10.1192/bjp.174.4.330Access the most recent version at DOI:
1999, 174:330-338. BJP
Fichter, H van Oyen, M Roelands, I Skoog, C Turrina and J R Copeland
M J Prince, F Reischies, A T Beekman, R Fuhrer, C Jonker, S L Kivela, B A Lawlor, A Lobo, H Magnusson, M
compare symptoms of depression in 14 European centres.
Development of the EURO-D scale--a European, Union initiative to
This article cites 0 articles, 0 of which you can access for free at:
To obtain reprints or permission to reproduce material from this paper, please write
to this article at
You can respond
The Royal College of PsychiatristsPublished by
on March 5, 2013http://bjp.rcpsych.org/
go to: The British Journal of PsychiatryTo subscribe to
BRITISH J O U R N A L OF PSYCHIATRY (1999). 174, 330-338
E U R O D E P S T U D Y
Development of the EURO-D scale - a European
Union initiative to compare symptoms of
depression in 1 4 European centres
M. J. PRINCE, F. REISCHIES, A.T. F . BEEKMAN, R. FUHRER, C. JONKER,
S.-L. KIVELA, B. A. LAWLOR, A. LOBO. H. MAGNUSSON, M. FICHTER,
H. VAN OYEN, M. ROELANDS, I. SKOOG, C. TURRINA and J. R. M. COPELAND
Background In an I I-country
European collaboration, 1 4 population-
The EURODEP Concerted Action Pro-
gramme is a consortium of 14 research
groups from 11 European countries all
engaged in population-based research into
the epidemiology of late-life depression.
The aims of the consortium were three-
fold: to compare the prevalence of late-
life depression in different European
countries and cultures; to assess the
homogeneity or heterogeneity of aeti-
ological associations with late-life depres-
sion between centres; and (in so far as
patterns of association were homogeneous
between centres) to pool data in order to
increase the power and precision of these
The collaboration was initially in-
tended to include nine European centres,
all of which had used either the Geriatric
Mental State-AGECAT (GMS-AGECAT;
Copeland et al, 1986) or the very similar
SHORT-CARE (Gurland et al, 1984) as
their index of clinical case-level depression,
allowing direct comparison of prevalence
between centres (Table 1). Data from these
centres are reported elsewhere (Copeland et
al, 1999a,b, this issue). However, five large
European studies had used other measures
of depression (Kivela et al, 1988; Barberger
Gateau et al, 1992; Skoog et al, 1993; Roe-
lands et al, 1994; Beekman et al, 1995) (see
also Table 1). Three centres had used the
Centre for Epidemiological Studies Depres-
sion scale (CESD) (Radloff, 1977), one the
Zung Self-Rating Depression Scale (ZSDS)
(Zung, 1965) and one the Comprehensive
Psychopathological Rating scale (CPRS)
(Asberg et al, 1978). Incorporation of data
from these centres into the analyses would
include four countries (France, Belgium,
Sweden and Finland) not otherwise repre-
sented, and would increase the overall Sam-
ple by 8427, to 21 724 subjects. Our task
was, therefore, to find a way of harmonis-
ing data from these five depression
measures in such a way as to allow
meaningful comparison, while retaining
based surveys included 21 724 subjects
aged 2 65 years. Most participating
centres used the Geriatric Mental State
(GMS), but other measures were also
Aims To derive from these instruments
a common depression symptoms scale, the
EURO- D, to allow comparison of risk
factor profiles between centres.
Method Common items were
identified from the instruments.
Algorithms for fitting items to GMS were
derived bv observation of item
correspondence or expert opinion.The
resulting 12-item scale was checked for
internal consistency, criterion validity and
uniformity offactor-analytic profile.
Results The EURO- D is internally
consistent, capturing the essence of its
parent instrument. Atwo-factor solution
seemed appropriate: depression,
tearfulness and wishing to die loaded on
the first factor (affective suffering), and
loss of interest, poor concentration and
lack of enjoyment on the second
Conclusions The EURO- D scale
should permit valid comparison of risk-
factor associations between centres, even
if between-centre variation remains
Declaration of interest The
European Commission BIOMED I
initiative funded this Concerted Action
GMSAGECAT is a semi-structured clinical
interview schedule which can be adminis-
tered by medically qualified or trained lay
interviewers. Identification of symptoms
requires some judgement to be exercised
by the interviewer, who rates the severity
of symptoms and also makes observations
regarding the subject's demeanour and
behaviour. Diagnoses are allocated by the
computer program AGECAT, using a hier-
archical typological algorithm.
