Doyle Patrick J, Hula William D, McNeil Malcolm R, Mikolic Joseph M, Matthews Christine
Geriatric Research Education & Clinical Center, VA Pittsburgh Healthcare System, Pittsburgh, PA 15206, USA.
J Speech Lang Hear Res. 2005 Dec;48(6):1412-28. doi: 10.1044/1092-4388(2005/098).
The purposes of this investigation were to examine the construct dimensionality and range of ability effectively measured by 28 assessment items obtained from 3 different patient-reported scales of communicative functioning, and to provide a demonstration of how the Rasch approach to measurement may contribute to the definition of latent constructs and the development of instruments to measure them.
Item responses obtained from 421 stroke survivors with and without communication disorders were examined using the Rasch partial credit model. The dimensionality of the item pool was evaluated by (a) examining correlations of Rasch person ability scores obtained separately from each of the 3 scales, (b) iteratively excluding items exceeding mean square model fit criteria, and (c) using principal-components analysis of Rasch model residuals. The range of ability effectively measured by the item pool was examined by comparing item difficulty and category threshold calibrations to the distribution of person ability scores and by plotting the modeled standard error of person ability estimates as a function of person ability level.
The results indicate that most assessment items fit a unidimensional measurement model, with the notable exception of items relating to the use of written communication. The results also suggest that the range of ability that could be reliably measured by the current item pool was restricted relative to the range of ability observed in the patient sample.
It is concluded that (a) a mature understanding of communicative functioning as a measurement construct will require further research, (b) patients with stroke-related communication disorders will be better served by the development of instruments measuring a wider range of communicative functioning ability, and (c) the theoretical and methodological tools provided by the Rasch family of measurement models may be productively applied to these efforts.
本研究旨在检验从3种不同的患者报告的沟通功能量表中获得的28个评估项目有效测量的结构维度和能力范围，并展示拉施测量方法如何有助于潜在结构的定义以及测量这些结构的工具的开发。
使用拉施部分计分模型检查从421名有或无沟通障碍的中风幸存者获得的项目反应。通过以下方式评估项目库的维度：(a) 检查从3个量表中分别获得的拉施个体能力得分的相关性；(b) 迭代排除超过均方模型拟合标准的项目；(c) 对拉施模型残差进行主成分分析。通过比较项目难度和类别阈值校准与个体能力得分的分布，并绘制个体能力估计的建模标准误差作为个体能力水平的函数，来检查项目库有效测量的能力范围。
结果表明，大多数评估项目符合单维测量模型，但与书面沟通使用相关的项目明显除外。结果还表明，相对于患者样本中观察到的能力范围，当前项目库能够可靠测量的能力范围受到限制。
得出以下结论：(a) 要对作为测量结构的沟通功能有成熟的理解，还需要进一步研究；(b) 开发能够测量更广泛沟通功能能力的工具，将更有利于中风相关沟通障碍患者；(c) 拉施测量模型家族提供的理论和方法工具可有效地应用于这些工作。
Zotero 插件