Keywords postpartum depression - severe psychological distress - prenatal marijuana use - very
preterm infants
Postpartum depression (PPD) and severe psychological distress (SPD) affect 10 to 15%
of mothers,[1 ] placing them at risk for recurring symptoms,[2 ]
[3 ]
[4 ] threats to positive relationships with their infants,[5 ]
[6 ]
[7 ] and increasing risk for children's poor developmental outcomes.[8 ]
[9 ]
[10 ] PPD and SPD have been associated with a history of depression, anxiety, perinatal
loss, obstetric complications, traumatic events, and low social support.[1 ]
[11 ]
[12 ] These risk factors are also associated with preterm birth,[13 ] and among mothers who deliver preterm, reported rates of PPD and SPD range from
18 to 43%.[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ] At greatest risk are mothers of infants born <30 weeks of gestation, where the infants'
immaturity and illness create persistent challenges to maternal coping processes.[17 ]
[20 ]
Previous studies of mothers of preterm infants highlighted the need to detect early,
potentially treatable risk factors for PPD and SPD.[6 ]
[16 ]
[21 ] Characterizing birthing parents' risk factors and needs at admission to the neonatal
intensive care unit (NICU) could expedite the initiation and continuation of supportive
interventions beginning early in the infant's stay in the NICU.[14 ]
[15 ]
[22 ] Concurrent care for birthing parents and their infants would also benefit from an
illumination of specific neonatal medical conditions that may initiate or exacerbate
maternal psychological disequilibrium and present cumulative threats to adaptation
over time.[8 ]
[9 ]
[10 ]
Toward the goal of informing family-centered practices, the present study was designed
to address evidence gaps about birthing parents' needs at NICU admission and discharge
by examining in concert the parent, infant, and socioenvironmental characteristics associated with postpartum
psychological challenges in a large and diverse multicenter sample.
Our own previous work described adverse medical and socioenvironmental conditions,
including prenatal depression, anxiety, and substance use associated with neurobehavioral
regulation and epigenetic processes in infants born <30 weeks postmenstrual age (PMA).[23 ]
[24 ]
[25 ] The present study extends this work to identify risk factors associated with PPD
and SPD at NICU discharge, a point at which important findings about antecedent conditions
could inform both preventive and targeted NICU interventions for those with significant
risk factors and demonstrated needs. Further, we sought to address a knowledge gap
by identifying maternal needs that emerge in the time between routine prenatal care
assessments, PPD screening in the context of routine postpartum obstetric follow-up,
and in follow-along pediatric care. Building on earlier studies,[17 ]
[26 ] we hypothesized that PPD and SPD would be associated with prior depression and/or
anxiety, prenatal substance use, and more extensive medical and socioenvironmental
adversities. To our knowledge, this is the first multicenter study to examine these
risk factors in concert in a large and diverse sample of mothers who delivered a very
preterm infant (<30 weeks gestation), and to describe the relative associations between
psychological history, substance use, and socioenvironmental conditions, as well as
maternal and infant medical complications with PPD and SPD at NICU discharge.[11 ]
[12 ]
[17 ]
[20 ]
[27 ]
Materials and Methods
Participants
From April 2014 through June 2016, the birthing parents who self-identified as mothers
of 704 infants were enrolled from nine Level IV NICUs affiliated with six universities
across the United States for participation in the Neonatal Neurobehavior and Outcomes
in Very Preterm Infants Study (NOVI), a longitudinal multicenter study. NICUs included
seven with delivery services and two in children's hospitals, all of which were participating
in the Vermont-Oxford Network.[28 ] NOVI's primary aims have focused on the early and long-term outcomes among very
preterm infants and their families, and the present study focuses on early maternal
outcomes at NICU discharge.
The sample included 562 mothers of 641 infants with complete maternal assessments
([Fig. 1 ]). Inclusion criteria were: (1) birth <30 weeks PMA determined by available information
in order of the following hierarchy: assisted reproduction dates, fetal ultrasonogram,
postmenstrual dates, or neonatal physical exam at NICU admission[29 ]; (2) maternal ability to read and speak English or Spanish (based on discrepancies
in requisite interpreter availability across sites); and (3) residence within 3 hours
of the NICU and follow-up clinic. Infants were excluded for major congenital anomalies,[30 ] NICU death, and maternal inability to consent for their infants due to maternal
age <18 years, cognitive impairment, or death. Based on longitudinal follow-up aims,
parents were invited to participate at 30 to 31 weeks PMA, when survival to NICU discharge
was determined to be likely by the attending neonatologist. Informed consent was obtained
in accordance with approvals by each site's institutional review board. The present
study followed Strengthening the Reporting of Observational Studies in Epidemiology
(STROBE) guidelines for observational studies.
Fig. 1 Study flow and eligibility.
