Journal of Affective Disorders 173 (2015) 239–244

Contents lists available at ScienceDirect

Journal of Affective Disorders

journal homepage: www.elsevier.com/locate/jad

Research Report

Predictors of post-natal depression are shaped distinctly by the

measure of ‘depression’
Gordon B. Parker a,b,n, Bronwyn Hegarty a,b, Amelia Paterson a,b, Dusan Hadzi-Pavlovic a,b,
Isabelle Granville-Smith a,b, Aniela Gokiert a,b
a
School of Psychiatry, University of New South Wales, Australia
b
Black Dog Institute, Australia

art ic l e i nf o a b s t r a c t

Article history: Background: Many variables have been proposed as predictive of post-natal depression (PND).
Received 2 May 2014 Aims: To investigate and refine PND risk variables.
Received in revised form Method: We recruited a large sample and employed two measures of PND (the dimensional Edinburgh
29 October 2014
Postnatal Depression Scale or EPDS, and DSM-defined major depression).
Accepted 30 October 2014
Available online 20 November 2014
Results: High levels of stress in the post-natal period, previous depression and higher depression scores
during pregnancy were the only consistent predictors across measures. Those exceeding the EPDS cut-off
Keywords: had additional psychosocial risk factors while those meeting criteria for major depression were strongly
Post-natal depression predicted by a past history of depression as well as higher pre-natal state depression scores.
Depression
Limitations: The EPDS has been used with variable cut off scores across multiple studies. We used only
Measure
nine of the 10 EPDS items, electing to exclude the self-harm related question, but preserving the
recommended EPDS cut-off score, and which might have impacted on predictions.
Conclusions: Study results generated a refined set of predictors of PND but, more importantly, identified
that predictors of PND status are distinctly influenced by the measure of PND. Such inconsistencies are
intrinsically noteworthy and of potential key importance in shaping intervention strategies.
& 2014 Elsevier B.V. All rights reserved.

1. Introduction
Predicting those likely to develop post-natal depression (PND)
is an important public health and clinical priority, with the
condition quantified as affecting 10–15% of mothers in interna-
tional reviews (Markhus et al., 2013) and 8–12% in Australian
studies of respective non-indigenous and indigenous women
(Bowen et al., 2014). Post-natal depression is commonly diagnosed
using DSM criteria with the specifier “with peripartum onset,” or
by use of specifically designed dimensional measures. Our cur-
rently reported study used DSM-IV criteria as that manual was
current at the time of study design. Those criteria include either a
depressed mood and/or a loss of interest or pleasure, with at least
three other symptoms (four if only one of depressed mood/loss of
pleasure is endorsed) including weight change, sleep alterations,
psychomotor agitation or retardation, fatigue, worthlessness or
guilt, diminished ability to think or concentrate and recurrent
n
Corresponding author at: School of Psychiatry, University of New South Wales,
Australia. Tel.: þ61 2 9382 4372.
E-mail address: g.parker@unsw.edu.au (G.B. Parker).
thoughts of death for two weeks or more. The Edinburgh Postnatal
Depression Rating Scale or EPDS (Cox et al., 1987) is the most
commonly used dimensional measure of PND and is a 10-item
self-report measure containing items such as “I have blamed
myself unnecessarily” and “I have been anxious or worried for
no good reason”.
The current study sought to refine a range of previously
identified and suggested risk factors for PND. Candidate risk
factors were derived by reference to several key sources. First, a
systematic review of published studies reviewed by the Australian
National Health and Medical Research Council (NHMRC) (Pope
et al., 2000), which quantified PND risk factors by their consis-
tency of identification across studies. ‘Confirmed’ factors (defined
as being identified in 75% of studies) included a personal history of
depression, depression during the pregnancy, marital difficulties,
lack of support and stressful life events. ‘Probable’ factors (i.e.
identified in 40–60% of studies) included a family history of
psychopathology, single parenthood, severe maternity blues, per-
sonality nuances (e.g. neuroticism, interpersonal sensitivity),
negative cognitive style, birth experiences and obstetric complica-
tions, partner's level of depression, as well as the infant's health
and temperament. ‘Possible’ risk factors, in the sense of providing

