REGULAR ARTICLE

Expressed Emotion in Eating Disorders Assessed via

Self-Report: An Examination of Factors Associated with
Expressed Emotion in Carers of People with Anorexia
Nervosa in Comparison to Control Families
whereas 47% (71/151) of carers showed
Olivia Kyriacou, MA* ABSTRACT
high criticism (CC) compared with 15%
Objective: High expressed emotion (EE)
Janet Treasure, PhD, FRCP, (14/93) of controls. The most significant
has been linked to outcome in anorexia
FRCPsych nervosa (AN). Traditional interview meas-
variable related to CC was ‘‘negative/diffi-
Ulrike Schmidt, MRCPsych, MD, cult behaviors,’’ and anxiety, for EOI.
ures of EE require considerable resources.
PhD The aim was to compare EE, measured with Conclusion: The FQ was sensitive both
a self report tool, in parents of people with to CC and EOI. Interventions aimed at
AN (carers; N 5 151) with that of parents of reducing EE may need to focus on
healthy people (controls; N 5 93), and to decreasing carers’ anxiety and teaching
explore factors associated with EE. skills to moderate difficult behaviors.
V
C 2007 by Wiley Periodicals, Inc.
Method: A cross-sectional design was
used to assess EE (Family Questionnaire)
Keywords: expressed emotion; eating
and psychological distress (Hospital Anxi-
disorders; carers; families
ety and Depression Scale).

Results: Over 60% (96/151) of carers

showed high emotional over-involvement
(EOI) compared with 3% (3/93) of controls, (Int J Eat Disord 2008; 41:37–46)

there are the stresses of caring for someone who

Introduction
Caregiving in the area of mental illness has been
well documented across a diverse range of disor-
ders such as schizophrenia and Alzheimer’s disease
and has been associated with both physical and
psychological distress of carers.1 Eating disorders
(ED) and particularly anorexia nervosa (AN) are
thought to present unique demands on carers,
stemming from the severity and extended course of
the illness. These include the strain of a develop-
mental ‘‘inappropriateness’’ involved in caring for
increasingly dependent adult children.2 Moreover,
Accepted 15 July 2007
Part of this work has been presented as a poster in the
International Conference of Eating Disorders of the Academy for
Eating Disorders, in June 2006 in Barcelona, Spain.
Supported by 284286 from Nina Jackson fellowship by Research
Into Eating Disorders in association with the Psychiatry Research
Trust, UK.
*Correspondence to: Ms. Olivia Kyriacou, Section of Eating
Disorders, PO 59, Institute of Psychiatry, King’s College London,
De Crespigny Park, London SE5 8AF, UK.
E-mail: o.kyriacou@iop.kcl.ac.uk
Section of Eating Disorders, Institute of Psychiatry, King’s
College London, London, United Kingdom
Published online 5 October 2007 in Wiley InterScience
(www.interscience.wiley.com). DOI: 10.1002/eat.20469
V
C 2007 Wiley Periodicals, Inc.
may deny the seriousness of their illness and be
ambivalent or resistant to treatment.3,4 In addition,
the entire family unit is known to be disrupted as
parents may experience shame and self-blame for
the development of the illness while family mem-
bers may struggle to understand, feel stigmatised
and isolated as a result of social misperceptions of
ED, as well as experience anxiety, loss, anger, grief,
and helplessness at the relentless, controlling and
all-encompassing grip of the illness.5–11 Further-
more, carers face difficulties accessing services and
often feel excluded, misunderstood, or blamed by
professionals, and report unmet needs for informa-
tion on treatment, prognosis, and coping strategies,
as well as lack of social and professional sup-
port.6,7,12,13 In light of the severity of the illness and
the tremendous impact on carers and families, it is
becoming increasingly recognized that professio-
nals need to address carer needs and provide fami-
lies with the resources necessary to enable caregiv-
ing as well as help to create a familial environment,
which is conducive to adaptive changes in family
functioning.13,14
The concept of caregivers’ expressed emotion
(EE) has been of interest in psychiatry because
overprotection, criticism (CC), and hostility influ-

