Smoking Cessation and Relapse Among Pregnant African-American Smokers in Washington, DC

Ayman A. E. El-Mohandes • M. Nabil El-Khorazaty •

Michele Kiely • Marie G. Gantz

Published online: 8 June 2011

� Springer Science+Business Media, LLC (outside the USA) 2011

Abstract Smoking is the single most preventable cause

of perinatal morbidity. This study examines smoking

behaviors during pregnancy in a high risk population of

African Americans. The study also examines risk factors

associated with smoking behaviors and cessation in

response to a cognitive behavioral therapy (CBT) inter-

vention. This study is a secondary analysis of data from a

randomized controlled trial addressing multiple risks dur-

ing pregnancy. Five hundred African-American Washing-

ton, DC residents who reported smoking in the 6 months

preceding pregnancy were randomized to a CBT inter-

vention. Psycho-social and behavioral data were collected.

Self-reported smoking and salivary cotinine levels were

measured prenatally and postpartum to assess changes in

smoking behavior. Comparisons were made between active

smokers and those abstaining at baseline and follow-up in

pregnancy and postpartum. Sixty percent of participants

reported quitting spontaneously during pregnancy. In

regression models, smoking at baseline was associated with

older age,\a high school education and illicit drug use. At follow-up closest to delivery, smoking was associated with

lower education, smoking and cotinine level at baseline

and depression. At postpartum, there was a relapse of 34%.

Smokers postpartum were significantly more likely to

smoke at baseline and use illicit drugs in pregnancy.

Mothers in the CBT intervention were less likely to

relapse. African-American women had a high spontaneous

quit rate and no response to a CBT intervention during

pregnancy. Postpartum mothers’ resolve to maintain a quit

status seems to wane despite their prolonged period of

cessation. CBT reduced postpartum relapse rates.

Keywords Smoking � Pregnancy � African-Americans � Washington, DC

Introduction

Smoking during pregnancy is associated with problems of

placentation [1], low birthweight [2–8], prematurity [3, 9,

10], sudden infant death [10–12], infant mortality [12],

and later physical [13], developmental [13] and behav-

ioral [10, 14, 15] problems. A significant percentage of

smokers continue to smoke during pregnancy and are

unable to quit on their own despite their knowledge of the

risks involved. Existing behavioral interventions are only

modestly successful, having an attributable benefit of no

more than 10% above spontaneous quit rates among

pregnant women [16].

M. Nabil El-Khorazaty: Deceased.

Clinical trial registration: ClinicalTrials.gov, www.clinicaltrials.gov,

NCT00381823.

A. A. E. El-Mohandes

College of Public Health, University of Nebraska Medical

Center, Omaha, NE, USA

M. N. El-Khorazaty � M. G. Gantz RTI International (RTI International is a trade name of Research

Triangle Institute), Rockville, MD, USA

M. Kiely

National Institutes of Health, Bethesda, MD, USA

M. Kiely (&) Epidemiology Branch, Division of Epidemiology, Statistics and

Prevention Research, Eunice Kennedy Shriver National Institute

of Child Health and Human Development, National Institutes

of Health, 6100 Executive Blvd, Rm. 7B-05, Rockville,

MD 20852-7510, USA

e-mail: [email protected]

123

Matern Child Health J (2011) 15:S96–S105

DOI 10.1007/s10995-011-0825-6

Success in smoking cessation during pregnancy may be

different among different ethnic groups. In one study,

Mexican–Americans had three times higher cessation rates

than non-Hispanic whites [17]. Only a few studies in the

literature describe African-American smoking behaviors

during pregnancy and postpartum, and even fewer tested the

efficacy of smoking cessation interventions programs in that

population. Although smoking rates during pregnancy are

lower among African Americans, genetically mediated dif-

ferences in nicotine metabolism are associated with higher

nicotine levels among African Americans compared to

whites [18].

High rates of low birthweight and prematurity in Afri-

can-Americans may be partly attributable to smoking in

pregnancy [19], either independently or as a complicating

factor for other medical risks including chronic hyperten-

sion. Although contested by some authors in the literature

[20], the effect of smoking on poor birth outcomes has been

estimated as high as 14.4% among black births [21]. To

improve birth outcomes among African-Americans, there

is a need to better understand their smoking behaviors in

pregnancy and postpartum, including spontaneous cessa-

tion and relapse rates, associated variables impacting on

these rates, and responses to behavioral interventions.

This study is a secondary analysis of a larger randomized

controlled trial (RCT) addressing multiple risks during

pregnancy. The main results of the RCT have been published

elsewhere [22, 23]. This paper investigates smoking cessa-

tion and relapse rates among African American women

reporting smoking in the 6 months preceding pregnancy in

Washington, DC Women were recruited during pregnancy

and followed through the postpartum period and randomized

to an integrated cognitive behavioral intervention addressing

smoking, environmental tobacco smoke exposure (ETSE),

depression, and intimate partner violence (IPV).

Method

Study Population

The population recruited to this study was part of a larger

cohort recruited to the District of Columbia Healthy Out-

comes of Pregnancy Education (DC-HOPE), under the

umbrella of the National Institutes of Health-District of

Columbia Initiative to Reduce Infant Mortality in Minority

Populations. DC-HOPE was a randomized controlled trial

evaluating the efficacy of an integrated cognitive behav-

ioral intervention targeting cigarette smoking, environ-

mental tobacco smoke exposure (ETSE), intimate partner

violence (IPV) and depression during pregnancy. Mothers

were eligible if they were 18 years or older, English-

speaking, less than 29 weeks gestation and Washington,

DC residents. Women were recruited from six prenatal care

sites and were screened using an audio-computer assisted

self-interview (A-CASI) (For details see El-Khorazaty

et al. [24]). Recruitment occurred between July 2001 and

October 2003 and followed through July 2004. Baseline

interviews for eligible women occurred on average 9 days

after screening. IRB approval was obtained from all par-

ticipating institutions.

There were 2,913 women screened and 1,515 were

ineligible. Of the 1,398 eligible women, 1,070 enrolled as

eligible minority participants. (See Fig. 1) 1,044 women

were included in these analyses; they self-identified as

African-American. Eligible women consented for ran-

domization into the intervention or usual care group. Per-

muted block randomization was site- and risk-specific. The

field staff were blinded with respect to the block size. Eight

women (6 intervention and 2 usual care) were identified as

suicidal during intervention or data collection and were

referred immediately to mental health care and excluded

from further participation. Five hundred women were

screened into the study as having smoked a puff of a cig-

arette or more in the 6 months preceding pregnancy. This

level was chosen to be as inclusive as possible because

these women were at risk for continuing to smoke or

relapsing if they had quit early in pregnancy.

Data and Saliva Sample Collection

Data on sociodemographic and behavioral risk were col-

lected during a baseline telephone interview, on average

within 9 days of screening. Follow-up telephone interviews

were conducted during the second and third trimesters

(22–26 weeks and 30–34 weeks, respectively) and 8–10

weeks postpartum to evaluate changes in the psycho-

behavioral risks. Interviewers were blinded to whether

women were in the intervention or usual care group.

Smoking risks during pregnancy and postpartum were

measured based on self-report. Saliva samples were col-

lected at the prenatal care site on average within 19 days

following the baseline interview, within a week from the

follow-up telephone interview and 23 days following the

postpartum interview. Salivary cotinine was measured by a

radio-immune assay using gas chromatography-mass

spectrometry (GC/MS) with lower detection limits of

10 ng/ml. IPV was measured using the Revised Conflict

Tactics Scale physical assault and sexual coercion sub-

scales [25]. Depression was measured using the 20-item

Hopkins Symptom Checklist-Depression Scale [26].

Intervention

Of the 500 women included in these analyses, 262 were

randomized to the intervention group and 238 were

Matern Child Health J (2011) 15:S96–S105 S97

123

randomized to usual care. This integrated intervention was

based on a conceptual framework of overlapping and

interactive behavioral risks. Such risk factors are known to

co-occur within a population of urban African-Americans

living in communities with high poverty rates. The risks

selected are all associated with poor pregnancy outcomes

[27]. The smoking intervention was delivered to women

who self-reported as smokers and not on a cutoff cotinine

level since randomization was based on the initial response

to the A-CASI.

The 10-session intervention was delivered during pre-

natal (8 sessions) and postpartum (2 booster sessions) care

visits. Four prenatal sessions were considered minimal

adherence. The session duration was approximately

35 min. The smoking intervention was consistent with the

Smoking Cessation or Reduction in Pregnancy Trial

(SCRIPT) and the Counseling and Behavioral Interventions

Work Group of the United States Preventive Services Task

Force recommendations, a five-step behavioral counseling

approach [28, 29]. The intervention was tailored to the

woman’s stage of change. Women were encouraged to

avoid triggers and to use alternative coping and behavioral

change strategies. The intervention included content to

address both active smoking and ETSE, whether or not

they met criteria for ETSE.

The intervention sessions also addressed the other

associated risks. For depression, the intervention focused

on secondary prevention of symptoms in pregnancy and

extended into the postpartum period. Cognitive behavioral

therapy strategies for mood management, increasing posi-

tive social interactions, and pleasurable activities were

emphasized. The IPV interventions used the Parker’s

model to address the role of a negative partner support

[30]. Danger assessment to identify risks for harm and

prevention options were considered along with the devel-

opment of a safety plan. (For more details see Katz et al.

[27]). All measures were based on validated questionnaires.

Statistical Analysis

Women who were screened as having smoked a puff of a

cigarette or more in the 6 months preceding pregnancy

were compared according to their self-reported smoking

status at baseline interview, last follow-up interview

prior to delivery, and postpartum interview conducted

8–10 weeks after delivery. Comparisons were conducted

by means of Chi-square tests for binary variables and t tests

for continuous variables.

Based on the results of these bivariate comparisons, we

used logistic regression procedures to model the probabil-

ity of cigarette smoking at each of the three time points

controlling for covariates with P value \0.10 in the bivariate analyses. For control variables with a strong

colinearity, we selected the variables with the highest level

Fig. 1 Profile of project DC-HOPE randomized controlled trial

S98 Matern Child Health J (2011) 15:S96–S105

123

of significance or the greatest biological plausibility.