Procedures
Maternal Medical and Socioenvironmental Data Collection
Study Coordinators received training in the reliable implementation of standardized
consenting, interviewing, and medical record extraction. In addition, each site's
study neonatologist provided oversight to Study Coordinators who completed standardized
reviews of infant medical records and maternal delivery admission notes. We also conducted
discussions to consensus among the six NOVI Study neonatologists regarding questions
about specific variables charted: prenatal and intrapartum conditions, prescribed
medications, and the use of alcohol, tobacco, marijuana, and other psychoactive substances
including cocaine, opioids, methamphetamine, hallucinogens, and nonprescribed pharmaceuticals
(e.g., amphetamines, benzodiazepines). Standardized maternal interviews were administered
to collect information about race, ethnicity, partner and cohabitation status, insurance
type, education, occupation, and diagnoses of anxiety and depression both (1) prior
to and (2) during the pregnancy, as well as counseling and/or medication for either
or both of these conditions during either time epoch. For analyses, we identified
mothers with a prior diagnosis and treatment of anxiety and/or depression by including
either medical record notations or answers to interview questions regarding prior
diagnoses, counseling, and/or medication for anxiety and depression.
Maternal Postpartum Depression and Severe Psychological Distress Assessments
To avoid potential variations due to literacy, Study Coordinators were trained to
reliably administer the Edinburgh Postnatal Depression Scale (EPDS)[31 ] to assess PPD risk, and the Brief Symptom Inventory (BSI)[32 ] to assess SPD risk. Mothers completed the EPDS and BSI by interview during the week
of their infant's discharge from the NICU. This assessment point provided information
at a standardized point in time that was feasibly integrated into all nine NICUs'
routine care protocols, and at a time following maternal transitions through their
infants' varied acute and/or chronic medical conditions.
The well-validated EPDS[31 ] is a self-report screening instrument comprising 10 items related to risk for depression
and anxiety. Each symptom is scored on a severity scale ranging from 0 to 3, with
“0 = hardly at all/not at all/never” to “3 = quite a lot/most of the time.” Total
scores range from 0 to 30, where scores ≥12 indicate major depressive disorder risk;
scores of 10 to 12 indicate probable mild depression risk requiring monitoring.[33 ] We adopted the criteria of EPDS scores of ≥101 to capture a range of severity of PPD risk from mild to severe.[34 ] The anxiety subscale of the EPDS, EPDS-3A, is the sum of questions 3, 4, and 5 with
a range of 0 to 9 and a cut-off score of 6 suggesting symptoms of anxiety, and is
distinct from overall EPDS scores among childbearing individuals.[35 ]
The BSI[32 ] is a screening instrument composed of 53 item with ratings of 0 to 5 reflecting
the extent of symptom distress (i.e., 0 = “not at all” to 5 = “extremely”). Items
are grouped into symptom scales (internal consistency >0.70; test–retest reliability
0.68–0.91): Somatization, Obsessive-Compulsive, Interpersonal Sensitivity, Depression,
Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation, and Psychoticism. Validated
criteria for a positive BSI screen were employed to identify SPD risk using composite
measures of overall psychological distress, where alpha reliability coefficients ≥0.90
exceeded reliability of individual symptom scales: (1) a Global Severity Index T-score
of ≥63, or (2) T-scores ≥63 on any two primary symptom scales.[32 ]
[36 ]
[37 ]
Both the EPDS and the BSI include thoughts of self-harm and other symptoms experienced
during the previous week. Maternal assessments were administered during the week of
NICU discharge, at a time designed to (1) avoid potentially stressful questions and
limit measurement variability during periods of the infants' severe and wide-ranging
illnesses, (2) standardize the developmental stage for maternal screening to the time
at which infants were stable enough for discharge, and (3) identify the range of both
maternal postdischarge needs and early and ongoing neonatal risk factors that preceded
PPD and SPD. The present study's protocol ensured psychological support referrals
for mothers who reported thoughts of self-harm or extreme symptoms on the EPDS or
BSI, became distraught completing the questionnaires, or requested help. All data
collected for the study remained confidential, including when we implemented referrals
for mothers who requested additional services.
Neonatal Medical Complications
Medical record reviews ascertained neonatal characteristics and complications using
Vermont-Oxford Network definitions and criteria[38 ] about infections, retinopathy of prematurity (ROP), and respiratory, renal, neurologic,
cardiac, genetic, hematologic, and gastrointestinal morbidities. A validated neonatal
medical complications index[39 ] included one point each for (1) severe brain injury: periventricular leukomalacia,
moderate–severe ventricular dilatation, or parenchymal echodensity identified by centralized
readings and consensus agreement on diagnoses; (2) culture-positive sepsis; (3) severe
ROP; (4) necrotizing enterocolitis (NEC); and (5) bronchopulmonary dysplasia (BPD;
supplemental oxygen requirement at 36 weeks PMA).