International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat 37

KYRIACOU ET AL.
ence the course of and predict relapse of many dis-
orders, including ED.15 In the Maudsley model of
maintenance of AN carers’ high EE is a key compo-
nent of the interpersonal domain that is thought to
impact on outcome.16
The mechanisms underpinning the development
of high EE are not fully understood. Parents can
become critical and/or over protective because of
how they understand aspects of the illness or how
they perceive the individual with the illness.17
For example, in schizophrenia, critical relatives
assumed that the patients had more volitional con-
trol and choice over their behaviors whereas over-
involved relatives assumed that the patient was a
passive, impotent victim of the illness.18 In this
case the parents took over control and responsibil-
ity from the patient. In a qualitative exploration of
the experience of caregiving in AN, carer misper-
ceptions were associated with CC and hostility
towards the person with AN, whereas fear of the
consequences of the illness was found to contrib-
ute to over-protectiveness and even collusion with
symptomatic behaviors, especially in mothers.19
The level of EE may be related to the duration of
the disorder. In adolescents with AN with an aver-
age illness duration of less than a year only 6% of
parents were rated as high EE using standard cut
off scores.20 The level of EE increased in the group
with a poor outcome and decreased in those with a
good outcome.
High levels of EE may result from mood disorders
in the parents. Many studies across different disor-
ders have found that carers with high EE have poor
mental health, greater caregiving strain, in addition
to endorsing unhelpful attributions about the ill-
ness.21–25 Parents of people with AN with high EE
blamed themselves for the illness.26 This guilt may
be a symptom or a cause of depression and anxiety.
Qualitative evidence19 suggests that mothers caring
for daughters with ED experience high distress and
particularly anxiety, which typically manifest as
self-blame and helplessness in response to the ill-
ness, and are accompanied by high emotional
over-involvement (EOI).19 Overall CC in ED fami-
lies has been found to be lower than in families
with schizophrenia27 and EE levels in ED families
have been described as low.28 It is possible that in
ED, and especially in AN families, EOI rather than
CC is a marked and more characteristic response as
parents, and especially mothers, are likely to feel
highly anxious because of the severity of the illness,
whereas in families of people with bulimia nervosa,
frustration and CC may be more common.27 There
is some empirical evidence to suggest higher
maternal EOI in AN, as compared with paternal
38 International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
EOI, but differences may not be statistically signifi-
cant.28–30 Based on this information, we would
therefore expect mothers in AN families to report
higher EOI and distress than fathers. Part of the
aims of the current study was to further explore
these hypothesized gender differences.
EE is traditionally assessed from an audiotaped
semistructured interview, the Camberwell Family
Interview (CFI).31 However, this method requires a
high level of resources. The interview takes 1–2 h to
conduct and must be administered by a trained
interviewer, while coding by an experienced rater
may require an additional 3 h. The average training
period for rater trainings is 2 weeks, hence curtail-
ing the applicability of the method to routine clini-
cal and research purposes. The Five Minute Speech
Sample is a shorter alternative but it also requires
the training of interviewers and raters.32 The FMSS
has been found to have correspondence rates of
80–90% with CFI ratings of CC and hostility, how-
ever, it has been found to underestimate high
EOI.33,34 In an ED sample the results from the
FMSS differed from those obtained with the CFI.35
These authors argued that the FMSS is not an
adequate alternative for evaluating EE in families of
individuals with ED.
Self-report measures of EE may be preferable to
interview methods, which may preclude or inhibit
the free expression of CC or concern by relatives of
patients. Two questionnaire measures of EE, the
Level of Expressed Emotion Scale (LEE)36 and the
Influential Relationships Questionnaire (IRQ),37
administered to patients with schizophrenia and
their carers have shown agreement and consistency
with the Camberwell Family Interviews (CFI).38
Relatives’ Warmth and Critical Comments scores
on the CFI correlated with the LEE Total and Atti-
tude Toward Illness scores (r 5 2.41, p \ .05, for
CFI Warmth and LEE Attitude Toward Illness;
r 5 .41, p \ .05, for CFI CC and LEE Attitude To-
ward Illness; r 5 .38, p \ .05, for CFI CC and LEE
Total score). The only significant correlation
between both Patient and Relative versions of the
IRQ scores and the CFI was for IRQ CC scale of the
Patient Version and CFI EOI (r 5 .48, p \ .05), with
a near-significant correlation between the IRQ
Overprotection scores and CFI EOI ratings (r 5 .39;
r 5 .41 required for significance).38 An adapted ver-
sion of the LEE has been used in people with ED. In
a sample of adolescents with bulimia nervosa,
patient ratings of EE predicted a negative experi-
ence of caregiving by the parents.39 The Family
Questionnaire (FQ)40 was also created as a more ef-
ficient and research-applicable alternative measure
of EE in schizophrenia. It measures CC and EOI.

