Dementia caregiver burden: reliability of the Brazilian version of the Zarit caregiver burden interview

The object of this article is to examine the reliability of the Brazilian version of the Zarit Caregiver Burden Interview (ZBI). The instrument is a 22-item scale assessing the extent to which caregivers view their responsibilities as having an adverse impact on their social life, health, emotional well-being, and finances. We assessed 50 primary informal caregivers of demented patients coming from 3 different health care centers, using the test-retest method. Analysis of the results showed an intraclass reliability coefficient of 0.88, while Cronbach's coefficient alpha was 0.77 for the test and 0.80 for the retest items. The Brazilian version of ZBI shows sufficient reliability, comparable to the original version.


A b s t r a c t
The object of this article is to examine the re l i ability of the Brazilian version of the Zarit Ca reg i ver Bu rden In t e rv i ew (ZBI). The instrument is a 22-item scale assessing the extent to which c a re g i vers view their responsibilities as having an adverse impact on their social life, h e a l t h , emotional well-being, and finances. We assessed 50 primary informal care g i vers of demented patients coming from 3 different health care cent e r s , using the test-retest method. Analysis of the results showed an intraclass reliability coefficient of 0.88, while Cro n b a c h's coefficient alpha was 0.77 for the test and 0.80 for the retest items.

The Brazilian version of ZBI shows sufficient rel i a b i l i t y, c o m p a rable to the original ve r s i o n .
De m e n t i a ; Ca re g i ve r ; Home Nu r s i n g

I n t ro d u c t i o n
Patients with dementia demand a great deal of c a re; depending on the seve rity of the disord e r, significant changes in the family stru c t u re may be re q u i red. Ve ry often the family is forced to reexamine the allotment of roles and obligations in daily tasks. The patient now has req u i rements that may range from just a little help to complete assistance to perf o rm ro u t i n e tasks such as dressing, feeding, moving, etc. A f requent occurrence during this moment of c risis is that a care g i ver is chosen within the family circ l e. Howe ve r, this new role is often taken on quite suddenly by the care g i ve r, who may be unpre p a red or inexperienced, leading to an emotional burden 1 , 2 , 3 .
The care g i ver burden should thus be cons i d e red an important dimension of Alzheimer's d i s e a s e, and as such calls for in-depth inve s t igation. It is there f o re essential to use objective scales to measure the care g i ver burden and to design and evaluate therapeutic interve n t i o n s. The object of the present study is to ve rify the reliability of the Brazilian version of the Za ri t Ca re g i ver Bu rden In t e rv i e w.

I n s t r u m e n t s
The Za rit Ca re g i ver Bu rden In t e rview (ZBI) was d e veloped by Za rit and cow o rkers in 1985 4 a n d c o m p rises 22 questions graded on a scale fro m 0 to 4, according to the presence or intensity of an affirm a t i ve re s p o n s e. The questions refer to the care g i ver/patient relationship and eva l u a t e the care g i ve r's health condition, psyc h o l o g i c a l well-being, finances, and social life. The careg i ver burden is evaluated by means of the total s c o re obtained from the sum total of questions. The reliability of the original version was exc e llent ICC (intraclass correlation coefficient = 0.71; alpha = 0.91).
The ZBI was adapted to seve ral languages, s h owing a perf o rmance similar to the ori g i n a l version. The reliability indicators measured by the ICC we re in the 0.71 to 0.89 ra n g e, where a s i n t e rnal consistency ranged from 0.85 to 0.9393 5 , 6 , 7 . In the validation studies, the corre l a t i o n was r = 0.41 with the Brief Symptoms In ve n t o ry and r = 0.71 with the Global Bu rden Index 8 .

P ro c e d u re s
Ca re g i vers we re identified in three of the foll owing health care centers: (a) the Ou t p a t i e n t Ne u ro p s ychology Clinic of the Mental He a l t h Unit at Hospital Is raelita Albert Einstein, São Those who we re willing to participate signed a term of informed consent and we re submitted to another interv i e w within 3 to 6 days.