http://dx.doi.org/10.1016/j.jad.2014.10.066
0165-0327/& 2014 Elsevier B.V. All rights reserved.
240 G.B. Parker et al. / Journal of Affective Disorders 173 (2015) 239–244
little evidential support, included thyroid dysfunction, hormonal
changes, early discharge from hospital, premature delivery, breast-
feeding, poor relationship with parents, maternal age and parity.
The authors noted that the status of risk factors (i.e. ‘confirmed’,
‘probable’ and ‘possible’) was influenced by the actual measure of
PND, an issue central to our study. A more recent national guide-
line published by Healthcare Improvement Scotland (Scottish
Intercollegiate Guidelines Network (SIGN), 2012) nominated most
of the variables listed in the NHMRC report as risk factors but also
listed domestic violence, unplanned pregnancy, unemployment,
inability to breast feed, longer time to conception and having two
or more children.
We elected to assess the most commonly nominated factors
within those summary documents. While many studies have been
undertaken to identify PND risk factors, our study had three key
advantages – a large sample, a substantive set of potential
predictors and two principal outcome measures of PND – presence
of a DSM-IV-defined episode of major depression and rating
positively above a defined cut-off score on the EPDS (Cox et al.,
1987). We report analyses quantifying low consistency of predic-
tors across the two measures. Although many studies have used
both the EPDS and DSM criteria, to our knowledge only one study
(Yonkers et al., 2001) has directly examined the predictors of a
high EPDS score and of a DSM-IV diagnosis of major depression. In
that study, an EPDS diagnosis of depression was more likely for
more highly educated women and less likely for women who were
breastfeeding. By contrast, a DSM diagnosis of major depression
was predicted by higher previous scores on the EPDS and the
Inventory of Depressive Symptoms (Rush et al., 1986), and living at
home with an extended family. Interestingly, although the pre-
dictors of PND have not been compared using self-report and
interview methods, it is recognized that interview methods result
in much lower levels of reported depression (O’Hara and Swain,
1996).
2. Materials and method
2.1. Participants
We recruited women between 34 and 37 weeks of their
pregnancy from obstetric units based in one large Sydney hospital
and one coastal hospital north of Sydney, with their contrasting
regional profiles ensuring a broad socioeconomic range of parti-
cipants. Inclusion criteria were: age over 18 years, proficiency in
English and the ability to provide informed consent. Potentially
eligible women were invited by midwives or research assistants to
take part in the study and detailed study hypotheses and compo-
nents. We did not record the number of women approached or any
reasons for declining. The study was formally approved by the
Sydney South West Area Health Service Human Research Ethics
Committee and ratified by the University of New South Wales
Human Research Ethics Committee.
2.2. Measures
Socio-economic variables were assessed via a self-report ques-
tionnaire. These included age, education level (7 categories from
primary school to postgraduate degree), current employment
(8 categories including full-time, part-time, student, home duties
and receiving benefits) as well as the number of hours worked,
marital status (6 categories) and family income (3 categories,
less than $750 per week, $751–1596 per week and more than
$1596 per week). A medical history was also assessed via the self-
report questionnaire, including the number of prior children,
previous stillbirths (yes/no), previous terminations (yes/no), previous
miscarriages (yes/no), thyroid difficulties (yes/no), food allergies
(yes/no) and a history of premenstrual syndrome (yes/no and their
level of impairment from 1 ¼ no impairment to 5 ¼ severely
impaired). Food consumption habits were also assessed, and
included coffee consumption (cups per day), alcohol consumption
(standard drinks per week), smoking status (yes/no) and cigarettes
per day, as well as any illicit drug use (yes/no). Mood history was
assessed by questioning if the women had ever had a severe,
impairing depressive episode lasting for two weeks or more (yes/
no) and whether it was unipolar or bipolar. Levels of stress during
pregnancy were rated by the participant from 1 (no stress) to 10
(severely stressed). Medication use was assessed by asking if
antidepressants or benzodiazepines were currently or previously
used (yes/no).
The subject's levels of introversion and neuroticism was quan-
tified by subscales of the self-report Temperament and Personality
Scale (Parker et al., 2006). The impact of life events was assessed
by 47 items of the 79-item Life Events Questionnaire (LEQ)
(Norbeck, 1984; Sarason et al., 1978) – with irrelevant or inap-
propriate items deleted (e.g. pregnancy and addition of a new
family member). Participants were asked to record whether they
experienced any of the events in the previous year and, if so,
whether each event was essentially positive or negative (and
whether it had ‘no’, ‘some’, ‘moderate’ or a ‘great’ effect; being
respectively scored 0, 1, 2 or 3). Data were analyzed on the
separate positive and negative impact scores. Mood levels were
assessed using the Costello–Comrey Anxiety scale (Costello and
Comrey, 1967) as well as the Depression, Anxiety and Stress (DASS)
(Lovibond and Lovibond, 1995) state depression and anxiety sub-
scales (for the previous week). The quality of interpersonal
relationships was assessed using the Interpersonal Relationship
Inventory (IPRI) (Tilden et al., 1994) which generates ‘supportive
relationship’ and ‘conflictual relationship’ scores in relation to the
individual’s support figures (whether partner, family members or
others).
We had two principal outcome measures of depression. Firstly,