The scale shows sensitivity and successfully pre-
dicts both high and low EE and is more cost-effec-
tive than the FMSS. The FQ has also shown better
agreement with the CFI on EOI than other short EE
questionnaires.40 The authors report a correspon-
dence of 74% (combined CC and EOI subscales of
the FQ) with CFI EE ratings in both their original
sample and their validation sample (Pearson’s cor-
relations of FQ CC to CFI CC: r 5 .44, p \ .001;
Pearson’s correlations of FQ EOI to CFI EOI: r 5 .42,
p \ .001).40 The measure has been successfully
used in predicting stress of carers of people hospi-
talized for schizophrenia or depression.41,42 This
instrument has not yet been used to measure EE in
the families of people with ED.
The aim of this study was to compare EE meas-
ured with a self report tool (FQ) in parents of peo-
ple with AN with that of parents of healthy young
people. Our hypothesis was that parents of people
with AN would show high levels of EOI and a mod-
erate increase in CC. An additional aim was to
examine factors associated with EE and specifically,
whether the level of EE was related to the experi-
ence of burden from caregiving and/or anxiety and
depression in the parents. Our hypothesis was that
carer-related and illness-related characteristics
would be associated with parental depression and
anxiety, which in turn would be related to over-
involvement or CC. An additional hypothesis was
that mothers would show higher emotional over
involvement than fathers in the AN carers group.
Method
Design
This was a cross-sectional study using self-report
questionnaires to examine EE in parents of people with
AN and a comparison group. As the study was cross-sec-
tional it precluded the examination of the exact mecha-
nisms and mediators leading to the development of
parental distress and of EE; such a ‘‘mechanism hypothe-
sis’’ would require a longitudinal design. Regression anal-
yses as used in the study are meant to assess level of sig-
nificance of association between the variables under ex-
ploration, and are not meant to imply causality or
chronological direction and antecedence. The parents
gave information about the eating disorder symptoms
and history, in addition to demographic information for
parents and patients. Ethical approval and registration of
the study protocol was obtained from the Institute of
Psychiatry Research Ethics Committee and Research and
Development Department (Ref. No. 317/03).
International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
DEVELOPMENT OF EE IN CARERS OF PEOPLE WITH AN
Participants
Carers. Parents on the Carers Volunteer Database main-
tained by the eating disorder unit (EDU) of the IOP and
Maudsley Hospital were offered the opportunity to par-
ticipate in this project. The study was also advertised on
the website of the EDU and in the EDU newsletter. At the
time of recruitment for this project, the Carers Volunteer
Database consisted of 197 carers.a Parents who con-
firmed interest and willingness to participate were then
sent the full set of questionnaires, along with an informa-
tion and consent sheet and a prepaid return envelope.
Participants were not offered any incentives for partici-
pation.
The inclusion criteria for parents were that they were
currently caring for a child with AN (siblings and partners
were excluded). One hundred and twenty-nine carers (a
response rate of 65%) from the database and 36 parents
from current EDU patients offered to participate giving a
total of 165 carers. Fourteen carers were excluded
because their responses were incomplete and/or the
patient was currently well. The final sample consisted of
151 carers (91 mothers and 60 fathers).
Controls. Control parents (ascertained from contacts of
the ED carers and via e-mail advertisements to King’s
College London staff and students) were recruited. The
inclusion criteria were parents of female offspring cur-
rently or until recently living at home, with no history of
an ED in the parents or daughters. The final sample con-
sisted of 93 parents (53 mothers and 40 fathers).
Assessment Measures
Family Questionnaire. Created as a more efficient self-
report alternative to the CFI, the FQ40 has shown good in-
ternal consistency of subscales (ranging from 0.78 to 0.80
for EOI and from 0.91 to 0.92 for CC. The measure con-
sists of 20 items, 10 for each subscale, asking carers to
rate the applicability of each item in describing their feel-
ings and thoughts of the patient, such as ‘‘I can’t sleep
because of him/her’’ (EOI), and, ‘‘He/she does not appre-
a
Carers in this database come from a variety of sources, such as
Beat [beat is a national charity (Registered Charity No. 801343)
based in the UK providing information, help, and support for
people affected by eating disorders and, in particular, anorexia
and bulimia nervosa and their carers, families, etc]. All of the 197
carers were eligible to participate, however, 68 did not respond to
the study invitation. In addition, 36 carers came from current EDU
inpatients. This sample was self-selected and we are unable to
accurately comment on how many of the parents of inpatients
were initially approached over the period of recruitment, how
many refused to participate or what their exact reasons were for
nonparticipation. This raises the issue of representativeness as the
self-selected sample of carers may be skewed in terms of over-
representation of actively involved and ‘‘empowered’’ carers.
Possible limitations as a result of the sample are addressed in the
‘‘Discussion’’ section of this article.
39
KYRIACOU ET AL.
ciate what I do for him/her’’ (CC). Responses range from
‘‘never/very rarely,’’ with a score of 1, to ‘‘very often,’’
with a score of 4, and higher total score indicates higher
EE. The authors give a cut-off point of 23 for CC as an
indication of high CC, and 27 for EOI.
Hospital Anxiety and Depression Scale. The Hospital
Anxiety and Depression Scale (HADS)43 is a reliable, self-
report measure of anxiety and depression. The subscales
have shown high internal consistency, with Cronbach’s a
of 0.80–0.93 for anxiety and 0.81–0.90 for depression. A
score of 11 or higher for each subscale is indicative of the
corresponding mood disorder, and a score of 8–10 is
suggestive.44
Experience of Caregiving Inventory. The Experience of
Caregiving Inventory (ECI)45 is a 66-item self-report
questionnaire, which assesses carers’ perceptions of their
experience of caregiving. Items ask carers to rate the fre-
quency of experiences they have had over the past
month, such as, ‘‘How often in the past month have you
thought about whether she/he will ever get well?’’ on a 5-
point Likert scale ranging from 0, ‘‘Never’’ to 4, ‘‘Nearly
always,’’ with higher scores indicating greater frequency
or severity. The measure is comprised of 10 subscales, 8
of them being negative and 2 delving into positive
aspects of caregiving, with the aim of assessing an over-
all, all-encompassing experience of caregiving. The scale
can be used as a combined score of positive and negative
caregiving, or assessed per individual subscale. The ECI
has shown high internal reliability (Cronbach’s a values
ranging from 0.74 to 0.91) as well as strong construct
validity.45,46 It has been shown to predict a large portion
of the variability of carers’ psychological distress and its
use with carers of people with AN has been found to be
valid and effective.2 For the purpose of this study only
two of the negative scales of the ECI were used to isolate
specific negative aspects of caregiving in ED, while simul-
taneously allowing the inclusion of items that would be
suitable for both carers and parents of healthy controls.
Difficult Behaviors (e.g., him/her being unpredictable)
and Negative Symptoms (e.g., him/her being withdrawn)
were used as a combined score (i.e., Negative/difficult
behaviors).
Anorectic Behavior Observation Scale. The Anorectic
Behavior Observation Scale47 was developed as a self-
report questionnaire specifically aimed at parents’ obser-
vations of their son/daughter’s behaviors that may be
symptomatic for AN or BN. The measure has been found
to obtain reliable information from parents and to have
specificity and sensitivity in capturing parents’ observa-
tions of both behaviors and attitudes in their children
that may be symptomatic of ED. The complete scale con-
sists of 30 items, asking parents to respond based on their
40 International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
observations of their child over the past month, answer-
ing yes, no, or unsure as to whether the behavior
occurred. The scale is scored as follows: ‘‘yes’’ (2 points),
‘‘?’’ (1 point), and ‘‘no’’ (0 points), with a higher score
indicating higher frequency of symptomatic or patholog-
ical behavior. Parents are instructed to rate an item, such
as ‘‘Patient shows obvious signs of tension at mealtimes,’’
as ‘‘unsure’’ if they did not have the opportunity to
observe the behavior directly, if they were told it
occurred, or if they can only suppose it did. The scale has
a cut-off score of 1947 to indicate the presence of ano-
rexic symptoms. The scale has demonstrated three sepa-
rate factors: (1) eating behavior, concern with weight and
food, denial of problems; (2) bulimic-like behavior; and
(3) hyperactivity, with sufficient internal consistency
(Cronbach’s a values of 0.80, 0.69, and 0.69, respectively).
The scale was used to yield a combined score indicative
of AN symptomatology as reported by carers.
Analyses
Parametric tests were first carried out to check for the
homogeneity of the variance and normal distribution of
the data. Data were entered and checked for normality in
SPSS 13.48 The analyses were performed in STATA 9.49 A
series of simple (univariate) and multiple (multivariate)
regressions were conducted, to test gender and group dif-
ferences, as well as the degree of association between
variables. Comparisons between mothers and fathers in
the carer group included only parents from the same
families.
There was clustering within families as 120 of the 151
carers were from the same families and caring for the
same individual with AN (i.e., 60 pairs of parents (60
mothers, and 60 fathers), were from the same families
and caring for the same individual within each family).
This violates the first assumption required for analysis of
variance, that is, independence of observations. Robust
cluster regression controls for minor deviations from nor-
mal distributions and heterogeneity in the data, as well
as allows for analyses to be conducted using a sample
that may contain ‘‘intraclass’’ correlations within obser-
vations. Huber-white sandwich estimators of variance50,51
using the robust cluster option in STATA were thus used.
The robust cluster option in STATA does not provide
standardized beta coefficients, therefore the regression
coefficients presented in this study are unstandardized.
Furthermore, when measuring the effect of an independ-
ent variable on an outcome, it is suggested that unstan-
dardized regression coefficients are superior and prefera-
ble to correlation coefficients.52
A series of univariate regressions were first carried out
to identify the association between demographic and
patient/illness/carer-related characteristics and anxiety
and depression, as well as EE (CC and EOI). Variables
reaching significance of p \ .01 were then explored for