Statistical Analyses
We compared maternal characteristics for mothers included in analyses based on complete
data versus those not included due to missing EPDS and BSI data using one-way analysis
of variance (ANOVA) for continuous measures or Chi-square for categorical measures.
Unadjusted analyses were conducted to describe maternal and infant correlates of the
separate EPDS and BSI outcomes at NICU discharge. To build on findings from previous
studies and identify specific risks for PPD and SPD in multivariable model testing,
we examined individual covariates based on univariate associations presented in [Tables 2 ] and [3 ] for this sample, as well as on significant predictors of PPD and SPD identified
in prior studies.[1 ]
[27 ]
[40 ]
[41 ]
[42 ]
Table 2
Maternal and infant characteristics by positive screen on the Edinburgh Postnatal
Depression Scale (EPDS)
EPDS > 10 (n = 76)
n (%) or mean ± SD
EPDS < 10 (n = 486)
n (%) or mean ± SD
p
Maternal characteristics (n = 562)
Psychological history
Prepregnancy or prenatal depression and/or anxiety[a ]
34 (44.7)
147 (30.6)
0.01
Socioenvironmental risk factors
No partner
20 (26.3)
118 (24.3)
0.70
Low income-based public assistance/health insurance or uninsured
55 (72.4)
304 (62.6)
0.10
Education < college
63 (82.9)
372 (76.5)
0.22
Member of a minority race or ethnic group[b ]
48 (63.2)
277 (57.0)
0.31
Prenatal substance use
Tobacco
12 (15.8)
66 (13.6)
0.61
Alcohol
2 (2.6)
14 (2.9)
0.62
Marijuana
11 (14.5)
40 (8.2)
0.08
Recreational/illicit substances[c ]
2 (2.6)
18 (3.7)
0.48
Prenatal medical risk factors
Maternal age >35 years
12 (15.8)
85 (17.5)
0.72
Diabetes
4 (5.3)
30 (6.2)
0.51
Vaginal or urinary infection
6 (8.0)
54 (11.1)
0.42
Chronic or pregnancy-induced hypertension
16 (21.1)
138 (28.5)
0.18
Sexually transmitted infection or human immunodeficiency virus
4 (5.3)
33 (6.8)
0.43
Body mass index ≥30 kg/m2
20 (26.7)
173 (35.6)
0.13
Underweight <18.5 kg/m2
4 (5.3)
23 (4.7)
0.50
Fetal congenital anomaly
4 (5.3)
28 (5.8)
0.57
Neonatal characteristics (n = 641)
(n = 83)
(n = 558)
p
Sex, male
55 (66.3)
303 (54.3)
0.04
Postmenstrual age at birth (wk)
26.7 ± 1.9
27.1 ± 1.9
0.06
Postmenstrual age at discharge >40 wk
43 (51.8)
221 (39.7)
0.04
Multiple gestation
18 (21.7)
154 (27.6)
0.25
Study site
0.11
1
16 (19.3)
90 (16.1)
2
11 (13.3)
115 (20.6)
3
11 (13.3)
57 (10.2)
4
22 (26.5)
94 (16.8)
5
15 (18.1)
111 (19.9)
6
8 (9.6)
91 (16.3)
Severe neonatal medical conditions
Brain ultrasound abnormality[d ]
14 (16.9)
67 (12.1)
0.22
Bronchopulmonary dysplasia
51 (61.4)
270 (48.4)
0.03
Severe retinopathy of prematurity
10 (12.0)
30 (5.4)
0.02
Necrotizing enterocolitis or sepsis
9 (10.8)
108 (19.4)
0.06
Abbreviation: SD, standard deviation.
a Prepregnancy and prenatal depression and anxiety variables were combined due to significant
overlap for these diagnoses.
b Self-reported member of the following race or ethnic group: American Indian, Asian,
Black, Hispanic, Native Hawaiian, Pacific Islander, or “other” self-specified.
c Maternal recreational/illicit substance use included cocaine, methamphetamine, hallucinogens,
and nonprescribed pharmaceuticals, e.g., opioids and benzodiazepines.
d Cranial ultrasound abnormality: periventricular leukomalacia, moderate–severe ventricular
dilatation, or parenchymal echodensity identified by centralized readings and consensus
agreement on diagnoses.