the first nine items of the ten-item Edinburgh Postnatal Depres-
sion Scale or EPDS (Cox et al., 1987). We elected to exclude the
final self-harming question as those administering the EPDS were
untrained in mental health crisis management and not necessarily
able to provide the necessary services to women who might have
affirmed this item. We imposed a cut-off score (then and post-
natally) of 10 or more, reflecting judgments made by Matthey et al.
(2006) that not only was this the cut-off score recommended by its
developers (albeit for the 10-item measure) but had also been
confirmed in other studies, while additionally optimizing sensi-
tivity. Our second measure assessed whether they met DSM-IV
criteria for major depression, assessed via the MINI International
Neuropsychiatric Interview (Sheehan et al., 1998).
At follow up, food consumption, the quality of interpersonal
relationships and outcome measures were again assessed. Addi-
tional information as to whether the baby had any settling
problems (yes/no) was assessed during a telephone interview
and participants were asked if they had commenced taking
antidepressants (yes/no).
2.3. Baseline assessment
Participants were asked to complete a number of question-
naires, as detailed above and all participants were assessed for
mood using both the EPDS and MINI via telephone interviews.
2.4. Follow-up assessments
The follow-up questionnaire was completed by study partici-
pants three months post-natally. Respondents were requested to
 G.B. Parker et al. / Journal of Affective Disorders 173 (2015) 239–244
complete the EPDS in relation to their functioning over the
previous week, with those scoring 10 or more (EPDS cases)
telephoned and a MINI interview undertaken to determine if they
met DSM-IV criteria for major depression.
A supplementary telephone interview was undertaken 6–8
months (mean ¼225.6, SD ¼125.9 days) later for all participants
– with the interviewer asking mothers if (i) they had experienced
any period of feeling “depressed or down” for more than a few
days in the first three months after their child's birth or (ii) if they
had commenced taking antidepressants over that post-natal
interval. If either was affirmed, the MINI was administered to
determine if their depressive symptoms met DSM-IV criteria for a
major depressive disorder. MINI ‘cases’ were those who met DSM
criteria at either review in the three-month post-natal period.
2.5. Statistical analyses
We principally report data analyses in relation to our two
primary outcome measures of PND. We also examine predictors of
incident ‘cases’ (i.e. those who did not meet case criteria at the
baseline pre-natal assessment but were cases at the post-natal
reviews), and also examine predictors of depression status on a
composite outcome measure (MINI diagnosed cases and/or taking
an antidepressant).
3. Results
3.1. Sample features
We enrolled 1232 women (577 from the Sydney and 655 from
the Central Coast regions), but the sample number was reduced to
756 (as detailed in Fig. 1) as a consequence of many not taking part
in all data collection procedures.
At follow up, 122 women did not provide an EPDS score. They
were significantly more depressed (mean ¼7.19, SD ¼5.35) than
those who did provide a follow up EPDS (mean ¼ 5.64, SD¼ 4.07,
t¼ -3.72, p o0.001).
3.2. Depression status at pre-natal baseline assessment
At 36 weeks of pregnancy, 24 (3.2%) were MINI cases in
meeting DSM-IV criteria for a current major depressive episode
while 140 (18.5%) rated as positive on the EPDS. No women
meeting MINI criteria for depression at baseline scored o10 on
the EPDS.
Fig. 1. Flow chart.
241
3.3. Post-natal depression status
Numbers and percentages of participants diagnosed as depressed
using each measure as well as whether they were new ‘incident’
cases of depression or had had their depression persist since baseline
assessment are reported in Table 1. EPDS ‘case’ rates were distinctly
higher than DSM-derived case rates. Only 42 (5.6%) were assigned as
‘cases’ by both measures (agreement in assigning cases was 84.0%,
reflecting the high percentages of non-cases assigned by each
measure, while the kappa was 0.36). The rate of new incident case
status in the post-natal period was also higher when assessed by the
MINI than by the EPDS. At follow up, three women were diagnosed
as depressed using the MINI who scored o10 on the EPDS.
3.4. Predictors of PND
After undertaking univariate analyses, statistically significantly
differing variables were analyzed in multivariate sets according to
the variable ‘type’ or domain. Thus, the ‘socio-demographic/pre-
morbid’ variable set comprised age, any previous miscarriage(s),
any termination, severity of any premenstrual symptoms, any
lifetime mood disorder, and neuroticism, introversion and Cost-
ello–Comrey trait anxiety scale scores. Two variables predicted
MINI case status – previous termination and a past episode of
depression. While there were three predictors of EPDS case status,
only one (previous termination) was consistent across both out-
come measures, while severity of premenstrual symptoms and
higher neuroticism scores were also EPDS case predictors. For
those who were both MINI and EPDS cases there were three
predictors – previous termination, a higher neuroticism score and
lifetime mood disorder. Odds ratios and Wald statistics for
predictive variables are reported in Table 2.
Our pregnancy-related data set domain included EPDS score at
baseline, negative life event scores, baseline ‘supportive’ and
‘conflictual’ relationship scores, baseline state DASS anxiety and
Table 1
Post-natal assessment of depression by the MINI measure of DSM major depression
and EPDS.
MINI n (%) EPDS n (%)
MINI cases 58 (7.7) EPDS cases 138 (18.3)
Baseline cases 13 (54.2) Baseline cases 58 (41.4)
New MINI cases 45 (77.6) New EPDS cases 80 (58)
242 G.B. Parker et al. / Journal of Affective Disorders 173 (2015) 239–244

