ＧＢＳスケール日本版の信頼性と妥当性の検討

The Japanese version of the GBS Scale (GBSS-J) was administered to 246 elderly patients with dementia in a geriatric hospital to test the inter-rater reliability and validity of the items and the subscales of the GBSS-J. The inter-rater reliability of the 26 items was tested by rating 20 patients. After the raters conferred with each other how to rate the items of the scale, the items were rated independently. Pearson's correlation coefficients indicating the agreement between the rates were as good as more than. 660 except for the items of impaired wakefulness and confusion. The validity of the 28 items was examined by comparing them with the stages of Functional Assessment Staging (FAST). Hasegawa's dementia scale (HDS) was employed as an external standard to test the items of the subscale of intellectual functions. Physical disability was checked by using the Rapid Disability Rating Scale (RDRS). Mean scores of the 20 items except for the 6 items included in the subscale of different symptoms common in dementia significantly increased as the FAST stage advanced. These results indicated that the 20 items of the GBSS-J measured the severity of dementia with sufficient validity. In addition, according to the results of analysis of variance and covariance, the items of motor insufficiency in taking food, impaired physical activity, impaired wakefulness and irritability were related to the severity of physical disability rather than that of dementia. Aslo, the 10 items of the subscale of intellectual functions showed significant correlations with the scores of HDS, while the item of impaired wakefulness showed a significant, although fairly low correlation. In the partial correlation controlled with eating, walking, bathing, dressing and excretion in RDRS, no significant correlation with scores of HDS was obtained with the item of impaired wakefulness. This finding supports the above mentioned results. Cronbach's α was used to test the internal reliability of the 4 subscales of the GBSS-J. Cronbach's αs of the subscales of motor functions, intellectual functions and emotional functions were as high as .938, .958 and .818, respectively, while that of different symptoms common in dementia was as low as .716. These findings indicated the 3 subscales other than that of different symptoms common in dementia showed sufficiently high internal reliability. As for the relation between the total scores of the 4 subscales and the stages of FAST, Spearman's correlation coefficients were as high as .791 in the subscale of motor functions, .987 in intellectual functions, and .769 in emotional functions, while that of different symptoms common in dementia was as low as .213. Pearson's correlation between the scores of HDS and the total scores of intellectual functions was as high as -.866. In factor anlaysis with 26 items to test the construct validity of the 4 subscales, 5 factors emerged. The original 4 factors, i.e., 4 subscales were not confirmed. The above mentioned results indicated that the reliability and validity of the items comprising the GBSS-J were sufficient apart from the several items, to measure the severity of dementia. However, the construct validity of the original 4 subscales which is very important in developing scales for the aged was not confirmed in the present study.