Hofer S M, Piccinin A M, Hershey D
University of Southern California, USA.
J Clin Psychol. 1996 Jul;52(4):395-409. doi: 10.1002/(SICI)1097-4679(199607)52:4<395::AID-JCLP4>3.0.CO;2-P.
The study examines the factor structure and provides test of the discriminative properties of the 38-item Mattis Dementia Rating Scale (MDRS). The MDRS was designed a priori to measure five broad domains of cognitive abilities: attention, initiation/perseveration, conceptualization, construction, and memory. Complete item level data were collected at the USC Alzheimer Disease Research Center from 19 probable Alzheimer's Disease (AD) patients, 17 cases with dementia of various etiologies (e.g., multiple infarct), and 49 contrast subjects. Factor analyses, with rotation to equamax criterion, were performed on education partialled data. Five and six factor solutions accounted for most of the reliable variance and permitted simple structure theoretical description for separate subscales. These factors, similar to Mattis' design, can be characterized as Memory (Recall)/Verbal Fluency, Construction, Memory (short-term), Initiation/Perseveration, and Simple Commands. Cross-validated discriminant analyses performed on five unit-weighted composite variables derived from factor analysis provided better classification (72% vs 62%) than the 38 Mattis items alone.
该研究考察了38项马蒂斯痴呆评定量表（MDRS）的因子结构并对其区分性特征进行了检验。MDRS在设计之初旨在测量认知能力的五个广泛领域：注意力、启动/持续、概念化、构图和记忆。完整的项目级数据是在美国南加州大学阿尔茨海默病研究中心收集的，来自19名可能患有阿尔茨海默病（AD）的患者、17名患有各种病因（如多发性梗死）痴呆的病例以及49名对照对象。对剔除教育因素的数据进行了采用等量最大旋转标准的因子分析。五因素和六因素解决方案解释了大部分可靠方差，并允许对各个子量表进行简单结构的理论描述。这些因子与马蒂斯的设计类似，可被描述为记忆（回忆）/语言流畅性、构图、记忆（短期）、启动/持续和简单指令。对从因子分析得出的五个单位加权复合变量进行的交叉验证判别分析提供了比单独使用38个马蒂斯项目更好的分类（72%对62%）。
