Expert answer:see all attachements
article_1.pdf
article_2.pdf
_research_article_critique_paper_rubric.docx
article_3.pdf
guidelines.docx
Unformatted Attachment Preview
JOURNAL OF WOMEN’S HEALTH
Volume 23, Number 9, 2014
ª Mary Ann Liebert, Inc.
DOI: 10.1089/jwh.2014.4824
Predictors of Postpartum Depression
Wayne Katon, MD,1 Joan Russo, PhD,1 and Amelia Gavin, PhD 2
Abstract
Objective: To examine sociodemographic factors, pregnancy-associated psychosocial stress and depression,
health risk behaviors, prepregnancy medical and psychiatric illness, pregnancy-related illnesses, and birth
outcomes as risk factors for post-partum depression (PPD).
Methods: A prospective cohort study screened women at 4 and 8 months of pregnancy and used hierarchical
logistic regression analyses to examine predictors of PPD. The study sample include 1,423 pregnant women at a
university-based high risk obstetrics clinic. A score of ‡ 10 on the Patient Health Questionnaire-9 (PHQ-9)
indicated clinically significant depressive symptoms.
Results: Compared with women without significant postpartum depressive symptoms, women with PPD were
significantly younger ( p < 0.0001), more likely to be unemployed ( p = 0.04), had more pregnancy associated
depressive symptoms ( p < 0.0001) and psychosocial stress ( p < 0.0001), were more likely to be smokers
(p < 0.0001), were more likely to be taking antidepressants (ADs) during pregnancy ( p = 0.002), were less likely
to drink any alcohol during pregnancy ( p = 0.02), and were more likely to have prepregnancy medical illnesses,
including diabetes ( p = 0.02) and neurologic conditions ( p = 0.02).
Conclusion: Specific sociodemographic and clinical risk factors for PPD were identified that could help
physicians target depression case finding for pregnant women.
Introduction
T
he post-partum period is a high-risk time for development of major depressive episodes. Two systematic
reviews have found that 7%–13% of women will experience a
serious episode of postpartum depression (PPD).1,2 Women
who experience PPD have an increased risk of future depressive episodes and resulting functional impairment.3,4
PPD has been shown to adversely affect maternal functioning
and is a risk factor for poor mother–infant bonding, subsequent delayed child developmental milestones,5–7 and child
and adolescent mental health disorders.8,9
Several systematic reviews have examined risk factors for
development of PPD.1,2 Risk factors found to be associated
with moderate to high risk of PPD include depression or
anxiety during pregnancy, stressful life events, low levels of
social support, previous history of depression, and the personality factor of neuroticism.1,2 Pregnancy-related complications such as preeclampsia, premature labor, and other
labor-related complications were associated with significant
but lower level of risk in most studies.1 Markers of lower
socioeconomic status such as unemployment and lower educational attainment have also been associated with significant but lower risk of PPD.1,2
1
2
The systematic reviews found that a limitation of the literature was that few studies included the full wide range of
potential risk factors for PPD, such as sociodemographic
factors; prepregnancy medical illness; health risk behaviors
such as smoking, drug and alcohol use; depression history
prior to and during pregnancy; psychosocial stress; intimate
partner violence during pregnancy; pregnancy-related complications such as gestational diabetes and pregnancy-related
hypertension; and adverse birth outcomes such as preterm
birth, low birth weight, and fetal death.1,2
The purpose of this study was to examine a wide range of
socio-demographic factors, health risk behaviors, depression
history, prepregnancy medical illness, pregnancy-related
illnesses, and birth outcomes as risk factors for PPD.
Materials and Methods
Participants in this study were women receiving prenatal
care at the University Obstetrics Clinic between January 2004
and June 2011, who delivered at the University of Washington
Hospital. The university’s Obstetrics and Gynecology Clinic
and Obstetrics Inpatient Service have linked electronic records. Questionnaires assessing mood and other important
sociodemographic, medical, and behavioral information were
Department of Psychiatry and Behavioral Sciences, University of Washington School of Medicine, Seattle, Washington.
University of Washington School of Social Work, Seattle, Washington.
