Background:: The Zarit Burden Interview (ZBI) is a commonly used self-report to assess caregiver burden. A 12-item short form of the ZBI has been developed; however, its measurement invariance has not been examined across some different demographics. It is unclear whether different genders and educational levels of a population interpret the ZBI items similarly. Therefore, this study aimed to examine the measurement invariance of the 12-item ZBI across gender and educational levels in a Taiwanese sample.

Methods:: Caregivers who had a family member with dementia (n = 270) completed the ZBI through telephone interviews. Three confirmatory factor analysis (CFA) models were conducted: Model 1 was the configural model, Model 2 constrained all factor loadings, Model 3 constrained all factor loadings and item intercepts. Multiple group CFAs and the differential item functioning (DIF) contrast under Rasch analyses were used to detect measurement invariance across males (n = 100) and females (n = 170) and across educational levels of junior high schools and below (n = 86) and senior high schools and above (n = 183).

Results:: The fit index differences between models supported the measurement invariance across gender and across educational levels (∆ comparative fit index (CFI) = −0.010 and 0.003; ∆ root mean square error of approximation (RMSEA) = −0.006 to 0.004). No substantial DIF contrast was found across gender and educational levels (value = −0.36 to 0.29).

Conclusions:: The ZBI is appropriate for combined use and for comparisons in caregivers across gender and different educational levels in Taiwan.