Abstract

Background: The Novak Caregiver Burden Inventory (CBI) is commonly used to assess caregiver burden in caregivers of people with dementia. Although the CBI's five-factor structure has been supported in previous studies, the unidimensionality of each individual domain requires further validation, particularly in culturally specific populations and disease contexts. This study aimed to examine the domainlevel unidimensionality of the CBI among caregivers of people with dementia in Taiwan. Internal consistency and ceiling and floor effects were also examined.

Methods: A total of 200 caregivers of people with dementia completed the CBI. The mean ages of the caregivers was 55.9 years (standard deviation = 13.4), with the most were female (70.0%), and an average caregiving duration was 5.2 years (standard deviation = 4.2). Confirmatory factor analysis (CFA) and Rasch analysis were performed to examine the unidimensionality of each domain. Cronbach's alpha was applied to examine internal consistency. The percentages of participants with maximum and minimum scores in each domain were used to estimate ceiling and floor effects, respectively.

Results: The four CFA fit indices of the five domains were: comparative fit index = 0.98–1.00, Tucker-Lewis index = 0.95–1.00, root mean square error of approximation = 0.000–0.170, and standardized root mean square residual = 0.023–0.055. Based on Rasch analysis, Infit and outfit mean squares of the five domains were 0.61–1.69 and 0.61–1.43, respectively. The alpha values of the five domains were 0.79–0.92. Ceiling effects ranged from 2.5% to 17.5% and floor effects ranged from 1.0% to 9.5%.

Conclusion: Our results supported the unidimensionality, demonstrated good internal consistency, and showed no notable ceiling or floor effects in the five CBI domains for caregivers of people with dementia. The total score for each domain can be utilized to indicate domain-specific caregiver burden.