Modernising measurement in psychiatry: item banks and computerised adaptive testing.
Modernising measurement in psychiatry: item banks and computerised adaptive testing.
复制标题
精神病学测量现代化：项目库和计算机化适应性测试。
DOI：
10.1016/s2215-0366(21)00041-9
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Stochl J
中科院分区：
文献类型：
--
作者：
Stochl J
评论：柳叶刀。COM/精神病学第8卷2021年5月355，因此研究人员经常开发更多据称在测量这些结构方面更好的工具。不那么果断的同事可能会为同一结构选择两种工具；多重分析和选择性报告加深了可再现性危机。4最后，它使测量工具的选择变得复杂，支持可比性的合理愿望产生了使用众所周知的工具的压力，而不管其结构是否有效。当来自过多量表的组成部分与关于每个单独项目的反应如何与感兴趣的结构和其余项目相关的心理测量信息一起展开并汇集在一起时，就创建了所谓的题库。这个题库是一个强大的工具，可以缓解任何测量领域定义模糊的问题，因为这样一个题库的覆盖范围比任何特定工具都要广泛得多。项目不再与天平联系在一起，消除了选择哪种测量工具的棘手问题。管理题库中的所有项目是不必要的(也是不切实际的)；对一个项目的反应将高度预测对其他人的反应，通常情况下，一个子集将为相应的心理健康结构提供有效和准确的分数。几十年来，计算机化的自适应测试(CAT)涉及到一种实现这种平衡的算法。在其他领域已经司空见惯，5，6猫在精神病学中仍然不常见。CAT使用题库作为从目标人群样本中对其组成部分项目(所谓的校准)的反应模式的信息储存库。心理测量项目参数结合接受评估的个人对之前项目的反应，立即确定下一个最具信息量的问题；避免相关性较低的项目，因为这些项目几乎不会带来任何好处。该CAT过程一直持续到达到停止标准为止，该停止标准通常是通过最终分数的测量误差评估的预定精度。因此，对于任何受访者，CAT使用预先指定的、固定的待测量构件的精度，仅管理此人所需数量的物品。这种方法与传统的测量方法不同，传统的测量方法使用固定数量的项目，以不同的人以不同的精度估计结构。具有CAT的经过良好校准的题库通常管理比标准少75%的项目，
Comment www. thelancet. com/psychiatry Vol 8 May 2021 355 validity of mental health conditions, so that researchers frequently develop yet more instruments that are claimed to be better at measuring these constructs. Less decisive colleagues might choose two instruments for the same construct; and multiple analyses and selective reporting deepen the reproducibility crisis. 4 Finally, it makes the choice of measurement instrument complex and a reasonable desire to support comparability creates pressure to use a well known instrument, regardless of its construct validity.When component items from the plethora of scales are unwrapped and pooled alongside psychometric information on how the responses to each individual item relate to the construct of interest and the remaining items, a so-called item bank is created. This item bank is a powerful tool that mitigates the problem of vague definition of any measured domain because the coverage of such an item bank is much wider than any particular instrument. Items are no longer affiliated with scales, negating the vexed question as to which measurement instrument to select. It is unnecessary (and impractical) to administer all the items from an item bank; responses to one will be highly predictive of responses to others and usually, a subset will provide a valid and precise score for the corresponding mental health construct. Available for a few decades, computerised adaptive testing (CAT) involves an algorithm to achieve this balance. Already routine in other fields, 5, 6 CAT is still not common in psychiatry. CAT uses item banks as a repository of information about patterns of responses to their component items (their so-called calibration) from samples of the target population. Psychometric item parameters combined with responses to previous items from an individual being assessed instantly identifies the next, most informative question; less relevant items that will deliver little or no gain are avoided. This CAT process continues until a stopping criterion is achieved, typically a predefined precision assessed by the measurement error of the final score. Thus, for any respondent, CAT uses a prespecified, fixed precision of the construct to be measured, administering only the required number of items for that person. This approach contrasts with the traditional measurement approach that uses a fixed number of items that estimates the construct with different precision in different people. A well calibrated item bank with CAT typically administers up to 75% fewer items than standard,
影响因子：
7.2
作者：
Wahl, Inka;Loewe, Bernd;Rose, Matthias
通讯作者：
Rose, Matthias
人群心理困扰的计算机化适应性测试：GHQ-30 的模拟评估
影响因子：
4.4
作者：
J. Stochl;J. Böhnke;K. Pickett;T. Croudace
通讯作者：
T. Croudace