Table 3
Maternal and infant characteristics by positive screen on the Brief Symptom Inventory
(BSI)
Positive BSI screen = yes
n (%) or mean ± SD
Positive BSI screen = no
n (%) or mean ± SD
p
Maternal characteristics (n = 562)
(n = 102)
(n = 460)
Psychological history
Prepregnancy or prenatal depression and/or anxiety[a ]
42 (41.6)
139 (30.5)
0.03
Socioenvironmental risk factors
No partner
29 (28.4)
109 (23.7)
0.32
Low income-based public assistance/health insurance or uninsured
71 (69.6)
288 (62.6)
0.18
Education < college
85 (83.3)
350 (76.1)
0.11
Member of a minority race or ethnic group[b ]
65 (63.7)
260 (56.5)
0.18
Prenatal substance use
Tobacco
15(14.7)
63 (13.8)
0.80
Alcohol
3 (2.9)
13 (2.8)
0.95
Marijuana
17 (16.7)
34 (7.4)
<0.01
Recreational/illicit substances[c ]
4 (3.9)
16 (3.5)
0.77
Prenatal medical complications
Maternal age >35 years
18 (17.6)
79 (17.2)
0.91
Diabetes
8 (7.9)
26 (5.7)
0.39
Vaginal or urinary infection
9 (8.9)
51 (11.1)
0.52
Chronic or pregnancy-induced hypertension
17 (16.8)
137 (29.8)
<0.01
Sexually transmitted infection or human immunodeficiency virus
5 (5.0)
32 (7.0)
0.46
Body mass index ≥30 kg/m2
31 (30.7)
162 (35.2)
0.39
Underweight <18.5 kg/m2
5 (5.0)
22 (4.8)
0.94
Fetal congenital anomaly
8 (7.9)
24 (5.2)
0.29
Neonatal characteristics (n = 641)
(n = 114)
(n = 527)
P
Sex, male
67 (58.8)
291 (55.2)
0.49
Postmenstrual age at birth (mean ± SD)
26.6 ± 1.9
27.1 ± 1.9
0.01
Postmenstrual age at discharge >40 weeks
47 (41.6)
160 (30.7)
0.02
Multiple gestation
28 (24.6)
144 (27.4)
0.54
Study site
0.02
1
15 (13,2)
91 (17.3)
2
19 (16.7)
107 (20.3)
3
19 (16.7)
49 (9.3)
4
29 (25.4)
87 (16.5)
5
21 (18.4)
105 (19.9)
6
11 (9.6)
88 (16.7)
Severe neonatal medical conditions
Brain ultrasound abnormality[d ]
18 (15.9)
63 (12.0)
0.25
Bronchopulmonary dysplasia
68 (59.6)
253 (48.0)
0.02
Severe retinopathy of prematurity
11 (9.6)
29 (5.5)
0.10
Necrotizing enterocolitis or sepsis
17 (14.9)
100 (19.0)
0.31
Abbreviation: SD, standard deviation.
a Prepregnancy and prenatal depression and anxiety variables were combined due to significant
overlap for these diagnoses.
b Self-reported member of the following race or ethnic group: American Indian, Asian,
Black, Hispanic, Native Hawaiian, Pacific Islander, or “other” self-specified.
c Maternal recreational/illicit substance use included cocaine, methamphetamine, hallucinogens,
and nonprescribed pharmaceuticals, e.g., opioids and benzodiazepines.
d Cranial ultrasound abnormality: periventricular leukomalacia, moderate–severe ventricular
dilatation, or parenchymal echodensity identified by centralized readings and consensus
agreement on diagnoses
We examined the multivariable influences of maternal, infant, and socioenvironmental
characteristics on maternal EPDS and BSI outcomes using generalized estimating equations
(GEEs), where infants from multiple births were nested within families and within
site to examine positive EPDS and BSI total scores, and the anxiety symptom subscores
for the EPDS (3A) and the BSI separately. Models examined the following risk factors:
prepregnancy/prenatal depression/anxiety diagnoses, as well as the influences of socioenvironmental
adversities, substance use, maternal and neonatal medical complications, infant sex,
as well as PMA at birth and discharge as indicators of the effects of immaturity and
duration of illness. GEE procedures using a log link function to obtain beta weights
that can be converted to relative risks (RRs) directly produced model convergence
issues. Odds ratios (ORs) from the GEE models were converted to RR using the method
described by Zhang and Yu[43 ]: RR = OR/((1 − P0) + (P0 * OR)), where P0 is the observed prevalence within the
reference group. A priori comparisons were also implemented to determine if prenatal
substance use was more prevalent in mothers with a diagnosis of prenatal depression
or anxiety.
Prior studies reported disparities and discrimination associated with postpartum distress
and depressive symptoms reported by members of racial or ethnic minority groups.[44 ]
[45 ]
[46 ]
[47 ] To examine these influences in our sample, model testing examined self-reported
race and ethnicity. Data were analyzed using SPSS 24.0 (Chicago, IL: SPSS, Inc.).
