Validation of the Burden Index of Caregivers (BIC), a multidimensional short care burden scale from Japan

Participants were caregivers providing home healthcare to patients with intractable neurological conditions and to stroke patients between November, 2003 and May, 2004. A self-rating questionnaire was mailed to all caregivers of patients registered as having the intractable neurological conditions PD, SCD, MSA, and ALS in Mie Prefecture, Japan.

The participants were requested to answer the questionnaire and return the answer sheets. As the qualitative interview study progressed, characteristics of the care burden for neuromuscular disease were found to be similar to those for other diseases, including cerebrovascular accident (CVA), which is the leading cause of bedridden elderly in Japan. Therefore, in consideration of the possible application of the instrument to diseases other than intractable neurological conditions, we also included caregivers of patients with CVA. For this CVA component, the questionnaire was sent to outpatient, day care, rehabilitation, and visiting nurse station departments of the Brain Attack Center Oota Memorial Hospital, a neurological hospital in Hiroshima Prefecture, Japan, for distribution to caregivers who provide nursing care at home to patients following cerebral infarction, cerebral hemorrhage, and subarachnoid hemorrhage. These caregivers were similarly asked to answer the questionnaire and return the answer sheets by mail. To examine reliability, two-week test-retesting of the questionnaire was done in consenting caregivers of stroke patients.

Before the survey, a qualitative study was conducted to collect data and interview primary caregivers of patients with PD, ALS, and CVA. A total of 22 caregivers were interviewed and analyzed. This study used the theoretical sampling method. Caregivers were interviewed for 1.5 h at home under written consent with questions such as "what is the hardest thing about caring for the patient?" The data were analyzed using the grounded theory approach and constant comparative method by means of open, axial and selective coding [16, 17]. Data were continually analyzed as they were collected, with comparison between newly gathered data and previous data to examine similarities and differences. Analysis was supervised by a qualitative researcher and results were discussed among neurologists and nurses experienced with neurology patients to examine validity.

Characteristic results revealed care burden items such as "difficulty with finding meaning in prolonging life" (anguish of caregivers in continuing care) and difficulty with finding "a sense of satisfaction" due to a lack of appreciation from other family members, relatives, and others. These items had not been included in the conventional scale. In addition, the qualitative study also extracted an item that reflected the current situation of troubled caregivers whose patients resented receiving public nursing care services, despite having agreed to do so before the services were made available. We then combined our qualitative research results and literature review [1‐9] to prepare a preliminary group of 28 care burden scale items. A number of items, including overall burden, were adopted from pre-existing care burden scale items and their wording was revised. The 28 care burden items comprised 27 items of specific burdens (e.g. "I do not have enough time for myself," "I am completely distressed by caregiving", et al), assessed using a 5-point Likert scale from 0: never, 1: almost never, 2: sometimes, 3: often, to 4:always; and 1 item of overall burden i.e. "how much burden do you think providing care is to you?" To examine content validity, three caregivers, two neurologists, and two visiting home nurses checked the questionnaire.

The following instruments were also measured to examine construct and concurrent validity:

Regarding demographic factors, we collected information on caregiver's age, gender, relationship to patient, working status, income, period of nursing, daily hours spent on nursing, hours required for close supervision of the patient, patient's age, level of ADL (as determined by the Ministry of Health, Labor and Welfare), and Barthel Index (for CVA only) [21].

We first prepared a multidimensional scale by selecting items from the 28 preliminary items. In the item-selection process, we performed item analysis to confirm that selected answers were not heavily weighted on either end of the 5-point scale (i.e. 80% or more responders selected points 0 and 1 or 4 and 5), and examined reproducibility with a weighted kappa coefficient to exclude items with κ value < 0.5. On the basis of these results, we confirmed multidimensionality by an explanatory factor analysis (principle method with a promax rotation), applied the item response theory (marginal maximum likelihood method, graded response model) to each factor, and determined the difficulty level and discrimination power of each item. We then adopted those items which included a difficulty level of 0 and a discrimination power of 2 or higher. Finally, we established a 5-dimensional Burden Index of Caregiver (BIC) that contains 11 items.

We then verified the validity and reliability of the BIC. We first calculated the descriptive statistics of scores in each domain, and examined internal consistency using Cronbach's alpha coefficient. We then performed explanatory factor analysis (principle method with a promax rotation) to examine factor validity. To address construct and concurrent validity, we calculated Pearson's correlation coefficient between the BIC and health-related QOL scale (SF-8), depression (CES-D), and preexisting ZBI. To address known groups validity, we summed BIC scores according to the hours spent on caregiving per day, and calculated Spearman's correlation coefficient, r [2, 22]. To evaluate reproducibility, we calculated intraclass correlation coefficient (ICC). In addition, we similarly confirmed the unidimensionality of overall care burden feelings, calculated Cronbach's alpha coefficient and ICC to examine the availability of a total score, and calculated correlation of the total score to confirm items of construct and concurrent validity.