753
754
introduced as a quality improvement initiative in January
2004.10 All women receiving obstetrical care and completing
at least one survey during their second or third trimester as well
as at 6-week postpartum follow-up were eligible for the study.
Women presented to this tertiary care clinic at different pregnancy trimesters, and therefore, some had only one questionnaire completed during pregnancy and others had two (i.e.,
4- and 8-month questionnaires). Clinic staff were responsible
for screening patients with the survey questionnaire and once
completed, obtaining written informed consent to link medical
records with survey results. Given the nature of a busy, urban
obstetrics clinic, staff were unable to get a small percentage of
questionnaires completed.
Exclusion criteria included being < 15 years of age at the
time of delivery or inability to complete the questionnaire due
to language difficulty or mental incapacity. All procedures
were approved by the University of Washington Human
Subjects Institutional Review Board.
Study variables and measures
The Patient Health Questionnaire-9 (PHQ-9) was used to
assess depressive symptoms during the second or third trimester and postpartum.11 A continuous PHQ-9 severity
measure was used as an independent variable. When questionnaires were filled out at both 4 and 8months, the mean
PHQ-9 score was used. A PHQ-9 of ‡ 10 was utilized as the
main outcome variable to define significant depressive
symptoms at the post-partum visit. A PHQ-9 of ‡ 10 has been
found in obstetrics and gynecology (Ob-Gyn) patients to have
the highest sensitivity (73%) and specificity (98%), compared to a structured psychiatric interview diagnosis of major
depression. Information about antidepressant (AD) use in
pregnancy was obtained from the self-report questionnaire.11
Self-report of AD use in pregnancy has been found to have
high concordance with pharmacy records.12
Sociodemographic information on age, marital status,
race/ethnicity, education, and employment, as well as general
health history, health risk behaviors, social history, and
psychosocial stressors were collected during either the second or third trimester. Chronic medical problems prior to
pregnancy were screened for with a standard list that included
hypertension, diabetes, asthma, thyroid disorders, migraines,
arthritis, seizure disorders, heart failure, cancer, and other
heart disease. Tobacco status was assessed using the SmokeFree Families prenatal screen that was developed to screen
for smoking during pregnancy.13 Women with any current
smoking were classified as smokers.
Diagnosis of pregnancy-induced/gestational hypertension
was based on outpatient and inpatient physician International
Classification of Diseases, Ninth Revision (ICD-9) diagnoses
of 642.3, 642.4, 642.5, 642.6, and 642.7, respectively.14 Diagnosis of gestational diabetes mellitus (GDM) was determined by a physician ICD-9 diagnosis of 648.8 in the
outpatient or inpatient medical record.15 GDM is clinically
defined as glucose intolerance with the first recognition or
onset in pregnancy;16 therefore, this diabetes category could
potentially include women with previously unrecognized
type 2 diabetes.
The Prenatal Psychosocial Profile Stress Scale is an 11item self-report scale that measures perceived current hassles
and stressors.17 Women indicate the extent to which each
KATON ET AL.
item is a current hassle or stressor on a 4-point Likert scale [1
(no stress) to 4 (severe stressor)] with a range of 11 to 44. It
has been shown to have high reliability and validity in
pregnant populations. The three-question Abuse Assessment
Screen has been validated in pregnant patients as a sensitive
and specific screen for intimate partner violence.18 Each item
is rated as ‘‘yes’’ or ‘‘no,’’ and the percentage of women with
at least one measure of intimate partner violence was described. The revised four item alcohol screening questionnaire,
the T-ACE (Take [number of drinks], Annoyed, Cut-down,
Eye-opener), was employed to assess use of alcohol during
pregnancy. The T-ACE modification of the alcohol screening
questionnaire, the CAGE (Cut-down, Annoyed, Guilt, Eyeopener), substitutes the guilt question with an alcohol tolerance
question.19 The T-ACE has been found to outperform obstetric
staff assessment of alcohol use by pregnant women.19 Women
with any use of alcohol during pregnancy were identified with
the T-ACE.