Statistical analysis
The analysis is divided into two parts: the des c ri p t i ve study (the care g i ve r s' sample demog raphic history and the patients' functional and clinical history) and reliability study. T h re e analytical pro c e d u res we re used to calculate the reliability indicators: (1) an estimate of the i n t raclass correlation coefficients of the instrum e n t's total score 9 ; (2) the internal consistency measured by Cro n b a c h's coefficient alpha 1 0 ; (3) an estimate of the Kappa values for each of the items in the instrument, measured by the f i ve levels (polychotomous) and dichotomize d into "p re s e n t" and "a b s e n t". In order to interp ret the Kappa values in the present study, they we re chara c t e ri zed according to the levels m e as u red by the degree of concordance coeffic i e n t s p roposed by Landis & Koch 1 1 , who suggested that values over 0.75 should be classified as "e xc e l l e n t" concord a n c e, whereas values below 0.40 would be considered "p o o r" and those in the 0.40 to 0.75 range as "s a t i s f a c t o ry" to "g o o d".

R e s u l t s
The ZBI was administered to a total of 50 caregivers, of whom 82% were women, ranging f ro m 23 to 81 years of age (ave rage age = 56 ye a r s ) , with 3 to 19 years of schooling (ave rage = 11 years; SD = 0.62). Some 78% we re married, 70% we re housewive s, and 56% we re the patients' d a u g h t e r s. A weekly ave rage of 111 hours was dedicated to caring for the patient; the minimum c a regiving time was 1 hour and the maximum 168 hours. Si x t y-four percent of care g i vers rec e i ved help from other family members, where a s 78% had the assistance of a hired pro f e s s i o n a l .
The most pre valent disorder was Alzheimer's disease (42%), followed by vascular dementia (8%). Ca re g i vers re p o rted that 98% of these patients had deficits in orientation, whereas 100% had memory deficits, and pro g re s s i ve aggra vation of symptoms was re p o rted in 96%. Language d i s o rders we re identified in 64% of the patients and 86% got lost in familiar places. As for functional chara c t e ri s t i c s, 56% re q u i red help taking a bath, 64% in getting dressed, 50% in going to the b a t h room, 42% needed some kind of assistance in getting out of bed, 56% had some degree of incontinence (fecal or uri n a ry) or occasional "a c c id e n t s", and 48% needed help feeding themselve s.
The ZBI showed an ICC of 0.88, while Cro nb a c h's coefficient alpha was 0.77 for the test and 0.80 for the retest items. Table 1 shows the re s u l t s of the reliability analysis for each of the questionn a i re's items. Kappa values for most of the questions with polychotomous measurements we re s a t i s f a c t o ry. The only exceptions we re questions 11 (k = 0.54), 15 (k = 0.59), and 22 (k = 0.57) with good indicators and questions 3 (k = 0.27), 7 (k = 0.39), 14 (k = 0.29), and 21 (k = 0.37) with poor ind i c a t o r s. The coefficients of questions 4, 5, 6, 13, and 18 could not be calculated due to the lack of va riability in the answe r s. The kappa values for most of the dichotomized questions we re "g o o d" and "e xc e l l e n t" in question 22 (k = 0.78). T h e questions that could not be evaluated on a polychotomous measurement level had coefficients ranging from "s a t i s f a c t o ry" to "g o o d".
Upon examining the distribution of the lowreliability answe r s, in question 3, "Do you feel s t ressed between caring for your re l a t i ve and t rying to meet other responsibilities for yo u r family or work?" the authors noted that the c a re g i ve r s' feelings changed from greater stre s s in the first interview to less stress in the second one (n = 20 to n = 7, ranging from "ve ry frequently" to "a lw a y s"). This alteration was mainly due to a greater number of answers expre s s i n g feelings of occasional stress ("s o m e t i m e s": n = 18 to n = 27), followed by "n e ve r" (n = 7 to n = 12). Two of the family care g i vers showed quite significant changes that va ried from feeling stre s s e d "ve ry frequently" and "almost always" to "n e ve r". This va riation in the answers seems to indicate a tendency tow a rds a decrease in the perc e p t i o n of stre s s, from the first to the second interv i e w s. In question 7, "A re you afraid what the fut u re holds for your re l a t i ve?" during the first int e rv i e w, the care g i vers we re inclined to keep their