Table 2
Post-natal depression predictive variables.

MINI depression predictors EPDS depression predictors MINI and EPDS predictors

Variable OR Wald P 95% CI Variable OR Wald p 95% CI Variable OR Wald p 95% CI

Sociodemographic variables
Previous termination 1.96 4.8 0.028 1.07– Previous termination 1.70 5.1 0.024 1.07– Previous termination 2.83 8.8 0.003 1.42–
3.59 2.71 5.64
Previous depression 2.61 9.3 0.002 1.41– Severity of PMS 1.28 4.1 0.043 1.01– Higher neuroticism 1.11 7.5 0.006 1.03–
4.84 1.62 1.20
Higher neuroticism 1.15 31.0 o 0.001 1.09– Previous depression 2.15 4.2 0.041 1.03–
1.20 4.47

Pregnancy related variables

DASS depression at 1.17 15.3 o 0.001 1.08– DASS depression at 1.08 5.2 0.023 1.01– DASS depression at 1.22 18.9 o0.001 1.11–
baseline 1.26 baseline 1.15 baseline 1.33
Higher stress during 1.15 3.8 0.05 1.00– Higher baseline EPDS 1.16 17.9 o 0.001 1.08–
pregnancy 1.32 1.25
Higher baseline 1.03 4.5 0.034 1.00–
conflict 1.06

Birth and postpartum variables

Higher stress since birth 3.18 21.5 o 0.001 1.95– Higher stress since 4.13 49.3 o 0.001 2.78– Higher stress since 4.86 26.4 o0.001 2.66–
5.18 birth 6.13 birth 8.88
Post-partum support 0.92 17.7 o 0.001 0.88– Post-partum support 0.94 14.2 o 0.001 0.91– Post-partum support 0.91 14.4 o0.001 0.86–
0.95 0.97 0.96
Post-partum conflict 1.04 5.5 0.018 1.01– Post-partum conflict 1.12 55.6 o 0.001 1.08– Post-partum conflict 1.06 7.8 0.005 1.02–
1.08 1.15 1.10