We used MULTILOG 7.03 (Assessment Systems Corporation, St. Paul, MN) for item response theory and SAS Version 9.1 (SAS Institute, Cary, NC) for all other analyses.

Before implementing this study, the protocol was reviewed by Ethical Committees at the Faculty of Medicine, the University of Tokyo, Mie University Hospital, and Brain Attack Center Oota Memorial Hospital. Each subject was informed by a written document that participation in the study was voluntary and their privacy would be strictly protected.

The questionnaire was sent to 1577 patients with intractable neurological diseases and 332 following stroke, and answer sheets were received from 785 (50%) and 220 (67%), respectively. The number of total responders (analysis set) who provided valid final responses was 646 (PD, 279; SCD, 78; MSA, 39; ALS, 30; and CVA, 220).

The average age of caregivers was 64.3 ± 11.6 years, and 64.9% were women (Table 1). The relationship between patient and caregiver was spouse in 72.3%. The employed population was 23%. Among all responders, annual income was less than 3 million yen (US$X) for 41.4% and less than 5 million yen (US$X) for 73.0%. Average duration of caregiving was 5.2 ± 4.4 years, and average time spent on care was 6.0 ± 6.3 hours daily. According to ADL levels determined by the Ministry of Health, Labor and Welfare, 14.3% of patients were Level C (bed rest throughout the day), 34.2% were Level B (need help for daily activity), 34.2% were Level A (independent at home but need some help outside the house), and 17.2% were Level J (almost completely independent) or physically independent.

Care burden items were selected according to the procedures described in Methods. The finalized BIC comprise 11 items, as follows: 1. I do not have enough time for myself because of caregiving, 2. I cannot freely leave the house because of caregiving, 3. I am completely distressed by caregiving, 4. I want to delegate the care to someone else, 5. I am experiencing hardship because caregiving does not give me a sense of satisfaction, 6. Caregiving is hard because I cannot find the meaning of nursing, 7. My body aches when nursing, 8. I have ruined my health because of nursing, 9. I have a hard time because patients resent receiving public nursing care service, 10. It is a burden that public nursing care service personnel enter our house; and regarding the total care burden, 11. All things considered, how much burden do you think providing care is to you?

We assessed the face validity of the survey items among nurses who had experience in caring for patients with either intractable neurological diseases or stroke.

Five factors were extracted as a result of the exploratory factor analysis performed on the 10 items, excluding the total care burden of the BIC, and named "Time-dependent burden", "Emotional burden", "Existential burden", "Physical burden", and "Service-related burden" (Table 2). Service-related burden demonstrated a slight deviation in distribution and its Cronbach's alpha coefficient was low (alpha = 0.68). For the other domains, however, distributions showed no deviation and internal consistency was high (0.84 to 0.89). The average total score of the BIC was 16.8 (± 8.9) and its Cronbach's alpha coefficient was 0.91.

In terms of construct and concurrent validity, the correlation between BIC and SF-8 was rational as a whole, and was strongest for "Time-dependent burden" and Social Functioning (r = -0.59), "Emotional burden" and "Existential burden" and Mental Health (r = -0.58, -0.55, respectively), and "Physical burden" and Bodily Pain (r = -0.64) (Table 3). Correlation was weak between "Service-related burden" and the SF-8. The correlation coefficient between the BIC total score and each SF-8 item ranged from -0.47 to -0.65. In contrast, that between each domain in the BIC and CES-D ranged from0.32 to 0.54; that of the BIC total score and CES-D was 0.62; that between each domain in the BIC and ZBI ranged from 0.38 to 0.76; and that for BIC total score and ZBI was high, at 0.84.

Known groups validity was examined using the correlation between BIC total score and actual hours spent on daily caregiving (Fig. 1). The correlation between actual caregiving hours and a feeling of care burden has been previously raised in other studies [2, 22]. On average, the BIC total score was 12.3 for less than 3 hours, 18.1 for 3 to less than 6 hours, 20.1 for 6 to less than 12 hours, and 22.6 for 12 hours or more. Spearman's ρ was 0.47, and indicated a statistically significant correlation (p < 0.001).

ICC, an indicator of reliability, was 0.77 for "Time-dependent burden", 0.73 for "Emotional burden", 0.70 for "Existential burden", 0.80 for "Physical burden", 0.68 for "Service-related burden", and 0.70 for "Total care burden" (Table 4). The ICC was 0.83 for the BIC total score, indicating sufficient reliability.