answers directed tow a rds "ve ry fre q u e n tly" or "nearly always" afraid (n = 32). On the other hand, in the second interv i e w, the ans wers we re often altered to "s o m e t i m e s" (n = 7), or they would change from "ve ry frequently" to "a l w a y s" afraid and vice-versa (n = 7). Only 3 subjects came up with a contra ry altera t i o n , f rom "s o m e t i m e s" to "a l w a y s". Only one of the family care g i vers showed a significant change, f rom "nearly always" to "ra re l y ". The va ri a t i o n in the answers to this question seems to indicate that from the first to the second interv i e w t h e re is a tendency tow a rds becoming less a f raid about what could happen to their re l at i ve in the future.
In question 10, "Do you feel your health has s u f f e red because of your invo l vement with yo u r re l a t i ve ? ", in the first interview the care g i ve r s we re inclined to concentrate their answers on "n e ve r" having felt that caregiving had affected their health (n = 31). Howe ve r, in the second int e rview alterations in the answers tended to s h ow that this feeling had at the most, inc reased to "s o m e t i m e s". Howe ve r, there we re no significant changes in the answe r s. He re, it may be noted that the family care g i vers did tend to feel that their health had been affected more.
In question 12, "Do you feel that your social life has suffered because you are caring for your re l a t i ve ? ", in the first interview the ans wers we re distributed between "n e ve r" and "s o m e t i m e s" feeling that social activities had been affected (n = 35). Only a few changes we re Table 1 C o n c o rd a n c e / d i v e rgence distribution and the kappa reliability coefficient for each of the questions of the caregiver burden instrument as measured in a test-retest study with 50 patients.
1. Do you feel that your relative asks for more help than he/she needs? 0 . 6  o b s e rved in the second interv i e w, with answe r s tending to concentrate more on "n e ve r" (n = 41 "s o m e t i m e s", "ra re l y ", and "n e ve r"). T h e re we re two significant dive rg e n c e s. In one, the careg i ver answe red "ra rely" in the first interv i e w and "almost always" in the second, whereas in the other the care g i ver changed the answe r f rom "a l w a y s" to "n e ve r".
Question 14, "Do you feel that your re l a t i ve seems to expect you to take care of him/her, as if you we re the only one he/she could depend on?" shows an even distribution of the answe r s and the greatest dive rgences we re those found in the questions related to the feeling that the family re l a t i ve sometimes expects to be care d for exc l u s i vely by the care g i ve r. The six subjects in the first interview changed to 10 subjects in the second one. In other word s, from the family re l a t i ves that chose the answer "s o m e t i m e s" in the first interv i e w, 3 answe red "n e ve r" and "ra rely" and 3 answe red "ve ry frequently" and "almost always" in the second interv i e w. The 10 subjects who marked this answer in the second i n t e rview chose answers ranging from "n e ve r" to "almost always" in the first interv i e w, although none of them chose "s o m e t i m e s". In a ddition, there we re two significant dive rg e n c e s, in which the answer "n e ve r" changed to "a lmost always". It became evident that question 14 had not been fully understood in the distribution analysis of the answers on the polychotomous measurement level. He re, the va riations did not follow any justifiable pattern , such as a change of feeling as to the phenomenon. On the other hand, this did not occur in other low -reliability questions.
In question 21, "Do you feel you could do a better job at caring for your re l a t i ve?" duri n g the first interview the answers tended to be c o n c e n t rated on the belief that caregiving c o u l d not be improved (n = 36: "s o m e t i m e s", "ra re l y ", and "n e ve r"), whereas in the second interv i e w the scores tended to incre a s e, thus consideri n g such a possibility. This tendency was not obs e rved by five subjects, one of whom changed the answer from "ra re l y ", to "n e ve r" and four of whom changed their answers from "nearly alw a y s" to "n e ve r" and "ra rely" In general, it was noted that the family care g i ver was eve n t u a l l y inclined to consider that his or her care g i v i n g tasks could be improve d .