Results
Maternal and Newborn Characteristics
Thirty-seven mothers with missing EPDS and BSI data were excluded from analyses, as
were the only two mothers who received prenatal antipsychotic medication for unidentified
conditions. Compared with the 562 included mothers, excluded mothers had higher rates
of prenatal marijuana and illicit substance use ([Supplementary Table S1 ], available in the online version).
The included sample ([Table 1 ]) was ethno-racially diverse, with a mean maternal age of 29 years (standard deviation
[SD] = 6.3); 25% were neither married nor living with a partner. Thirteen percent
of mothers did not complete high school, 29% were high school graduates, 35% attended
some college, and 23% graduated from college; however, 64% reported receiving income-based
public assistance or health insurance. Prepregnancy or prenatal anxiety or depression
diagnoses were reported by 32% of mothers, 61% of whom reported receiving counseling
and/or pharmacological treatment ([Table 1 ]).
Table 1
Maternal and infant characteristics
Maternal characteristics (n = 562)
n (%) or mean ± SD
Psychological history
Prepregnancy or prenatal depression and/or anxiety[a ]
181 (32.2)
Socioenvironmental risk factors
No partner
138 (24.6)
Low income-based public assistance/health insurance or uninsured
359 (63.9)
Education < college
435 (77.4)
Race/ethnicity
Hispanic, any race
114 (20.3)
Non-Hispanic American Indian/Native American
1 (0.2)
Non-Hispanic Asian
37 (6.6)
Non-Hispanic Black
116 (20.7)
Non- Hispanic Native Hawaiian or Pacific Islander
12 (2.1)
Non-Hispanic White
224 (39.8)
Non-Hispanic unknown or not reported
2 (0.4)
Non-Hispanic multiple race
56 (10.0)
Prenatal substance use
Tobacco
78 (13.9)
Alcohol
16 (2.8)
Marijuana
51 (9.1)
Recreational/illicit substances[b ]
20 (3.6)
Prenatal medical risk factors
Maternal age >35 years
97 (17.3)
Diabetes
34 (6.0)
Vaginal or urinary infection
60 (10.7)
Chronic or pregnancy-induced hypertension
154 (27.4)
Sexually transmitted infection or human immunodeficiency virus
37 (6.6)
Body mass index ≥30 kg/m2
193 (34.3)
Underweight <18.5 kg/m2
27 (4.8)
Fetal congenital anomaly
32 (5.7)
Neonatal characteristics (n = 641)
Sex, male
358 (55.9)
Postmenstrual age at birth (wk)
27.0 ± 1.9
Postmenstrual age at discharge >40 wk
264 (41.3)
Multiple gestation
172 (26.9)
Study site
N (%)
1
106 (16.5)
2
126 (19.7)
3
68 (10.6)
4
116 (18.1)
5
126 (19.7)
6
99 (15.4)
Severe neonatal medical conditions
Brain ultrasound abnormality[c ]
81 (12.7)
Bronchopulmonary dysplasia
321 (50.2)
Severe retinopathy of prematurity
40 (6.3)
Necrotizing enterocolitis or sepsis
117 (18.3)
Abbreviation: SD, standard deviation.
a Prepregnancy and prenatal depression and anxiety variables were combined due to significant
overlap for these diagnoses.
b Maternal recreational/illicit substance use included cocaine, methamphetamine, hallucinogens,
opioids, and benzodiazepines.
c Cranial ultrasound abnormality: periventricular leukomalacia, moderate–severe ventricular
dilatation, or parenchymal echodensity identified by centralized readings and consensus
agreement on diagnoses.
For 641 newborns born to 562 mothers, 44% were female; there were 492 (87.5%) singletons
and 26.9% multiple gestations (64 [11.4%] sets of twins and 6 [1.1%] sets of triplets
or more). Mean birth PMA was 27 weeks (SD = 2), mean birth weight was 950 g (SD = 279),
head circumference mean was 24.5 cm (SD = 2.4), and 7% had fetal growth restriction.
More than half (52%) of the infants received maternal breast milk during their NICU
stay. Thirteen percent of infants had severe brain injuries, 6.9% had NEC, 13.4% had
sepsis, and 50% had BPD (24% mild, 26% moderate/severe adapted from Jensen et al's
criteria[48 ]); 23% of infants had two or more of these severe conditions,[39 ] whereas 41% of the sample had none of these severe conditions ([Table 1 ]). Timing of the EPDS and BSI at NICU discharge reflects a median NICU stay of 85
days; mean PMA at discharge was 40 weeks.