SHORT-CARE is a very similar instru-
ment to GMS, with the majority of items
identically or nearly identically worded. It
is, however, more structured; interviewers
are not required to make judgements
regarding symptom severity and there are
no observational items. The diagnosis of
pervasive depression is made on the basis
of a simple algorithm requiring the presence
of six or more of 17 key symptoms of de-
pression. Like GMSAGECAT depression,
it is a broad diagnostic category signifying
depression of clinical case-level severity.
CES-D is a 20-item self-report depres-
sion symptom rating scale. Subjects are
asked to rate each symptom as being
present over the previous week (a) rarely
or none of the time, (b) some or a little of
the time, (c) occasionally, or (d) most or
all of the time. A cut-off point of 15/16 is
commonly used to identify a group with a
high probability of suffering from depres-
sion of clinical significance. This cut-off
point has been validated against a diagnosis
of clinical case-level depression in both
younger and older subjects (Breslau, 1985;
Beekman et al, 1994).
The modified ZSDS is a 20-item self-
report depression symptom rating scale.
Subjects are asked to rate each symptom
as being present over the previous two
weeks (a) never, (b) now and then, (c) quite
often, or (d) most of the time.
CPRS is a wide-ranging semi-structured
clinical psychiatric interview with operatio-
nalised symptom definitions and symptom
ratings. Sufficient information can be gen-
erated to make a variety of ICD-10 (World
Health Organization, 1993) or DSM-IV
(American Psychiatric Association, 1994)
Harmonisation of scales
GMSAGECAT, SHORT-CARE and CPRS
are interviewer-administered and generate
EURODEP STUDY 5
We l The EURODEP consortium - studies and subjects
Centre Number dsubjects
Percentage of subjects
Gahenbug. Sweden (CPRS)
Ahtari, Finland (ZSDS)
GI%. Geriurk hul State; CES-D. Cenm for E p i d e m i i Studies Depresdon Scale; CPRS, Comprehensive Psychopmhdogiul Rating Sale; ZSDS. Zung Self-Rating Depres-
I. LASA. Longkdid Ageing Study Amsterdam
2. PAQUID. F'ersonms A g k QUID.
clinical diagnoses, whereas CES-D and
ZSDS are simple self-report instruments
assessing the presence and pervasiveness
of depression symptoms. We decided,
therefore, to develop a depression symptom
scale by selecting those symptoms from the
clinical instruments which were also com-
mon to the CES-D and the ZSDS, rather
than attempting the unrealistic goal of a
common categorical criterion.
Fitting CES- D , ZSDS and CPRS items to GMS
We next attempted to fit the CES-D, ZSDS
and CPRS items to the GMS items, with the
0, 1, 2 GMS item scores (symptom not
present, present to a moderate degree, pre-
sent to a marked degree) collapsed into 0
for not present and 1 for present. Two
methods were used.
5, how good was the fit between the CES-D
item or items and the GMS item.
We reviewed each instrument to identify
items that were common to all or most of
the five, and had good face validity for
late-life depression. We selected 12 items:
depression, pessimism, wishing death, guilt,
sleep, interest, irritability, appetite, fatigue,
concentration, enjoyment, and tearfulness
(Table 2). All were present in GMSAGE-
CAT (although some centres used a version
in which enjoyment was not assessed) and
in ZSDS, 11 were present in SHORT-
CARE, and 10 in CES-D and CPRS (Table
2). In some cases, we judged that more than
one item from CES-D related to a single
Method I - expert opinion. Six old-age psy-
chiatrists with experience of epidemiologi-
cal research, familiar with the use of the
five instruments, but independent of the
EURODEP consortium, were asked to
advise. For each item on each questionnaire
they had three judgements to make. In the
case of the CES-D, for example, they had
to judge (a) in cases where the GMS item
could be addressed by two items on CES-
D, whether the GMS symptom be consid-
ered to be present if the subject scored
above a threshold (i) on both CES-D items,
(ii) on either item, or (iii) on only one of the
two (the other adjudged to be superfluous);
(b) at what level of frequency or pervasive-
ness of the CES-D symptom would the
symptom be considered to be present (to
match with the dichotomous coding of the
GMS item); (c) overall, on a scale of 1 to
Method 2 - probabilistic modelling. One
centre, Berlin, had administered both
GMSAGECAT and CES-D to the same
488 subjects. Thus, it was possible to exam-
ine directly the correspondence between
items from the two instruments. The
EURO-D item score was calculated as the
positive predictive value (PPV) for a GMS
item score of 1 at each possible level of
CES-D (0, 1, 2 or 3). Where the fit with
GMS could be improved by combining
information from two CES-D items, PPVs
were calculated for all 16 possible combi-
nations of CES-D score from the two items.
