Journal of Clinical Epidemiology 63 (2010) 535e542
ORIGINAL ARTICLE
Short-form Zarit Caregiver Burden Interviews were valid
in advanced conditions
Irene J. Higginsona,*, Wei Gaoa, Diana Jacksona, Joanna Murrayb, Richard Hardinga
a
King’s College London, Department of Palliative Care, Policy and Rehabilitation, School of Medicine at Guy’s, King’s College and St Thomas’ Hospitals,
London, United Kingdom
b
King’s College London, Health Service and Population Research Department, Institute of Psychiatry, David Goldberg Centre, London, United Kingdom
Accepted 3 June 2009
Abstract
Objectives: To assess six short-form versions of Zarit Burden Interview (ZBI-12, ZBI-8, ZBI-7, ZBI-6, ZBI-4, and ZBI-1) among three
caregiving populations.
Study Design and Setting: Secondary analysis of carers’ surveys in advanced cancer (n 5 105), dementia (n 5 131), and acquired
brain injury (n 5 215). All completed demographic information and the ZBI-22 were used. Validity was assessed by Spearman correlations
and internal consistency using Cronbach’s alpha. Overall discrimination ability was evaluated using the area under the receiver operating
characteristic curve (AUC).
Results: All short-form versions, except the ZBI-1 in advanced cancer (rho 5 0.63), displayed good correlations (rho 5 0.74e0.97)
with the ZBI-22. Cronbach’s alphas suggested high internal consistency (range: 0.69e0.89) even for the ZBI-4. Discriminative ability
was good for all short forms (AUC range: 0.90e0.99); the best AUC was for ZBI-12 (0.99; 95% confidence interval [CI]: 0.98e0.99)
and the second best for ZBI-7 (0.98; 95% CI: 0.96e0.98) and ZBI-6 (0.98; 95% CI: 0.97e0.99).
Conclusions: All six short-form ZBI have very good validity, internal consistency, and discriminative ability. ZBI-12 is endorsed as the
best short-form version; ZBI-7 and ZBI-6 show almost equal properties and are suitable when a fewer-question version is needed. ZBI-4 and
ZBI-1 are suitable for screening, but ZBI-1 may be less valid in cancer. Ó 2010 Elsevier Inc. All rights reserved.
Keywords: Carer; Outcome; Palliative; Burden; Aging; Validity
1. Background
Informal carers are the primary resource for patient care
and are known to have high needs for support and psychological morbidity [1e4]. Although there are many suggested interventions seeking to improve their overall
well-being, there is little evaluative research into the
efficacy of such interventions [5,6]. Measurement of appropriate carer outcomes is essential for such studies. Although
there are many measures to assess caregiver burden, strain,
well-being, or other outcomes in specific disease, such as
stroke or mental illness [7,8], there are fewer measures
targeted for the carers of patients with advanced disease.
Mularski et al. in a major systematic review of measures
Competing interests: None.
* Corresponding author. Department of Palliative Care, Policy and Rehabilitation, School of Medicine at Guy’s, King’s College and St Thomas’
Hospitals, King’s College London, Weston Education Centre, 3rd Floor,
Cutcombe Road, Denmark Hill, London SE5 9RJ, United Kingdom.
Tel.: þ44-0-20-7848-5516; fax: þ44-0-20-7848-5517.
E-mail address: irene.higginson@kcl.ac.uk (I.J. Higginson).
0895-4356/10/$ e see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi: 10.1016/j.jclinepi.2009.06.014
for use toward the end of life for the National Institute of
Health (USA) highlighted ‘‘significant gaps’’ in measuring
caregiver outcomes, identifying only two measures in their
literature search of 24,423 citations [9].
Caregiver burden is closely aligned to the goals of many
interventions and is associated with negative health outcomes in carers of people with common conditions, such
as dementia, stroke, and cancer [8,10,11]. Moreover, perceived burden had been shown to predict anxiety and
depression in carers of patients with these conditions
[12e14]. Caregiver burden had been defined as a contextspecific negative affective outcome, occurring as a result
of perceived inability to contend with role demands [15].
There is general agreement that caregiver burden is a multidimensional concept affected by objective elements related
to the nature and time of the practical tasks undertaken by
carers and subjective elements arising from the perceived
emotional, social, and relationship stresses that can accompany this role [8,16]. Therefore, it would seem appropriate
to measure caregiver burden as an outcome in advanced
disease.
536
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
2. Data and methods
What is new?
Key findings
The 12-item Zarit Burden Interview (ZBI-12) was
suitable in all situations; the ZBI-7 and the ZBI-6
were almost equally good and may be suitable for
palliative care settings; the ZBI-4 and ZBI-1 were
useful when a very short screening instrument was
needed.