Maternal and Infant Correlates of Postpartum Depression and Severe Psychological Distress
The average EPDS score was 4.6 (SD = 4.1; range 0–21) and 76 mothers (13.5%) had a
positive screen (EPDS ≥10). The mean BSI score was 0.29 (SD = 0.34; range 0–2.4);
102 mothers (18.1%) had a positive BSI screen. The average EPDS-3A score was 2.1 (SD = 2.0;
range 0–9) and 47 mothers (8.4%) had a positive anxiety screen (EPDS-3A >6). The mean
BSI score was 0.29 (SD = 0.34; range 0–2.4) and the mean BSI anxiety symptom dimension
was 0.30 (SD = 0.45; range 0–3); 102 mothers (18.1%) had a positive BSI screen and
40 mothers had a positive BSI anxiety screen (7.1%). Fifty-four mothers (9%) screened
positive on both the EPDS and the BSI. Thoughts of self-harm were reported by 14 mothers
(2.5%); eight (57.1%) of those 14 screened positive on the EPDS or BSI.
Unadjusted analyses indicated that individuals with a positive screen on the EPDS
([Table 2 ]) or the BSI ([Table 3 ]) were more likely to report prepregnancy/prenatal depression and/or anxiety than
those with no prior diagnosis. Compared to individuals with a negative BSI screen,
those with a positive screen had lower rates of chronic or pregnancy-induced hypertension,
but higher rates of prenatal marijuana use. Follow-up comparisons examined marijuana
use to determine if rates differed in those with versus without preexisting depression
or anxiety. Prenatal marijuana use was more prevalent in individuals with prenatal
depression or anxiety (15.7%) compared to those without prenatal depression or anxiety
(8.7%), but this difference was not significant (χ
2 = 2.65, p < 0.10).
Compared to infants born to mothers with negative EPDS or BSI screens, those whose
mothers had positive screens were more likely to have been born at younger gestational
ages, with BPD prevalence that was 13 and 11% higher, respectively, among infants
born to mothers with positive EPDS and BSI screens. Individuals with either positive
screen had infants whose NICU stays averaged 14 days longer, and discharge after 40
weeks PMA was 14% more prevalent. There were no significant differences in EPDS or
BSI screen rates associated with any adverse socioenvironmental conditions, nor with
BPD severity.[48 ]
Multivariable Analyses of Maternal and Infant Risk Factors
GEE accounted for correlations among multiple births by nesting infants within families
and within site to examine EPDS and BSI outcomes tested in the following models separately.
Variables were identified based on prior research and the univariate results above:
prepregnancy/prenatal depression or anxiety diagnosis, prenatal marijuana use, receipt
of income-based public assistance, minority ethnicity, maternal education, and infant
PMA at birth, discharge >40 weeks PMA, BPD, and severe ROP. As presented in [Table 4 ], a positive EPDS screen was associated with prepregnancy/prenatal depression and/or
anxiety (RR = 1.6, 95% CI: 1.02,2.1), and male sex (RR = 1.7, 95% CI: 1.1,2.4). A
positive BSI screen was associated with prepregnancy/prenatal depression and/or anxiety
(RR = 1.6, 95% CI: 1.1,2.2) and prenatal marijuana use (RR = 1.9, 95% CI: 1.1,2.9).
Postnatal anxiety as measured by the EPDS-3A was associated only with prepregnancy/prenatal
depression and/or anxiety (RR = 1.9, 95% CI: 1.03,3.6) and there were no associations
found using the clinical cutoff for the BSI anxiety symptom dimension. As with unadjusted
results, there were no significant differences in EPDS or BSI outcomes associated
with adverse socioenvironmental conditions.
Table 4
Multivariable test statistics for maternal outcome models on the Edinburgh Postnatal
Depression Scale (EPDS) and Brief Symptom Inventory (BSI)
Variable
EPDS positive screen (score > 10)
BSI positive screen (positive screen criteria = yes)
B
SE
RR (95% CI)
B
SE
RR (95% CI)
Maternal characteristics (n = 562)
Prepregnancy or prenatal depression and/or anxiety[a ]
0.57
0.28
1.6 (1.02–2.1)
0.57
0.25
1.6 (1.1–2.2)
Low income-based public assistance/health insurance or uninsured
0.26
0.33
1.3 (0.7–2.1)
0.07
0.30
1.1 (0.6–1.7)
Minority race or ethnicity[b ]
0.8
0.31
1.4 (0.8–2.3)
0.36
0.31
1.3 (0.8–2.1)
Education < college
0.20
0.39
1.2 (0.6–2.2)
0.19
0.34
1.2 (0.6–1.9)
Prenatal marijuana use
0.46
0.41
1.5 (0.7–2.7)
0.82
0.35
1.9 (1.1–2.9)
Infant characteristics (n = 641)
Sex, male
0.57
0.26
1.7 (1.1–2.4)
0.10
0.22
1.1 (0.7–1.5)
Postmenstrual age at discharge >40 weeks
0.10
0.33
1.1 (0.6–2.1)
0.13
0.27
1.1 (0.7–1.7)
Bronchopulmonary dysplasia
0.44
0.30
1.5 (0.9–2.3)
0.38
0.26
1.4 (0.9–2.0)
Severe retinopathy of prematurity
0.77
0.42
1.9 (0.9–3.5)
0.40
0.39
1.4 (0.7–2.4)
Abbreviation: CI, confidence interval; RR, risk ratio; SE, standard error.
a Prepregnancy and prenatal depression and anxiety variables were combined due to significant
overlap for these diagnoses.
b Self-reported member of the following race or ethnic group: American Indian, Asian,
Black, Hispanic, Native Hawaiian, Pacific Islander, or “other” self-specified.