lmputotion of missing data
Subjects with missing data for more than
one relevant GMS, SHORT-CARE, CES-
D, CPRS or ZSDS item were excluded from
the analysis. We used statistical imputation
to estimate missing data in two situations:
where individual subjects were missing only
one item, and where all subjects in one
centre were missing only one item, and
where all subjects in one centre were
missing (by design) one or two items. For
T d e 2 EURO-D scale items
Symptom GMS-item CES-D item
EUROI l Enjoyment
Have you been sad (depressed, miserable, in low spirits, blue)
How do you see your future? (Pessimistic, empty
expectations or bleak future)
Have you ever felt that you would rather be dead? (Has ever
felt suicidal or wished to be dead)
Do you tend to blame yourself or feel guilty about anything?
(Obvious guilt or self blame)
Have you had trouble sleeping recently? (Trouble with sleep
or recent change in pattern)
What is your interest in things? (Less interest than is usual)
Have you been irritable recently?
What has your appetite been like? (Diminution in the desire
Have you had too little energy (to do the things you want to
do)? (Listlessness or subjective energy restriction)
How is your concentration? (Diculty in concentrating on
entertainment or reading)
What have you enjoyed doing recently? (Almost nothing
Have you cried at all?
I felt depressed; and I felt that I could not shake &the M w
wen with the help o f my family o r friends
I feel hopeful about my future
I felt as good as other people; and. I thought my life had been
My sleep was restless
I was bothered by things which don't usually bother me
I did not like eating; my appetite was poor
I felt that everything I did was an effort
I had trouble keeping my mind on what I was doing
I enjoyed life
I had crying spells
I. Expen panel's m n
rating dovedl match with GMS item, scored on a scale of I to 5.
single subjects we imputed the missing item
score by applying to that subject's other
item scores the relevant coefficients from
the logistic regression (GMS and SHORT-
CARE) or multiple regression (CES-D,
ZSDS and CPRS) model predicting the
missing item score, derived from all subjects
in that centre with complete data. Similarly,
where an item was missing by design from
an entire centre, we imputed the item scores
using regression models derived from all
other centres with complete data. In the
case of the CES-D centres, where EURO-
D item scores were non-integer probabil-
ities, Berlin CES-D item scores were used
to predict, using logistic regression, GMS
EUR03 and EUR06 scores on the same
subjects. We then used the coefficients from
the resulting logistic model to predict for
each subject from the Longitudinal Ageing
Study Amsterdam (LASA) (Netherlands),
PAQUID (Personnes Agks QUID, France)
and Antwerp centres a probability of an
item score of 1 for the missing E m 0 3
and EUR06 items.
for the 12 contributing items. For the CES-
D centres we calculated the scale according
to two methods: for the expert opinion
method, by summing the 0,l scores derived
from the panel's recommendations; and, for
the probabilistic modelling method, by sum-
ming the PPVs calculated from algorithms
derived from the Berlin data set.
We assessed the internal consistency of the
resulting scales from each centre by calcu-
lating the inter-item correlations, the item-
total correlations and the standardised
alpha value. For these analyses, we excluded
imputed items so as not to bias upwards the
estimates of internal consistency. We
tested whether internal consistency (judged
by the standardised alpha) could be
improved by omitting items with low
The criterion validity of the EURO-D scales
could be assessed directly in Berlin, where
the EURO-D derived from the GMS could
be compared with the CES-D, and in the
Netherlands LASA study where the
EURO-D derived from CES-D could be
For the GMS centres, we constructed the
EURO-D scale by summing the 0 or 1 scores
compared with Composite International
Diagnostic Interview (CIDI; World Health
Organization, 1990) diagnoses of major
depression. In other centres, comparison of
EURO-D with GMSAGECAT depression
diagnoses or CES-D scale scores involved
an element of circularity, in that the
EURO-D scale had been derived from items
from the instrument which provided the cri-
terion measure. Agreement with continuous
measures was assessed by Spearman non-
parametric correlations, and for dichoto-
mous measures by the area under the recei-
ver operating characteristic curve.
