Investigating Psychometric Properties of the Thai Version of Zarit Burden Interview using Rasch Model

Objective: The Zarit Burden Interview (ZBI) has been widely used to assess caregiver’s burden. Research investigating the Thai version of the ZBI is few. The study aimed to examine the psychometric properties of both the full length (ZBI-22) and short versions (ZBI-12) using Rasch analysis among a sample of Alzheimer’s disease caregivers. Results: The ZBI-22 fitted the Rasch measurement model regarding unidimensionality but not for ZBI-12. Five items from ZBI-22, and two items from ZBI-12 were shown to be misfitting items. The model of ZBI-12 was improved when item “should do more” and “could do a better job caring” were removed. Reliability was good for both forms of the ZBI (a = 0.86 - 0.92). Significant correlations were found with caregiver’s perceived stress and caregiver's depression. Significant correlation with subscales of anxiety/depression, pain and mobility were indicative of discriminant validity but not with self-care and usual activity (p > 0.05). To conclude, the Thai version ZBI was supported for the reliability and validity in both the full length and 12 short forms among Alzheimer’s disease caregivers; however, some misfitting items of the ZBI undermined the unidimensionality of the scale, and need revision.


Introduction
Feeling burden is stressor-related and varies from one to another. Studies showed that burden can ultimately lead to depression [1]. Caregiving especially for elderly with dementia usually causes burden among caregivers, so identifying feeling of burden to have early as possible intervention is important.
One of the oldest and most common measurements to assess caregiving burden is the Zarit Burden Interview (ZBI) [2,3]. It currently has two forms, a long form, consisting of 22 items, and a short form.
Recently, the ZBI was tested using item response theory, yielding a different set of items for the short scale as compared with the former 12-item ZBI [14].
Thailand is a becoming ageing society, so burden of the caregivers is important and associated with mental health problems. To quantify the level of burden would allow caregivers an opportunity to be helped in a timely fashion. Thus, the ZBI is one vital tool to gauge the level of burden as it shows good psychometric properties. However, the ZBI has never been tested for psychometric property in a Thai population using Rasch measurement model. We aimed to examine its construct by means of convergent, discriminant and concurrent validity, using both Rasch analysis and confirmatory factor analysis.

Population
Maharaj Nakorn Chiang Mai Hospital, participated in the study. All gave written informed consent before completing the questionnaires. Sociodemographic data and records related to caregiving were obtained in addition to specific measurement.

ZBI
The ZBI is a caregiver-reported questionnaire measuring the burden the respondent feels in providing caregiving to the patient. Currently, it has two forms, long (22 items) and short (12 items), with a Likert scoring scale between 0 (never) and 4 (nearly always). The ZBI offers the interpretation of score as follows; 0 to 20, little or no burden, 21 to 40, mild to moderate burden, 41 to 60, moderate to severe burden and 61 to 80, severe burden [4,15].
The Thai version of the ZBI was allowed to be used for the present study by Professor Zarit and Mapi Research Trust [16]. The study sample showed a Cronbach's alpha of .921 for the full-length version, and .865 for the short version.

Perceived stress scale (PSS)
The PSS is a self-reporting, 10-item questionnaire measuring the extent to which individuals perceived stress [17]. The 4-response Likert scale, ranges from 0 (not at all) to 4 (the most); the total scores ranges from 0 to 40 and the higher the score, the higher the level of feeling stress.
The Thai version showed good psychometric properties [18]. The study sample showed a Cronbach's alpha of .850.
Patient Health Questionnaire (PHQ-9) The PHQ-9 is a self-reporting, 9-item questionnaire measuring the extent to which an individual feels bothered due to depressive symptoms over the past two weeks [19]. The 4-response Likert scale ranges from 0 (not at all) to 3 (nearly every day); the total scores ranges from 0 to 27 and the higher the score, the higher the level of feeling depression. The Thai version showed good psychometric properties [20]. The study sample showed a Cronbach's alpha of .849.

EQ-5D
The EQ5D is a self-reporting questionnaire measuring health-related quality of life [21]. It comprises a 5-item questionnaire assessing 5 domains of health state: mobility, self-care, usual activities, pain and anxiety/depression, with a 5-response type ranging from 1 (no problem) to 5 (severe problem).
All 5 aspects were calculated to an index score, with the maximum of 1.000, the higher the score, the better the quality of life. The utility values for EQ-5D health states were estimated from the Thai general population [22]. The study sample showed a Cronbach's alpha was .723.

Statistical Analysis
Sociodemographic data was analyzed using descriptive statistics. Pearson or Spearman rank was used for correlational analysis. The same items are present in both tests, leading to an overestimate of the "true" correlation, so a corrected correlation was made between the full length and short forms Based on measurement theory, a scale should demonstrate that all items contribute to the same construct, and has monotonically increasing steps. All these properties can be illustrated by the Rasch model. The following approach was conducted for analysis.