Control variables included alcohol and illicit drug use

during pregnancy, depression, IPV, prior smoking status,

cotinine levels, and the intervention. Variables descriptive

of demographic and socioeconomic status (maternal age,

education level, and Medicaid enrollment) were included

as covariates so their cumulative effects could be accoun-

ted for in the final logistic model. We used the LOGISTIC

procedure in SAS version 9.1.3 (SAS Institute, Cary, NC)

to conduct the analysis.

Results

Of the 500 mothers reporting cigarette smoking at

screening, data were available on 396 at a follow-up

interview prior to delivery and 384 mothers were inter-

viewed in the postpartum period. No significant differences

between the 500 and the 396 or 384 were seen in any

sociodemographic or behavioral characteristics at baseline.

(Data not shown) No significant differences were noted

between the intervention and usual care groups regarding

sociodemographic or behavioral characteristics at baseline.

Among the 500 women who reported smoking at A-CASI

screening and were included in these analyses, 39%

reported active smoking at baseline. An earlier paper from

our study showed that women who reported smoking at

A-CASI screening were significantly less likely to resolve

risk (smoking, ETSE, depression and IPV) during preg-

nancy [22].

A significant difference was noted in salivary cotinine

levels collected at baseline between mothers who self-

identified as smokers and non-smokers (179 ± 156 ng/ml

vs. 32 ± 59 ng/ml, P \ 0.001). At baseline 86.4% of women who reported themselves as non-smokers had a

salivary cotinine level \50 ng/ml and 90.3% \100 ng/ml. In a logistic regression model, the factors that remain

significantly associated with smoking at baseline are

reviewed in Table 4A. Older maternal age, education at

less than high school level and illicit drug use as reported

by mothers at baseline were the factors significantly asso-

ciated with smoking at baseline (Table 1).

At follow-up prior to delivery, 34% of mothers reported

smoking. A significant difference was noted in salivary

cotinine levels collected at that time between those

reporting smoking or not (135 ± 145 ng/ml vs.

28 ± 57 ng/ml, P \ 0.001). 83.5% of the women who reported themselves as non-smokers had a salivary cotinine

level\50 ng/ml and 88.5% had a level\100 ng/ml. Of the women reporting smoking during the follow-up interview,

13.0% had not been smoking at baseline and represented a

relapse. Similarly, 12.8% of non-smokers had smoked at

baseline but quit at a later stage of pregnancy. There was

no significant interventional effect on smoking behavior as

reported in the follow-up interviews during pregnancy.

Women who continued to smoke during pregnancy were

significantly older, had a lower level of education attain-

ment, and had higher rates of enrollment in Medicaid.

These women were also more likely to have reported

alcohol and illicit drug use during the baseline interview

and higher baseline cotinine levels. Depression at baseline

was a predictor of smoking at follow-up, while IPV was

not. Depression and IPV confirmed during the follow-up

interview were associated with smoking. No significant

differences were seen between the characteristics of

smokers randomized to the intervention group and those in

usual care (Table 2).

In a logistic regression model (Table 4B), the factors

that retained significant association with continued smok-

ing at follow-up were active smoking at baseline and sal-

ivary cotinine levels at baseline. Depression at the follow-

up period preceding delivery was also predictive of active

smoking during the same time period.

In the postpartum period, 50% of participants self-

reported as actively smoking. Salivary cotinine levels were

significantly higher in women reporting active smoking

(249 ± 176 ng/ml vs. 109 ± 149 ng/ml, P \ 0.001). Only 60.4% of the women who reported themselves as non-

smokers had a salivary cotinine level \50 ng/ml, and 64.4% had cotinine levels \100 ng/ml. The intensity of

Table 1 Women screening positive for smoking before pregnancy: baseline assessment

Characteristic Smoking

(n = 195)

Not smoking

(n = 305)

P value

Maternal age

(mean ± SD)

26.9 ± 6.3 23.6 ± 4.5 \0.001

Pregnancies (incl. current)

(mean ± SD)

4.8 ± 2.9 3.5 ± 2.2 \0.001

Previous live births

(mean ± SD)

2.2 ± 1.9 1.2 ± 1.4 \0.001

Relationship status: 0.770

Single/separated/

widowed/divorced

152 (78.0%) 241 (79.0%)

Married/living with

partner

43 (22.0%) 64 (21.0%)

Education level: 0.002

\High school 91 (46.7%) 97 (31.8%) High school/GED 79 (40.5%) 147 (48.2%)

Some college or more 25 (12.8%) 61 (20%)

Medicaid recipient 174 (89.2%) 241 (79.3%) 0.004

Alcohol use 68 (34.9%) 72 (23.7%) 0.007

Illicit drug use 54 (27.7%) 47 (15.4%) \0.001 Depression 102 (52.3%) 129 (42.3%) 0.029

Intimate partner violence 73 (37.4%) 89 (29.2%) 0.054

Matern Child Health J (2011) 15:S96–S105 S99

123

smoking in the postpartum was significantly higher than

during the two preceding time points as confirmed by

cotinine levels. Among postpartum smokers salivary coti-

nine levels were significantly higher than the baseline

levels (P \ 0.001) or the levels at follow-up closest to delivery (P \ 0.001). Of the women reporting smoking postpartum, 34% had not reported smoking during the

follow-up interview during pregnancy. This group was

considered postpartum relapsers. A much smaller per-

centage (7.4%) of women not reporting smoking post-

partum had reported smoking during pregnancy. A higher

likelihood of smoking postpartum was associated in

bivariate analysis with older age, higher gravidity and

parity, lower educational attainment, higher Medicaid

enrollment, other substance use, active smoking at baseline

and follow-up (Table 3). Depression documented at base-

line, during follow-up interviews or in the postpartum was

significantly associated with active smoking. IPV did not

show a similar association at any of the three time points.

The intervention for the first time showed an association

with reported smoking abstinence in the postpartum period,

at a P value of 0.053.

In a logistic regression model, factors that increased the

likelihood of reported smoking in the postpartum were

active smoking as reported by mothers and cotinine levels

at baseline and illicit drug use during pregnancy. The

intervention had a significant protective effect against

smoking in the postpartum period (Table 4C).

Discussion

The results of this study confirm the difficulty pregnant

mothers who smoke have in quitting during pregnancy.

Mothers included in our study that were less educated,

depressed or using illicit substances were least likely to

quit. The literature emphasizes the underlying demographic

and psychosocial factors that impact smoking behaviors

among African-American women [31]. In spite of findings

that African-Americans were significantly more likely than

Table 2 Women screening positive for smoking before pregnancy: baseline smokers randomized to intervention versus usual care

Characteristic Intervention (n = 105) Usual care (n = 90) P value

Maternal age (mean ± SD) 26.9 ± 6.5 26.8 ± 6.1 0.930

Pregnancies (incl. current) (mean ± SD) 4.8 ± 3.1 4.7 ± 2.6 0.948

Previous live births (mean ± SD) 2.2 ± 2.1 2.2 ± 1.7 0.991

Relationship status: 0.522

Single/separated/widowed/divorced 80 (76.2%) 72 (80.0%)

Married/living with partner 25 (23.8%) 18 (20.0%)

Education level: 0.765

\High school 51 (48.6%) 40 (44.4%) High school/GED 42 (40.0%) 37 (41.1%)

Some college or more 12 (11.4%) 13 (14.4%)

Medicaid recipient 95 (90.5%) 79 (87.8%) 0.545

Alcohol use 39 (37.1%) 29 (32.2%) 0.472

Illicit drug use 28 (26.7%) 26 (28.9%) 0.730

Active smoking at follow-up 62 (74.7%) 54 (78.3%) 0.607

Active smoking at postpartum 64 (83.1%) 67 (91.8%) 0.111

Cotinine level at baseline (mean ± SD) 192.9 ± 165.0 162.4 ± 144.6 0.216

Cotinine level at follow-up (mean ± SD) 146.0 ± 139.4 131.9 ± 117.6 0.528

Cotinine level at postpartum (mean ± SD) 290.8 ± 182.7 236.1 ± 162.2 0.103

ETSE at baseline 89 (87.3%) 75 (84.3%) 0.555

ETSE at follow-up 66 (80.5%) 52 (75.4%) 0.448

ETSE at postpartum 58 (74.4%) 56 (78.9%) 0.516

Depression at baseline 57 (54.3%) 45 (50.0%) 0.550

Depression at follow-up 39 (47.0%) 34 (49.3%) 0.779

Depression at follow-up postpartum 23 (29.9%) 23 (31.5%) 0.828

Intimate partner violence at baseline 38 (36.2%) 35 (38.9%) 0.698

Intimate partner violence at follow-up 8 (9.8%) 6 (8.7%) 0.823

Intimate partner violence at postpartum 6 (7.8%) 7 (9.6%) 0.696

S100 Matern Child Health J (2011) 15:S96–S105

123

whites to express a desire to quit smoking [32] it is not

clear what barriers prevent them from quitting. None of the

studies meeting the guidelines for inclusion in the Public

Health Service Report [33] specified abstinence rates by

racial/ethnic group [34]. The only studies that report on the

results of smoking cessation interventions during preg-

nancy by race come to opposite conclusions [35, 36].

Studies examining smoking cessation interventions during

pregnancy published since the Public Health Service

Report did not report their spontaneous cessation and

relapse rates in pregnancy and postpartum by race/ethnic-

ity, or found no differences in rates by race [37–41].

Our results agree with previous authors showing a sig-

nificant spontaneous cessation rate among smokers who

become pregnant [42, 43]. The quit rate of more than 60%

in our population of urban African-American pregnant

women experiencing other socioeconomic and psycholog-

ical stressors is encouraging. Notably, amongst this popu-

lation of smokers women who reported quitting during

pregnancy experienced a high rate of depression (42%) and

IPV (29%) throughout the pregnancy. Almost one-third of

these women who reported quitting on their own had not

completed high school, and the majority were Medicaid

enrollees. It is hard to determine whether the social desir-

ability of quitting during pregnancy within this community,

and/or the knowledge of the detrimental effects of smoking

on the fetus could have been the main driving force.