degree of association via multivariate regressions using
forward selection procedures; variables were added to
the model of ascending p-value. R2 partitioning of
variance was then conducted to ascertain the unique
variance that could be attributed to each of the variables
by fitting a series of models.
Results
Psychometric Properties of Instruments
Internal consistency of ECI and the FQ for carers
and controls was assessed using Cronbach’s a.b
Results show high consistency. On the combined
Difficult Behaviors and Negative Symptoms sub-
scales of the ECI, Cronbach’s a was 0.92 for carers
and 0.89 for the comparison parents. On the FQ,
Cronbach’s a values for the CC subscale were 0.90
for carers and 0.88 for comparison parents, and for
the EOI subscale, Cronbach’s a was 0.76 for carers
and 0.67 for comparison parents.
Demographic and Clinical Information
Demographic and clinical information is pre-
sented in Table 1. Mean age of carers was 54.3 years
(SD 5 7.2) and 48.9 for control parents (SD 5 5.4).
Women with AN had an average age of 23.3 (SD 5
6.6) and comparison daughters had an average age
of 17.3 (SD 5 4.6). Average duration of illness was
7.8 years (SD 5 5.5). The comparison group were
somewhat younger and had younger children as it
was unusual to have children over the age of 20 still
living at home. Gender comparisons within the
carers group are shown in Table 2.
Anxiety and Depression
The level of depression and anxiety in carers and
especially in mothers (m) was much higher than in
the comparison population. Over 70% of carers (n
5 107, m 5 77) scored at or above threshold (score
8) on the Anxiety subscale, and 38% (n 5 54, m 5
42) scored at or above threshold on the Depression
subscale, compared with 40% (n 5 37, m 5 24) at
or above threshold on Anxiety and 7% (n 5 7, m 5
5) at or above threshold on the Depression subscale
in the control group. Over 50% of carers (n 5 78, m
5 60) scored at or above the clinical threshold
(score 11) for Anxiety, compared with 15% (n 5
14, m 5 8) of the comparison group, and 13% (n 5
20, m 5 15) scored at or above the clinical threshold
b
For original Cronbach’s a values of these measures please see
‘‘Assessment Measures’’ in the ‘‘Method’’ section of this article.
International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
DEVELOPMENT OF EE IN CARERS OF PEOPLE WITH AN
for Depression compared with 1% (n 5 1, m 5 1)
in the control group.
Expressed Emotion
Higher levels of over involvement and CC were
present in the carers group. Over 60% (n 5 96, m 5
70) of carers had high levels of EOI (using the cut-
off score of 27), compared with 3% (n 5 3, m 5 2)
of the comparison group. Forty-seven percent of
carers (n 5 71, m 5 48) had high levels of CC (using
the cut-off score of 23), as opposed to 15% of con-
trols (n 5 14, m 5 8).
Regression Analyses of Variables Associated
with CC
The variables showing an association at p \ .01
significance with CC in univariate regressions (i.e.,
anxiety, depression, AN symptomatology, rejection
of carer help, ambivalence in accepting illness, neg-
ative/difficult behaviors) were simultaneously
entered into a forward selection regression model.
The model accounted for 52% of the variance of
CC, with two variables showing higher significance
over all remaining variables being adjusted for.
According to the order of significance provided by
the model, R2 partitioning was then used to assess
the unique portion of the variance attributable
to the variables associated with CC through a series
of models. The most significant variable was nega-
tive/difficult behaviors, which accounted for 50% of
the variance at the first step of R2 partitioning (p \
.001). At the second and final step rejection of
carers’ help accounted for a further 2% of the
variance. The results are shown in Table 3.
Regression Analyses of Variables Associated
with EOI
The variables showing an association at p \ .01
significance with EOI via univariate regressions
(i.e., anxiety, depression, AN symptomatology,
rejection of carer help, ambivalence in accepting
illness, negative/difficult behaviors) were simulta-
neously entered into a forward selection regression
model. The model accounted for 63% of the var-
iance of EOI, with three variables showing higher
significance over all remaining variables being
adjusted for. According to the order of significance
provided by the model, R2 partitioning was then
used to assess the unique portion of the variance
attributable to the variables associated with EOI
through a series of models. The most significant
variable was anxiety, which accounted for 57% of
the variance at the first step of R2 partitioning (p \
.001). At the second step depression accounted for a
further 4% of variance and at the third and final
41
 TABLE 1. Clinical and demographic information

42
No. of No. of Total No./Mean Total No./Mean
Carers Controls Frequencies (s.d.) Carers % (s.d.) Controls %

Demographic characteristics (possible score range)

Parental sexa,b,c 151 93 Female 91 53
Male 60 40
KYRIACOU ET AL.