Combined variables
Higher stress since birth 3.25 20.9 o 0.001 1.96– Higher stress since 3.62 42.7 o 0.001 2.46– Higher stress since 4.83 23.2 o0.001 2.54–
5.38 birth 5.32 birth 9.16
DASS depression at 1.14 13.1 o 0.001 1.06– Higher neuroticism 1.10 15.2 o 0.001 1.05– DASS depression at 1.18 14.8 o0.001 1.08–
baseline 1.22 1.16 baseline 1.28
Previous depression 3.23 11.48 0.001 1.64– Post-partum conflict 1.10 37.4 o 0.001 1.07– Previous depression 2.64 5.0 0.025 1.13–
6.37 1.15 6.16
Post-partum support 0.93 8.8 0.003 0.89– Baseline conflict 0.97 5.9 0.015 0.94–
0.98 0.99
Post-partum support 0.94 9.1 0.003 0.91–
0.98

depression scale scores, pregnancy stress levels and coffee intake.
MINI case status was predicted only by higher state DASS depres-
sion levels at baseline and higher levels of stress during pregnancy.
EPDS case status was also predicted by higher baseline state DASS
depression scores, higher baseline EPDS scores and higher baseline
conflictual relationship scores. For those who were both MINI and
EPDS cases, only higher baseline DASS state depression scores
were predictive.
Our birth and post-partum data set domain comprised baby
settling problems, maternal stress at birth and, following the birth
of the baby, supportive and conflictual relationship scores. Three
were consistent significant predictors across all three outcome
measures (i.e. MINI, EPDS, MINI þEPDS). Higher stress since birth
predicted MINI case status, EPDS case status and combination
MINI and EPDS case status. Lower supportive relationship scores
were predictive of all three outcome measures as were higher
conflictual relationship scores.
We then entered all significant set predictors of individual
outcome measures into a multiple regression analysis to deter-
mine a refined set of overall significant predictors. MINI case status
was predicted by higher levels of stress following birth of the baby,
a higher DASS state depression score at baseline, having had a
lifetime mood disorder and lower supportive relationship scores in
the post-natal period. EPDS case status was predicted by higher
levels of post-birth stress, higher neuroticism levels and higher
post-partum conflict, but was less likely with higher baseline
conflict and higher post-partum support. For those who were
both MINI and EPDS cases, there were three predictors – higher
levels of post-natal stress, a higher baseline DASS depression score
and having had a lifetime mood disorder.
As our estimates of caseness may have been confounded by
treatment, we analyzed data in relation to a further outcome
variable – those who were MINI cases and/or were taking
antidepressant medication at the post-partum review, as receipt
of such medication (assuming its effectiveness) might have effec-
tively allocated some potentially MINI cases to MINI non-case
status. Analyses of the available data set contrasted 76 positive
‘cases’ (10.3% of the sample) with the residual group of 665
subjects. In the final multivariate analyses of all identified indivi-
dual set variables, only four were formally significant – having had
a lifetime mood disorder (OR¼ 4.40, Wald or W¼28.5, p o0.001,
95%CI ¼2.56–7.59), stress at birth (OR ¼1.47, W¼ 6.8, p ¼0.009, 95%
CI ¼1.10–1.96), a history of significant pre-menstrual symptoms
(OR¼ 1.4, W¼5.7, p ¼0.017, 95%CI ¼1.06–1.86) and baseline EPDS
depression (OR ¼1.09, W¼ 5.4, p ¼0.02, 95%CI ¼ 1.01–1.18).
3.5. Predictors of incident PND status
We repeated our analytic approach in relation to those who
were non-cases at baseline but had become cases in the post-natal
period to identify a set of refined predictors of new onset
(incident) PND. So-defined MINI cases were significantly predicted
by higher levels of post-natal stress and having had a lifetime
mood disorder. So-defined EPDS cases were predicted by higher
levels of post-natal stress, higher post-birth conflictual relation-
ship scores, not smoking cigarettes during pregnancy and lower
baseline conflictual scores. While there were three predictors of
PND for both MINI and EPDS cases, only one predictor was
quantified for those who were both MINI and EPDS incident cases
– higher levels of post-natal stress (Table 3).
