10.1192/bjp.150.6.782Access the most recent version at doi:
1987, 150:782-786. BJP
J L Cox, J M Holden and R Sagovsky
Edinburgh Postnatal Depression Scale.
Detection of postnatal depression. Development of the 10-item
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 July 29, 2011 http://bjp.rcpsych.org/
go to: The British Journal of PsychiatryTo subscribe to
British Journal of Psychiatry (1987), 150, 782â€”786
Detection of Postnatal Depression
Development of the 10-item Edinburgh Postnatal Depression Scale
J. L. COX, J. M. HOLDENand R. SAGOVSKY
Thedevelopmentof a 10-item self-reportscale(EPDS)to screenfor PostnatalDepression
in the community is described.After extensive pilot interviews a validationstudy was
carriedout on 84 mothersusingthe ResearchDiagnosticCriteriafor depressiveillness
obtainedfrom Goldberg'sStandardisedPsychiatricInterview. The EPDSwas found to
of depressionovertime. The scalecanbecompletedinabout 5 minutesandhasa simple
method of scoring. The use of the EPDS
in the secondary prevention of Postnatal
In the last decade several studies (Paykel eta!, 1980;
Cox et a!, 1982; Kumar & Robson, 1984; O'Hara
et a!, 1984; Watsonet a!, 1984) have provided
Pitt (1968) that the months following childbirth are
frequently characterised by psychiatric disorder, and
that at least 10-15% of mothers experience a marked
depressive illness at this time.
Furthermore,the results of a follow-up study
accurately recalled by the mothers 3years later; and
that at least half of the depressed mothers had not
recovered by the end of the first postpartum year
(Cox et a!, 1984). The finding that the children of
such depressed mothers
disturbance at 3 years (Wrate eta!, 1985) or cognitive
defects at 4 years (Cogill et a!, 1986) suggests that
postnatal depression may have a long-term negative
impact on the family.
In our earlier study (Cox et a!, 1982), we found
that 13of the 101women interviewed had a marked
post-natal depressive illness and yet the majority of
these depressed mothers
sustained treatment from their primary care workers
nor had they been referred to a psychiatrist; the three
women who had been referred were not those who
were depressed. This failure to identify depression
in the puerperium, especially when such mothers
were usually known to their GP,
Midwife or Health Visitor, was obviously a cause for
much clinical concern.
A further difficulty
mothers is that screening scales for depression appear
to have a number of limitations when used on
childbearingwomen. In our earlier study from
Edinburgh, for example, the Anxiety and Depression
for the earlier fmding of
had not received any
Scale (SAD) of Bedford & Foulds (1978) was found
to have uncertain validity; of the 13pregnant women
who scored 6+ (Foulds'
illness) only three had any form of psychiatric
disorder, while four had minor symptoms only and
six had no psychiatricillness
1983). The 30-item General Health Questionnaire
Goldberg et a! (1970) has been assessed by Nott &
Cutts (1982) for possible use in the puerperium and
was alsofound to requiresome
These authors reported
puerperal women scored highly on the 30-item GHQ,
but only 37(18Â°/a) were found to be psychiatric cases.
Similar difficulties with the Beck Depression Inventory
in the puerperium were reported by O'Hara et a!
(1983). In this study only 11 of the 19 women who
scored above the cut-off score fulfilled Research
Diagnostic Criteria (RDC) for Depression, and of
the 23 who scored below the cut off there were four
false negatives. Furthermore,
depression as measured by the Beck Scale were not
confirmed when a clinical syndrome diagnosis of
depression was made using Research Diagnostic
Criteria suggests that studies using this scale as the
only measure of depression following childbirth may
give misleading results (O'Hara et a!, 1984).
The possible explanation
limitations of well established scales when used on
childbearing women include their emphasis on the
somatic symptoms of psychiatric disorder which
may be caused by normal physiological changes
associated with childbearing, as wellas the reluctance
of community workers to use questionnaires which
may be regardedas time-consuming
threshold for personal
whatsoever (Cox eta!,
89 (45Â°/a)of 200
for these apparent
appear to lack face validity. These limitations may
also be relevant to a consideration
To be useful as a screening test for depression
following childbirth, therefore, a self-report scale
must be fully acceptable to women who may not
regard themselves as unwell, or as in need of medical
help. The scale needs also to be simple to complete,
and not require the health worker to have any
specialist knowledge of psychiatry. It must have
satisfactory reliability and validity.
