Assessing sensitivity to change: choosing the appropriate change coefficient.
Assessing sensitivity to change: choosing the appropriate change coefficient.
复制标题
DOI：
10.1186/1477-7525-3-23
复制
发表时间：
2005-04-05
影响因子：
3.6
通讯作者：
Riddle, Daniel L
中科院分区：
文献类型：
--
作者：
Stratford, Paul W;Riddle, Daniel L
在过去的20年里，许多健康状况衡量标准都得到了发展和评估。与那些进行和报告临床干预试验的人所要求的高标准不同，检验健康状况指标对变化的敏感性的研究的方法严格性要求较低。造成这一缺陷的原因很可能是缺乏改变健康状况的标准标准。为了增加对这些类型研究结果的信心，研究人员经常计算同一患者样本的多个变化系数。这份报告的目的是确定当多个变化系数应用于同一患者样本时出现的冲突。根据关于样本组成的不同假设，确定了三个变化系数家族：(1)样本关于变化是均匀的；(2)存在真实变化程度不同的患者亚组；(3)个别患者，其中许多人真实变化程度不同。我们提供了几个分析，这些分析说明了一个主要的概念冲突：这些系数中的一些系数的信号(测量的真正检测变化的能力)出现在其他系数的噪声项(测量误差)中。我们推测，这种两难境地的发生是由于准备工作不足，如初步研究，以确定感兴趣的患者群体可能的变化特征。通过开展试点研究以确定感兴趣人口的可能变化特征，可以克服选择变化系数的不确定性。一旦确定了人口的变化特征，就应该明确变化系数的选择。
The past 20-years have seen the development and evaluation of many health status measures. Unlike the high standards demanded of those who conduct and report clinical intervention trials, the methodological rigor for studies examining the sensitivity to change of health status measures are less demanding. It is likely that the absence of a criterion standard for change in health status contributes to this shortcoming. To increase confidence in the results of these types of studies investigators have often calculated multiple change coefficients for the same patient sample. The purpose of this report is to identify the conflict that arises when multiple change coefficients are applied to the same patient sample. Three families of change coefficients based on different assumptions concerning the sample composition are identified: (1) the sample is homogeneous with respect to change; (2) subgroups of patients who truly change by different amounts exist; (3) individual patients, many of whom truly change by different amounts exist. We present several analyses which illustrate a major conceptual conflict: the signal (a measure's true ability to detect change) for some of these coefficients appears in the noise term (measurement error) of the others. We speculate that this dilemma occurs as a result of insufficient preparatory work such as pilot studies to establish the likely change characteristic of the patient population of interest. Uncertainty in the choice of change coefficient could be overcome by conducting pilot studies to ascertain the likely change characteristic of the population of interest. Once the population's change characteristic is identified, the choice of change coefficient should be clear.