Offspring birth weight, gestational age at birth, and fetal
death were obtained from study participants’ computerized
medical records. Low birth weight was based on a gestational
weight threshold of 2,500 g. Pre-term birth was determined as
less than 37 weeks of completed gestation.
Statistical analyses
We selected the study sample from the entire screening
sample (N = 3,039). The 1,423 women whose data were included in the study were compared with the 1,616 women
who were excluded based on missing key data on baseline
demographic and clinical variables using Fischer exact tests
and t-tests. Due to significant inclusion group differences,
non-response propensity scores were created using baseline
variables (demographics, medical conditions, health risk
behaviors, pregnancy variables, and depression). The inverse
of the probability of not responding was used to weight our
regression analyses.
We compared baseline variables for the women with and
without PPD (having a postpartum PHQ-9 score ‡ 10) using
Fischer Exact tests and t-tests. Hierarchical logistic regression
analyses were used to predict the odds of PPD. We first examined the unadjusted association between prepregnancy
PHQ-9 scores and PPD. In the next model, we added demographic variables and reevaluated the odds for prepregnancy
depression. The third model contained the demographics and
medical conditions. The fourth model added health-related
behaviors to the model, and the fifth model added pregnancyrelated variables. Lastly, we calculated a final model predicting PPD, including all odds ratios and their 95% confidence
intervals for all the predictor variables.
Because an increase of five points on the PHQ-9 is associated with significant clinical change,20 we recalculated the
odds of PPD based on pregnancy PHQ-9 total scores formed
by creating groups with five-point intervals, adjusting for all
other study variables. Lastly, we conducted a sensitivity
analysis by examining the final model without the use of the
propensity weights.
Results
A total of 3,039 women were screened either at four
months or eight months (or at both time periods) of pregnancy. Of these, 1,515 women were excluded due to lack of a
PREDICTORS OF POSTPARTUM DEPRESSION
postpartum assessment (the vast majority of these women
attended post-partum visits at clinics closer to their homes
rather than the university high-risk obstetrics clinic); 84 were
excluded due to only filling out the 8-month questionnaire
which had no questions on medical history and 17 were excluded due to lack of data on birth outcomes (preterm labor or
low birth weight), leaving a study sample of 1,423.
Univariate analyses comparing the pregnancy data for
those women who were and were not included in this study
revealed significant group differences. Those women not
eligible for this study were slightly younger ( p < 0.001, although the difference in means was only 1.3 years), less
likely to be college educated ( p < 0.001), and more likely to
be single ( p < 0.001), non-white ( p < 0.001), and unemployed ( p < 0.001) than the women included in the study
sample. In addition, ineligible women reported being slightly
more depressed ( p < 0.003), 0.5 difference on PHQ-9), were
more likely to have a baby die ( p < 0.01), and less likely to
755
have GDM ( p < 0.006) or preeclampsia ( p < 0.02) than those
women retained for analysis. Due to these differences, we
created nonresponse propensity weights utilizing the variables in Table 1, and have applied these weights to our regression analyses.