We assessed whether a common factor
structure existed across all the centres by
carrying out a principalcomponents analy-
sis of the EURO-D scale items (with vari-
max rotation) separately for each centre.
Harmonisation - the expert
opinion and the probabilistic
The expert panel's recommendations were
remarkably consistent. For nearly all of
the items on each of the three scales there
was a clear majority for one cut-off point
and, where appropriate, for a single
approach in combining two items. There
was also good agreement on the extent to
which the non-GMS items could be
matched to those from the GMS. The CPRS
items were all felt to match adequately. For
the CES-D, three items - numbers 2 (pessi-
mism), 4 (guilt) and 7 (irritability) - were
considered to give poor matches; for the
ZSDS, there were five such items - numbers
2 (pessimism), 3 (wishing death), 6 (inter-
est), 8 (appetite) and 10 (concentration).
The accuracy of the panel's judgement on
matching could be checked for CES-D using
data from Berlin, where the correlation
between the CES-D score and the GMS
item score was measured directly. The
Spearman correlation between the panel's
rating and the observed correlation for 10
items was only 0.14 (P=0.7).
Table 3 compares the Berlin CES-D
EURO-D calculated using the experts'
opinion and the probabilistic model with
the GMS EURO-D from the same centre.
The panel's opinion led to the CES-D
EURO-D underestimating the prevalence
of the depression, wishing death, sleep
and tearfulness items (with reference to
GMS EURO-D from the same centre),
and overestimating the prevalence of the
pessimism and enjoyment items. The net
result was a similar mean and standard
deviation for the total scale score. The
probabilist model inevitably resulted in
item 'prevalences' which were almost
identical to those observed for the GMS
EURO-D. The scale mean was also very
similar to that for the GMS EURO-D.
The probabilistic model, when applied to
CES-D data from the LASA study, also
seemed to result in a more plausible pattern
of item prevalences (cf. Table 4) than did
the expert panel's algorithm.
The EURO-D item prevalences, and scale
means and standard deviations for all
centres, are given in Table 4. In each centre
the scale distribution was positively
skewed. The median value, 1 for most cen-
tres, 2 for London and Berlin and 3 for Mu-
nich, was always less than the mean, which
ranged from 1.34 to 3.58. When the three
outlying centres were excluded (Dublin,
mean 1.34; Ahtari, Finland 3.17; Munich
3.58), the mean EURO-D scores for the
other nine centres ranged only from 1.79
to 2.54. The standard deviations were
much lower for the CES-D EURO-D
(range 0.94-1.13) than for the GMS
EURO-D (1.83-2.66). Within each instru-
ment category, however, the standard
deviations were remarkably constant.