What was known?
The ZBI-22 is a widely used outcome measure of
caregiver burden and has been validated in diverse
caregiving samples. Several short-form versions of
ZBI have been developed, but little is known about
how well they perform in diverse populations.
What does this study add?
For the first time, six short-form versions of ZBI have
been comprehensively and systematically evaluated
in diverse populations. Evidence-based recommendations have been provided for choosing the best shortform version in various settings. The ZBI-6, a further
improvement to the ZBI-7, has been developed for
palliative care settings.
What is the implication and what should change
now?
Stop using the ZBI-8; use ZBI-6 in palliative care;
consider using the ZBI-1 when rapid screening is
needed.
Although there are other measures, the 22-item Zarit
Burden Interview (ZBI) is the most widely used tool for
measuring the level of subjective burden among carers
[17,18]. Several shorter versions of the ZBI have been developed, including Bedard et al.’s 12-, 8-, 7-, and 4-item
screening versions [19e21]. However, the factorial structures of ZBI were established among the carers of patients
with dementia whose concerns may be different from carers
of patients with cancer or sudden onset illness. Analysis of
four abridged versions of the ZBI in 503 carers of people
with dementia suggested that the 12-item version was optimal [15]. However, in some clinical settings, such as intensive care units, palliative care, and care of older people, the
12-item ZBI can still be a heavy assessment burden. There
is a need to test shorter forms of the ZBI with the carers of
people with advanced or progressive illness. Therefore, we
designed this study to investigate the validity and internal
consistency of six shorter forms of ZBI (ZBI-12, ZBI-8,
ZBI-7, ZBI-6, ZBI-4, and ZBI-1) among informal carers
of patients with three different conditions compared with
the 22-item version as the gold standard.
2.1. Design and data sources
This is a secondary analysis using data pooled from four
studies.
1. Baseline data from a multicenter evaluation of palliative day care for cancer patients involving six centers
across the south of England [22,23];
2. Baseline data from a two-center evaluation of the ‘‘90
Minute Group,’’ a supportive intervention for the
carers of cancer palliative care patients [6];
3. A national postal questionnaire survey of caregiver
experiences of acquired brain injury (ABI) [24]; and
4. Baseline data from a prospective longitudinal cohort
study of caregiver burden in dementia involving
participants from South East London [25].
All studies collected data from informal carers using the
self-reported 22-item ZBI (ZBI-22), with interviewers present in the cancer and dementia studies to provide support to
respondents during data collection if needed. In addition to
ZBI, the data set contains basic demographic data, including age, sex, and relationship, and clinical data regarding
the patients.
2.2. Short-form versions of the Zarit Burden Interview
Several short-form versions have been developed. The
three most common short-form versions of ZBI are the
12-item version (ZBI-12) by Bedard et al. [19], the eightitem version (ZBI-8) by Arai et al. [21], and the four-item
version (ZBI-4) by Bedard et al. [19]. The ZBI-12 and the
ZBI-4 were reported in the same study [19]. The ZBI-22
data were factor analyzed using a principal component
analysis and revealed a two-factor structure. The items
for the ZBI-12 were selected through a combination of high
factor loading and high itemetotal correlations across all
six situations, and the ZBI-4 screening items were selected
based on the itemetotal correlations while keeping the
three-to-one item ratio between factors 1 and 2 [19]. The
ZBI-8 items were chosen in terms of their factor loadings
(>0.65) on a two-factor structure [21].
A new seven-item version (ZBI-7) proposed specifically
for palliative care was included for evaluation; the items
extracted were decided by an expert committee [20]. However, ZBI-7 included item 22 of the full scale, a global
question to assess overall subjective burden. This is
unusual, because short-form versions do not usually contain
the global question, especially for the measures of subjective burden [8,16,26]. Therefore, we derived a six-item version of ZBI (ZBI-6), excluding this global question. We
also tested item 22 on its own (ZBI-1) to understand
whether the single global question (‘‘Overall how burdened
do you feel’’) would be useful as a screening tool. Items included in each of the short-item versions are listed in Table
1. The responses to every item are in 5-point Likert scale
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
537
Table 1
Items included in the short-form ZBI
Item
Q1
Q2
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Q10
Q11
Q12
Q13
Q14
Q15
Q16
Q17
Q18
Q19
Q20
Q21
Q22
Question (do
you feel/wish.)
Your relative asks for more
help than he/she needs?
You don’t have enough time
for yourself?
Stressed between caring and
meeting other
responsibilities?
Embarrassed over behaviors?
Angry when around your
relative?
Your relative affects your
relationship with others in
a negative way?
Afraid of what the future
holds for relative?
Your relative is dependent on
you?
Strained when are around
your relative?