Rasch analysis
The Rasch model belongs to the item-response latent trait models, probabilistic logistic model that predicts that the response to a particular item is influenced by the quality of both person and item. To investigate the ZBI data, the partial credit Rasch model was used [24], with the following criteria.

Convergent, discriminant and concurrent validity
We tested the ZBI against the EQ5D subscale, hypothesizing that ZBI should relate more to anxiety/depression than mobility. We expected to find a positive correlation between ZBI and PSS and PHQ-9 to demonstrate concurrent validity.

Confirmatory factor analysis (CFA)
To test how data were well modeled with the unidimensional construct, CFA was performed for both

Results
The average age of the caregiver sample was 55 years (S.D.= 12.9); most were women (77.5%), with a high school level of education. The details are shown in Table 1.
According to ZBI level, the sample was reported to have little or no burden (< 21). However, nearly 10% felt they had moderate to severe burden. The quality of life index score was quite high on average (0.88), while perceived stress and depressive symptoms were low (Table 1).
Correlation analysis showed that ZBI-22 had a coefficient of 0.855 with ZBI-12 for the uncorrected correlation, and 0.784 for correlation when error variance due to sharing items was corrected. Both forms of ZBI significantly related to the EQ5D index score, PHQ-9 and PSS. For the EQ5D subscale, ZBI significantly related to subscale mobility, pain and anxiety/depression, but not with self-care and usual activity ( Table 2).
For the distribution of the ZBI-items. Items 10, 13, 16, 17 and 18 for ZBI-22 and items 4, 6, 8 and 9 for ZBI-12 had unacceptable kurtosis (>±3). The kurtosis contributed to the high frequency of zero categories on these respective items (Table S1).
Rasch analysis of the ZBI showed that PCA of the residual yielded 46% variance, explained by measure for ZBI-22, and 48.7% by ZBI-12. The unexplained variance in the first contrast yielded eigen values of 2.52 and 3.03, indicating that 2 to 3 items formed another dimension, i.e., items 20 (should do more) and 21(could do a better job caring) for ZBI-22, likewise for ZBI-12. However, based on disattenuated correlation between person measure, ZBI-22 was shown to be more unidimensional than ZBI-12 (all coefficients > 0.7). Items 1, 14, 15, 18 and 21 for ZBI-22, and items 11 and 12 of ZBI-12 were shown to be misfitting items. In addition, 3 pairs of items from ZBI-22 and two pairs of items from ZBI-12 had standardized residual correlations above 0.3, indicating item dependency of both forms of ZBI (data not shown). Item and person separation and reliability were shown to be in an acceptable range. For category function, one-half of ZBI-22 and 4 of 12 of ZBI-12 were found to be disordered category or threshold. From all, three was common between ZBI-22 and ZBI-12 (Table 3).

Discussion
The present study aimed to evaluate the psychometric properties of the full length and short form of the Thai version of the ZBI among caregivers of the patients with Alzheimer's disease. In general, both forms of ZBI demonstrated a valid and reliable scale.
According to Rasch analysis and CFA, both ZBI-22 and ZBI-12 did not; however, demonstrate a unidimensional scale as expected using the measurement model, even though the ZBI-22 seemed to be favored over ZBI-12.
Both Rasch analysis and CFA identified items 11 (should do more) and 12 (could do a better job caring) as problematic for the scale. In addition to these two items, CFA suggested more pairs of error variances to be correlated to make the model an acceptable fit for ZBI-12. These included items 5 (feel strained) vs. 3 (feel angry); 8 (lack of social life) vs.7 (lack of privacy) and 8 (lack of social life) vs. 9 (lost control of life). This was indicative of item dependence, which was usually ignored in CFA, but unacceptable by the measurement model according to the Rasch model.
The fact that disattenuated correlation (>0.70) in ZBI-22 indicated that it could be sufficient unidimensional, another real dimension did not exist. Our results showed that all acceptable items of ZBI-22 were embedded in ZBI-12. Hence, the ZBI-12, except for items 11 and 12, looked promising. Disordered category and threshold; however, indicated that the response might need to be collapsed to improve measurement ability.
Our results were consistent with Ballesteros et al. [14] using the same item response theory; items 11 (should do more) and 12 (could do a better job caring) were excluded from the new 12-item ZBI.
This was interesting because it might go against the investigators' intention to make the scale shorter for practical use while maintaining acceptable unidimensionality, or else the notion to use the ZBI-12 by sum or total score would be unjustified. From these findings, ZBI-10 (excluding items 11 and 12) could be promising for the short ZBI rather than the ZBI-12 version by Bedard et al. [4].
In conclusion, the Thai version of ZBI showed adequate psychometric properties to measure burden among caregivers of patient with Alzheimer's disease. The ZBI-12, despite being shorter, showed less unidimensionality than the original ZBI-22, and some items were suggested to be removed. All patients provided written informed consent to the study,

Consent for publication
Consent for publication is not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.