In each centre, the EURO-D seemed to be
adequately internally consistent, although
W e 3 EURO-D item pmakncet and scale scores from Berlin (GMS and CES-D) and LASA (Longitudinal
Ageing Study Arnsterdan; CES-D). derived according to t h m different methods
Symptom Berlin' Berlin2 Berlin3
1 0 Concentration
~ ~ i n b d d d g d y d n r f r o m h ~ ~
3. Non-integer item rcasr akuhted from the CES-D M
to positive predictive values for GMS items derived
the inter-item and item-total correlations
and the standardised alpha value were
higher for the CES-D EURO-D than for
the GMS EURO-D (Table 5). In most
centres the internal consistency of the scale
could not be improved upon; however, in
Verona, Italy the omission of EUR07 (irrit-
ability), and in Zaragoza, Spain EUR04
(guilt) marginally increased the standardised
The strong associations between the
EURO-D and its parent instruments,
measured as correlations for continuous
measures and as areas under the receiver
operating characteristic curves for the di-
chotomous measures, suggest that the
EURO-D has captured the essence of the
instruments from which it was derived (Ta-
ble 6). Evidence for the validity of compar-
ing EURO-D results derived from different
parent instruments comes from Berlin,
where the GMS EURO-D had correlations
of 0.72 with the CES-D EURO-D and 0.70
with CES-D. Our only evidence for true
criterion validity of EURO-D comes from
LASA, where CES-D EURO-D gave an
area under the curve of 0.93 for the predic-
tion of DSM major depression (Diagnostic
Interview Schedule (DIS); Robins et al,
1981). The optimal cut-off point on the
EURO-D scale for prediction of GMS de-
pression (DN or DP 3 and above) and
SHORT-CARE pervasive depression was
generally 314; the predictive characteristics
for these diagnoses at this cut-off point
are also given in Table 6. In Dublin a 213
cut-off point may have been more appropri-
ate, with a large gain in sensitivity in return
for small falls in specificity and positive
predictive value. For the CES-D centres, a
score of 2.5 or above on the EURO-D
seemed to correspond best to the conven-
tional CES-D cut-off point of 15/16; in
the LASA study this cut-off point was asso-
ciated with 94% sensitivity, 90% specifi-
city and a PPV of 64%.
Principal components analysis (Table 7)
generated two or more factors with eigen-
values over one from each centre. Inspec-
tion of the items loading on these factors
suggested that two factors were common
to nearly every centre. The depression and
tearfulness items, and less consistently the
pessimism and wishing death items, tended
to load on the first factor (which we shall
call depressed affect), while the interest,
concentration and enjoyment items tended
PRINCE E l AL
ToMe 4 Item prevllences and scale score according to centre
SY m~tom W A 1 3 PAQUID'.' Antwerp' Amsterdam London Liverpod Berlin Dublin Munich Iceland Zaragoza4 Verona5 Gothenbug' Ahuri7
I I Enjayment
1 7 21
6 1 9
*. Inlpumd hems.
I. Non-integer hem scores &laced
scams far tha centre.
2. Longitudinal Agdng Study Amsterdam (LASA). Nethatudr.
3. pAQUD (Rrsonnes A*
4. i k a p a . Spin.
5. &ronh ldy.
6 . Gotharbeg. Sweden.
Z Ahtarl. Finland.
from CES-D according m paritiw predictive nknr for GMS hs
derived from Badin dam s e t . Im'pmdewduc in fact thc mean im
ToMe 5 Reliability
Inter-item correlations Item-total
correlation - range value
Alpha Alpha d u e if item
M A , ' Netherlands
0.14 - 0.02-0.46
to load on the second factor (motivation).
Other, relatively distinct factors emerged
in a minority of centres: a somatic factor
(sleep, appetite and fatigue) in Amsterdam,
Iceland, Gothenburg and Ahtari; an irrit-
ability factor in Berlin, London, Iceland
and Verona; and a guilt factor in London,
Zaragoza and Verona. When a two-factor
solution was forced, a recognisable de-
pressed effect and motivation factor was
generated in each centre (Table 7). When
data were pooled from all centres, a scree
plot again favoured a two-factor solution,
the depressed affect factor accounting for
24.6% of the variance in the items, and
the motivation factor for 12.1%. The
depression, tearfulness, sleep and wishing
death items loaded on the depressed effect
factor, with smaller contributions (0.4-
0.5) from pessimism, appetite and fatigue.
The motivation factor was clearly charac-
terised, with large contributions from the
interest, concentration and enjoyment items
but negligible loadings from other items.
0.06-0.47 0.64 0.560.64
For dl theseandps, imputed items hm been exduded.CES-RGnmfor Epklemiologid Studies Dcpradon Sale;
~ b m ~ ~ R u i n g S u l a ; Z S D S . Z u n g S d f - R t o i n g D c p r t t k n S u k ; G M S . C e r k a k
I. LASA, Longitudinal Age~ng Study Amsterdam.
2. PAQUD. bonnes A*
The principal aim of this exercise was to
harmonise data derived from a variety of
EURODEC STUDY 5
Spearman correlation Area under ROC c u m Values at 3 1 4 c u t 4 point
Sensitivity Specificity P W
0 . 9 2 '
0 . 9 3 '
0 . 9 2 '
0 . 9 7 '
0 . 9 6 '
0 . 9 8 '
0 . 8 3 '
0 . 9 3 '
0 . 9 2 '
0 . 7 @
0 . 8 4 5
0 . 9 2 '
0 . 9 3 '
0 . 7 9 1
0 . 9 4 5
0 . 9 5 "
0 . 8 8 5
0 . 8 5 '
0.m 0 . 8 5 ' 8 3
a . CES-D total score.
h SHORT-CARE dtprcsdoo dbgnortic SC& score.