Your health has suffered
because of your
involvement with your
relative?
You don’t have as much
privacy as you would like,
because of your relative?
Your social life has suffered
because you are caring for
your relative?
Uncomfortable about having
friends over because of
your relative?
Your relative seems to expect
you to take care of him/her,
as if you were the only one
he/she could depend on?
You don’t have enough
money to care for your
relative, in addition to the
rest of your expenses?
You will be unable to take
care of your relative much
longer?
You have lost control of your
life since your relative’s
illness?
You could just leave the care
of your relative to someone
else?
Uncertain about what to do
about relative?
You should be doing more for
your relative?
You could do a better job in
caring for your relative?
Overall, how burdened do
you feel in caring for your
relative?
Gort et al.’s
ZBI-7 [20]
Higginson
et al.’s ZBI-6
(this paper)
Bedard et al.’s
ZBI-4 [19]
Y
Y
Y
Y
Y
Y
Y
Y
Bedard et al.’s
ZBI-12 [19]
Arai et al.’s
ZBI-8 [21]
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Higginson
et al.’s ZBI-1
(this paper)
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Possible responses to Q1eQ21: 0, never; 1, rarely; 2, sometimes; 3, quite frequently; 4, nearly always.
Possible responses to Q22: 0, not at all; 1, a little; 2, moderately; 3, quite a bit; 4, extremely.
Abbreviations: ZBI, Zarit Burden Interview; Y, yes.
Y
538
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
from 0 (never) to 4 (nearly always). The overall burden is
assessed by the total score of all items, with a higher score
representing a greater caregiver burden.
2.3. Data analysis
Demographic characteristics were described, and differences between diagnostic groups were compared using oneway ANOVA (for age) and chi-square test (for sex and
relationship to patient). Total scores of ZBI-22, ZBI-12,
ZBI-8, ZBI-7, ZBI-6, ZBI-4, and ZBI-1 were summarized
using descriptive statistics. The score differences between
three diagnoses were examined for overall differences using KruskaleWallis test followed by Wilcoxon two-sample
test if the overall difference was significant. The Bonferroni
procedure was used to adjust raw P values (multiplying 3)
to control the type 1 error in multiple testing [27].
We planned to examine the subscales of ZBI, but found inconsistencies in the literature regarding the content of the
subscales. For example, the original article by Whitlatch
et al. [28] presented two subscales: role and personal strain
with six and 12 items, respectively. Hebert et al. [29] used
these in their factor analysis producing a 12-item version.
However, Knight et al. [30], Bedard et al. [19], and O’Rourke
et al. [31], when conducting exploratory and confirmatory
factor analysis, revealed that different items were included
in the role and personal subscales, even though they gave
them the same names. Because of the inconsistencies, we
did not proceed with the subscale analysis.
Validity was assessed by testing for correlations of the
ZBI-12, ZBI-8, ZBI-7, ZBI-6, ZBI-4, and ZBI-1 with the
ZBI-22 (as the gold standard) using Spearman rank order correlation. Terwee et al. [32] suggest that correlations of 0.7 or
more are required. Internal consistency was examined with
Cronbach’s alpha, and a value in the range of 0.7e0.9 is good
[32]. An alpha value of greater than 0.9 suggests redundant
items. For small scales (e.g., four items), values of 0.6 are
good, because alpha tends to underestimate internal consistency when the number of items is small [32,33].
Using a total burden score of 21 on the ZBI-22 as the cutoff
point for high burden [18], the discriminatory performance of
various short-form versions of ZBI was assessed and compared with the receiver operating characteristic (ROC) curve
[34], which was constructed by plotting sensitivity against
1 specificity. Each point in the ROC plot represents a sensitivity/1 specificity pair corresponding to a particular cutoff
value. A test with perfect discrimination has an ROC plot that
passes through the upper-left corner (100% sensitivity and
100% specificity). Therefore, the closer a ROC plot is to
the upper-left corner, the higher the overall accuracy of the
test. The point closest to (0, 1) on the curve was used to determine the most optimal combination of sensitivity and
specificity [34]. The areas under the curves (AUC) were calculated using the trapezoidal method [34,35]. The AUC represents the overall discriminative ability of a test, that is, the
ability to correctly classify those with and without burden.
The range of the AUC is 0.5e1.0. A discriminative test is
considered perfect if AUC 5 1.0, good if AUC 5 0.8e1.0,
moderate if AUC 5 0.6e0.8, and poor if AUC 5 0.5e0.6;
an area of 0.5 reflects a random rating model [35]. 95% Confidence intervals of AUCs were computed. A P value below
0.05 was considered as statistically significant. All analyses
were carried out using SAS 9.1 package (SAS Institute
Inc., Cary, NC, USA).