 G.B. Parker et al. / Journal of Affective Disorders 173 (2015) 239–244
Table 3
Incident depression predictive variables.
MINI depression predictors EPDS depression predictors
Variable OR Wald p 95% CI Variable
Higher stress since 4.93 38.3 o 0.001 2.98– Higher stress since
birth 8.17 birth
Previous Depression 3.00 10.2 0.001 1.53– Post-partum conflict
5.89
Smoking during
pregnancy
Baseline conflict
4. Discussion
We sought to identify a refined set of perinatal variables
predictive of PND, and after analyzing socio-demographic, pre-
natal, pregnancy related and post-natal variable sub-sets, we
undertook an additional multivariate analysis of variables identi-
fied as significant in those initial sub-set analyses. The study had
the advantages of comprising a large sample size, a broad socio-
economic spectrum, a large range of previously suggested pre-
dictive variables and was able to contrast two quite differing
measures of post-natal depression – an analysis which seemingly
has only been undertaken once previously (Yonkers et al., 2001).
We suggest that the major study finding was that the identifica-
tion of predictors was influenced sharply by the actual measure of
PND, a conclusion with important implications.
PND defined by meeting DSM-IV criteria for major depression
generated a prevalence rate of 6.5% in comparison to 15.4%
quantified by the EPDS. The first is similar to rates reported in a
review study of 6.5% for major depression and 14.5% for major and
minor depression combined (Gavin et al., 2005), as well as rates
reported in other Australian research studies (Bowen et al., 2014).
When a composite variable was created (major depression present
and/or in receipt of an antidepressant medication and therefore
capturing women who had been depressed/distressed but pre-
sumably responded to medication) the prevalence rate of 10.6%
was in the PND range noted in the Section 1 of 10–15%. The
dissonance in prevalence estimates by our two principal outcome
measures might have been reconciled if we had expanded the
DSM-defined category to include minor DSM depressive states
and/or increased the cut-off score for the EPDS. In relation to the
last, the Beyond Blue clinical practice guideline document (Beyond
Blue, 2011) considered seven “detection tools” for identifying
“depression in the perinatal period” and stated that the EPDS
“can be considered an appropriate tool” as the majority of 14
evaluative studies had quantified it as having high sensitivity and
specificity. Further, they observed that the “optimal score for major
or minor depression” had been identified in differing studies as
ranging from 44 to 13 or more, a range that is concerning at face
value (although such low cut-off scores may perhaps be artefacts
of culture, or use of abbreviated EPDS measures) and further
concerning in that such a range would generate widely varying
prevalence estimates. They did observe, however, that scores of
13–15 were most commonly used for “detecting possible major
depression.” As noted in Section 1, we preserved the EPDS cut-off
recommended by its developers and one that optimized sensitiv-
ity, but note that the study by Yonkers et al. (2001) used a cut-off
of 12 or more. Preserving the original cut-off score in our modified
9-item EPDS may, however, be problematic, especially as the
impact of this alteration cannot be ascertained and is thus a study
limitation.
Such findings also raise the question of what is meant by
‘clinical depression’ (or here ‘post-natal clinical depression’). The
243
MINI and EPDS predictors
OR Wald P 95% CI Variable OR Wald P 95% CI
3.19 34.2 o 0.001 2.16–4.71 Higher stress since 7.89 29.2 o 0.001 3.73–
birth 16.67
1.07 19.9 o 0.001 1.04–1.11
0.34 5.8 o 0.025 0.14–0.82
0.97 5.0 0.025 0.95–1.00
EPDS rates mood state severity dimensionally. The imposition of
any cut-off score on a dimensional measure risks both false
positive and false negative assignment, while design objectives
for the measure (e.g. case definition, screening) will dictate
determining a cut-off score for prioritizing sensitivity or specifi-
city. DSM criteria for major depression have a severity component
(i.e. it is a major rather than minor depressive state), but also
