We read with interest the study by Smith et al. (Reference Knight, Fox and Chou2018) among 110 family caregivers of persons with dementia that affirmed the multidimensionality of burden as measured by the Zarit Burden Interview (ZBI) (Lau et al., Reference Ankri, Andrieu, Beaufils, Grand and Henrard2015). Factor analysis identified three factors (53.45% of total variance), namely direct impact of caregiving, uncertainty over the future, and frustration/embarrassment.

We suggest that these findings need to be interpreted in light of the extant literature. Prior studies in diverse Asian and Caucasian populations that examined the factor structure of ZBI report a range of three to five factors, accounting for 52.4 to 62.8% of total variance (Table 1). A large proportion consists of factors depicted by role and personal strain. In addition, a distinct factor has consistently been described that involves items 20 (“doing more”) and 21 (“doing better”) and independently explains 5.0 to 14.8% of total variance. This factor represents a conceptual continuum of worry about caregiving performance (WaP), ranging from “inadequacy” and “worry” through to more severe degrees of “self-criticism” and “guilt.”

Table 1. Summary of factor structure of the Zarit Burden Interview (ZBI)

All figures correct to 1 decimal place. (%): Amount of variance explained by each factor. [ ]: Question numbers from the ZBI questionnaire.

Recent studies corroborate the existence of WaP as a distinct dimension of burden among adult children caregivers, which may be particularly salient in Asian societies that are often heavily influenced by notions of filial piety and obligatory care (Cheah et al., Reference Al-Rawashdeh, Lennie and Chung2012). Unlike role and personal strain, WaP is poorly correlated with other factor scores, significantly endorsed even in milder stages of cognitive impairment and is not predicted by “conventional” factors such as functional impairment and behavioral problems (Lim et al., Reference Cheah, Han, Chong, Anthony and Lim2014). In support of this, a recent confirmatory factor analysis revealed that ZBI factor models premised on the three key dimensions of role strain, personal strain, and WaP were superior to one-factor (total score) and two-factor (role/personal strain) models (Li et al., Reference Chattat2018).

We postulate that WaP may have been missed in Smith’s study due to small sample size and inadvertent spectrum bias. In smaller samples, the correlation coefficients among variables are less reliable, affecting the generalizability of these factors. For small samples, one would be more confident if factor solutions have several high loading variables above 0.80 – which is not the case for “uncertainty” or “frustration” factors. The response rate of 21.5% inadvertently selected a group of predominantly spousal caregivers (80.9%) who likely provided direct hands-on care and for whom WaP would understandably be less relevant compared to adult children caregivers who may not be the ones providing direct care. The greater dementia severity (mean DEX: 43.6) of care recipients along with concomitant behavioral, functional, and cognitive issues would also contribute to the correspondingly higher level of burden (mean ZBI: 41.6) compared to that observed in earlier studies (Table 1). This spectrum bias may have resulted in the problematic correlation values and omission of five items including items 20 and 21.

We are thus inclined to conclude that the three derived factors reported in the study by Smith et al. (Reference Knight, Fox and Chou2018) may in fact represent variants of role strain (impact and uncertainty) and personal strain (frustration/embarrassment) with the omission of WaP. Taken together, we therefore caution against premature conclusions about the omission of WaP from the multidimensionality of ZBI burden.