3. Results
One hundred and five, 131, and 215 informal carers for
patients with advanced cancer, dementia, and ABI respectively, were recruited, making a total sample of 451. Most
carers were womend81% for ABI, 72% for cancer, and
72% for dementia (c2df52 5 5.50, P 5 0.06). The carers of
ABI patients were the youngestdmean age (standard deviation) of 54 (11) compared with 66 (12) for cancer and
62 (13) for dementia (F(2,448) 5 42.8, P ! 0.0001).
Spouse/partner carers were the most commond59% for
ABI, 82% for cancer, 37% for dementiadfollowed by parent (37% for ABI, 4% for cancer, and 0% for dementia) or
son/daughter (44% for dementia, 11% for cancer, and 0%
for ABI) (c26 5 227.6, P ! 0.0001).
Table 2 shows the descriptive statistics of the total score
for various short-form versions of ZBI in each diagnostic
group. All scores showed highly significant overall difference across three groups (all P values ! 0.001). In all
groups, there was a wide range of burden scores. The disease-specific burden pattern reflected in the full-scale ZBI
was well captured by all short-form versions. Subjective
caregiver burden were lowest for the carers of cancer patients and highest for those of patients with ABI. However,
the significance of the difference shown in ZBI-22 (ABI vs.
cancer: z 5 7.66, Padj ! 0.0001; ABI vs. dementia:
z 5 5.62, Padj ! 0.0001; cancer vs. dementia: z 5 2.03,
Padj 5 0.12) was only satisfactorily revealed by ZBI-12,
ZBI-6, and ZBI-4. ZBI-1 was among the worst in all the
short-form versions for misjudging two pairwise comparisons of ABI vs. dementia (z 5 2.15, Padj 5 0.10) and cancer
vs. dementia (z 5 4.30, P ! 0.0001), whereas ZBI-7
(z 5 2.76, Padj 5 0.018) and ZBI-8 (z 5 4.24, P ! 0.0001)
did not reflect the difference in comparing cancer with
dementia, as was the case in ZBI-22.
High correlation coefficients were found between the
full version and short forms (Table 3), with correlations
well above our criteria (O0.7) for the full scale (range:
0.88e0.97) using all short-form versions except for the
ZBI-1. Even with the one-item version (ZBI-1), satisfactory
correlation with ZBI-22 was obtained in dementia
(rho 5 0.74) and ABI (rho 5 0.78) groups. The full ZBI
showed a high Cronbach’s alpha in all three diagnostic
groups, ranging from 0.88 to 0.93. The alpha values met
our internal consistency criteria of good for all the shortform versions, ranging from 0.69 (for a four-item scale in
cancer) to 0.90. Correlations and internal consistencies
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
539
Table 2
Descriptive statistics of scores for six short-form and full-scale versions of the ZBI
Diagnosis
a
Version
Statistics
Cancer
Dementia
ABI
Multiple comparisonsb
ZBI-12
Mean (SD)**
95% CI
Median (min, max)
12.0 (8.5)
10.4e13.7
10 (0, 36)
15.1 (10.0)
13.4e16.8
15 (0, 43)
21.7 (10.1)
20.4e23.1
22 (0, 46)
B**, C**
ZBI-8
Mean (SD)**
95% CI
Median (min, max)
5.5 (4.7)
4.6e6.4
4 (0, 24)
8.8 (6.1)
7.7e9.8
9 (0, 28)
11.5 (7.2)
10.5e12.4
11 (0, 31)
A**, B**, C**
ZBI-7
Mean (SD)**
95% CI
Median (min, max)
7.4 (5.6)
6.3e8.5
7 (0, 21)
9.9 (6.8)
8.7e11.1
10 (0, 27)
14.3 (7.0)
13.4e15.3
14 (0, 28)
A*, B**, C**
ZBI-6
Mean (SD)**
95% CI
Median (min, max)
6.4 (4.9)
5.4e7.3
6 (0, 19)
8.2 (5.8)
7.2e9.2
8 (0, 23)
12.3 (6.0)
11.5e13.1
13 (0, 24)
B**, C**
ZBI-4
Mean (SD)**
95% CI
Median (min, max)
4.8 (3.5)
4.2e5.5
5 (0, 12)
6.1 (4.1)
5.4e6.8
6 (0, 16)
7.9 (3.8)
7.3e8.4
8 (0, 16)
B**, C**
ZBI-1
Mean (SD)**
95% CI
Median (min, max)
1.0 (1.2)
0.8e1.3
1 (0, 4)
1.7 (1.3)
1.5e2.0
2 (0, 4)
2.1 (1.3)
1.9e2.2
2 (0, 4)
A**, C**
ZBI-22
Mean (SD)**
95% CI
Median (min, max)
23.2 (13.4)
20.7e25.8
22 (0, 66)
27.9 (16.4)
25.0e30.7
26 (0, 73)
39.1 (17.3)
36.8e41.4
39 (5, 80)
B**, C**
*P ! 0.05, **P ! 0.01.