appropriately emphasised that rating scales which
had been validated on hospital samples must be
revalidated if they are used in community popula
tions where the differences between psychiatric illness
and normality is often less distinct. The earlier work
of Snaith (1981, 1983) in this regard was also
important, as he clearly recognised the need to
modify existing scales of depression for use in new
specific clinical situations and in particular was aware
of the need to develop a screening questionnaire to
detect postnatal depression. Others, such as Kumar
(1982)and Cox eta! (1983),had also emphasised that
this task was an important
Spurred on by these observations
as well as by the pressing
workersto have practical
natal depression, we therefore
depressed following childbirth.
of the Zung
1965) for use in the
eta! (1980) have
current research priority.
of our colleagues
need for primary
help in identifying
decided to develop a
the Hospital Anxiety and Depression Scale (HAD) (Ziginond
& Snaith, 1983), and theAnxietyand Depression Scale of
selected 21 items, including several of our own construction,
which we thoughttobe appropriate forthedetection of
postnatal depression.These items were then tested during
extensive pilot interviews with mothers of young babies.
The detailed wording of items, their acceptabilityto mothers
and healthworkers,aswellastheirlikelihoodof detecting
postnatal depressionwas then carefully evaluated.
items were eventually selected as being those most likely
to detectpostnatal depression;sevenof thesewereitems
constructedby ourselvesand the other six were adapted
from the IDA and the HAD.
The validityof this 13-itemscalewasthen established
on a sampleof 63purperalwomenwhoattendeda health
centre in Livingston (seeCox, 1986for details). This study
showedthat these 13itemsdistinguishedclearlybetween
analysis of the suitability
of the Irritability,
Scale (IDA) (Snaith et a!, 1978),
from items the
suggestedthatthespecificity of thescalemightbeincreased
by omittingthesethree items, wedecidedto carry out a
further validationstudy using only the 10 items which were
more clearly related to depression.
also have the advantage of taking less time to complete.
The validation study of the 10-itemEPDS to be reported
in this paper was carried out on 84 mothers living in
Edinburgh or at Livingston new town. Most of the mothers,
who were taking part in a study to determine the
effectiveness of counselling by health visitors in the
treatmentof postnataldepression,had beenidentifiedby
theirhealthvisitorsat about 6 weeksfollowing deliveryas
beingpotentiallydepressed.The healthvisitorshad been
asked to indicate whether, in their opinion, these mothers
wereâ€˜¿?normal',â€˜¿?depressed',or wereconsideredas having
â€˜¿?problems'.As we envisagedthat a useful function of the
scale would be to confirm the diagnosis
being possibly depressed, this sample was particularly
whether the scale would satisfactorily identify postnatal
optimum settingin whichto validatethe 10-itemscale;home
visits by health visitors being regarded as an important
betweenthe assessmentof puerperal mothers and other
members of the primary care team. Mothers in our sample
were interviewedby R.S. using Goldberg's Standardised
majority of such interviews took place in the mothers own
home (SPI-l). At this home visit the EPDS was first
completedby the mother and was then placed in a sealed
envelopesothattheinterviewer remainedblindto thescore
possible bias effect caused by the interviewer knowing that
the subjectmayhavebeenregardedby the health visitor
as being â€˜¿?depressed' or as having problems, 12normal
for the diagnosisof a depressiveillnesswerethe Research
DiagnosticCriteria of Spitzer
wereobservedto havea depressedmoodbut whodid not
meet full RDC criteria for depression
separatelyidentified. As recruitmentof subjects into the
studywas slowerthanexpected, a further 12women were
interviewedby J.C. at a localhealth clinic.Both interviewers
werejointlydiscussed.Thevalidationof the 10-itemEPDS
was determined for the total sample by comparing the
EPDS scoreswith the RDC clinicaldiagnosis of depression.
was often identified
of the IDA,
as a separate
As this analysis
with an item
of our data had
This 10-item scale would
eta!,1970) and the
the SPI. To prevent any
et a! (1975). Motherswho
were, however, also
The mean age of womenwas26years, and that of their
in some clinical or research settings for actual cases of
depression not be missed,our data suggest that the failed
detection of cases can be reduced to under 10% with a cut
off score of 9/10.