Discussion
Building on earlier research, this study examined maternal, socioenvironmental, and
neonatal risk factors for PPD and severe distress in a large, diverse multicenter
sample of mothers of very premature newborns. We found higher PPD and/or SPD rates
associated with prior depression or anxiety, prenatal marijuana use, and male sex.
In addition, socioenvironmental adversities typically associated with PPD/SPD—i.e.,
single, low income, or education[11 ]
[12 ]
[20 ]
[49 ]—while still useful information for screening, were not significant when tested in
this study's risk assessment model that also included specific infant medical complications.
Notably, 42 to 46% of mothers who screened positive on either the EPDS or BSI had
a prior diagnosis of depression or anxiety. In addition, 8% of mothers had a positive
screen on the EPDS anxiety subscale and were significantly more likely to have experienced
prior depression or anxiety. These positive screens for mothers with a prior diagnosis
are indicative of a high-risk group identifiable in the context of preconceptional
and prenatal care. Conversely, >54% were not identifiable during pregnancy, suggesting
additional postnatal needs assessment is warranted, and may be incorporated in the
context of family-centered NICU care. Although the present sample's rate of thoughts
of self-harm rates at 2.5% was lower than in other samples,[34 ]
[50 ] 43% of mothers reporting thoughts of self-harm did not have positive screens on
either the EPDS or BSI. As such, an additional review of self-harm item responses
may be warranted to ensure indicated referrals.
Monitoring also appears relevant to pregnant individuals facing multiple risk factors
documented here. In particular, the 9% rate of prenatal marijuana use in this sample
(confidence interval = 6.8–11.7%) was substantially higher than the reported 4 to
6% in the general population.[51 ]
[52 ] Despite the limitation of potentially underreported use due to being ascertained
from medical records rather than toxicology screens, this higher than general use
rate may indicate an additional support need in pregnant individuals at risk for SPD
associated with varied constellations of complications.[53 ] Paradoxically, whether marijuana is a form of self-medication used to manage stress,
mood, and/or nausea during pregnancy, it has shown no benefit for depression, may
contribute to panic when accompanied with anxiety, and safer antinausea treatment
alternatives have been available.[52 ]
[54 ]
[55 ] Importantly, these contradictions and the findings presented here highlight salient
opportunities to overcome biases in substance use assessments and to enhance supportive,
nonjudgmental, and widely accessible mental health and substance use services prior
to conception, and throughout prenatal and early postpartum care.
Another key finding suggests the importance of measuring both PPD and SPD, as there
was only a 9% overlap in positive screens on both the EPDS assessment of depression
risk and the BSI assessment of severe distress. Further, these two postpartum outcome
measures were predicted by different combinations of prenatal and neonatal conditions.
Consequently, each measure appears to offer unique risk detection value concerning
both the sources and symptoms of depression and severe distress among birthing parents
of very preterm infants.
These multicenter findings expand upon prior work that identified indicators of risk
for PPD and SPD. Similar to findings among parents of full-term infants, rates of
PPD/SPD were predicted by prior depression or anxiety.[3 ]
[33 ]
[56 ] However, this sample's >40% rates of positive EPDS and BSI screens for mothers with
prior depression or anxiety are higher than the reported 10 to 25% in the general
population.[3 ] This 15 to 30% elevation of positive screen rates suggests that a prior history
combined with severe infant medical complications in this very preterm sample increases
vulnerability to recurring symptoms.
Compared to findings from studies involving a wide range of medical variation due
to the selection of all “premature” or “NICU” samples, our 13% rate of PPD and 18%
rate of SPD are lower than the 18 to 43% previously reported.[14 ]
[16 ]
[49 ]
[57 ] This difference may have resulted from our decision to complete the BSI and EPDS
by interview in order to overcome literacy challenges. However, it is also possible
that higher rates of positive screens occurred in more general “NICU samples” where
infants had a wider range of major anomalies and acute illnesses with well-established
risk for mortality and/or a poor long-term prognosis. Also in contrast to prior work,
our rates of PPD and SPD are slightly lower than the 20[20 ]
[27 ] and 27%[19 ]
[58 ] rates reported at comparable postpartum periods, and may reflect study follow-up
aims requiring inclusion of infants likely to survive to NICU discharge.