Abbreviations: ABI, acquired brain injury; SD, standard deviation; CI, confidence interval; min, minimum; max, maximum;
a
The mean difference between three groups was tested using KruskaleWallis nonparametric ANOVA.
b
Multiple comparisons were made by Wilcoxon two-sample test, and type 1 error was controlled using Bonferroni method. A: cancer vs. dementia;
B: cancer vs. ABI; C: dementia vs. ABI; only significant pairwise differences were presented.
were similar to those in Table 3 separately for men and
women, for those older and younger than 70 years, and
for higher and lower burdened carers.
The most optimal combination of sensitivity and specificity, as visualized from ROC curves (Fig. 1), was 92% and
94% for ZBI-12 (cutoff score: 12), 82% and 92% for ZBI-8
(cutoff score: 6), 95% and 86% for ZBI-7 (cutoff score: 7),
91% and 91% for ZBI-6 (cutoff score: 6), 88% and 85% for
ZBI-4 (cutoff score: 4), and 91% and 53% for ZBI-1 (cutoff
score: 1). All shorter versions were overall successful in differentiating low- and high-burden individuals with all AUCs
well above 0.90. The short-form version with the best discriminative ability was ZBI-12 and that with the lowest
was ZBI-1. ZBI-6 performed slightly better than ZBI-8 and
to the same level as the ZBI-7.
4. Discussion
We tested and validated six short forms of the ZBI in
three caregiving populations. In all groups, there was a wide
range of scores for the ZBI-22; therefore, the short forms
were tested in samples reporting varying caregiver burden.
However, the highest burden scores were in the dementia
and the ABI groups, and we were not able to test burden
scores above 34 in the advanced cancer group. It may be
that caregiver burden using ZBI was lower in advanced cancer compared with dementia and ABI because of the lower
levels of cognitive disturbance or greater specialist palliative support for the cancer carers (the cancer carers were
sampled from palliative care services) [4,22,24,25]. However, it may equally be because of ZBI failing to measure
some aspects of caregiver burden in advanced cancer. Caregiver burden is a complex construct. It has been described
as having physical, social, financial, and emotional components, as well as leading to relationship and personal strain
[16]. Some measures seek to capture ‘‘burden’’ and others
‘‘strain’’ [7e9]. More recently measures to capture carer
positivity and satisfaction have been developed [36]. Which
components of burden are present and whether these are
different across different conditions are important questions
and need to be the subject of future research.
Three key findings emerge from our analysis. First, we
found high levels of validity (with correlations ranging
from 0.74 to 0.97) for all the short forms compared with
ZBI-22 in three diagnostic groups with the only exception
of the one-item version in cancer. The ZBI-12 has the highest validity (rho 5 0.95e0.97), and this is consistent across
advanced cancer, dementia, and ABI samples. In our populations, the performance of the ZBI-8 and the ZBI-4 are almost identical and slightly worse than that of ZBI-6 in both
540
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
Table 3
Spearman correlation coefficients (95% CI) between full and short-form versions of ZBI and Cronbach’s alpha of all versions of ZBI
Cancer
Dementia
ABI
Version
Rho (95% CI)
Alpha
Rho (95% CI)
Alpha
Rho (95% CI)
Alpha
ZBI-12
ZBI-8
ZBI-7
ZBI-6
ZBI-4
ZBI-1
ZBI-22
0.95 (0.92e0.96)
0.86 (0.80e0.90)
0.90 (0.86e0.93)
0.89 (0.84e0.93)
0.88 (0.82e0.91)
0.63 (0.50e0.73)
d
0.85
0.74
0.82
0.78
0.69
NA
0.88
0.96 (0.94e0.97)
0.90 (0.87e0.93)
0.94 (0.92e0.96)
0.94 (0.91e0.95)
0.89 (0.85e0.92)
0.74 (0.66e0.81)
d
0.87
0.80
0.86
0.83
0.78
NA
0.91
0.97(0.97e0.98)
0.93 (0.90e0.94)
0.95 (0.94e0.96)
0.95 (0.93e0.96)
0.92 (0.89e0.93)
0.78 (0.73e0.83)
d
0.89
0.89
0.90
0.88
0.79
NA
0.93
Abbreviation: NA, not applicable.
correlation and ROC analysis. Although the correlations
between short and long forms appeared to be slightly and
consistently stronger for the ABI group, and lowest in the
advanced cancer group, the differences were marginal,
not significant, and could be an artifact of the narrow range
of scores for the cancer group.