include other parameters such as impairment and persistence over
a defined period. As a consequence, the DSM approach is likely to
position PND categorically as ‘clinical depression’ compared to the
dimensional approach integral to the EPDS.
A second concern in relation to the EPDS is that it is not limited
to depression items, with two items assessing anxiety/worry and
scared/panicky, as well as a general stress (things “getting on top
of me”) item. While such domains may be correlates of PND they
do not quantify depression per se and may therefore compromise
total EPDS scores (in relation to measuring ‘depression’) and thus
consequentially risk compromising ‘case’ allocation in studies
pursuing predictors of PND.
As PND has a relatively low prevalence, we recruited a large
sample and one distinctly larger than employed in most predictor
studies of this nature. Due to such a large sample size many
significant study predictors generated very small odds ratios.
While all significant associations have been reported it should be
noted that some may not be clinically meaningful as they only
fractionally increase the likelihood of developing PND. Whether
PND was defined by DSM criteria or EPDS status (and when
limited to incident cases only) the only consistent predictor was
high levels of stress in the post-partum period. It is possible that
stress for the mother in the postnatal period impacts on the
likelihood of depression and, in turn, that depression increases the
likelihood of experiencing stress in the postnatal period. As such,
the utility of such a predictive factor is limited (despite its
plausibility) by the overlapping constructs. The inconsistent find-
ings across our two outcome measures are perhaps worthy of
greater emphasis. The key and highly significant predictors of PND
as defined by DSM criteria were a previous mood disorder, high
levels of depression at baseline, and as previously discussed,
stress. As clinical depression is commonly persistent, chronic or
recurrent this finding would seem hardly surprising, but is never-
theless of great utility in terms of its clinical application for
screening women at high risk of PND. However, and of key
importance, a previous history of depression was not a significant
predictor of EPDS-defined PND. The latter was more distinctly
predicted by high baseline neuroticism and DASS state depression
scores. It has long been recognized that neuroticism predisposes
strongly to both anxiety and depression states (Hirschfeld et al.,
1989) and, as the EPDS measures anxiety as well as depression
constructs (Stuart et al., 1998), such a finding suggests either some
confounding of the predictor and outcome measures or that the
EPDS measure is better viewed as capturing post-natal distress
rather than depression per se. Indeed in a development study for a
244 G.B. Parker et al. / Journal of Affective Disorders 173 (2015) 239–244
measure of Postpartum Distress, the correlation between the final
measure of distress and the EPDS was 0.92 (Allison et al., 2011).
This is not, of necessity, a downside of the EPDS measure, as many
practitioners are equally concerned by ‘distress’ post-natally, be it
anxiety, depression or another source of distress.
As PND is a public health and clinical priority, study findings
demonstrating the influence of differing measures on predicting
PND have implications for screening and shaping the focus of
clinical interventions. Of relevance, an analysis in the United
Kingdom judged that the EPDS did not meet criteria for an
effective screening mechanism, while no other screening tool
currently has enough research to support its use as an alternative
(Hill, 2010). Obtaining a history of a previous depressive episode or
reports of a distinctly depressed mood late in pregnancy as well as
a report on the mother's stress since the birth of the child might
therefore be more efficient signals of a high risk to PND and of
distinct clinical utility in being readily operationalised and
included in the mother's ante-natal clinical assessment.
Role of funding source
This study was funded by an Australian Department of Health and Aging Grant
and an Australian National Health and Medical Research Council Program Grant
(103716).
