Unidimensional 12-item Zarit Caregiver Burden Interview for the assessment of dementia caregivers' burden obtained by item response theory

Abstract

Objectives: To develop a one-dimensional version of the 22-item Zarit Caregiver Burden Interview (ZBI) by applying item response theory approaches.

Methods: The answers to the 22-item ZBI of 241 caregivers participating in a clinical trial were analyzed 1) with a Mokken nonparametric item response theory analysis to ascertain the dimensional structure underlying the scale and obtain a one-dimensional reduced version, 2) with the Samejima's graded response model to assess the item characteristics of the reduced version, and 3) with confirmatory factor analysis to confirm the unidimensionality of the reduced ZBI version and assess the item loadings to the burden latent variable.

Results: Mokken analysis resulted in a major one-dimensional scale comprising 12 items directly related with burden. All items showed scalability indices over 0.30. The scalability for the overall scale was 0.44 defining a medium scale according to Mokken's criteria. An unconstrained Samejima's graded response model showed appropriate fit, and most items of the reduced 12-item ZBI presented pertinent difficulty and discrimination parameters. The results of the 12-item ZBI confirmatory factor analysis fitted to a one-dimensional latent structure for burden (comparative fit index=0.975; root-mean-square error of approximation=0.067; weighted root mean square residual=0.677). All factor lodgings were above 0.40 with items 9 (strained by the relative) and 22 (overall feeling of burden) presenting the highest loadings.

Conclusions: The reduced 12-item ZBI fits a one-dimensional latent variable of burden. Further psychometric studies, focusing on its equivalence for different populations, sensitivity to change, and minimal important difference are warranted.