The underlying depressive symptoms in our study pop-

ulation may have interfered with their ability to control

their smoking. There is a growing awareness of the prev-

alence of depressive symptoms within the smoking popu-

lation, with a range of 22–61% amongst those entering

smoking cessation programs [44–46]. The literature is

mixed regarding the effect of depression on the success of

smoking cessation. One may infer that depression inter-

feres with short term quit rates and not long-term success in

smoking cessation [47, 48]. In the logistic analyses we

conducted at three time points, depression was predictive

Table 3 Women screening positive for smoking before pregnancy: postpartum assessment

Characteristic Smoking (n = 191) Not smoking (n = 193) P value

Maternal age (mean ± SD) 25.9 ± 6.0 24.0 ± 4.9 \0.001 Pregnancies (incl. current) (mean ± SD) 4.5 ± 2.6 3.3 ± 2.2 \0.001 Previous live births (mean ± SD) 2.0 ± 1.9 1.2 ± 1.4 \0.001 Relationship status: 0.957

Single/separated/widowed/divorced 149 (78.0%) 151 (78.2%)

Married/living with partner 42 (22.0%) 42 (21.8%)

Education level: \0.001 \High school 89 (46.6%) 55 (28.5%) High school/GED 79 (41.4%) 96 (49.7%)

Some college or more 23 (12.0%) 42 (21.8%)

Medicaid recipient 172 (90.1%) 149 (77.6%) \0.001 Alcohol use 67 (35.1%) 43 (22.4%) 0.006

Illicit drug use 57 (29.8%) 23(11.9%) \0.001 Active smoking at baseline 131 (68.6%) 19 (9.8%) \0.001 Active smoking at follow-up 100 (64.5%) 8 (4.7%) \0.001 ETSE at baseline 159 (85.0%) 134 (70.5%) \0.001 ETSE at follow-up 115 (74.7%) 95 (55.6%) \0.001 ETSE at postpartum 147 (79.0%) 97 (51.1%) \0.001 Cotinine level at baseline (mean ± SD) 143.3 ± 155.4 36.3 ± 73.1 \0.001 Cotinine level at follow-up (mean ± SD) 122.3 ± 128.6 33.7 ± 75.0 \0.001 Cotinine level at postpartum (mean ± SD) 248.8 ± 176.0 109.3 ± 149.2 \0.001 Depression at baseline 98 (51.3%) 82 (42.5%) 0.008

Depression at follow-up 78 (50.3%) 59 (34.3%) 0.003

Depression at postpartum 62 (32.6%) 42 (21.8%) 0.017

Intimate partner violence at baseline 73 (38.2%) 57 (29.5%) 0.072

Intimate partner violence at follow-up 16 (10.3%) 11 (6.4%) 0.200

Intimate partner violence at postpartum 22 (11.6%) 15 (7.8%) 0.210

Intervention group 88 (46.1%) 108 (56.0%) 0.053

Matern Child Health J (2011) 15:S96–S105 S101

123

of smoking during the follow-up period during pregnancy

but neither at baseline nor postpartum.

Other studies have confirmed the effect of psychosocial

challenges as a mediator of smoking in pregnancy [49, 50].

Alcohol and illicit drug use in the Washington, DC residents

we recruited may have also interfered with their ability to

quit. The relationship between alcohol use and smoking

during pregnancy has been previously confirmed, although

not in an exclusively African-American population [51].

Studies have shown smoking cessation in alcohol drinkers

to be more difficult due to reactivity between alcohol and

nicotine withdrawal [52, 53]. In our logistic models, alcohol

effect was not seen to be significantly associated with

smoking at any of the three time points. Illicit drug use

significantly increased the chances of smoking at baseline

and during the postpartum period. Illicit drug use may have

served as a surrogate for the severity of addiction to nico-

tine, a reliable marker for maintenance of smoking and

failure of cessation attempts among African-Americans

[54]. This is confirmed by our findings that active smoking

at baseline and cotinine level at baseline, markers for

intensity of smoking, were both predictors of smoking

during the follow-up interview closest to delivery and

postpartum. Other studies show similar results using

reported number of cigarettes smoked early in pregnancy as

a marker for intensity of smoking [40].

The literature shows an association between poverty and

smoking during pregnancy and postpartum [51, 55, 56]. In

this study we used Medicaid enrollment as a marker for

poverty. In bivariate analyses, Medicaid was a significant

predictor of smoking during the three time points. In the

logistic models, Medicaid lost its significance and a low

level of education was only significantly associated with

smoking at baseline. However, other studies have shown

that education is negatively associated with smoking dur-

ing pregnancy and with relapse after delivery [57]. It is

plausible that level of education may influence the

knowledge base mothers draw upon in their decision

making during pregnancy. A more compelling argument

would be a high resilience in mothers attaining higher

educational levels under challenging living conditions in

environments of urban poverty. Such women may also

possess a higher level of self-efficacy proven to impact

significantly on successful smoking cessation [40]. Women

with higher levels of education may be products of a more

supportive social environment, which is known to influence

successful quitting during pregnancy as well [39].

Some women may quit early in pregnancy due to

physical aversion to tobacco smoke during the first tri-

mester [58]; these pregnant mothers may then be suscep-

tible to relapse at a later stage. Our results show relapse

during pregnancy as reported by the population we studied

(13%) to be lower than previously reported in the Canadian

study (21%) [43]. In fact, the reported smoking rates in our

population declined from 39% at baseline to 33% during

follow-up.

The postpartum period represents a different challenge,

where a high percentage of women resume smoking after a

prolonged period of cessation. The literature cites media-

tors to resumed smoking such as postpartum depression

and concerns related to weight gain [59, 60]. Although this

has not been studied in populations that are predominately

Table 4 Logistic regression models to predict active

smoking among pregnant

women at baseline, follow-up,

and postpartum

a This model also controlled for

Medicaid status, alcohol use

during pregnancy, depression at

baseline and IPV at baseline b This model also controlled for

maternal age, education,

Medicaid enrollment status,

alcohol and illicit drug use

during pregnancy and IPV at

follow-up c This model also controlled for

maternal age, education,

Medicaid enrollment status,

alcohol, depression at

postpartum and IPV at baseline

Characteristic Odds ratio 95% Confidence interval

(A) Active smoking at baselinea

Maternal age 1.14 1.10, 1.18

Education level:

\High school 2.43 1.30, 4.54 Completed high school or GED 1.36 0.75, 2.47

At least some college (reference) 1.00 –

Illicit drug use 2.09 1.27, 3.44

(B) Active smoking at follow-upb

Active smoking at baseline 18.54 8.63, 39.84

Cotinine level at baseline (10 ng/ml) 1.09 1.05, 1.13

Depression at follow-up prior to delivery 2.69 1.27, 5.68

(C) Active smoking at postpartumc

Active smoking at baseline 10.89 5.28, 22.47

Cotinine level at baseline (10 ng/ml) 1.04 1.01, 1.08

Illicit drug use 2.38 1.11, 5.12

Intervention group 0.45 0.25, 0.80

S102 Matern Child Health J (2011) 15:S96–S105

123

black, in a study based on the Pregnancy Risk Assessment

Monitoring system, postpartum relapse was significantly

more likely among black mothers [59]. It is also possible

mothers are less aware of the harms of ETSE to infants as

compared to in utero exposure. The added stress of par-

enting a newly born infant, especially among mainly single

mothers with limited social network and community sup-

port, may trigger the need for stress relief associated with

smoking [49]. Sixty-five percent of women who quit during

pregnancy will relapse by 3 months and an additional 10%

by 6 months postpartum [61]. Other studies show that of

those who quit smoking during pregnancy, half relapse at

2–6 months [62] and 60–70% relapse within 1 year [63]

after delivery. In our study, women who reported actively

smoking increased from 33 to 50% during our follow-up

period of 10 weeks postpartum.

Few studies address the efficacy of interventions tar-

geting reduction of postpartum relapse [64, 65]. We were

encouraged this integrated intervention did impact on

relapse rates reported postpartum. This could be explained

by a longer exposure to the intervention, but also the

emphasis on ETSE as a significant risk to the newborn

infant, which may have encouraged mothers to maintain

their quit status. In addition, emphasis on mood regulation

could have assisted mothers in dealing with postpartum

depression and the stress associated with caring for a

newborn. Previous studies showing similar success post-

partum emphasized interventions including partners and

close friends, and encouraging the social networks to

support the mother in her decision [66, 67]. Although our

intervention did not address either of these strategies

directly, it encouraged women to establish a supportive

social network, and as such may have had similar effects.

Furthermore, studies emphasize the postpartum success of

interventions if they start earlier in pregnancy [67], which

was our case.

The strength and limitation of our study is that it was

conducted with high-risk African American women. The

results cannot be generalizable to other populations without

corroboration. Although the intervention did not influence

smoking behavior significantly during the pregnancy, it had

a protective effect against relapse during the postpartum.

This study also confirmed the importance of associated

addictions to illicit drugs and co-occurring depression as

important associations with smoking during pregnancy in

this population. More qualitative research to examine why

African American women may or may not be inclined to

stop smoking in pregnancy may inform research in the

future in the design of appropriate interventions with effi-

cacy in this population.

The results of this study support the importance of

screening early in pregnancy and providing mothers

with opportunities for behavioral modification through

culturally informed interventions. Behavioral interventions

for smoking should be available but cannot be relied upon

alone as the intervention of choice for mothers who con-

tinue to smoke during pregnancy. More research is needed

to test efficacy and safety of pharmacological therapy with

proven efficacy in non-pregnant populations. Studies such

as ours emphasize the importance of expanding prenatal

care beyond the medical model in order to respond to the

complex health risks of minority populations during

pregnancy.

Acknowledgments The authors wish to thank the field work staff, the interviewers, and data management staff. We wish to thank the

participants who welcomed us into their lives in hopes of helping

themselves and their children. This work was supported by grants no.

3U18HD030445; 3U18HD030447; 5U18HD31206; 3U18HD031919;

5U18HD036104, Eunice Kennedy Shriver National Institute of Child Health and Human Development and the National Center on Minority

Health and Health Disparities, National Institutes of Health, Depart-

ment of Health and Human Services. These analyses were supported,

in part, by the intramural program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development.

Conflict of interest None of the authors have any conflict of interests to declare.

References

1. Zdravkovic, T., Genbacev, O., McMaster, M. T., & Fisher, S. J.

(2005). The adverse effects of maternal smoking on the human

placenta: A review. Placenta, 26(Suppl A), S81–S86. 2. Bernstein, I. M., Mongeon, J. A., Badger, G. J., Solomon, L.,

Heil, S. H., & Higgins, S. T. (2005). Maternal smoking and its

association with birth weight. Obstetrics and Gynecology, 106(5), 986–991.