Our findings are similar to PPD[20 ] and distress[59 ] in mothers of preterm infants with lengthier ventilation and BPD, and to PPD associated
with younger PMA at birth,[17 ] neonatal medical complications, and lengthier NICU stays.[60 ] Regarding persisting BPD influences on maternal psychological difficulties in particular,
Singer et al[61 ] reported more prevalent maternal distress in the first 2 years of life, with persisting
parenting-specific stress at 3 years of age that was also related to the child's developmental
challenges. Importantly, at age 14 follow-up, although mothers of the high-risk children
with BPD showed persisting family stress scores, they also demonstrated increases
in adaptive coping skills.[59 ] These findings suggest that the interrelationships observed in our study contribute
important information for preventive and targeted interventions to promote multiple
aspects of maternal coping strategies and to address family support needs.
It is noteworthy that this sample reported high rates of concurrent prepregnancy and
prenatal anxiety and depression diagnoses. As a result, we analyzed global scores
for both the EPDS and BSI to examine associations with positive screens on either
measure. The EPDS and BSI are screening instruments, rather than diagnostic, and both
measures have anxiety symptoms embedded. The global scores were used here to focus
on elevated symptom levels that would identify mothers who required more detailed
follow-up with specific diagnostic assessments for anxiety, depression, and/or other
disorders.
Strengths and Limitations
Study strengths include the large and diverse multicenter sample, rigorously standardized
assessments and medical record reviews, with consensus definitions and systematic
oversight by site primary investigators, including immediate referrals for indicated
maternal needs. The timing of postpartum assessments enabled the identification of
both predischarge risk factors and postdischarge referral needs. This was done with
the recognition that the postpartum assessments at NICU discharge potentially measured
both anxiety around caring for the infant at home[62 ] and relief at the infant's survival.
Generalizability was enhanced by including mothers of outborn infants admitted to
NICUs in children's hospitals. Since prenatal medical records are not uniformly available
for infants transported to children's hospitals, and data collection standardization
was achieved by abstracting prenatal and intrapartum medical and substance use information
from the neonatal medical records including delivery and admission notes, histories,
and physicals.
A limitation of the study was that postnatal enrollment was required to determine
eligibility, and that timing precluded prenatal maternal interviews and toxicology
screens to ascertain prenatal substance use. Postnatal enrollment at varied postnatal
ages of outborn infants also precluded an earlier administration of the EPDS and BSI
at a standardized time point. However, the statistically significant association between
marijuana use and BSI severe distress demonstrates the strength of that association,
despite underreported use and without an assessment prior to the week of discharge.
Another issue underlying findings about marijuana use is that legalization of medicinal
and/or recreational marijuana use varied by state of residence from 2013 through 2016
during the study of pregnancies in our six sites.[63 ]
[64 ] While legalization may have influenced use, our statistical models adjusted for
six sites located in six states prior to examining associations between marijuana
use and EPDS and BSI outcomes. This procedure would be expected to adequately adjust
for state differences in legalization.
EPDS and BSI positive screen rates may also have been based on potentially underreported
symptoms using the interview format to avoid literacy bias. Results may differ from
rates obtained by a self-reported questionnaires used in studies that included less
diverse samples. Despite these limitations, findings demonstrated the significant
strength of relationships between PPD and/or SPD, charted marijuana use, and self-reported
prior depression and anxiety diagnoses.
Conclusion
Our findings demonstrate the potential value of incorporating screening and support
for psychological well-being that begin in the preconceptional period, with continued
services focused on perinatal depression, anxiety, and marijuana use.[14 ]
[65 ]
[66 ]
[67 ]
[68 ] Further, extended services could address ongoing needs evidenced by postpartum individuals
whose infants have lengthy hospital stays and complex conditions.[69 ]
[70 ]
[71 ] Addressing needs for safer alternatives to marijuana use and personalized approaches
to services for pregnant individuals at increased risk for depression and severe distress
has demonstrated benefits when integrated into family-centered care.[62 ]
[72 ] These efforts have the potential to improve outcomes for pregnant and postpartum
individuals, their infants, and families facing multiple challenges to their well-being
and relationships.[73 ]
[74 ]
This study identified early and persisting prepregnancy and prenatal characteristics,
and neonatal complications associated with PPD and SPD in a vulnerable sample that
represents a population of over 59,000 U.S. families annually,[75 ] and numbers in the millions worldwide.[76 ] Programmatic efforts are particularly relevant in light of the current increases
in prenatal cannabis use, depression, and anxiety during the on-going coronavirus
pandemic, and the cumulative burden of these stressors.[77 ]
[78 ]
[79 ] Addressing risk is clearly warranted,[80 ] and targeting demonstrated needs continuously and concurrent with infant care has
demonstrated potential.