Amount of contact with daughter per week 144 92 \21 h 89 62 54 59

[21 h 55 38 38 41
Parental age in years 145 93 54.3 (7.2) 48.9 (5.4)
Daughter age in years 142 93 23.3 (6.6) 17.3 (4.6)
Parental marital status 151 93 Single 1 1
Married/living together
Divorced/separated/widowed 133 88
17 4
Parental education level 143 93 Up to A level 87 42
University/equivalent 56 51
Parental employment statusb 145 93 Full-time 58 52
Not full time 87 41
Residing with daughter 151 93 Yes 93 74
No 58 19
Patient/illness-related characteristics
Anorexic behaviours observation scale (ABOS) (0–60)a,b,c,d 149 — 27.5 (10.6) N/A
Duration of illness in years 140 — 7.8 (5.5) N/A
Acceptance/rejection of eating disorder by suffererc,d 149 — Denies/ambivalent 33 22 N/A
Acceptance of ED 116 78
a,b,c,d
Rejection of carers’ help by sufferer 151 — Yes 102 67 N/A
No 49 32
Additional diagnosis/comorbidity of daughter 150 93 Yes 87 93
No 63 0
Sufferer’s average number of hospital admissions 110 — 2.9 (2.7) N/A
Sufferer currently in treatment 149 — Yes 112 N/A
No 37
Age of diagnosis 137 — 16.7 (4.4) N/A
Age of onset 140 — 15.2 (4.7) N/A
Parent-related characteristics
Hospital anxiety & depression scale (HADS)
Total of combined anxiety and ression scales (0–42)c,d* 150 93 16.6 (7.9) 39.5 9.9 (5.6) 23.6
Anxiety scale (0–21)c,d* 150 93 10.4 (4.9) 49.5 6.8 (3.8) 32.4
Depression Scale (0–21)c,d* 150 93 6.2 (3.7) 29.5 3.1 (2.6) 14.8
Experience of caregiving inventory (ECI)
Total of combined difficult behaviours and 150 93 24.9 (12.3) 44 11.1 (7.4) 20
negative symptoms scales (0–56)a,b,c,d*
Family questionnaire (FQ) Criticism (10–40)* 149 93 22.5 (6.1) 56 17.3 (4.5) 43
Emotional over-involvement (10–40)* 149 93 27.9 (4.5) 70 19.7 (5.6) 49
Notes: Percentage out of total score enables quick comparisons between subscales. This transforms the ratio of mean value to maximum total value into a percentage value for each individual subscale, show-
ing level of subscale endorsement.
Regressions results for Carers Group:
a
Univariate association with anxiety significant at p \ .01.
b
Univariate association with depression significant at p \ .01.
c
Univariate association with emotional over-involvement significant at p \ .01.
d
Univariate association with criticism significant at p \ .01.
Group comparisons:

International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat

* Significant group difference between carers and controls at p \ .001.
 DEVELOPMENT OF EE IN CARERS OF PEOPLE WITH AN

TABLE 2. Gender comparisons via regressions (carers group)

Measure (Possible Score Range) Mother, N Mother, Mean (s.d.) % Father, N Father, Mean (s.d.) %

Rejection of carers’ help 60 Yes: 43 71.6 60 Yes: 37 62

No: 17 28.4 No: 23 38
Amount of weekly contact 53 60
\21 h 28 53 45 75
[21 h 25 47 15 25
Hospital anxiety and depression scale (HADS)
Total of combined anxiety and 60 18.7 (7) 44.5 59 13.2 (8.1) 31
depression scales (0–42)*
Anxiety subscale (0–21)* 60 11.7 (4.3) 55.7 59 8.3 (5.1) 39.5
Depression subscale (0–21)* 60 7 (3.5) 33.3 59 5 (3.6) 23.8
Experience of caregiving inventory (ECI)
Total of combined difficult behaviours and 60 26.3 (12) 47 59 22.8 (12.5) 41
negative symptoms scales (0–56)
Family questionnaire
Criticism (10–40) 60 23 (6.4) 57.5 58 21.6 (5.7) 54
Emotional over-involvement (10–40)* 60 29 (3.9) 72.5 58 26.2 (4.8) 65.5
Anorexic behaviours observation scale (ABOS)
Total score (0–60) 59 27.9 (10.8) 46.5 60 27 (10.5) 45
Note: Percentage out of total score enables quick comparisons between subscales. This transforms the ratio of mean value to maximum total value into
a percentage value for each individual subscale, showing level of subscale endorsement.
* Significant gender difference at p \ .001.

TABLE 3. Multivariate regression of variables associated with criticism

Forward Selection Model Reg. Coef. Rob. Stand. Error t-Test Statistic (DF) p-Value R2

Model 1
Negative/difficult behaviours .35 .02 15.16 (1, 114) \.001 .495
Model 2
Negative/difficult behaviours .30 .03 9.14 \.001 .519
Rejection of carer’s help 2.44 .97 2.52 (2, 114) .013
At final step F(2, 114) 5 116.02, p \ .001. Degrees of Freedom given for clusters, not individual cases. Reference groups as follows: patient accepts carer’s
help. All remaining variables are continuous.

TABLE 4. Multivariate regression of variables associated with emotional over-involvement

Forward Selection Model Reg. Coef. Rob. Stand. Error t-Test Statistic (DF) p-Value R2

Model 1
Anxiety .69 .05 13.39 (1, 114) \.001 .572
Model 2
Anxiety .52 .06 8.62 \.001 .611
Depression .33 .08 4.11 (2, 115) \.001
Model 3
Anxiety .45 .06 7.31 \.001 .628
Depression .31 .08 3.93 \.001
Negative/difficult behaviours .07 .02 3.10 (3, 114) .002
At final step F(3, 114) 5 73.87, p \ .001. Degrees of freedom given for clusters, not individual cases.

step negative/difficult behaviors accounted for an
additional 2% of the variance. The results are
shown in Table 4.
Conclusion
The aim of this study was to compare EE measured
with a self report tool in parents of people with AN
with that of parents of healthy young people. The
results confirmed our first hypothesis in that over
60% of carers had high levels of EOI compared with
3% parents from the comparison group. Forty-
seven percent of carers had high levels of CC as
opposed to 15% of the comparison group. Mothers
in particular had high levels of EOI. Mothers also
had significantly higher levels of anxiety and
depression and EOI as compared with fathers in
the carers group. Carers of people with AN also
showed significantly higher levels of anxiety and
depression than parents of healthy controls. An