Second, we found high internal consistency of all versions of the ZBI, suggesting that some items are redundant
and short forms can be used. Only the ZBI-4 had lower, but
still good internal consistency (Cronbach’s alpha:
0.69e0.79) in the cancer group. Third, as we planned our
analysis, we found confusions in the categorization of the
role and personal strain subscales. Given the variability regarding the subscales and the high internal consistency of
the full Zarit scale and all the short forms, we doubt that
using the ‘‘personal’’ and ‘‘role’’ strain subscales has either
face or psychometric validity [30].
The choice of ZBI version should be based on the specific aims of the research. For most situations, 12-item ZBI
should have comparable performance with the full version
with the differentiating capacity close to 1. For situations
requiring rapid identification of caregiver burden, for example, screening for assessment or referral, four-item and
even one-item versions will be the ideal choice, given their
simplicity and optimal combination of high sensitivity
(O80%) and high specificity (O50%) as evident by the
ROC curves. Given that burden is multidimensional, the
success of the single- and four-item versions surprised us.
Although the ZBI-8 has more items, it did not exhibit
any superiority over the ZBI-7 or the ZBI-6 in psychometric characteristics and differential ability. An earlier small
study reported a perfect performance of the ZBI-7 with
100% sensitivity and 100% specificity in the palliative care
setting [20]. The validation of the ZBI-7 in our larger study
was close to thisd90% specificity at the sensitivity level of
90% (Fig. 1). However, we also found that, compared with
the ZBI-7, excluding the ‘‘global burden question’’ did not
compromise the ability of the ZBI-6 to distinguish carers
with burden from those without, and results were almost
equal to ZBI-12 (Fig. 1). Therefore, when investigators
need to keep the number of questions short, the ZBI-6 appears to be a good choice. Short forms to assess caregiver
1.0
0.9
0.8
Sensitivity
0.7
0.6
0.5
0.4
0.3
AUC (95%CI)
0.2
ZBI- 1: 0.90(0.87-0.93)
ZBI- 6: 0.98(0.97-0.99)
ZBI- 8: 0.97(0.96-0.98)
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
ZBI- 4: 0.94(0.92-0.96)
ZBI- 7: 0.98(0.96-0.98)
ZBI-12: 0.99(0.98-0.99)
0.6
0.7
0.8
0.9
1.0
1-Spectificity
Fig. 1. Receiver operating characteristic curves for various short-form versions of the Zarit Burden Interview (ZBI) and areas under the curve (AUC, 95%
confidence interval [CI]). A total score of 21 on the full-scale ZBI as the cutoff value between low and high burden [18].
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
burden may be important to ensure feasible collection of
datadburdened carers often focus on the needs of the patients whom they care for and not their own experiences
and stresses related to caring [37] and may not wish
to spend time completing anything but the briefest
questionnaire.
When using short-form versions of the ZBI as a screening test, sensitivity and specificity are standard measures
for the diagnostic performance compared with the gold
standarddZBI-22 [38]. However, these two measures are
inversely related. Increasing one measure (by changing
the cutoff value) results in a decrease in the other. Which
one is more important is a question that can only be
answered in the context in which it is used. Often a balance
is needed. For example, sensitivity is important when identifying highly burdened carers (as measured by the summary score of the ZBI), because they can be offered
more support that is unlikely to do harm [39]. However,
when resources are scarce or if carers felt they were ‘‘labeled’’ as not coping by false-positive screening results,
specificity is more important than sensitivity. It should be
noted however that, in this study, we were primarily testing
the performance of short forms of the ZBI as a screening
tool. In clinical practice, a wider exploration of the components contributing to burden may be needed. Ideally, qualitative or cognitive interviewing would be needed to
establish if all relevant aspects for caregiver burden are
included.
We recognize several limitations for this study. First, the
performance comparisons were evaluated with cross-sectional data; therefore, they provide no information on short
forms’ responsiveness to change (an essential psychometric
property in intervention and longitudinal studies) [40].
Three of the original studies collected data at several time
points; we are planning to use these data to assess the adequacy of short-form ZBI to detect change. Second, our
analyses were restricted to the limited number of common
demographic variables in the pooled data set; therefore, we
could not make detailed performance comparisons across
subsamples. Third, our validation was based on a comparison between short forms of ZBI with the full 22-item version, and thus makes an assumption that ZBI-22 accurately
captures caregiver burden. Ideally, we would have assessed
the short-form versions with other measures of burden or
against clinical findings, but this would have required more
intensive data collection among burdened carers, which
may not have been feasible. Our data only allow conclusions to be drawn about the short-form versions of ZBI
compared with the full version.
