The ZBI is frequently used to measure the subjective burden of care. There are several versions of the ZBI, that is, those with 29 items [5
], 20 items [6
], and 22 items [7
]. In a review, Hérbert cited two papers that showed test-retest reliability of 0.71 and 0.89 and Cronbach’s alpha of 0.91 and 0.85, respectively [23
]. Validity was assured by an association with psychological well-being [24
], health-related quality of life [25
], frequency and difficulty of caregivers’ problems [26
], adherence ratings [27
], categories of problems of caregivers in relation to various parameters [26
], and the Brief Symptoms Inventory or Global Index of Burden [23
]. Arai proposed a Japanese version of the ZBI (Z-JBI) through the review process of translation and reverse-translation and established its high reliability and validity [12
]. Since it is very time consuming to answer the 22 items on the most frequently used ZBI, Arai made a short version of the Japanese version of the ZBI (J-ZBI_8) [17
]. The reliability and validity of J-ZBI_8 were as follows. The correlation coefficient
between ZBI and J-ZBI_8 was 0.93 among 691 people with disability in a Japanese community [17
] and 0.92 among 169 people who used care services in the community [18
]. To assess adequately the degree of the burden of care, the score of the J-ZBI_8 can be converted to the standard ZBI by multiplying by 2.75 (22 divided by 8).
Factors influencing the burden of the caregiver can be classified into two groups: those related to caregivers and those related to care recipients. Caregivers’ emotional factors such as depression would be elicited by factors related to care recipients. The cause-and-effect logic may be complicated; however, factors related to recipients would be first in the chain of events affecting the caregiver. In this study, to avoid confusion. we discussed factors only related to care recipients such as their ADL level, instrumental ADL level and the depressive state. Factors related to impairment, such as paralysis or range of motion, are neglected in this study since the degree of impairment is reflected in the ADL or instrumental ADL in the chronic
The factor most related to the ZBI in this study was depression. The JSS-D, which we used, is not popular internationally. The items on the JSS-D were selected from items on 11 existing scales, DSM-IV and ICD-10. Conjoint analysis, which used 18 hypothetical patients and involved 100 specialists belonging to the Japan Stroke Society, was performed to decide the proper weight of each of the selected items [21
]. The inter-rater reliability of each item on the JSS-D was expressed by the value of the kappa coefficient ranging from 0.65 to 0.89 and the intra-rater reliability was 0.80 to 0.97 [21
]. We chose the JSS-D because of its validity, reliability and ease of use.
With regard to the relationship between the ZBI and depression scales, Arai reported a correlation coefficient of 0.67 between the ZBI and Center for Epidemiologic Studies Depression Scale in 45 Japanese elderly in a rural community [13
]. Since a depressive state not only lowers the activities of those with the depression but also makes the caregiver overly careful not to worsen the depression, the burden on the caregiver becomes great.
In this study, the ZBI score had little relation to ADL or instrumental ADL. Opinions on such a relationship in previous papers were controversial, with some reports showing no relationship. There was no relation between Physical Self Management of demented patients in an ADL instrument and the total score of Relatives’ Stress Scale [10
]. Scholte showed a 14% contribution of the FAI to predict results of the Sense of Competence Questionnaire for partners of stroke survivors; however, no contribution of the Barthel Index was found [28
]. The fact that more than half of Scholte’s patients were independent would contribute to the negative result. Zarit showed a non-significant correlation coefficient of -0.06 between the ZBI and ADL scores and -0.20 between the ZBI and Lawton’s instrumental ADL scores [5
]. Arai showed no differences in the mean ZBI score of caregivers when caregivers were divided into two groups according to the patients’ Barthel Index score (more than 60; 60 or less) [12
On the other hand, Fitting noted that severe functional impairment was related to the severe burden expressed by the ZBI in spouses of patients with dementia [15
]. Although this impairment
scale includes both cognitive disturbance and physical disability, this is one proof of the contribution of ADL status to the burden of care. The Katz ADL index showed a small but significant correlation to the ZBI score among caregivers of demented patients [29
Why did these discrepancies arise? Detection power of evaluation instruments to find small but important differences may change the result. Since the FIM uses a 7-point scale for each item, the use of the FIM can contribute to refining such a relationship. The FIM is the most frequently used ADL instrument in rehabilitation. Its reliability [30
] and validity [30
] are well established. Only two papers mentioned both the ZBI and the FIM [26
]; however, these papers separately described the relation between the caregivers’ problems and the ZBI score and the relation between the caregivers’ problems and the recipients’ FIM. Therefore, we employed the FIM in this study.