International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat 43

KYRIACOU ET AL.
additional aim was to examine whether aspects of
EE were related to caregiving burden and/or paren-
tal anxiety and depression. Negative/difficult
behaviors and rejection of carers’ help were the
most significant variables associated with CC,
accounting for over 50% of the variance. Parental
anxiety, depression and the negative/difficult
behaviors of the individual with an eating disorder
accounted for over 60% of the variance of EOI.
These findings suggest that this self report tool is
clearly sensitive both to CC and EOI in families of
people with ED. Indeed it is possible that it is a
more sensitive measure of CC than the CFI for use
in ED Previous work using the CFI did not find dif-
ferences in CC between families with an eating dis-
order and a control group.53 Also, earlier studies
that have reported low CC have tended to use the
conventional cut-off thresholds used in schizo-
phrenia (e.g., Refs. 20, 53), whereas use of a lower
threshold, as sometimes used in depression
research, is found to warrant the classification of
high EE in a greater number of carers. Furthermore,
the difference between carers and controls in high
EOI classification is far more striking compared
with CC, which is consistent with previous infer-
ences according to which overall, CC is lower in
eating disorder families as compared with schizo-
phrenia and other disorders.27,30
Furthermore, we found that the level of EOI was
related to parental anxiety and depression and
aspects of the illness. This is in agreement with the
findings of Blair et al., who found that the levels of
EE were higher than that seen in the comparison
population but were similar to those of parents car-
ing for an offspring with cystic fibrosis.53 Higher
EOI, as Blair et al. suggest, may be a more charac-
teristic familial reaction than CC in response to a
serious illness in a child. In terms of gender differ-
ences furthermore, mothers in the carers group
showed higher EOI, which is somewhat consistent
with previous findings, as Szmukler et al.,29 found
that fathers generally had lower scores on EE than
mothers, however, mothers did not differ in CC as
reported elsewhere.28 It is possible that CC is more
characteristic of families with BN as identified by
Winn et al.,39 whereas EOI may be a more salient
factor in carers of AN.
In terms of comparability of psychological dis-
tress of the current carer sample to carers in other
areas of mental health, it appears that these AN
carers experience significant psychological morbid-
ity, and particularly anxiety, similar to or even
greater than in other areas. In comparison to HADS
scores of carers in Alzheimer’s disease, for example,
44 International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
(Anxiety: M 5 7.4, SD 5 4.4, 23.5% of carers;
Depression: M 5 5.1, SD 5 4, 10.5% of carers)54 our
carers show higher psychological morbidity. This
difference is consistent with evidence reported by
Treasure et al.,2 who compared caregivers’ mental
wellbeing in psychosis and AN and found that AN
carers had significantly higher GHQ scores (in psy-
chosis, mean GHQ score was 16.4 (SD 5 8), which
is suggestive of clinical psychological distress,
whereas in AN the mean GHQ score was 27.7 (SD
5 6.2), meeting clinical the clinical caseness crite-
rion. Our findings therefore offer additional evi-
dence of the considerable psychological morbidity
experienced by carers in ED.
Notably, no significant gender differences were
found in the control group on any of the measures.
This suggests that the salient gender differences
found in our group of carers are likely to be inher-
ent in the caregiving role and that the impact on
carers’ distress is different for fathers and mothers
caring for a daughter with AN.
Limitations
The current sample of carers was self-selected as
participants from an existing database were sought,
which raises the question of representativeness of
the respondents. Although some of the carers on this
database are parents of the unit’s inpatients (36
parents in the current sample), they may also come
from a variety of sources, such as Beat, and may dif-
fer in terms of experience of caregiving and familiar-
ity with services and resources to carers, which may
impact on physical and psychological wellbeing. It
could be argued, furthermore, that these carers are
more empowered and actively involved in seeking
treatment and support for themselves and their
loved ones, which could lead to a skewed sample.
We have explored main demographic and clinical
characteristics of nonrespondents in comparison to
those of responders from the database. Although our
comparison has yielded no significant differences we
do acknowledge a limitation as regards the represen-
tativeness of the database sample.
As the current study was cross-sectional in design,
results suggest the degree of association between EE
and the variables under study, however, no causal
links can be identified. A longitudinal study of carer
distress and EE from the early stages over the course
of an eating disorder and exploring the mechanisms
contributing to both carer distress and EE would be
needed to elucidate the exact process leading to the
development of these constructs. No inferences can
be made as to the direction of the relationship
between EE and carers’ distress. It has been sug-