5. Conclusions
We found strong validity and internal consistency for
each of the short-form versions in all three samples. The
541
ZBI-12 is suitable in all situations, whereas the ZBI-7 or
the ZBI-6 is suitable when a fewer-question version is
needed, for example, in palliative care setting. The ZBI-7
is equivalent to the ZBI-6 although with one more question.
The ZBI-4 and ZBI-1 may be useful when a very short
screening instrument is needed, but the ZBI-1 may be less
valid in cancer.
Acknowledgments
We thank the patients, carers, staff, and volunteers who
participated in the original studies, including (1) six day
and home hospice and palliative care services which recruited and interviewed patients and carers and Danielle
Goodwin and other interviewers in the study; (2) two home
palliative care services in London, Celia Leam and Liz Taylor who worked with us to recruit patients, and Alison
Pearce (research assistant); (3) Research assistants Shehla
Kazim, Amanda Tadrous, and Joel Sheridon and representatives of Headway, the Encephalitis Society and the Meningitis Trust, who helped to disseminate information about
the ABI study to carer participants; (4) Community Mental
Health Teams for Older Adults in the South London and
Maudsley Mental Health Trust and the research workers
Beth Foley and Louise Atkins.
In addition, we thank the funders of the original studies:
the NHS Executive (London and South East) for funding
projects 1 and 2, the Department of Health (R&D grant
030/0066) for project 3, and the Department of Health, Policy Research Programme for project 4. Dr. Gao Wei is 50%
supported by the National Cancer Research Institute, UK,
a part of the ‘‘COMPlex interventions: Assessment, trialS
and implementation of Services in Supportive and Palliative Care (COMPASS)’’ collaborative. We also thank Professors Peter Fayers, Gordon Murray, and Julia Brown for
their helpful comments to improve this manuscript.
References
[1] Charlton R. Palliative care in non-cancer patients and the neglected
caregiver. J Clin Epidemiol 1992;45:1447e9.
[2] Hung SY, Pickard AS, Witt WP, Lambert BL. Pain and depression in
caregivers affected their perception of pain in stroke patients. J Clin
Epidemiol 2007;60:963e70.
[3] Maher J, Green H. Carers 2000. Office for National Statistics. London: The Stationary Office; 2002.
[4] The Resource Implications Study Group of the MRC Study of Cognitive Function and Aging (RIS MRC CFAS). Psychological morbidity among informal caregivers of older people: a 2-year follow-up
study. The Resource Implications Study Group of the MRC study
of cognitive function and ageing (RIS MRC CFAS). Psychol Med
2000;30:943e55.
[5] Harding R, Higginson IJ. What is the best way to help caregivers in
cancer and palliative care? A systematic literature review of interventions and their effectiveness. Palliat Med 2002;17:63e71.
[6] Harding R, Higginson IJ, Leam C, Donaldson N, Pearce A, George R,
et al. Evaluation of a short-term group intervention for informal
542
I.J. Higginson et al. / Journal of Clinical Epidemiology 63 (2010) 535e542
carers of patients attending a home palliative care service. J Pain
Symptom Manage 2004;27:396e408.
[7] Harvey K, Catty J, Langman A, Winfield H, Clement S, Burns E,
et al. A review of instruments developed to measure outcomes for
carers of people with mental health problems. Acta Psychiatr Scand
2008;117:164e76.
[8] Visser-Meily JM, Post MW, Riphagen II, Lindeman E. Measures
used to assess burden among caregivers of stroke patients: a review.
Clin Rehabil 2004;18:601e23.
[9] Mularski RA, Dy SM, Shugarman LR, Wilkinson AM, Lynn J,
Shekelle PG, et al. A systematic review of measures of end-of-life
care and its outcomes 3. Health Serv Res 2007;42:1848e70.
[10] Goldstein NE, Concato J, Fried TR, Kasl SV, Johnson-Hurzeler R,
Bradley EH. Factors associated with caregiver burden among caregivers
of terminally ill patients with cancer. J Palliat Care 2004;20:38e43.
[11] Schneider J, Hallam A, Murray J, Foley B, Atkin L, Banerjee S, et al.
Formal and informal care for people with dementia: factors associated with service receipt. Aging Ment Health 2002;6:255e65.
[12] Alvarez-Ude F, Valdes C, Estebanez C, Rebollo P. Health-related
quality of life of family caregivers of dialysis patients. J Nephrol
2004;17:841e50.
[13] Zarit SH, Reever KE, Bach-Peterson J. Relatives of the impaired elderly: correlates of feelings of burden. Gerontologist 1980;20:
649e55.
[14] Gort AM, Mingot M, Gomez X, Soler T, Torres G, Sacristan O, et al.