D i s c u s s i o n
The Brazilian version of the ZBI proved to be easy to administrate to the dementia patients' c a re g i vers and also re l i a b l e, with ove rall re l i a-bility rates that we re comparable to those of p revious studies 5 , 6 , 7 . This is supported by the fact that the profile of the selected care g i ve r was similar to that of the original study 1 2 . In the latter, two-thirds of the patients needed the c a re g i ve r s' assistance to perf o rm at least one daily activity, which re q u i red intensive care. Fu rt h e rm o re, by selecting subjects from services with different socioeconomic standard s among the assisted patients, it was possible to obtain va riability in the levels of acceptable ans wers to each of the questions. The authors theref o re observed that only 6 questions of the 22 contained in the instrument could not be evaluated on the polychotomous measurement leve l .
The present study also shows the re l i a b i l i t y rates for each of the questions. Although this is not usually presented in the available literat u re, the authors believe that it is an import a n t aspect that could be useful in the elabora t i o n of this version of the instrument. In genera l , the instrument demonstrates "g o o d" re l i a b i l i t y, with the exception to six questions. The aut h o r s b e l i e ve that this outcome resulted from two p ro b l e m s, namely, va riation in the inform a t i o n because the question had not been fully understood and occasional va riations arising fro m the chosen study design.
Question 14 was the one that showed " l ow" re l i a b i l i t y, probably due to problems in the t ranslation. A good translation for this question can be found in the study by Scazufca 1 3 .
The authors believe that the " l ow" re l i a b i l i t y of the remaining questions was due to occasional va ri a t i o n s. This is because the pattern of alt e rations in the answers can be explained by a va riation in the phenomenon discussed in this s t u d y. For example, to diminish the stre s s, as in question 3; to feel less afraid of what might happen to the family member in the future, in question 7; to feel that one's health has been affected, in question 10; only ra rely feeling that one's social activities have been affected, in question 12; and being inclined to feel that the care g i v i n g tasks could be improved, in question 21.
The subject's interaction with this kind of i n s t rument could lead to reflections on the questions and lead to va riation in the abovementioned phenomenon. Family members in the present sample we re care g i vers of patients with a high degree of dependence, as show n b e f o re. After the interv i e w, most of the careg i vers requested orientation for dealing with the patients. This shows that the interv i e w' had raised doubts and careful consideration and may have altered their evaluation of the impact of caregiving. Co n s e q u e n t l y, being able to discuss the problems invo l ved in the care g i v i n g task could have diminished stress and fear, simply because of the feeling that such pro blems had been understood. Ca reful consideration of this issue may also have resulted in the a w a reness of some aspects that had not been noticed pre v i o u s l y, such as anxiety about not being able to do a good job. These considerations have raised two important issues: the possible impact of this type of interv i e w, especially with subjects who have already been exposed to these re f l e c t i o n s, and the need to provide support and information to these people.
Fi n a l l y, the authors conclude that this ve rsion of the ZBI has proven reliable for measuring care g i ver burden with dementia patients. In some of the questions the reliability was l owe r, for seve ral possible re a s o n s, including c u l t u ral ones. Since the reliability study of each of the questions had not been previously studied for other versions of this scale, the authors cannot compare their re s u l t s.

C o n c l u s i o n
The Brazilian version of the ZBI employed in the present study is a reliable instrument to be used as a measure of the dementia care g i v i n g b u rden. The rate of reliability proved to be c o m p a rable to the original version and is a useful instrument to be applied both in re s e a rc h and clinical pra c t i c e.