gested that EE is not a static construct but one that
changes over time and generally decreases over the
treatment period.28
The control group used in the current study is
similar in amount of weekly contact between par-
ent and offspring, however, both parents and off-
spring in the comparison group were younger. We
did not have the resources to simultaneously mea-
sure EE in this group using the original methodol-
ogy the CFI, neither were we able to examine the
validity of the constructs as measured by this
instrument in terms of whether they have an
impact of the outcome of AN. This work needs to
be done next. Also, replication using more sophisti-
cated methods of analysis, such as structural equa-
tion modeling, in larger samples is also to be rec-
ommended. Finally, additional comparisons of EE
in carers of people with BN as well as carers with a
different type of chronic illness, e.g., diabetes,
cystic fibrosis or some other mental health prob-
lem, are needed.
Clinical Implications
The FQ appears to be a useful and cost-effective
method to assess family functioning in AN. The
precise nature and direction of the relationship
between EE dimensions and eating disorder course,
outcome and impact on carer wellbeing requires
further research to be deciphered. Specifically,
future research is needed to explore whether CC
and EOI can act as outcome and relapse markers in
AN and BN. The finding that parental depression
and anxiety are related to high EOI suggests that
interventions aimed at reducing EOI may need to
focus on strategies for improving the mood and
quality of life for carers. Interventions which help
carers manage difficult behaviors in the face of the
individual with an eating disorder rejecting help
may reduce the level of CC. Educational and/or
skills based interventions for carers of people with
ED can reduce parental distress, burden and EE 55–60
and it will be interesting to note whether these
improve the outcome of the eating disorder itself.
We would like to thank all the parents and professionals
for their valuable help with this research. Special thanks
to Dr. Sabine Landau for statistical advice and to Frankie
Bishopp for helping with recruitment of control families.
References
1. Baronet A. Factors associated with caregiver burden in mental
illness: A critical review of the research literature. Clin Psychol
Rev 1999;19:819–841.
International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
DEVELOPMENT OF EE IN CARERS OF PEOPLE WITH AN
2. Treasure J, Murphy T, Szmukler G, Todd G, Gavan K, Joyce J.
The experience of caregiving for severe mental illness: A com-
parison between anorexia nervosa and psychosis. Soc Psychia-
try Psychiatr Epidemiol 2001;36:343–347.
3. Zucker NL, Marcus M, Bulik C. A group parent-training pro-
gram: A novel approach for eating disorder management. Eat
Weight Disord 2006;11:78–82.
4. Strober M. Managing the chronic, treatment-resistant patient
with anorexia nervosa. Int J Eat Disord 2004;36:245–255.
5. Perkins S, Winn S, Murray J, Murphy R, Schmidt U. A qualitative
study of the experience of caring for a person with bulimia
nervosa, Part 1: The emotional impact of caring. Int J Eat Dis-
ord 2004;36:256–268.
6. Haigh R, Treasure J. Investigating the needs of carers in the
area of eating disorders: Development of the carers’ needs
assessment measure (CaNAM). Eur Eat Disord Rev 2003;11:125–
141.
7. Treasure J, Gavan K, Todd G, Schmidt U. Changing the environ-
ment in eating disorders: Working with carers/families to
improve motivation and facilitate change. Eur Eat Disord Rev
2003;11:25–37.
8. Murphy F, Troop NA, Treasure JL. Differential environmental
factors in anorexia nervosa: A sibling pair study. Br J Clin Psy-
chol 2000;39 (Part 2):193–203.
9. Cottee-Lane D, Pistrang N, Bryant-Waugh R. Childhood onset
anorexia nervosa: The experience of parents. Eur Eat Disord
Rev 2004;12:169–177.
10. Highet N, Thompson M, King RM. The Experience of living with
a person with an eating disorder: The impact on the carers. Eat
Disord 2005;13:327–344.
11. De La Rie SM, van Furth EF, De Koning A, Noordenbos G,
Donker MCH. The quality of life of family caregivers of eating
disorder patients. Eat Disord 2005;13:345–351.
12. Winn S, Perkins S, Murray J, Murphy R, Schmidt U. A qualitative
study of the experience of caring for a person with bulimia
nervosa, Part 2: Carers’ needs and experience of services and
other support. Int J Eat Disord 2004;36:269–279.
13. McMaster R, Beale B, Hillege S, Nagy S. The parent experience
of eating disorders: Interactions with health professionals. Int J
Ment Health Nursing 2004;13:67–73.
14. Zucker NL, Ferriter C, Best S, Brantley A. Group parent training:
A novel approach for the treatment of eating disorders. Eat Dis-
ord 2005;13:391–405.
15. Butzlaff RL, Hooley JM. Expressed emotion and psychiatric
relapse: A meta-analysis. Arch Gen Psychiatry 1998;55:547–
552.
16. Schmidt U, Treasure J. Anorexia nervosa: Valued and visible. A
cognitive-interpersonal maintenance model and its implica-
tions for research and practice. Br J Clin Psychol 2006;45:343–
366.
17. Whitney JB, Haigh R, Weinman J, Treasure J. Caring for people
with eating disorders: Factors associated with psychological dis-
tress and negative caregiving appraisals in carers of people
with eating disorders. Br J Clin Psychol (in press).
18. van Os J, Marcelis M, Germeys I, Graven S, Delespaul P. High
expressed emotion: Marker for a caring family? Compr Psychia-
try 2001;42:504–507.
19. Whitney J, Murray J, Gavan K, Todd G, Whitaker W, Treasure J.
Experience of caring for someone with anorexia nervosa: Quali-
tative study. Br J Psychiatry 2005;187:444–449.
20. le Grange D, Eisler I, Dare C, Hodes M. Family criticism and self-
starvation: A study of expressed emotion. J Fam Ther 1992;
14:177–192.
21. Scazufca M, Kuipers E. Links between expressed emotion and
burden of care in relatives of patients with schizophrenia. Br
J Psychiatry 1996;168:580–587.
45
KYRIACOU ET AL.
22. Wagner AW, Logsdon RG, Pearson JL, Teri L. Caregiver expressed
emotion and depression in Alzheimer’s disease. Aging Ment
Health 1997;1:132–139.
23. Barrowclough C, Tarrier N, Johnston M. Distress, expressed
emotion, and attributions in relatives of schizophrenia
patients. Schizophr Bull 1996;22:691–702.
24. Wearden AJ, Terrier N, Barrowclouch C, Zastowny TR, Rahill AA.
A review of expressed emotion research in health care. Clin
Psychol Rev 2000;20:633–666.