Use of the Zarit scale for assessing caregiver burden and collapse in
caregiving at home in dementias. Int J Geriatr Psychiatry 2007;22:
957e62.
[15] O’Rourke N, Tuokko HA. The relative utility of four abridged versions of the Zarit Burden Interview. J Ment Health Aging 2003;9:
55e64.
[16] Chenier MC. Review and analysis of caregiver burden and nursing
home placement. Geriatr Nurs 1997;18:121e6.
[17] Bachner YG, O’Rourke N. Reliability generalization of responses by
care providers to the Zarit Burden Interview. Aging Ment Health
2007;11:678e85.
[18] Zarit SH, Orr RD, Zarit JM. The hidden victims of Alzheimer’s disease:
families under stress. New York: New York University Press; 1985.
[19] Bedard M, Molloy DW, Squire L, Dubois S, Lever JA, O’Donnell M.
The Zarit Burden Interview: a new short version and screening version. Gerontologist 2001;41:652e7.
[20] Gort AM, March J, Gomez X, de MM, Mazarico S, Balleste J. [Short
Zarit scale in palliative care]. Med Clin (Barc) 2005;124:651e3.
[21] Arai Y, Tamiya N, Yano E. [The short version of the Japanese version
of the Zarit Caregiver Burden Interview (J-ZBI_8): its reliability and
validity]. Nippon Ronen Igakkai Zasshi 2003;40:497e503.
[22] Goodwin DM, Higginson IJ, Myers K, Douglas H-R, Normand C. Effectiveness of palliative day care in improving pain, symptom control
and quality of life. J Pain Symptom Manage 2003;25:202e12.
[23] Higginson IJ, Donaldson N. Relationship between three palliative
care outcome scales. Health Qual Life Outcomes 2004;2:68e75.
[24] Jackson D, Turner-Stokes L, Murray J, Leese M, McPherson KM.
Acquired brain injury and dementia: a comparison of carer experiences. Brain Inj 2009;23:1e12.
[25] Banerjee S, Murray J, Foley B, Atkins L, Schneider J, Mann A.
Predictors of institutionalisation in people with dementia. J Neurol
Neurosurg Psychiatry 2003;74:1315e6.
[26] Knowles ES, Condon CA. Why people say ‘‘Yes’’: a dual-process
theory of acquiescence. J Pers Soc Psychol 1999;77:379e86.
[27] Shaffer JP. Multiple hypothesis-testing. Annu Rev Psychol 1995;46:
561e84.
[28] Whitlatch CJ, Zarit SH, von EA. Efficacy of interventions with caregivers: a reanalysis. Gerontologist 1991;31:9e14.
[29] Hebert R, Bravo G, Preville M. Reliability, validity and reference
values of the Zarit Burden Interview for assessing informal caregivers
of community-dwelling older persons with dementia. Can J Aging
2000;19:494e507.
[30] Knight BG, Fox LS, Chou CP. Factor structure of the burden interview. J Clin Geropsychol 2000;6:249e58.
[31] O’Rourke N, Tuokko HA. Psychometric properties of an abridged
version of The Zarit Burden Interview within a representative Canadian caregiver sample. Gerontologist 2003;4:121e7.
[32] Terwee CB, Bot SD, de Boer MR, van der Windt DA, Knol DL,
Dekker J, et al. Quality criteria were proposed for measurement properties of health status questionnaires. J Clin Epidemiol 2007;60:
34e42.
[33] Nunnally JC, Bernstein IH. Psychometric theory. New York:
McGraw-Hill; 1994.
[34] Coffin M, Sukhatme S. Receiver operating characteristic studies and
measurement errors. Biometrics 1997;53:823e37.
[35] Weinstein MC, Fineberg HV, Elstein AS, Frazier HS, Neuhauser D,
Neutra RR, et al. Clinical decision analysis. Philadelphia: WB Saunders; 1980.
[36] Lim JW, Zebrack B. Caring for family members with chronic physical illness: a critical review of caregiver literature. Health Qual Life
Outcomes 2004;2:50.
[37] Harding R, Higginson I. Working with ambivalence: informal caregivers of patients at the end of life. Support Care Cancer 2001;9:
642e5.
[38] Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP,
Irwig LM, et al. Towards complete and accurate reporting of studies
of diagnostic accuracy: the STARD initiative. Ann Intern Med
2003;138:40e4.
[39] Fletcher RH, Fletchher SW. Prevention. In: Fletcher RH,
Fletchher SW, editors. Clinical epidemiology: the essentials. Baltimore, MD: Lippincott Williams & Wilkins; 2005. p. 158e61.
[40] Kirshner B, Guyatt G. A methodological framework for assessing
health indices. J Chronic Dis 1985;38:27e36.