Table 1 displays the descriptive data for the groups with
and without PPD. A total of 6.7% of women had a PHQ-9
score of ‡ 10 during pregnancy, and 5.8% had a score of ‡ 10
at the postpartum check. Women with PPD reported significantly more depressive symptoms during pregnancy than
women without PPD. In addition, women with PPD were
significantly younger, less likely to be married, less educated,
and more likely to report being unemployed than women
without PPD. In terms of medical conditions, women with
PPD in comparison with women without PPD reported higher
rates of diabetes, migraines, and a trend ( p = 0.06) for hypertension and neurological conditions (0.07). Women with
PPD were more likely to report current smoking, taking an
Table 1. Descriptive Data for Women With and Without Post-Partum Depression (PPD)
Variablesa
mean (SD) or percent (n)
Pregnancy PHQ-9 Total Scoreb
Pregnancy PHQ-9 ( ‡ 10)c
Pregnancy minor depression PHQ-9 (5–9)c
Ageb
Racec
White
African American
Hispanic
Other
Marriedc
At least some collegec
Unemployedc
Total sample
N = 1,423
PHQ-9 (0–9)
no PPD
n = 1,340
Depression
3.4 (3.4)
3.1 (3.1)
6.7 (95)
4.9 (66)
4.4 (62)
4.1 (55)
Demographic variables
31.5 (5.9)
31.7 (5.9)
73.8 (1006)
74.1 (953)
5.6 (76)
5.4 (70)
4.6 (63)
44 (56)
16.0 (218)
16.1 (207)
89.4 (1257)
90.5 (1197)
85.5 (1200)
86.6 (1143)
40.1 (564)
38.7 (511)
Prepregnancy medical conditions
Asthmac
10.7 (152)
10.4 (139)
Diabetesc
7.3 (104)
6.9 (92)
7.2 (103)
7.0 (93)
GI disordersc
4.7 (67)
4.5 (60)
Heart conditionsc
Hypertensionc
6.6 (94)
6.3 (84)
Migrainec
14.1 (200)
13.2 (176)
2.0 (28)
1.8 (24)
Neurological conditionsc
7.0 (99)
6.8 (91)
Thyroid problemsc
Health-related behaviors
5.5 (78)
4.0 (53)
Current cigarette smokingc
Taking an antidepressantc
7.0 (100)
6.2 (83)
14.1 (201)
14.3 (190)
Drinking alcohol during pregnancyc
2.5 (36)
2.3 (30)
Intimate partner violence during the past yearc
Stress total scoreb
14.3 (3.4)
14.0 (3.1)
Pregnancy-related variables
GDMc
20.7 (294)
20.7 (278)
Preeclampsiac
21.2 (301)
20.6 (277)
Low birth weightc
11.5 (164)
10.8 (145)
Preterm birthc
15.0 (214)
14.2 (190)
Fetal deaths
0.35 (5)
0.37 (5)
a
PHQ-9 ( ‡ 10)
PPD
n = 83
p Value from t-test
or Fisher’s Exact test
7.8 (5.3)
34.9 (29)
8.4 (7)
0.0001
0.0001
0.09
28.5 (6.3)
0.0001
68.8
7.8
9.1
14.2
72.3
68.7
63.9
(53)
(6)
(7)
(11)
(60)
(57)
(53)
0.17
15.7
14.5
12.0
8.4
12.0
29.6
4.9
9.6
(13)
(12)
(10)
(7)
(10)
(24)
(4)
(8)
0.14
0.02
0.12
0.11
0.06
0.0001
0.07
0.37
30.5
20.5
11.6
7.2
18.3
(25)
(17)
(11)
(6)
(4.9)
0.0001
0.0001
0.54
0.02
0.0001
19.3
28.9
22.9
28.9
0
(16)
(24)
(19)
(24)
(0)
0.89
0.10
0.002
0.001
1.0
0.0001
0.0001
0.0001
Only a mean of 2.7% of the model variables were missing.
Mean (standard deviation [SD]).
Percent (n).
GI, gastrointestinal; GDM, gestational diabetes mellitus; PHQ-9, Patient Health Questionnaire-9; PPD, post-partum depression.
b
c
756
KATON ET AL.
AD [women taking an AD had higher PHQ-9 scores than
women not treated with ADs: PHQ-9 = 6.1 (4.5) versus 3.1
(3.2), p < 0.001], being a victim of intimate partner violence
in the past year, and had higher stress scores than women
without PPD. The percentage of women with GDM and
preeclampsia during the pregnancy did not differ between the
depression groups. However, having an infant with low birth
weight or having a preterm birth were both significantly more
prevalent in the women with PPD. There were five fetal
deaths, all in women without PPD.