25. Tarrier N, Barrowclough C, Ward J, Donaldson C, Burns A, Gregg
L. Expressed emotion and attributions in the carers of patients
with Alzheimer’s disease: The effect on carer burden. J Abnorm
Psychol 2002;111:340–349.
26. Besharat MA, Eisler I, Dare C. The self- and other-blame scale
(SOBS). The background and presentation of a new instrument
for measuring blame in families. J Fam Ther 2001;23:208–223.
27. Hodes M, Le Grange D. Expressed emotion in the investigation
of eating disorders: A review. Int J Eat Disord 1993;13:279–288.
28. van Furth EF, van Strien DC, Martina LML, van Son MJM,
Hendrick JJP, van England H. Expressed emotion and the pre-
diction of outcome in adolescent eating disorders. Int J Eat Dis-
ord 1996;20:19–31.
29. Szmukler GI, Eisler I, Russell GF, Dare C. Anorexia nervosa, pa-
rental ‘expressed emotion’ and dropping out of treatment. Br
J Psychiatry 1985;147:265–271.
30. Hodes M, Dare C, Dodge E, Eisler I. The Assessment of expressed
emotion in a standardised family interview. J Child Psychol Psy-
chiatry 1999;40:617–625.
31. Vaughn C, Leff J. The measurement of expressed emotion in
the families of psychiatric patients. Br J Soc Clin Psychol
1976;15:157–165.
32. Magana AB, Goldstein MJ, Karno M, Miklowitz DJ. A brief
method for assessing expressed emotion in relatives of psychi-
atric patients. Psychiatry Res 1986;17:203–212.
33. Miklowitz DJ, Goldstein MJ. Mapping the intrafamilial environ-
ment of the schizophrenic patient. In: Cromwell RL, Snyder CR,
editors. Schizophrenia: Origins, Processes, Treatment and
Outcome. New York: Oxford University Press, 1993, pp. 313–
332.
34. Glynn SM, Randolph ET, Eth S, Paz GG, Leong GB, Shaner AL,
et al. Patient psychopathology and expressed emotion in schiz-
ophrenia. Br J Psychiatry 1990;157:877–880.
35. van Furth EF, van Strien DC, van Son MJ, van Engeland H. The
validity of the five-minute speech sample as an index of
expressed emotion in parents of eating disorder patients. J Child
Psychol Psychiatry 1993;34:1253–1260.
36. Cole JD, Kazarian SS. The level of expressed emotion scale: A
new measure of expressed emotion. J Clin Psychol 1988;
44:392–397.
37. Kazarian SS, Baker B. Influential relationships questionnaire: Data
from a nonclinical population. Psychol Rep 1987;61:511–514.
38. Kazarian SS, Malla AK, Cole JD, Baker B. Comparisons of two
expressed emotion scales with the Camberwell family inter-
view. J Clin Psychol 1990;46:306–309.
39. Winn S, Perkins S, Walwyn R, Schmidt U, Eisler I, Treasure J,
et al. Predictors of mental health problems and negative care-
giving experiences in carers of adolescents with bulimia nerv-
osa. Int J Eat Disord 2007;40:171–178.
40. Wiedemann G, Rayki O, Feinstein E, Hahlweg K. The family
questionnaire: Development and validation of a new self-
report scale for assessing expressed emotion. Psychiatry Res
2002;109:265–279.
41. Moller-Leimkuhler AM. Burden of relatives and predictors of
burden. Baseline results from the Munich 5-year-follow-up
46 International Journal of Eating Disorders 41:1 37–46 2008—DOI 10.1002/eat
study on relatives of first hospitalized patients with schizophre-
nia or depression. Eur Arch Psychiatry Clin Neurosci 2005;
255:223–231.
42. Moller-Leimkuhler AM. Multivariate prediction of relatives’
stress outcome one year after first hospitalization of schizo-
phrenic and depressed patients. Eur Arch Psychiatry Clin Neu-
rosci 2006;256:122–130.
43. Zigmond AS, Snaith RP. The hospital anxiety and depression
scale. Acta Psychiatr Scand 1983;67:361–370.
44. Snaith RP. The hospital anxiety and depression scale. Health
Qual Life Outcomes 2003;1:29.
45. Szmukler GI, Burgess P, Herrman H, Benson A, Colusa S, Bloch
S. Caring for relatives with serious mental illness: The develop-
ment of the experience of caregiving inventory. Soc Psychiatry
Psychiatr Epidemiol 1996;31:137–148.
46. Joyce J, Leese M, Szmukler G. The experience of caregiving in-
ventory: Further evidence. Soc Psychiatry Psychiatr Epidemiol
2000;35:185–189.
47. Vandereycken W. Validity and reliability of the anorectic behav-
ior observation scale for parents. Acta Psychiatr Scand 1992;
85:163–166.
48. SPSS. SPSS; Version 13.0. Chicago, Illinois: SPSS, 2005.
49. STATA. Stata statistical software. In: Station C, editor. Release 9.
TX: StataCorp LP, 2005.
50. Huber PJ. The behavior of maximum likelihood estimates
under nonstandard conditions. In: Proceedings of the Fifth
Berkeley Symposium on Mathematical Statistics and Probabil-
ity. Berkeley, CA: University of California Press, 1967, pp. 221–
233.
51. White H. A heteroskedasticity-consistent covariance matrix esti-
mator and a direct test for heteroskedasticity. Econometrica
1980;48:817–830.
52. Baron RM, Kenny DA. The moderator-mediator variable distinc-
tion in social psychological research: Conceptual, strategic, and
statistical considerations. J Pers Soc Psychol 1986;51:1173–
1182.
53. Blair C, Freeman C, Cull A. The families of anorexia nervosa and
cystic fibrosis patients. Psychol Med 1995;25:985–993.
54. Mahoney R, Regan C, Katona C, Livingston G. Anxiety and
depression in family caregivers of people with alzheimer dis-
ease: The LASER-AD study. Am J Geriatr Psychiatry 2005;13:
795–801.
55. Eisler I, Dare C, Hodes M, Russell G, Dodge E, Le Grange D. Fam-
ily therapy for adolescent anorexia nervosa: The results of a
controlled comparison of two family interventions. J Child Psy-
chol Psychiatry 2000;41:727–736.
56. Eisler I, Dare C, Russel GFM, Szmukler G, Le Grange D, Dodge E.
Family and individual therapy in anorexia nervosa: A five year
follow-up. Arch Gen Psychiatry 1997;54:1025–1030.
57. Eisler I. The empirical and theoretical base of family therapy
and multiple family day therapy for adolescent anorexia
nervosa. J Fam Ther 2005;27:104–131.
58. Uehara T, Kawashima Y, Goto M, Tasaki S, Someya T. Psycho-
education for the families of patients with eating disorders and
changes in expressed emotion: A preliminary study. Compr
Psychiatry 2001;42:132–138.
59. Eisler I, le Grange D, Asen E. Family interventions. In: Treasure
J, Schmidt U, van Furth EF, editors. Handbook of Eating Disor-
ders, 2nd ed. Chichester: Wiley, 2003, pp. 291–311.
60. Treasure J, Sepulveda AR, Whitaker W, Todd G, Lopez C, Whit-
ney J. Collaborative care between professionals and non-profes-
sionals in the management of eating disorders: A description
of workshops focussed on interpersonal maintaining factors.
Eur Eat Disord Rev 2007;15:24–34.