C. EURO-D dKived frwn CES-R
d. Zung total score.
Area under ROC curves:
I. CES-D 216.
2. ryor dsprardon (am).
3. Major daprardon, dysthymia and atypical dtprcsdoo (DSM-Ill dinkd d
4. S H O R T - C r W p c m r i v c ~ .
5. GMS-AGECAT deprrrJon (DN3+
ROC, rsaiva + n g
characteristic; PPV, poritivc predktive dues; LASA, Longitudinal Ageing Study Amsterdam;
PAQWR Penonnu A*
be of questionable validity if the panel's
judgement on calibrating items to GMS
was as flawed as it proved to be for CES-
D. There may be a particular problem with
the ZSDS, as the panel rated five of the 12
items as matching only weakly with their
GMS analogues; however, their judgement
on the adequacy of the match turned out
to be inaccurate for the CES-D. Also, the
standard deviation of the CES-D EURO-
D calculated by this method was much low-
er than for the GMS EURO-D. This was
almost certainly an artefact of the item-
scoring method for the two scales. CES-D
EURO-D used non-integer PPVs ranging
between 0 and 1, but for most items scores
of 0 and 1 were impossible. Thus, the range
of possible total scores was 1.09-7.89
instead of 0-12 for the GMS EURO-D.
There are three potential uses for the
EURO-D in the EURODEP collaboration:
(a) comparison of EURO-D item preva-
lence between centres;
(b) comparison of EURO-D scale distri-
bution between centres; and
(c) comparison of effect sizes for associ-
EURO-D score between centres.
Use (a) is likely to be valid for GMS
centres and CES-D centres, to the extent to
which the pattern of prediction observed in
Berlin pertains elsewhere. Use (b) is again
likely to be valid for GMS centres, and poss-
ibly for CES-D centres. However, even
though the central tendency of the CES-D
EURO-D may accurately reflect that which
would have been measured with the GMS
EURO-D, the artefactual constraint of the
variance of the former will be problematic
when making statistical inferences about
observed differences. One solution might
be to standardise the variance of the scales
in each centre by dividing by the standard
deviation adding the mean and subtracting
meantstandard deviation. This transfonna-
tion would leave the means (and the be-
tween-centres differences in mean) intact,
but each scale would have a standard devia-
tion of one. Use (c) may be valid for all ver-
sions of the EURO-D, particularly if we
focus away from betweencentre main
effects by z-scoring all scales (subtracting
the mean and dividing by the standard de-
viation). Each centre's scale would then have
a mean of zero and a standard deviation of
one. Effect sizes - for example, the differ-
ence between the mean EURO-D scores
for men and women - could then be com-
pared directly between centres, and
risk factors and
depression measures for the purposes of a
multicentre collaborative analysis of puta-
tive risk factors for late-life depression.
For this aim, we can claim a qualified
success. The EURO-D, whether derived
from GMS, SHORT-CARE, CES-D, CPRS
or ZSDS, is an internally consistent scale,
which seems to capture the essence of its
parent instruments, and which has a
common factor structure whatever its origins.
The expert panel gave us a consistent
opinion on the fitting of CES-D, ZSDS
and CPRS items to the GMS items on
which the EURO-D was based. However,
application of the experts' algorithm gave
rise to some unusual item prevalences; for
example, only 3% of LASA subjects said
that they felt depressed or felt that they
could not shake off the blues for at least
3 4 days per week. The item prevalence
for the analogous GMS item "have you
been sad (depressed, miserable, in low
spirits, blue) recently?" ranged between
15% and 58%. Using the alternative meth-
od of assigning to each subject, on the basis
of their CES-D scores, the PPV observed in
Berlin for a GMS item score of 1 resulted in
an estimated item prevalence of 27%. The
item prevalences for sleep, fatigue and
tearfulness calculated from the panel's algo-
rithms were similarly aberrant, while again
those generated using the probabilistic
approach were more consistent with the
pattern observed in other centres. The
EURO-D was not developed to allow
meaningful comparison of item prevalences
between centres using different instru-
ments. Also, the aberrant item prevalences
would tend to even themselves out in the
total EURO-D scale score. However, the
Berlin data suggested that it was the vag-
aries of the panel's algorithm for calibrat-
ing CES-D to GMS, rather than any
genuine betweencentre difference in GMS
item prevalence, which were responsible
for the discrepancies. We felt that such
wide discrepancies were undesirable and
There were two disadvantages asso-
ciated with this decision. First, the method
could only be applied to CES-D, as no cen-
tres had used GMS with ZSDS or CPRS.