Whenanalysisof our data wascarriedout ononlythose
women (n=60) interviewedby R.S. (excludingthe 12
womenwithno previouslyidentifiedproblems,as wellas
the 12subjectsinterviewedby J.C. at the Health Centre)
larger sample, sensitivity 85Â°/a, specificity 77%, the positive
predictivevalue having increasedto 83Â°/a. The split-half
reliability of the scale was found to be 0.88, and the
Sensitivityto changein the severityof depressionover
EPDSscoreat the firstinterview(EPDS-l), whenmothers
obtainedat the li-weekfollow-upinterview.Atthissecond
home interview theEPDS was completedfora secondtime
(EPDS-2),and a repeat SPI (SPI-2)wascarried out, the
interviewer againremaining blind
depressed according to RDC criteria at both interviews
(n=15), showed no significantdifference between their
EPDS-1 (16.5) and EPDS-2 (15.38) mean scores on these
two occasions, whereasmothers who were depressedat
Interview 1 but not at Interview
of scorebetweenEPDS-l and EPDS-2whichwashighly
significant. (EPDS-i mean score= 15.8, EPDS-2 mean
score=9.8, t= 3.72, P= 0.002).The EPDS-2scorein all
but one subject fellto belowthe threshold of 12/13;the
mother whoseEPDS-2scoreincreased,but who wasnot
depressed,had a probable cancer of the cervixand was
diagnosedas havingan anxietyneurosis.
Analysisof thepossibleinfluenceonthe EPDSscoreof
another familymemberbeingpresentwhenthe scalewas
completed suggested that under these circumstances
tended either to exaggerate,or to minimise, their psychiatric
problems.Thus three subjectswho had the highestâ€˜¿?false
positive'score,andthreeof the four â€˜¿?false
not been alone whenthey wereinterviewed.
to boththe EPDS-land
2 (n = 16) had a reduction
COX ET AL
deliveries,15%Caesariansectionsand a further 10,1.had
13Â°!.had a permanentpartner.
parents. Social class distribution
or partner's occupation where one waspresent, or according
to the mother's previousoccupationin the caseof single
parents) was as follows: Social Class 11:7'!., 111:35Â°!.,
Only 6Â°/awere single
(according to husband's,
Validation of the 10-Item EPDS
The results of the validation of the 10-itemEPDS are shown
in Fig. 1.
A thresholdscoreof 12/13wasfound to identifyall of
the 21womenwith an RDC diagnosisof DefiniteMajor
Depressive Illness and two of the three women with
Probable Major DepressivelUness.Four of the 11women
with DefiniteMinor Depressionwerefalsenegatives,i.e.
although six of these 11 women had depressive
but did not meet full Research DiagnosticCriteria for
depression. The subject with the highest false positive score
(21) had a marked personalitydisorder; whilethe three
women with a psychiatric diagnoses other than depression
The sensitivityof the EPDS, the proportion of RDC
depressed women (n= 35)who were true positives (n= 30),
was 86Â°/a; the specificity, proportion of non-depressed
The positivepredictivevalue, the proportion of women
above threshold on the EPDS (n=41) who met RDC criteria
for depression (n=30), was 73Â°/a. As it is important
Our study has shown that the 10-item Postnatal
Depression Scale, which was derived from the earlier
work of Snaith, had satisfactory validity, split-half
reliability and was also sensitive to changes in the
severity of depression over time. Furthermore,
found that the scale was fully acceptable to the child
bearing women and was usually completed within
5 minutes. The simple method of scoring was an
advantage and the health visitors recognised that the
scale would greatly assist them in the detection of
mothers who were depressed postpartum.
We believe it to be a substantial advantage that
this validation study of the 10-item scale was carried
out in a community setting
and on women who were
No,ud other DepressedMvior
FIG.1 Validationof EdinburghPostnatal DepressionScale.
EDINBURGH POSTNATAL DEPRESSION SCALE
as close as possible to mothers regarded by their
primary care workers as having problems. Our data
nevertheless suggested that sensitivity and specificity
of the scale may be increased if it is completed when
other family members are not present.
It seems likely that the scale will be useful
in the routine work of community health workers
(e.g. health visitors, community psychiatric nurses
and General Practitioners)
postnatal depression in mothers thought by their
health worker to be at risk. It may also be of
use in treatment studies of postnatal depression when
carried out on mothers living in the community.