3. Jaddoe, V. W., Troe, E. J., Hofman, A., Mackenbach, J. P., Moll,

H. A., Steegers, E. A., et al. (2008). Active and passive maternal

smoking during pregnancy and the risks of low birth weight and

preterm: The generation R study. Paediatric and Perinatal Epi- demiology, 22, 162–171.

4. Li, C. Q., Windsor, R. A., Perkins, L., Goldenberg, R. L., &

Lowe, J. B. (1993). The impact on infant birth weight and ges-

tational age of cotinine-validated smoking reduction during

pregnancy. JAMA, 269, 1519–1524. 5. Magee, B. D., Hattis, D., & Kivel, N. M. (2004). Role of smoking

in low birth weight. Journal of Reproductive Medicine, 49, 23–27.

6. Peacock, J. L., Cook, D. G., Carey, I. M., Jarvis, M. J., Bryant, A.

E., Anderson, H. R., et al. (1998). Maternal cotinine level during

pregnancy and birth weight for gestational age. International Journal of Epidemiology, 27, 647–656.

7. Secker-Walker, R. H., Vacek, P. M., Flynn, B. S., & Mead, P. B.

(1997). Smoking in pregnancy, exhaled carbon monoxide, and

birth weight. Obstetrics and Gynecology, 89, 648–653. 8. Wang, X., Tager, I. B., Van Vunakis, H., Speizer, F. E., & Ha-

rahan, J. P. (1997). Maternal smoking during pregnancy, urine

cotinine concentrations, and birth outcomes. A prospective cohort

study. International Journal of Epidemiology, 26, 978–988. 9. Burns, L., Mattick, R. P., & Wallace, C. (2008). Smoking patterns

and outcomes in a population of pregnant women with other

Matern Child Health J (2011) 15:S96–S105 S103

123

substance use disorders. Nicotine & Tobacco Research, 10, 969–974.

10. Shea, A. K., & Steiner, M. (2008). Cigarette smoking during

pregnancy. Nicotine & Tobacco Research, 10, 267–278. 11. Hunt, C. E., & Hauck, F. R. (2006). Sudden infant death syn-

drome. CMAJ, 174(13), 1861–1869. 12. Salihu, H. M., & Wilson, R. E. (2007). Epidemiology of prenatal

smoking and perinatal outcomes. Early Human Development, 83, 713–720.

13. Jaakkola, J. J. K., & Gissler, M. (2004). Maternal smoking in

pregnancy, fetal development, and childhood asthma. American Journal of Public Health, 94(1), 136–140.

14. Day, N. L., Richardson, G. A., Goldschmidt, L., & Cornelius, M.

D. (2000). Effects of prenatal tobacco exposure on preschoolers’

behavior. Journal of Developmental and Behavioral Pediatrics, 21, 180–188.

15. Wakschlag, L. S., & Han, S. L. (2002). Maternal smoking during

pregnancy and conduct problems in high-risk youth: A develop-

mental framework. Development and Psychopathology, 14, 351–369.

16. Lumley, J., Chamberlain, C., Dowswell, T., Oliver, S. S., Oakley,

L., Watson, L. (2009). Interventions for promoting smoking

cessation during pregnancy. Cochrane Database Systems Review, 8(3), CD001055.

17. Camilli, A. E., McElroy, L. F., & Reed, K. L. (1994). Smoking

and pregnancy: A comparison of Mexican-American and non-

Hispanic white women. Obstetrics and Gynecology, 84, 1033–1037.

18. Wang, X., Zuckerman, B., Pearson, C., Kaufman, G., Chen, C.,

Wang, G., et al. (2002). Maternal cigarette smoking, metabolic

gene polymorphism, and infant birth weight. JAMA, 287, 195–202.

19. Moore, M. L., & Zaccaro, D. J. (2000). Cigarette smoking, low

birth weight, and preterm births in low-income African American

women. Journal of Perinatology, 3, 176–180. 20. Mathews, T. J., & MacDorman, M. F. (2006). Infant mortality

statistics from the 2003 period linked birth/infant death data set.

National Vital Statistics Reports, 54, 1–29. 21. Barnett, E. (1995). Race differences in the proportion of low birth

weight attributable to maternal cigarette smoking in a low-

income population. American Journal of Health Promotion, 10(2), 105–110.

22. Joseph, J. G., El-Mohandes, A. A. E., Kiely, M., El-Khorazaty,

M. N., Gantz, M. G., Johnson, A. A., et al. (2009). Reducing

psychosocial and behavioral pregnancy risk factors: Results of a

randomized clinical trial among high-risk pregnant African

American women. American Journal of Public Health, 99(6), 1053–1061.

23. El-Mohandes, A. A., Kiely, M., Joseph, J. G., Subramanian, S.,

Johnson, A. A., Blake, S. M., et al. (2008). An intervention to

improve postpartum outcomes in African-American mothers: A

randomized controlled trial. Obstetrics and Gynecology, 112(3), 611–620.

24. El-Khorazaty, M. N., Johnson, A. A., Kiely, M., El-Mohandes, A.

A. E., Subramanian, S., Laryea, H. A., et al. (2007). Recruitment

and retention of low-income minority women in a behavioral

intervention to reduce smoking, depression, and intimate partner

violence during pregnancy. BMC Public Health, 7, 233. 25. Straus, M. A., Hamby, S. L., Boney-McCoy, S., & Sugarman, D.

B. (1996). The Revised Conflict Tactics Scale (CTS2): Devel-

opment and preliminary psychometric data. Journal of Family Issues, 17, 283–316.

26. Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., &

Covi, L. (1974). The Hopkins Symptom Checklist (HSCL): A

self-report symptom inventory. Behavioral Science, 19, 1–15.

27. Katz, K. S., Blake, S. M., Milligan, R. A., Sharps, P. W., White,

D. B., Rodan, M. F., et al. (2008). The design, implementation

and acceptability of an integrated intervention to address multiple

behavioral and psychosocial risk factors among pregnant African

American women. BMC Pregnancy Childbirth, 8, 22. 28. Windsor, R. A. (2000). Counseling smokers in Medicaid mater-

nity care: The SCRIPT project. Tobacco Control, 9(Suppl 1), i62. 29. Whitlock, E. P., Orleans, C. T., Pender, N., & Allan, J. (2000).

Evaluating primary care behavioral counseling interventions: An

evidence-based approach. American Journal of Preventive Med- icine, 22(4), 267–284.

30. Parker, B., McFarlane, J., Soeken, K., Silva, C., & Reel, S.

(1999). Testing an intervention to prevent further abuse to

pregnant women. Research in Nursing and Health, 22, 59–66. 31. Webb, M. S., & Carey, M. P. (2008). Tobacco smoking among

low-income Black women: Demographic and psychosocial cor-

relates in a community sample. Nicotine & Tobacco Research, 10, 219–229.

32. Royce, J. M., Hymowitz, N., Corbett, K., Hartwell, T. D., &

Orlandi, M. A. (1993). Smoking cessation factors among African-

Americans and whites. COMMIT Research Group. American Journal of Public Health, 83, 220–226.

33. Fiore, M. C., Bailey, W. C., Cohen, S. J., Dorfman, S. F.,

Goldstein, M. G., Gritz, E. R., et al. (2000). Treating tobacco use and dependence. AHRQ publication no. 00–0032. Washington DC: US Department of Health and Human Services.

34. Piper, M. E., Fox, B. J., Welsch, S. K., Fiore, M. C., & Baker, T.

B. (2001). Gender and racial/ethnic differences in tobacco-

dependence treatment: A commentary and research recommen-

dations. Nicotine & Tobacco Research, 3, 291–297. 35. Gebauer, C., Kwo, C. Y., Haynes, E., & Wewers, M. (1998). A

nurse-managed smoking cessation intervention during pregnancy.

Journal of Obstetric, Gynecologic, and Neonatal Nursing, 27(1), 47–53.

36. Windsor, R. A., Lowe, J., Perkins, L., Smith-Yoder, D., Artz, L.,

Crawford, M., et al. (1993). Health education for pregnant

smokers: Its behavioral impact and cost benefit. American Jour- nal of Public Health, 83, 201–206.

37. Händel, G., Hannöver, W., Röske, K., Thyrian, J. R., Rumpf, H.

J., John, U., et al. (2009). Naturalistic changes in the readiness of

postpartum women to quit smoking. Drug and Alcohol Depen- dence, 101(3), 196–201.

38. Higgins, S. T., Heil, S. H., Badger, G. J., Skelly, J. M., Solomon,

L. J., & Bernstein, I. M. (2009). Educational disadvantage and

cigarette smoking during pregnancy. Drug and Alcohol Depen- dence, 104(Suppl 1), S100–S105.

39. Ma, Y., Goins, K. V., Pbert, L., & Ockene, J. K. (2005). Pre-

dictors of smoking cessation in pregnancy and maintenance

postpartum in low-income women. Maternal and Child Health Journal, 9, 393–402.

40. Morasco, B. J., Dornelas, E. A., Fischer, E. H., Oncken, C. A., &

Lando, H. A. (2006). Spontaneous smoking cessation during

pregnancy among ethnic minority women: A preliminary inves-

tigation. Addictive Behaviors, 31, 203–210. 41. Ruger, J. P., Weinstein, M. C., Hammond, S. K., Kearney, M. H.,

& Emmons, K. M. (2008). Cost-effectiveness of motivational

interviewing for smoking cessation and relapse prevention among

low-income pregnant women: A randomized controlled trial.

Value Health, 11(2), 191–198. 42. Canadian Task Force on the Periodic Health Examination.

(1994). The Canadian guide to clinical preventive health care. Ottawa: Canada Communication Group—Publishing, p. 28.

43. Kirkland, S. A., Dodds, L. A., & Brosky, G. (2000). The natural

history of smoking during pregnancy among women in Nova

Scotia. CMAJ, 163, 281–282.

S104 Matern Child Health J (2011) 15:S96–S105

123

44. Hall, S. M., Muñoz, R. F., & Reus, V. I. (1994). Cognitive-

behavioral intervention increases abstinence rates for depressive-

history smokers. Journal of Consulting and Clinical Psychology, 62, 141–146.

