[A neuro-psychological test (T-K-W test) for dementia based on working memory theory and item-response theory: its development and construction].

The purpose of the present study was to develop a new neuro-psychological test for evaluating the severity of dementia based on the newly-developed theory of working memory. In Study I, twelve candidate test items, including experimental tasks used in our previous developmental studies and test items newly developed for this study, were administered to twenty four demented patients. In Study II, the same twelve test items as used in Study I were administered to eighteen normative-aged volunteers to identify the baseline of difficulty of each test item. The normative-aged group, in marked contrast to the demented group, responded correctly to nearly all of these items. These items, therefore, proved to have a discriminating power between dementia patients and normative-aged. After examining thoroughly the result of factor analysis of test scores, the difficulty of the items, and the feasibility and communicability of instructions for each item, the authors finally chose five items to construct the test. Four of these five items were ones originally developed by ourselves. In Study III, the final version of the test composed of the above five items was administered to one hundred demented patients of various etiologies. The test scores were analyzed according to the method of principal factor analysis and the theory of working memory. The results of factor analysis showed the factor pattern of a typical homogeneous test that measures a single dimension of working memory capacity. The authors, therefore, decided to apply to the test scores an item-response theory that postulates test homogeneity as a prerequisite. Based on this theory, parameters of discrimination and difficulty of each item were estimated. In addition, the scale value of each examinee, i.e. working memory capacity, was estimated from his or her response pattern to the items. When the administration of the whole test is difficult in a clinical setting, this theory makes it possible to estimate the scale value of an examinee from even a single item, though inevitably with lower reliability. Owing to the adaptability of this test, its practical utility as a measurement tool was much improved. Further, the test items and their subitems were analyzed in terms of the theory of working memory. The resulting high agreement between item difficulty and working memory demand confirmed the validity of the theoretical analysis. In addition, the result of discriminant analysis with thirty-one subitem scores of five items showed that the test discriminated correctly 79.3% of DAT and VD.