Our data suggested that women who scored
above a threshold of 12/13 were most likely to be
suffering from a depressive illnessof varying severity,
and should therefore be further assessed by the
primary care worker to confirm
clinical depression is present. The EPDS is not
a substitutefor this clinical assessment, and a
score just below the cut-off should not be taken
to indicate the absence of depression, especially
if the health professional
consider this diagnosis. Our data also suggest that
a threshold of 9/10 might be appropriate if the scale
was considered for routine use by primary care
We now plan to validate the scale for possible
use during pregnancy and also to determine its
usefulness in other populations. The scale could, for
example, be administered
mother's visit to an antenatal or postnatal clinic,
and it may be useful as a more general screening
scale for depressive illness. The revalidation
the scale for use in these
must be carried out, however, before this wider
use is recommended.
and assist to identify
has other reasons to
by a computerat a
We are indebted to the many GPs and Health Visitors at
Livingstone and Edinburgh who collaborated with us. We thank
Mr R. J. McGuire for his statistical advice and for encouragement
at various stages of the study. The research was generously
supported by a grant from the Scottish Home and Health
Banroan, A. & Fouws, 0. (1978)Dehisions-Symptoms-States.State
of Anxiety anti Depression(Manual) Windsor: National
Bacic,A. T., WAJW,C. H. & MENDEISON, M. ci a! (1961)An
inventory for measuring depression. Archives of General
Psychiatry, 4, 561â€”571.
COGILL,S. R., CAPLAN.H. L., ALEXANDRA. H., ROBSON,K. M.
& Ku@t@a, R.(1986) Impact ofmaternal postnatal
on cognitive development of young children. British Medical
Journal, 292, 1165â€”1167.
Cox,J. L. (1986) Postnatal Depression
Professionals. Edinburgh: Churchill Livingstone.
â€”¿? A Guide forHealth
â€”¿?,Co@oR, Y.& KENDELL, R. E. (1982)Prospective
of the psychiatric disorders of childbirth.
Psychiatry, 140, 111â€”117.
â€”¿?,â€”¿?, HENDERSON,1., Mc0urna,R. J.& KENDELL,R. E.
(1983)Prospective study of the psychiatric disorders of childbirth
by self report questionnaire. Journal of Affective Disorders, 5,
â€”¿?, ROONEY,A., THOMAS, P.F.& W@m,R. W. (1984)How
accurately do mothers recall postnatal depression? Further data
from a 3 year follow-up study. Journal of Psychosomatic
Obstetrics and Gynaecology, 3, 185â€”189.
GOLDBERG,D. P. (1972) The Detection
Questionnaire. Maudsley Monograph,
21, Oxford: Oxford
â€”¿?. COOPER, B..
use in community surveys. British Journal of Preventive and
Social Medicine, 24, 18023.
Kui@u@a. R. (1982)Neuroticdisordersin childbearingwomen.In
Motherhood and Mental Illness, eds I. Brockington &R. Kumar.
â€”¿?& ROBSON,K. M. (1984)Aprospective studyof emotional
disorders in childbearing women. British Journal of Psychiatry,
Noi-r, P. N. & Cui-rs, 5. (1982)Validation of the 30-item General
Health Questionnaire in postpartum
O'H@a@,M. W., REHM,L. P. & CAMPBELL, S. B. (1983)
Postpartum depression:a roleforsocial networkand life
stress variables. Journal ofNervous andMental Disease,
postpartum prevalence, and
93, of AbnormalPsychology,
â€”¿?,E@tt@s, E. M.,FL@rcnea,J.&RASSABY,E.5.(1980)Life
events and social support in puerperal depression. British Journal
of Psychiatry, 136. 339-346.
Pirr, B. (1968) â€˜¿?Atypical'depression following childbirth. British
Journal of Psychiatry, 114, 1325â€”1335.
SNAITh.R. P. (1981) Rating scales. British Journal of Psychiatry,
â€”¿?(1983) Pregnancy-related psychiatric disorder.BritishJournal
of Hospital Medicine, 29, 450â€”456.
(1978) A clinical scale for the self-assessment of irritability.
British Journal of Psychiatry, 132, 164â€”171.
SPITZER. R.,ENDIcorr,J. & ROBINS, E. (1975)
Criteria.Instrument No. 58. New York: New York State
(1984) Psychiatric disorder in pregnancy and the first postnatal
year. British Journal of Psychiatry, 144, 453-462.
WILuAMS, P.,T@apioi'oisjcy, A. & H@,
assessment.Psychological Medicine,10, 101â€”114.
WSAm, R. M., ROONEY, A. C., THOMAS, P. F. & Cox, J. L. (1985)
Postnatal depression and child development: a three year follow
up study. British Journal of Psychiatry, 146, 622-627.
ZIGMOND, A. S. &R. P. (1983) The Hospital
Depression Scale. Acta Psychiatrica Scandinavica, 67, 361-370.
ZUNG.W. W. K. (1965)A self-ratingdepression scale.Archives
of General Psychiatry, 12, 63.
A. A., JARDINE,M. Y.& MCGUFFIN,P.
Ruoc,A.J.& BROUGH, D.I.
D. (1980) Casedefinition