（91%）的值来量化。第三，计算问卷Q上的最小可检测变化（SDC）（16.0），以考虑MIC值（10.5）是否福尔斯在测量误差之内或之外。这种解释伴随着不同形式的不确定性。为了理解这种不确定性，了解变化分数的基本分布是必不可少的。(C)2010年爱思唯尔公司All rights reserved.
Objective: Determining "minimally important change" (MIC) facilitates the interpretation of change scores on multi-item instruments. This article focuses on how MIC values should be interpreted when applied to individual patients.Study Design and Setting: The MIC value of a hypothetical questionnaire "Q" was determined in a sample of 400 patients who improved and 100 patients who did not improve, using the receiver operating characteristic (ROC) method, and three methods to quantify the uncertainty.Results: The MIC value on questionnaire Q was 10.5. Firstly, the 95% confidence interval (CI) of the MIC value (for questionnaire Q: 5.6-14.2) quantifies the uncertainty of the estimation of the MIC value. Secondly, "how sure we are that this MIC value holds for every patient" is quantified by the values for sensitivity (74%) and specificity (91%). Thirdly, the smallest detectable change (SDC) on questionnaire Q is calculated (16.0) to consider whether the MIC value (10.5) falls outside or within the measurement error.Conclusion: For application in clinical research and practice, MIC values are always considered at the individual level, but determined in groups of patients. The interpretation comes with different forms of uncertainty. To appreciate the uncertainty, knowledge of the underlying distributions of change scores is indispensable. (C) 2010 Elsevier Inc. All rights reserved.