45. Hall, S. M., Muñoz, R. F., Reus, V. I., Sees, K. L., Duncan, C.,

Humfleet, G. L., et al. (1996). Mood management and nicotine

gum in smoking treatment: A therapeutic contact and placebo-

controlled study. Journal of Consulting and Clinical Psychology, 64, 1003–1009.

46. Kinnunen, T., Doherty, K., Militello, F. S., & Garvey, A. J.

(1996). Depression and smoking cessation: Characteristics of

depressed smokers and effects of nicotine dependence. Journal of Consulting and Clinical Psychology, 64, 791–798.

47. Glassman, A. H., Covey, L. S., Dalack, G. W., Stetner, F., Rivelli,

S. K., Fleiss, J. F., et al. (1993). Smoking cessation, clonidine,

and vulnerability to nicotine among dependent smokers. Clinical Pharmacology and Therapeutics, 54, 670–679.

48. Ginsberg, D., Hall, S. M., Reus, V. I., & Muñoz, R. F. (1995).

Mood and depression diagnosis in smoking cessation. Experi- mental and Clinical Psychopharmacology, 3, 389–395.

49. Goedhart, G., van der Wal, M. F., Cuijpers, P., & Bonsel, G. J.

(2009). Psychosocial problems and continued smoking during

pregnancy. Addictive Behaviors, 34, 403–406. 50. al’Absi, M., Carr, S. B., & Bongard, S. (2007). Anger and psy-

chobiological changes during smoking abstinence and in response

to acute stress: Prediction of smoking relapse. International Journal of Psychophysiology, 66, 109–115.

51. Martin, L. T., McNamara, M., Milot, A., Bloch, M., Hair, E. C.,

& Halle, T. (2008). Correlates of smoking before, during, and

after pregnancy. American Journal of Health Behavior, 32, 272–282.

52. McKee, S. A., Krishnan-Sarin, S., Shi, J., Mase, T., & O’Malley,

S. S. (2006). Modeling the effect of alcohol on smoking lapse

behavior. Psychopharmacology (Berl), 189, 201–210. 53. Zimmerman, R. S., Warheit, G. J., & Ulbrich, P. M. (1990). The

relationship between alcohol use and attempts and success at

smoking cessation. Addictive Behaviors, 15(3), 197–207. 54. Choi, W. S., Okuyemi, K. S., Kaur, H., & Ahluwalia, J. S. (2004).

Comparison of smoking relapse curves among African-American

smokers. Addictive Behaviors, 29, 1679–1683. 55. Jun, H.-J., & Acevedo-Garcia, D. (2007). The effect of single

motherhood on smoking by socioeconomic status and race/eth-

nicity. Social Science and Medicine, 65, 653–666. 56. Everett-Murphy, K., Steyn, K., Mathews, C., Petersen, Z., Ode-

ndaal, H., Gwebushe, N., et al. (2010). The effectiveness of

adapted, best practice guidelines for smoking cessation counsel-

ing with disadvantaged, pregnant smokers attending public sector

antenatal clinics in Cape Town, South Africa. Acta Obstetricia et Gynecologica Scandinavica, 89, 478–489.

57. Kahn, R. S., Certain, L., & Whittaker, R. C. (2002). A reexam-

ination of smoking before, during, and after pregnancy. American Journal of Public Health, 92, 1801–1808.

58. Pletsch, P. K., Pollak, K. I., Peterson, B. L., Park, J., Oncken, C.

A., Swamy, G. K., et al. (2008). Olfactory and gustatory sensory

changes to tobacco smoke in pregnant smokers. Research in Nursing and Health, 31, 31–41.

59. Allen, A. M., Prince, C. B., & Dietz, P. M. (2009). Postpartum

depressive symptoms and smoking relapse. American Journal of Preventive Medicine, 36, 9–12.

60. Quinn, G., Ellison, B. B., Meade, C., Roach, C. N., Lopez, E.,

Albrecht, T., et al. (2006). Adapting smoking relapse–prevention

materials for pregnant and postpartum women: Formative

research. Maternal and Child Health Journal, 10, 235–245. 61. US Department of Health and Human Services. (1990). The

Health Benefits of Smoking Cessation. US Department of Health and Human Services, Public Health Service, Centers for Disease

Control, Center for Chronic Disease prevention and Health Pro-

motion, Office on Smoking and Health. DHHS Publication No.

(CDC) 90-8416.

62. Carmaichael, S. L., & Ahluwalia, I. B. (2000). Correlates of

postpartum smoking relapse: Results from Pregnancy Risk

Assessment Monitoring System (PRAMS). American Journal of Preventive Medicine, 19, 193–196.

63. Severson, H. H., Andrews, J. A., Lichtenstein, E., Wall, M., &

Akers, L. (1997). Reducing maternal smoking and relapse: Long-

term evaluation of a pediatric intervention. Preventive Medicine, 26, 120–130.

64. Fang, W. L., Goldstein, A. O., Butzen, A. Y., Hartsock, S. A.,

Hartmann, K. E., Helton, M., et al. (2004). Smoking cessation in

pregnancy: A review of postpartum relapse prevention strategies.

The Journal of the American Board of Family Practice, 17(4), 264–275.

65. Reitzel, L. R., Vidrine, J. I., Businelle, M. S., Kendzor, D. E.,

Costello, T. J., Li, Y., et al. (2101). Preventing postpartum

smoking relapse among diverse low-income women: A random-

ized clinical trial. Nicotine & Tobacco Research, 12, 326–335. 66. Gielen, A. C., Windsor, R. A., Faden, R., O’Campo, P., Repke, J.,

& Davis, M. (1997). Evaluation of smoking cessation interven-

tion for pregnant women in an urban prenatal clinic. Health Education Research, 12, 247–254.

67. Mullen, P. D., Richardson, M. A., Quinn, V. P., & Ershoff, D. H.

(1997). Postpartum return to smoking: Who is at risk and when.

American Journal of Health Promotion, 11, 323–330.

Matern Child Health J (2011) 15:S96–S105 S105

123

Copyright of Maternal & Child Health Journal is the property of Springer Science & Business Media B.V. and

its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.

Vol.:(0123456789)1 3

Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113 https://doi.org/10.1007/s10995-018-2537-7

Randomized Controlled Trial of Doula-Home-Visiting Services: Impact on Maternal and Infant Health

Sydney L. Hans1  · Renee C. Edwards1 · Yudong Zhang1

Published online: 31 May 2018 © The Author(s) 2018, Corrected publication August/2018

Abstract Introduction Although home-visiting programs typically engage families during pregnancy, few studies have examined mater- nal and child health outcomes during the antenatal and newborn period and fewer have demonstrated intervention impacts. Illinois has developed an innovative model in which programs utilizing evidence-based home-visiting models incorporate community doulas who focus on childbirth education, breastfeeding, pregnancy health, and newborn care. This randomized controlled trial (RCT) examines the impact of doula-home-visiting on birth outcomes, postpartum maternal and infant health, and newborn care practices. Methods 312 young (M = 18.4 years), pregnant women across four communities were randomly assigned to receive doula-home-visiting services or case management. Women were African American (45%), Latina (38%), white (8%), and multiracial/other (9%). They were interviewed during pregnancy and at 3-weeks and 3-months postpartum. Results Intervention-group mothers were more likely to attend childbirth-preparation classes (50 vs. 10%, OR = 9.82, p < .01), but there were no differences on Caesarean delivery, birthweight, prematurity, or postpartum depression. Intervention-group mothers were less likely to use epidural/pain medication during labor (72 vs. 83%; OR = 0.49, p < .01) and more likely to initiate breastfeeding (81 vs. 74%; OR = 1.72, p < .05), although the breastfeeding impact was not sustained over time. Intervention-group mothers were more likely to put infants on their backs to sleep (70 vs. 61%; OR = 1.64, p < .05) and utilize car-seats at three weeks (97 vs. 93%; OR = 3.16, p < .05). Conclusions for practices The doula-home-visiting intervention was associated with positive infant-care behaviors. Since few evidence-based home-visiting programs have shown health impacts in the postpartum months after birth, incorporating doula services may confer additional health benefits to families.

Keywords Doula · Home visiting · Breastfeeding · Safe sleep

Significance

What’s Known on This Subject Research has shown that home-visiting programs have positive impacts in varied domains of parent and child functioning. However, few stud- ies have examined maternal and child health at birth and during the newborn period.

What This Study Adds This study, evaluating a home- visiting model that incorporates community doulas into the intervention team, demonstrates improvements in childbirth preparation, breastfeeding initiation, safe sleep practices, and early car-seat use. The intervention was associated with less use of pharmacologic pain control during labor, but not

with other indicators of mother and newborn health at birth or improvements in maternal depression.

Introduction

Home Visiting and Maternal Child Health

Growing evidence shows that childhood home-visiting pro- grams for socially and economically vulnerable families can have impacts in multiple areas, including maternal and child health, parenting, child development, and family economic self-sufficiency (Paulsell et al. 2010). When federal sup- port for home visiting was dramatically increased in 2010 through the Maternal Infant Early Childhood Home Visit- ing (MIECHV) program (Thompson et al. 2011), the leg- islation set expectations that program should have impacts across multiple domains, including “improved maternal and

* Sydney L. Hans [email protected]

1 School of Social Service Administration, University of Chicago, 969 E 60th Street, Chicago, IL 60637, USA

S106 Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

newborn health” (“Patient Protection and Affordable Care Act”). Although MIECHV legislation did not prioritize specific maternal and newborn health outcomes, the U.S. Department of Health and Human Services’ national health blueprint, Healthy People 2020 (Office of Disease Preven- tion and Health Promotion 2014), identifies such priorities: mother health at birth and postpartum (including attendance at childbirth preparation classes, reduction in Caesarean deliveries, reduction in maternal postnatal medical compli- cations, and reduced postpartum depression), infant morbid- ity and mortality (including reduction in infant deaths, low birthweight and preterm birth), and infant care (including increased breastfeeding and increased proportion of infants put to sleep on their backs).