The ZSDS and CPRS EURO-D scales may
PRINCE ET AL
Table 7 Factor analysis
Number of factors
with eigenvalue > I
Content d factors not captured in
Factor I Factor 2
Sleep. appetite, fatigue 314
Interest, concentration, enjoyment I13
T d e 7 (continued)
Two-factor solution Number of factors
with eigenvalue > I
Content of factors not captured in
haw1 h a w 2
GMS centres (continued)
4 Irritability 314
(Sleep), fatigue, appetite 214
Guilt, suicidality 414
Guilt, concentration 313
Fatigue. appetite. sleep 314
Guilt, irritability 414
Suicidality, pessimism, guilt 213
Sleep, pessimism 215
Guilt, fatigue 415
R o m r n y p e : ) o a n r ~ u
Icdkr: hs bding at 0.55-0.9.
(Bndmts): lmpmd nrhbbs.
I. USA. ln@mdinrl Ageing Study Amrtardun.
k . k ; C P m
king kale; ZSDS. Zung Sdf-Rating DepnrQn k . k ;
GUS. CarhPk Pknul Strtc.
heterogeneity tested for by fitting centre-by- Download full-text
main-effect interaction tenns.
Discussion of the observed differences
in EURO-D scores between centres is
reserved for the accompanying paper on
the effect of age, gender and marital status
(Prince et al, 1999, this issue).
We are very grateful to the six senior old-age psy-
chiatrists who assisted in the development of the
EURO-D as members ofthe expert pad, advising
us on the optimal matching of items from different
LLmak.n Rychhrk *rod.rbn (1994) Diagmic
ond Stotirticol Manual of Mental Llisdes (4th edn)
(DSM - N). Washington. DC: APA.
CPRS: DHdopment and applications of a psychiatric
rating scde. Acta f'qchbtrica hmdinovico, 271 (suppl.)
~ G - . P , Q U r M . . L 1 . D v c b u a ,
a d (I=) Health measures awrelates in a French
elderly m n i t y population: the PAQUID study.
jwmd ofGemdog~ 47. S88-S95.
~ . I L T . F . . m U ~ J . . - k J . H . a
d (1994) Een screeningsinstrument voor depmsie bij
ouderen in de bevolking: de bruikbaarheid van de
Center for Epidemiologic Studies Depression scale
(CES- D) Ti#dschriP vow Gerontdogre en Geriovie. 25,
and minor depression in later life: a study of prevalence
and aaociated factors.)cinnd of AfPak Llisadn, 36.
J. H.. ~Tllburg,T, a d (1995) Major
self-reports on CES-D and resub from dinostic
interviews. Psychiatry fkemh. IS, 219-229.
H . M (1-6) Gmputaised psychiatric diagnostic
system and case nomenclature for elderly s & e & :
and AGECAT. ~ico
Dmbmn. AT. F., D.*nh M E., a a1 (I-)
Deprruion in Eumpe. Geographical distribution among
older people. British~rnnd ofl'sydkq 174,312-321.
- . - . - . a d (I-) Guss-cultural comparison of
depresshe symptoms in Eumpe does n d support
stereotypes of ageing. British jamd ofl'sydkq 174,
N. (1-9 Depressive symptoms, major
depmssh. and generalized anxiety: a comparison of
l Medicine, 1689-99.
GurM, &.GoMan, R. R..lhd. J. A. a d (1984)
an efti~ient instrument for the
asxswnent ofdepnssion dementia and disability. Jwmol
ofGemtdq)r 39 166-169.
