Psychometric Validation of The Cognitive Abilities Screening Instrument Using Rasch Analysis In People With Dementia
Abstract Background: The Cognitive Abilities Screening Instrument (CASI) is one of the most commonly used cognitive screening tools to assess overall cognitive function in people with dementia. However, the unidimensionliaty of the CASI using Rasch analysis has not been evaluated in people with dementia, limiting its utility in clinical and research settings. Unidimensionality verifies whether all items of a measure reflect a single theoretical construct, which is necessary to determine whether clinicians and researchers can appropriately use the sum scores of the CASI to describe overall cognitive function. This study aimed to examine unidimensionality of the CASI using Rasch analysis and estimate Rasch person reliability in people with dementia. Methods: CASI data of people with dementia was collected from medical records of one general hospital in northern Taiwan. A total of 506 people with dementia were recruited from the Department of Neurology. Unidimensionality was confirmed through two assumptions: (1) the infit and outfit mean square (MnSq) were 0.6-1.4, and (2) residual variance of the first principal component in principal component analysis was ≤ 20%. Rasch person reliability was estimated after undimensionality was supported. Results: One item from the list-generating fluency dimension was misfitted (outfit MnSq=1.42) and was deleted. The unidimensionality of the remaining 45 items (referred to as the CASI-45) was supported with an infit and outfit MnSq (0.85-1.24 and 0.84-1.28, respectively) and low residual variance of the first principal component (12.8%). The Rasch person reliability of the CASI-45 was 0.62. Conclusion: The CASI-45 showed a unidimensional construct and had acceptable Rasch person reliability in people with dementia.