The United States lags behind other developed nations with respect to infant mortality and low birthweight (Mac- Dorman et al. 2014; Wardlaw 2004), and there are dispari- ties in newborn and pregnancy outcomes related to maternal age, poverty and race (Bryant et al. 2010; Martin et al. 2017; Nagahawatte and Goldenberg 2008). Despite evidence that breastfeeding has advantages for mother and child health (Stuebe 2009), breastfeeding rates remain low in the US among young, low-income and African-American women (McDowell et al. 2008). Additionally, although the Ameri- can Academy of Pediatrics (Task Force on Sudden Infant Death Syndrome 2016) recommends that infants be placed in supine sleep positions in their own beds in order to reduce the risk of sleep-related infant deaths, infants born to young, low-income mothers have a relatively high risk for prone placement and for co-sleeping (Colson et al. 2009; Caraballo et al. 2016).

Despite many studies on infant and early childhood home visiting, few reports document impacts on maternal and newborn health or newborn care practices. Only a few home- visiting studies have examined maternal depression during the first postpartum months, and none have found program impacts reducing symptoms (e.g., Barlow et al. 2013; Carta et al. 2013). A few studies have shown impacts on preventing low birthweight and/or preterm birth (Lee et al. 2009; Wil- liams et al. 2017), but others have not (e.g., Kitzman et al. 1997; Olds et al. 1986). Most studies have not examined newborn health. Some home-visiting studies have reported impacts on early breastfeeding (Kitzman et al. 1997; Wen et al. 2011), but most have not found impacts (Green et al. 2014; Kemp et al. 2013; Mitchell-Herzfeld et al. 2005).

Community Doulas

Twenty years ago, early childhood advocates in Illinois were concerned about home-visiting programs having lim- ited impact on maternal and newborn health outcomes. A partnership between the Irving Harris Foundation, Health- Connect One and the Ounce of Prevention Fund developed

a model where doulas were integrated into home-visiting programs in order to enhance the quality of health-related services during pregnancy and the postnatal period (Glink 1998, 1999).

In the “community doula” model that resulted, doulas are community health workers who have training in pregnancy health, childbirth preparation, labor support, lactation coun- seling, and newborn care. They serve as specialized home visitors, providing home-based education and support dur- ing the last half of pregnancy and for 6 weeks postpartum. Doulas accompany laboring women to the hospital to pro- vide comfort measures and emotional support and to offer postpartum help around breastfeeding and bonding.

The rationale for including doulas within a home-based model drew from strong meta-analytic evidence that doula labor support is associated with improved health outcomes, including fewer Caesarean deliveries, decreased use of anal- gesia/anesthesia, shorter labors, and higher Apgar scores (Hodnett et al. 2013). One RCT examining the impact of a community doula model in which doulas provided home vis- its in addition to labor support found increases in breastfeed- ing initiation among young, low-income mothers (Edwards et al. 2013).

The goal of this RCT is to examine whether young, low- income families receiving doula-home-visiting services, compared to families receiving lower-intensity case-man- agement services, have improved maternal and child health outcomes during the period between birth and 3 months of age.

Methods

Study Sites, Enrollment, Randomization and Follow‑Up Procedures

Study recruitment took place between 2011 and 2015. Part- ners in the RCT were four agencies offering doula-home- visiting programs to young mothers in high-poverty Illinois communities. Two programs were located in a large city, and two in smaller urban areas. One served an African-American population, one served a Latinx population, and two served mixed-ethnic populations. Programs serving Latinx popu- lations provided services in English and Spanish. Each of the programs already was implementing an evidence-based home visiting model (see overview of evidence in Paul- sell et al. 2010), either Healthy Families America (HFA) (“Healthy Families America” 2015) or Parents as Teachers (PAT) (“Parents as Teachers” 2018). Programs were from a network of state-funded home-visiting programs and not demonstration programs for research purposes only.

Programs received information about young pregnant women from their usual referral networks—public health

S107Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

departments, WIC programs, health clinics, and schools. Program staff contacted women to determine eligibility and explain the program and research study. Women were told that the only way to participate in the doula-home-visiting program was to participate in the study. If they were not interested in the research, they received contact information for other community programs providing services for preg- nant women, including case management, home visiting, and parenting programs in hospitals and health clinics. To be eli- gible for the study, women needed to be under 26, less than 34 weeks gestation, living in the program geographic catch- ment area, planning to remain the area, and meeting sociode- mographic risk criteria used by the HFA or PAT models. Out of ethical concerns, pregnant women who were under 14, involved with the child welfare or juvenile justice systems, or had significant cognitive impairments were excluded from the study and offered home-visiting services.

After screening, the research team scheduled a baseline session with mothers that included a written-informed- consent procedure and a 2-h structured interview. At the end of this session, the interviewer opened a sealed opaque envelope that showed whether the participant was assigned to doula-home-visiting services (intervention condition) or case management (control condition). These envelopes had been prepared by the principal investigator before the beginning of the study. Randomization tables were created separately for each community.

At 37-weeks of pregnancy, 3-weeks postpartum and 3-months postpartum, mothers were re-interviewed. Fami- lies received modest monetary compensation at each session and a baby book and toy at each postpartum session. All study procedures were approved by the Institutional Review Board at The University of Chicago, and the study is regis- tered with clinicaltrials.gov [identifier NCT01947244].

Description of Group Conditions

Doula-Home-Visiting Intervention

After randomization to the intervention group, doula-home- visiting programs assigned families a home visitor (also called a Family Support Worker or Parent Educator) and a community doula. Doulas and home visitors all had deep roots in their communities. All home visitors and doulas had completed at least the foundational training required by their national models, and doulas had completed at least the basic training provided through the Ounce of Preven- tion Fund. During pregnancy and postpartum, mothers were visited weekly by a home visitor, doula, or both together. The doula worked with the mother more intensively during pregnancy and the first weeks postpartum, while the home visitor became the primary provider by 6 weeks postpartum.

Home visitors focused on the mother-infant relationship, child development, child safety, and educational-work plan- ning, as well as screening to make sure that family basic needs were being met. Doulas focused on issues related to pregnancy health, childbirth preparation, breastfeeding, newborn care, postpartum health, and early bonding. Doulas sometimes accompanied mothers to prenatal and postpartum medical visits. Doulas attended births at the hospital where they provided mothers with physical comfort, emotional sup- port, and advocacy during labor and delivery and breast- feeding counseling postpartum. Doulas also offered prenatal classes at the program sites. All programs conducted regular depression screenings and made referrals to mental health consultants.

Case Management Control

After randomization to the control group, mothers were pro- vided information about case management services in their communities, and case management providers were given mothers’ contact information. In some communities, moth- ers were referred to existing state-funded case-management providers; in other communities, social-service provid- ers were contracted to provide case management. It was expected that mothers would have at least two meetings with case managers—one during pregnancy and one after birth. Meetings could be in families’ homes, in agency offices, or occasionally by phone. Case managers determined whether families’ basic needs with respect to health, housing, food, employment, education, and childcare were being met, and if needed, made referrals. Case managers screened to identify needs for services regarding substance misuse, depression, and domestic violence.

Interviews

Outcomes were chosen based on Healthy People 2020 maternal and newborn health priorities and outcomes that have been reported in previous studies of doula interven- tions. Interviews were available in English and Spanish and administered in the mother’s preferred language. Interview- ers working in Latinx communities were bilingual. Inter- views were usually conducted in families’ homes.

At baseline, interviewers asked questions related to the pregnancy, health care, mental health, education and employment, and relationships with family. Baseline inter- view questions were used to check equivalence of the groups as randomized.

At all follow-up interviews, intervention-group mothers were asked about numbers of contacts with doulas and home visitors. All mothers were asked about childbirth preparation class attendance and any other pregnancy/parenting services.

S108 Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

At the 3-week postpartum interview (or 3-month inter- view if mother missed the earlier session), mothers reported on birth outcomes, including pain medication/epidural use during labor, vaginal versus Caesarean delivery, gestational age (GA) at delivery, infant birthweight, NICU admis- sion, length of hospital stay, and mother and/or newborn re-hospitalizations.

At the 3-week and 3-month interviews, mothers reported on breastfeeding. Breastfeeding initiation was defined as breastfeeding at least through the hospital stay. Mothers were asked how often they used a car-seat, the positions they used when laying down their infant to sleep and where the infant slept. Mothers reported on depressive symptoms using the Center for Epidemiological Studies-Depression Scale (CES- D) (Radloff 1977), dichotomized to identify mothers with clinically significant levels of depression (≥ 16).

Analytic Plan

First, the intervention and control groups were compared on multiple baseline maternal characteristics measured before randomization using t-tests and Chi square tests to check whether randomization was successful. Second, intent-to- treat logistic regression analyses were conducted to examine the impact of the doula-home-visiting intervention on out- comes measured at 37-weeks pregnancy, 3-week postpar- tum, and 3-months postpartum. Odds ratios, 95% confidence intervals, and one-tailed p-values were calculated for each outcome, using the control group as the reference group. Program site was used as a covariate in all analyses, and any baseline maternal variables that differed between the two groups were used as covariates.

Results

Sample Characteristics

Altogether 436 women were referred to the programs. 312 were enrolled in the sample and randomly assigned to the two conditions. Reasons families were not enrolled included inability to contact, women not wanting to participate in services or the study, women not meeting eligibility criteria, and women at high risk and referred to program services without randomization.

Interviews were completed for 256 mothers (82%) at 37-weeks of pregnancy, 283 mothers (91%) at 3-weeks and 278 mothers (89%) at 3-months. Sample attrition was unrelated to program site, race/ethnicity, age, education, co- residence, or prenatal depressive symptoms. There were no differences in sample attrition at either follow-up interview between the intervention and control groups. Figure 1 is a

CONSORT chart identifying the flow of subjects through the study.

Participants in the baseline interview were young and low-income, with almost half identifying as black/Afri- can American (45%, n = 140) and just over a third Latina/ Hispanic (38%, n = 117). 11% of mothers preferred to be interviewed in Spanish. Most mothers were in their second trimester of pregnancy and expecting their first child. Over two-thirds were partnered (coupled, engaged, married) with the father of the baby (71%, n = 220). Table 1 shows that the only baseline difference between groups was that more intervention-group mothers were living with a parent figure compared to control-group mothers (77 vs. 64%, p < .05). Co-residence with parent figure was a control variable in all analyses.