Table 2 presents the logistic regression results. The unadjusted odds ratio (OR) for PHQ-9 scores assessed during
pregnancy was 1.25 (95% confidence interval [95% CI] = 1.21–
1.29). The addition of the demographic variables and medical
conditions did little to change this result. The addition of the
health-related behaviors minimally reduced the odds to 1.10
(1.05–1.15). The addition of the pregnancy-related variables
did not change the odds ratio. Therefore, there is a 10% increase
in the odds of reporting PPD for every one-point increase in
PHQ-9 score assessed during pregnancy. A clinically significant five-point increase in PHQ-9 scores assessed during
pregnancy resulted in a 70% increase in the odds of PPD
[OR = 1.70, 95% CI = 1.34–2.16, p = .0001], after adjusting for
all the demographic, medical conditions, health-related behaviors, and pregnancy-related variables. The final model in
Table 2 shows the odds ratios and their 95% confidence intervals for all included variables. In addition to depression,
women with PPD were significantly more likely to be younger,
to be unemployed and to have prepregnancy diabetes and
neurological conditions, to be smokers; to report using Ads, and
Table 2. Risk Factors Associated with Post-Partum Depression
Model
Unadjusted total PHQ-9 score assessed during pregnancy
Total PHQ-9 Score assessed during pregnancy adjusted for demographic
variablesa
Total PHQ-9 Score assessed during pregnancy adjusted for demographic
variablesa and medical conditionsb
Total PHQ-9 Score assessed during pregnancy adjusted for demographic
variables,a medical conditions,b and health-related behaviorsc
Final multi-variable model
Total PHQ-9 Score assessed during pregnancy
Demographic variablesa
Age
White race
Married
Some college
Unemployment
Prepregnancy medical conditionsb
Asthma
Diabetes
GI disorders
Heart conditions
Hypertension
Migraines
Neurological conditions
Thyroid problems
Health-related behaviorsc
Current cigarette smoking
Taking an antidepressant
Drinking alcohol during pregnancy
Intimate partner violence within the past year
Stress total score
Pregnancy-related variablesd
GDM
Preeclampsia
Low birth weight
Preterm birth
Odds ratio (95% CI)
for total PHQ-9 Score (1 point) p Value
1.25 (1.21–1.29)
1.22 (1.18–1.26)
0.0001
0.0001
1.21 (1.17–1.25)
0.0001
1.10 (1.05–1.16)
0.0001
1.10 (1.05–1.15)
0.0001
0.94
1.01
1.14
1.18
1.50
(0.91–0.97)
(0.69–1.48)
(0.73–1.77)
(0.77–1.82)
(1.02–2.21)
0.0001
0.95
0.58
0.44
0.04
0.61
1.98
0.74
1.92
0.70
1.40
2.37
0.70
(0.37–1.00)
(1.12–3.52)
(0.42–1.32)
(0.98–3.79)
(0.39–1.25)
(0.93–2.11)
(1.12–5.02)
(0.34–1.46)
0.05
0.02
0.31
0.06
0.23
0.10
0.02
0.34
2.84
2.23
0.46
0.53
1.14
(1.80–4.48)
(1.35–3.68)
(0.24–0.90)
(0.24–1.13)
(1.09–1.19)
0.0001
0.002
0.02
0.10
0.0001
0.68
1.43
1.38
1.07
(0.40–1.13)
(0.95–2.17)
(0.79–2.43)
(0.64–1.81)
0.13
0.09
0.26
0.79
The above analyses are propensity weighted.
a
Demographic variables: age, race, marital status, education, unemployment.
b
Prepregnancy medical conditions: asthma, hypertension, diabetes, neurological condition, heart condition, GI condition, thyroid
problems, migraines.
c
Health-related behaviors: current cigarette smoking, taking an antidepressant, any domestic violence in the past year, drinking alcohol
during pregnancy, and stress total score.
d
Pregnancy-related variables: GDM, preeclampsia, low birth weight, and preterm birth.
95% CI, 95% confidence interval.
PREDICTORS OF POSTPARTUM DEPRESSION
to endorse more stress and less alcohol use during pregnancy in
comparison with the women without PPD.
For our sensitivity analysis, we ran the same complete model
without the propensity weights. The significance and odds ratios
for all the variables were similar for the propensity weighted and
unweighted models, except for four variables which became
nonsignificant: unemployment ( p = 0.37), asthma ( p = 0.41),
neurological problems ( p = 0.31), and alcohol consumption
during pregnancy ( p = 0.18). These variables were less robustly
associated with PPD in the weighted model, and all four of these
varia ...
Purchase answer to see full
attachment