In previous papers on the burden of caregiving, the most prevalent diagnosis of patients was dementia; however, our subjects included a mixture of physically disabled and demented patients. ADL can be lowered in two ways. In physically disabled patients, the impairment itself, such as paralysis, disturbs independence. Decreased initiative and confusion as to how to perform ADL are reasons for ADL dependency in demented patients. Thus, the proportion of the physical disabilities and intellectual problems will change the interpretation of the relationship between the burden of care and the ADL status.
In the present study population, the relationship between ZBI and FIM scores is rather complex and does not show a linear pattern as can be seen in the scattergram. Arai commented that those who looked after the disabled elderly with partially limited ADL tended to feel a heavier caregiver burden than those caring for the elderly who were totally dependent in ADL [13
]. This was reconfirmed in our study by the comparison of the FIMM among 3 groups that were formed according to the FIMM. This phenomenon can be explained as follows. Assisting movement of subjects who need a great deal of help is a harder burden of care than simply moving the subject without considering the subject’s desires. In order that a patient can perform an ADL a little, a caregiver must wait for long periods for the patient to execute the activity. Thus, the burden of the care of subjects who can perform a few ADL is more severe than the burden of care of completely bedridden subjects. On the other hand, the relationship between IADL and ZBI was rather simple. Subjects who performed instrumental ADL place only a slight burden on caregivers. It is because both good physical ability and good cognitive ability are needed to perform instrumental ADL.
The CART was used to find the cut-off point to discuss the above matter. In general, the beneficial points of the CART were the ability to treat non-linear relationships and simple results sufficiently to understand the logic of the result. Since the CART itself discards information on variables that are irrelevant for the purpose, we did not want to use the CART solely as a predictive tool but wanted to use it as a method to determine the cut-off point of the ADL.
Some reports stress the stronger relationship of the behavioral disturbance to the ZBI than the ADL [13
]. In our study, the cognitive subscore of the FIM can be recognized as a marker of behavioral disturbance because it contains items on social interaction and problem solving. Since the rank correlation coefficient between the ZBI and the cognitive subscore of the FIM is not very high, behavioral problems do not seem to be the main factor in a deteriorated ZBI score. The power of abnormal behavior changes according to the level of ADL. If patients with good physical ability have a behavior problem, they can go anywhere and do what they want. So the anticipated stress of an undesirable event will be large, such as in the case of dementia
. If the patient is almost dependent in ADL, the extent of damage by the abnormal behavior will be small because the area where an undesirable event occurs can be predicted. Thus, subjects with lower ADL status will weaken the correlation between the ZBI and FIMC in this study.
The age of people with disability showed low correlation to the J-ZBI_8 score. One reason is the narrow range of age distribution in this study. The average age was 81 with a standard deviation of 8. As a shortcoming of this study, profiles of caregivers were not analyzed. In a follow-up study, Arai found a difference in the change of the burden whether the caregiver was a spouse or a daughter-in-law [14
]. Furthermore, Fitting noted that in younger wives and older husbands scores indicating burden were higher with increasing levels of impairment; however, this did not hold true for older wives or younger husbands [15
]. The effects of characteristics of caregivers on the degree of burden of care is rather complicated. Although stratification by gender, age, or degree of kinship is needed and a fairly larger sample will be required to clarify the complicated structure, adding many cases is not easy and is an issue for studies in the future.