Intervention Participation

Virtually all mothers (99%, n = 153) assigned to the doula- home-visiting group received at least one home visit. Among mothers interviewed at 37 weeks, the average number of doula visits prior to 37 weeks was 8.9 (SD = 6.9) and the average number of visits from a home visitor was 5.8 (SD = 4.8). Doulas were present in the hospital for 75% (n = 106) of the births. By 3-months postpartum, 131 (92%) mothers had received at least one postpartum visit from their doula and 120 (84%) had received at least one postpartum visit from their home visitor.

Intervention Effects

Mother Birth and Postpartum Health

Results from logistic regression analyses using one-tailed hypothesis tests (Table 2) show that intervention-group mothers were more likely to attend a childbirth education class during pregnancy (OR 9.82, 95% CI 4.84–19.89) and less likely to use epidural or other pain medication during labor compared to control-group mothers (OR 0.47, 95% CI 0.25–0.88). The intervention was not associated with Caesarean deliveries, mother re-hospitalizations, or mother postpartum depressive symptoms.

Infant Mortality and Morbidity

The intervention was not associated with preterm births (GA < 37 weeks), low birthweight, NICU admission, length of newborn hospital stay, re-hospitalization of infants, hav- ing a pediatrician or pediatric clinic at 3 weeks, or having a well-baby check up by 3 months. Almost all families in both groups reported having a pediatrician for their infants (98%), and all mothers reported taking their infant in for at least one well-baby check up by 3 months of age.

S109Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

Health clinics, schools, and social service agencies identify young pregnant women and refer to home-visiting programs.

N=436

Informed consent and baseline interview N=312

Randomized to Doula Home-Visiting n=156

Randomized to Case Management n=156

3 week postpartum followup n=141

• Unable to contact 8 • Unable to schedule 2 • Declined 4 • Infant/fetal death 0 • Caregiver change 1

3 week postpartum followup n=142

• Unable to contact 3 • Unable to schedule 2 • Declined 7 • Infant/fetal death 2 • Caregiver change 0

3 month followup n=139

• Unable to contact 5 • Unable to schedule 2 • Declined 8 • Infant/fetal death 2 • Caregiver change 0

3 month followup n=139

• Unable to contact 11 • Unable to schedule 0 • Declined 5 • Infant/fetal death 0 • Caregiver change 1

• Couldn’t be reached: 44 • Didn’t meet eligibility criteria: 45 • Not interested: 23 • High risk and referred to program: 12

37 week pregnancy followup n=127

• Delivered before 37 weeks 8 • Unable to contact 2 • Unable to schedule 12 • Declined 6 • Fetal death 1

37 week pregnancy followup n=129

• Delivered before 37 weeks 5 • Unable to contact 6 • Unable to schedule 12 • Declined 4 • Infant death 0

Fig. 1 Study CONSORT diagram

S110 Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

Newborn Care Practices

Mothers in the intervention group were more likely to initiate breastfeeding in the hospital (OR 1.67, 95% CI 0.91–3.03). At 3 weeks, mothers in the intervention group were more likely to always place their infants on their backs for sleeping (OR 1.64, 95% CI 0.97–2.77) and to always put their infants in a car seat when traveling by car (OR 3.67, 95% CI 1.06–12.70). There was a non-statistically- significant trend for infants in the intervention group to have their own beds (OR 1.44, 95% CI 0.89–2.34, p = .07). There were no group differences on breastfeeding, sleeping or car seat use at 3 months.

Discussion

Although most early-childhood home-visiting programs begin working with families during pregnancy or soon after birth, relatively few evaluations have examined maternal and child health outcomes at birth or during the newborn period. The doula-home-visiting model, in which a com- munity doula partners with a home visitor during preg- nancy and through 6-weeks postpartum, provides greater emphasis on pregnancy health, childbirth, breastfeeding, and newborn health than most other home-visiting models, and additionally, offers hospital-based support during childbirth and agency-based childbirth preparation classes. This RCT shows that the doula-home-visiting intervention has impacts

on childbirth preparation, epidural/pain medication use dur- ing labor, breastfeeding, and safe newborn-care practices.

Mothers receiving the intervention were more likely to have attended a childbirth preparation class. Although vir- tually all mothers in the sample had opportunities to attend childbirth classes through prenatal clinics and hospitals, few control-group women took advantage of such opportunities. Half of the women in the intervention group participated in such classes either at clinics and hospitals or through weekly classes offered by their home-visiting programs. Moreover, all mothers who were visited by a doula also received indi- vidualized childbirth education at home. Perhaps as a result of this preparation and the presence of the doula during labor, mothers in the intervention group were less likely to use pharmacologic pain relief during labor, a finding similar to other studies of doula labor support (Hodnett et al. 2013). However, as with the few other home-visiting studies exam- ining birth outcomes, there were no intervention impacts on Caesarean deliveries, low birthweight, or preterm birth. Although other studies of labor-only doulas have found reductions of Caesarean rates (Hodnett et al. 2013), most of these studies limited samples to obstetrically low-risk moth- ers whose labors began spontaneously. The present sample of young, low-income mothers was likely more medically complex.

Mothers in the intervention were more likely to initiate breastfeeding, consistent with previous research on commu- nity doulas (Edwards et al. 2013). Few other home-visiting studies have found impacts on breastfeeding. Doulas, by

Table 1 Characteristics of doula-home-visiting intervention group and control group at enrollment

a Chi-square test shows significant difference between intervention and control groups at p < .05

Control group Doula/HV group n = 156 n = 156

Mother age in years M (SD) 18.3 (1.6) 18.5 (2.0) Mother years of school completed M (SD) 10.9 (1.5) 10.9 (1.5) Mother race/ethnicity n (%)  African American 72 (46.2%) 68 (43.6%)  Latina/Hispanic 56 (35.9%) 61 (39.1%)  White 13 (8.3%) 13 (8.3%)  Multiracial/other 15 (9.6%) 14 (9.0%)

Mother attends school n (%) 78 (50.0%) 86 (55.1%) Mother employed n (%) 28 (18.0%) 31 (19.9%) Mother expecting first child n (%) 154 (98.7%) 152 (97.4%) Baby gestational age in weeks M (SD) 25.7 (5.9) 25.5 (6.0) Mother has received prenatal care n (%) 154 (98.7%) 155 (99.4%) Mother receives public insurance (n = 305) n (%) 138 (90.8%) 140 (91.5%) Mother receives WIC n (%) 137 (87.8%) 131 (84.0%) Mother depressive symptoms (CES-D) M (SD) 14.2 (9.2) 13.8 (8.5) Co-residing with own mother or other parent figurea n (%) 100 (64.1%) 120 (76.9%) Co-residing with baby’s father n (%) 48 (30.8%) 39 (25.0%) Partnered with baby’s father n (%) 107 (68.6%) 113 (72.4%)

S111Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

offering skilled lactation counseling throughout pregnancy in mothers’ homes and postpartum in the hospital, increase breastfeeding initiation, even among populations that have traditionally low breastfeeding rates. However, the interven- tion impact on breastfeeding was not sustained, and only about 20% of mothers were breastfeeding at 3 months. Research is needed to understand why many mothers initi- ated breastfeeding but discontinued quickly postpartum (e.g., Rozga et al. 2015) and what strategies might be effective for supporting young mothers during that critical time. Never- theless, even brief periods of breastfeeding may have health benefits to infants by way of colostrum (e.g., Bardanzellu et al. 2017).

The doula-home-visiting intervention did not show impacts on postpartum maternal depressive symptoms, con- sistent with findings from most other home-visiting evalu- ations. Postpartum depression is powerfully influenced by a complex set of biological factors, chronic stress, trauma history, and instability in relationships with infants’ fathers

and have been challenging to prevent (Edwards et al. 2012; Grote et al. 2011). A systematic review found evidence that home-based services have the potential to be effective in preventing postpartum depression, but to date evidence is limited to intensive interventions delivered by professionals (Dennis and Dowswell 2013).

Mothers in the intervention were more likely to always place their newborns on their backs to sleep and always use a car-seat. Few previous home-visiting studies have looked at early infant safety practices. Although the pre- sent study does not address the manner in which mothers received these safety messages, previous research suggests that low-income mothers may reject infant sleep recom- mendations, for example, because of distrust of health pro- fessionals, reliance on advice from family members, and concern for infant comfort (Colson et al. 2005). Doulas have many opportunities during prenatal visits and through their intimate care during labor to become trusted advi- sors to young mothers. By being present in the hospital

Table 2 Intervention impacts on maternal health, newborn health, and newborn care outcomes

a Logistic regression analyses control for co-residence with parent figure at baseline and program site b A third infant from the control group died before age 4 months of age c Two infants were in the hospital continuously from birth through 3 weeks of age so were excluded from analyses on re-hospitalizations

Control group n (%)

Doula/HV group n (%)

OR [95% CI]a p-value (1-tailed)

Mother birth and postpartum health  Entered labor having attended childbirth preparation class (n = 255) 12 (9.5%) 64 (50.0%) 9.82 [4.84, 19.89] 0.00  C-section birth (n = 286) 31 (21.5%) 33 (23.2%) 1.04 [0.59, 1.84] ns  Epidural/pain medication use during labor (n = 268) 114 (83.2%) 94 (71.76%) 0.47 [0.25, 0.88] 0.01  Mother re-hospitalized within 3 weeks (n = 286) 3 (2.1%) 4 (2.8%) 1.53 [0.33, 7.21] ns  3 week high depressive symptoms (n = 282) 31 (21.8%) 31 (22.1%) 0.96 [0.53, 1.71] ns  3 month high depressive symptoms (n = 277) 21 (15.1%) 18 (13.0%) 0.95 [0.47, 1.91] 0.45

Infant morbidity and mortality  Fetal/newborn death (n = 286) 2b (1.3%) 0 (0.0%) – –  Preterm birth (GA < 37 weeks; n = 285) 12 (8.2%) 10 (6.7%) 0.57 [0.22, 1.46] 0.18  Low birthweight (n = 285) 13 (9.0%) 9 (6.4%) 0.64 [0.26, 1.59] 0.17  NICU admission (n = 286) 23 (16.0%) 21 (14.8%) 0.87 [0.45, 1.68] 0.34  Hospital stay ≥ 4 days (n = 286) 28 (19.4%) 25 (17.6%) 0.89 [0.48, 1.63] 0.35  Has pediatrician at 3 weeks (n = 282) 139 (97.9%) 138 (98.6%) 1.56 [0.25, 9.65] 0.32  Has pediatric checkup by 3 months of age (n = 278) 139 (100.0%) 139 (100.0%) – –  Infant re-hospitalized within 3 weeksc (n = 284) 5 (3.6%) 3 (1.4%) 0.45 [0.08, 2.48] 0.18