Conflict of interest
The authors report no conflict of interest.
Acknowledgments
We appreciate assistance from midwives and other staff at Wyong and Royal
North Shore Public Hospitals and North Shore Private Hospital whose assistance
made this project possible, and key study assistance from Julie Ho.
References
Allison, K.C., Wenzel, A., Kleiman, K., Sarwer, D.B., 2011. Development of a brief
measure of postpartum distress. J. Women's Health 20, 617–623.
Beyond Blue, 2011. Clinical Practice Guidelines: Depression and Related Disorders.
Beyond Blue.
Bowen, A., Duncan, V., Peacock, S., Bowen, R., Schwartz, L., Campbell, D., Muajarine,
N., 2014. Mood and anxiety problems in perinatal indigenous women in
Australia, New Zealand, Canada and the United States: a critical review of the
literature. Transcult. Psychiatry 51, 93–111.
Costello, C., Comrey, A., 1967. Scales for measuring depression and anxiety. J.
Psychol. 66, 303–313.
Cox, J.L., Holden, J.M., Sagovsky, R., 1987. Detection of postnatal depression:
development of the 10-item Edinburgh postnatal depression scale. Br. J.
Psychiatry 150, 782–786.
Gavin, N., Gaynes, B., Lohr, K., Meltzer-Brody, S., Gartlehner, G., Swinson, T., 2005.
Perinatal depression: a systematic review of prevalence and incidence. Obstet.
Gynecol. 106, 1071–1083.
Hill, C., 2010. An Evaluation of Screening for Postnatal Depression against NSC
Criteria. UK National Screening Committee.
Hirschfeld, R.M., Klerman, G.L., Lavori, P., Keller, M.B., Griffith, P., Coryell, W., 1989.
Premorbid personality assessments of first onset of major depression. Arch.
Gen. Psychiatry 46, 345–350.
Lovibond, S.H., Lovibond, P.F., 1995. Manual for the Depression Anxiety Stress
Scales. Psychology Foundation, Melbourne.
Markhus, M.W., Skotheim, S., Graff, I.E., Frøyland, L., Braarud, H.C., Stormark, K.M.,
Malde, M.K., 2013. Low omega-3 index in pregnancy is a possible biological risk
factor for postpartum depression. Plos One 8, e67617.
Matthey, S., Henshaw, C., Elliott, S., Barnett, B., 2006. Variability in use of cut-off
scores and formats on the Edinburgh postnatal depression scale – implications
for clinical and research practice. Arch. Women's Ment. Health 9, 309–315.
Norbeck, J.S., 1984. Modification of recent life event questionnaires for use with
female respondents. Res. Nurs. Health 7, 61–71.
O’Hara, M., Swain, A., 1996. Rates and risks of postpartum depression – a meta-
analysis. Int. Rev. Psychiatry 8, 37–54.
Parker, G., Manicavasagar, M., Crawford, J., Tully, L., Gladstone, G., 2006. Assessing
personality traits associated with depression: the utility of a tiered model.
Psychol. Med. 36, 1131–1139.
Pope, S., Watts, J., Evans, S., McDonald, S., Henderson, J., 2000. Postnatal Depres-
sion: A Systematic Review of Published Literature to 1999. National Health and
Medical Research Council.
Rush, A.J., Giles, D.E., Schlesser, M.A., Fulton, C.L., Weissenburger, J.E., Burns, C.T.,
1986. The inventory of depressive symptomatology (IDS): preliminary findings.
Psychiatry Res. 18, 65–87.
Sarason, I.G., Johnson, J.H., Siegel, J.M., 1978. Assessing the impact of life changes:
development of the life experiences survey. J. Consult. Clin. Psychol. 46,
932–946.
Scottish Intercollegiate Guidelines Network (SIGN), 2012. Management of Perinatal
Mood Disorders. SIGN.
Sheehan, D.V., Lecrubier, Y., Sheehan, K.H., Amorim, P., Janavs, J., Weiller, E.,
Hergueta, T., Baker, R., Dunbar, G.C., 1998. The mini-international neuropsy-
chiatric interview (M.I.N.I.): the development and validation of a structured
diagnostic psychiatric interview for DSM-IV and ICD-10. J. Clin. Psychiatry 59
(Suppl. 20), S22–S33.
Stuart, S., Couser, G., Schilder, K., O’Hara, M.W., Gorman, L., 1998. Postpartum
anxiety and depression: onset and comorbidity in a community sample. J. Nerv.
Ment. Dis. 186, 420–424.
Tilden, V.P., Hirsch, A.M., Nelson, C.A., 1994. The interpersonal relationship
inventory: continued psychometric evaluation. J. Nurs. Meas.2 (l), 63–78.
Yonkers, K., Ramin, S., Rush, A., Navarrete, C.A., Carmody, T., March, D., Heartwell, S.F.,
Leveno, K.J., 2001. Onset and persistence of postpartum depression in an inner city
mental health clinic system. Am. J. Psychiatry 158, 1856–1863.