I(M.. 9.-L. hhhh. K. & Wph. R ( 1 -
Prrvaknce of depression in an ekkrly Finnish
population. Acca Psychiotrica Swndinovica, 78,401-413.
Depression symptoms in late-life assessed using
the EURO-D scale. Effect of age, gender and marital
The subset o f GMS items used in the EURO-D scale has
and there is now an abunlnce of population-bused narmotive dPu fmm
depivsh revwity in comporotivs @&nbbgW studies.
T h e m a i n w i u t m o f t h e m - D m o y b e , o s h o s a m m d
The EURO-D seems to be cqable of identifying cppes ddeplrtskn, d d d
d i n g to a variety o f a-iterik with with rdsqurtey and rpedfidy. Hanravcr;
T h e E U R O - m ~ ~ l s s o r s n o t y c r m w s l l d i M u t h o s e o f
existing depression symptom scales validad for use in late-life, tueh er CES-D and
M. J. PRINCE. MD. Epidemiology Unit. London School of Hygiene and Tropical Medicine. London; F . REISCHIES.
Riv. Doz. Dr rned.. Freie Universitt Berlin. Department of Psychiatry, U n i i Klinikum Eschenallee 3.
Berlin, Germany; A. T . F . BEEKMAN, MD. Depart-
of Psychiatry, Wje Univwsitei Amsterdam.
The Netherlands; R. FUHRER, PhD. INSERM U360. Mpital de la Salpetriere. 75652 Paris. Cedex 1 3 , France;
C. JONKER. MD, Faculty XW/LASA.Wje Universiteii. De Boelelaan 1081C. 1081 Amsterdam.
The Netherlands; S.-L. KMLA, MD, Oulu University Hospital, Unit of General Practice PB5000.90401 Oulu.
Finland; B . A. LAWLOR, MRCPsych. Department of Psychiatry ofthe Elde*, St James's Hospital, Dublin.
Republic of Ireland: A. LOBQ Rof. Dr. Department of Rychiatry Hospital Clinico Universitario. Zaragoza,
Spain; H. MAGNUSSON, Prof. Dr. Heilsugaeslustod Grundarfjardar. Iceland; M. FICHTER. Pmf. Dr. Department
of Psychiatry. University of Munich. Munich, Germany; H.VAN UYEN. MD DrPH. Department of Epidemiology,
Scientific Institute of Public Health. J.Wytsmanstraat 4. 1050 Brussels, Belgium; M. ROELANDS, MD.
Department of BehaviourTherapy and Counselling. University of Gent, 8-9000 Gent. Belgium; I. SKOOG, MR
Fsykiatriska Kliniken, Sahlgrenska sjukhuset. 413 45 Goteborg. !jweden; C. TURRINA. MD, Clinica Richiica.
Ospendale C'vile. Bmchia. Italy; J. R. M. COPELAND, FRCRych. Department of Psychiiry, Institute of Human
Ageing, University of LNerpod. Liverpod
Correspondence: M. J. Prince. Epidemiology Unit. London School of Hygiene and Tropical Medicine.
Keppel Street. LondonWClE 7HT
(First received I April 1998. final revision 1 9 OctDber 1998. accepted 6 November 1998)
status in 1 4 Eumpean centres. British~ofpsvChiatry
RadaH, L S. (1977) The CES-D scale: a self repwt
depression scale for research in the general popJation
A#iM Rychdogicol Measurement. I, 385-401.
National Institute of Mental Health Diagnostic Interview
Schedule: its h i m , characteristics and validity. Archives
of Geneml Ps@ioa)c 38 381 -389.
m - .
prevalence of dementia in Belgium: a population-based
door-to-door survey in a rural mmunity.
Neumepdenwbgfi 1s 155161.
P.. Dan. H.. d (1994)
d i m
and the used psychotropic drugs in an 85-
year-old urban population. Internotional Psychcgerotncs.
r - k d
htem&ond Diognostic htenkv (CIDI. version 1.0).
(1993) The b t h Rmsion ofthe lncwnotiond
Umfiition ofDiseass ond Rdotedhlth Robkms
(KD-10): Diognoscic Giterio @r Research. Genm WHO
Zun&VYW. K. ( 1 w A self-rating & @ o n
Archives of Geneml Psychiav)c 12.62-70.