Aim: To validate the Croatian version of the Zarit Burden Interview (ZBI) and to investigate the predictors of perceived burden.; Methods: This cross-sectional study involved 131 dyads of one informal caregiver family member and one patient with dementia visiting primary care practices (Health Care Center Zagreb-West; 10/2017-9/2018). Patient-related data were collected with the Mini-Mental-State-Examination, Barthel-index, and Neuropsychiatric-Inventory-Questionnaire (NPI-Q); caregiver-related data with the ZBI, and general information on caregivers and patients with a structured questionnaire. Principal-axis-factoring with varimax-rotation was used for factor analysis.; Results: The caregivers' mean age was 62.1±13 years. They were mostly women (67.9%) and patients' children (51.1%). Four dimensions of ZBI corresponding to personal strain, frustration, embarrassment, and guilt were assessed and explained 56% variance of burden. Internal consistency of ZBI (α=0.87) and its dimensions (α1=0.88, α2=0.83, α3=0.72, α4=0.75) was good. Stronger cognitive and functional impairment of patients was associated only with personal strain, whereas more pronounced neuropsychiatric symptoms and the need for daily care were associated with more dimensions. Longer caregiver education suppressed embarrassment and promoted guilt. Guilt was higher in younger caregivers, caregivers of female patients, patients' children, and non-retired caregivers. In multivariate analysis significant predictors of higher overall burden were male sex of the patient, higher NPI-Q, the need for daily-care services, shorter duration of caregiving, non-spouse relationship, higher number of hours caring per-week, and anxious-depressive symptoms in a caregiver.; Conclusion: The Croatian version of ZBI is reliable and valid. Our data confirm that ZBI is a multidimensional construct. Caregivers may benefit from individually tailored interventions.