Newborn care practices  Breastfeeding initiation (n = 287) 107 (74.3%) 116 (81.1%) 1.67 [0.91, 3.03] 0.05  Breastfeeding at 3 months (n = 278) 31 (21.8%) 24 (16.9%) 0.85 [0.45, 1.60] ns  Always puts infant on back to sleep at 3 weeks (n = 282) 86 (60.6%) 98 (70.0%) 1.64 [0.97, 2.77] 0.03  Always puts infant on back to sleep at 3 months (n = 277) 83 (60.1%) 92 (66.2%) 1.34 [0.80, 2.23] 0.13  Infant sleeps in own bed at 3 weeks (n = 282) 63 (44.4%) 74 (52.9%) 1.44 [0.89, 2.34] 0.07  Infant sleeps in own bed at 3 months (n = 277) 67 (48.6%) 71 (51.1%) 1.19 [0.72, 1.95] 0.25  Always uses car seat at 3 weeks (n = 281) 132 (93.0%) 135 (97.1%) 3.67 [1.06, 12.70] 0.02  Always uses car seat at 3 months (n = 277) 126 (91.3%) 130 (93.5%) 1.28 [0.51, 3.20] 0.30

S112 Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

and the home during the earliest weeks when mothers first establish sleep practice, doulas may have unique oppor- tunities to explain to mothers and other family members the benefits of safe sleep practices and to offer mothers strategies for soothing babies who seem uncomfortable on their backs.

Notably, although doula-home-visiting impacts on new- born care practices were found in the first weeks postpartum, group differences diminished by 3 months. It may be that over time families chose infant feeding or sleeping prac- tices they felt were most effective for their family circum- stances or infant preferences. It may be that as home visitors took over intervention work from the doulas, the focus of the work shifted from feeding and sleep practices to other important areas such as responsive parenting, child develop- ment, and mother’s personal development.

Finally, although the present study has many methodo- logical strengths—a randomized design implemented within programs taken to scale in real agency settings, it also has limitations. The sample drew from only four programs in a single state and excluded adolescents at the most extreme levels of risk. Because data in this paper were provided through mother report and not administrative records, reli- able information on important medical procedures and out- comes during labor, such as qualifications of health pro- viders and Apgar scores, were unavailable. Because each mother in the intervention group was offered services from a doula and home visitor team, the independent contribution of the two different providers could not be determined. The sample was underpowered to detect important, but relatively rare maternal and child health problems, particularly infant mortality. Nevertheless, the study identified impacts on important maternal and newborn health outcomes that have rarely or never been found in other evaluations of home- visiting models. Future research should focus on a broader set of health outcomes, including outcomes documented through administrative records, and examine the processes through which doulas convey health information, contrasting their role to home visitors who are not doulas.

Consistent with an already strong evidence base regard- ing birth doula interventions and a smaller body of work on community doulas, the present study shows improved maternal and child health when mothers have access to doula services through community-based home-visiting programs. However, there are presently funding barriers to increas- ing low-income women’s access to doula services. Simple steps to improving access would be for states to recognize community doula services as evidence-based interventions eligible for existing home-visiting funding, as has been the case in Illinois, and also to develop state certification pro- cesses and other mechanisms for using Medicaid funds to reimburse for doula services, as has happened in Oregon and Minnesota (Gay 2016; Kozhimannil et al. 2014).

Acknowledgements This project was funded by award D89MC23146 from the MIECHV competitive grant program from the Health Resources and Services Administration (HRSA) to the State of Illinois Department of Human Services (IDHS). The contents of this publica- tion are solely the responsibility of the authors and do not represent the official views of HRSA or IDHS. The authors would like to thank their partners at the Ounce of Prevention Fund, the Illinois Governor’s Office of Early Childhood Development, and the agencies that imple- mented the doula home visiting and case management interventions. The authors thank project director, Linda Henson, data base manager, Marianne Brennan, and the research staff involved in collecting the data, including Tanya Auguste, Melissa Beckford, Ikesha Cain, Adri- ana Cintron, Nicole Dosie-Brown, Tytannie Harris, Morgan Johnson- Doyle, Katarina Klakus, Natasha Malone, Jasmine Nash, Erika Oslako- vic, Jillian Otto, Amy Pinkston, Magdalena Rivota, Rosa Sida-Nanez, Luz Silva, Caroline Taromino, Ardine Tennial, Maria Torres, Yadira Vieyra, and Lauren Wilder.

Compliance with Ethical Standards

Conflict of interest The authors declare that they have no conflict of interest.

Open Access This article is distributed under the terms of the Crea- tive Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

References

Bardanzellu, F., Fanos, V., & Reali, A. (2017). “Omics” in human colustrum and mature milk: Looking to old data with new eyes. Nutrients, 9, 843.

Barlow, A., Mullany, B., Neault, N., Compton, S., Carter, A., Hast- ings, R., … Walkup, J. T. (2013). Effect of a paraprofessional home-visiting intervention on American Indian teen mothers’ and infants’ behavioral risks: A randomized controlled trial. American Journal of Psychiatry, 170(1), 83–93.

Bryant, A. S., Worjoloh, A., Caughey, A. B., & Washington, A. E. (2010). Racial/ethnic disparities in obstetric outcomes and care: Prevalence and determinants. American Journal of Obstetrics & Gynecology, 202(4), 335–343.

Caraballo, M., Shimasaki, S., Johnston, K., Tung, G., Albright, K., & Halbower, A. C. (2016). Knowledge, attitudes, and risk for sud- den unexpected infant death in children of adolescent mothers: A qualitative study. Journal of Pediatrics, 174, 78–83.

Carta, J. J., Lefever, J. B., Bigelow, K., Borkowski, J., & Warren, S. F. (2013). Randomized trial of a cellular phone-enhanced home visitation parenting intervention. Pediatrics, 132(Supplement 2), S167–S173.

Colson, E. R., McCabe, L. K., Fox, K., Levenson, S., Colton, T., Lister, G., … Corwin, M. J. (2005). Barriers to following the back-to- sleep recommendations: Insights from focus groups with inner- city caregivers. Ambulatory Pediatrics, 5(6), 349–354.

Colson, E. R., Rybin, D., Smith, L. A., Colton, T., Lister, G., & Cor- win, M. J. (2009). Trends and factors associated with infant sleep- ing position: The National Infant Sleep Position Study, 1993– 2007. Archives of Pediatrics & Adolescent Medicine, 163(12), 1122–1128.

S113Maternal and Child Health Journal (2018) 22 (Suppl 1):S105–S113

1 3

Dennis, C.-L., & Dowswell, T. (2013). Psychosocial and psychologi- cal interventions for preventing postpartum depression. Cochrane Database of Systemic Reviews, 2, CD001134.

Edwards, R. C., Thullen, M. J., Isarowong, N., Shiu, C.-S., Henson, L., & Hans, S. L. (2012). Supportive relationships and the trajectory of depressive symptoms among young, Afrian American mothers. Journal of Family Psychology, 26(3), 585–594.

Edwards, R. C., Thullen, M. J., Korfmacher, J., Lantos, J. D., Henson, L. G., & Hans, S. L. (2013). Breastfeeding and complementary food: Randomized trial of community doula home visiting. Pedi- atrics, 132, S160-S166.

Gay, E. D. (2016). Insurance coverage of doula care would benefit patients and service providers alike. Rewire. Retrieved May 28, 2018, from https ://rewir e.news/artic le/2016/01/14/insur ance-cover age-doula -care-benefi t-patie nts-servi ce-provi ders-alike /.

Glink, P. (1998). The Chicago Doula Project: A collaborative effort in perinatal support for birthing teens. Zero to Three, 18, 44–50.

Glink, P. (1999). Engaging, educating, and empowering young moth- ers: The Chicago Doula Project. Zero to Three, 20, 41–44.

Green, B. L., Tarte, J. M., Harrison, P. M., Nygren, M., & Sanders, M. B. (2014). Results from a randomized trial of the Healthy Fami- lies Oregon accredited statewide program: Early program impacts on parenting. Children and Youth Services Review, 44, 288–298.

Grote, N. K., Bledsoe, S. E., Wellman, J., & Brown, C. (2011). Depres- sion in African American and White women with low incomes: The role of chronic stress. In L. E. Davis (Ed.), Racial disparity in mental health services: Why race still matters. Philadelphia, PA: Haworth Press.

Healthy Families America. (2015). Retrieved May 28, 2018, from http://www.healt hyfam ilies ameri ca.org.

Hodnett, E. D., Gates, S., Hofmeyr, G. J., & Sakala, C. (2013). Con- tinuous support for women during childbirth. Cochrane Library, Issue 7.

Kemp, L., Harris, E., McMahon, C., Matthey, S., Vimpani, G., Ander- son, T., … Aslam, H. (2013). Benefits of psychosocial interven- tion and continuity of care by child and family health nurses in the pre and postnatal period: process evaluation. Journal of Advanced Nursing, 69(8), 1850–1861.

Kitzman, H., Olds, D. L., Henderson, J., Hanks, C. R., Cole, C., Tate- lbaum, R. R., et al (1997). Effect of prenatal and infancy home visitation by nurses on pregnancy outcomes, childhood injuries, and repeated childbearing: A randomized controlled trial. JAMA, 278, 644–652.

Kozhimannil, K. B., Attanasio, L. B., Jou, J., Joarnt, L. K., Johnson, P. J., & Gjerdingen, D. K. (2014). Potential benefits of increased access to doula support during childbirth. American Journal of Managed Care, 20(8), e111–e121.

Lee, E., Mitchell-Herzfeld, S., Lowenfels, A. A., Greene, R., Dora- bawila, V., & DuMont, K. A. (2009). Reducing low birth weight through home visitation: A randomized controlled trial. American Journal of Preventive Medicine, 36(2), 154–160.

MacDorman, M. F., Matthews, T., Mohangoo, A. D., & Zeitlin, J. (2014). International comparisons of infant mortality and related factors: United States and Europe 2010. National vital statistics reports: from the Centers for Disease Control and Prevention, 63(5), 1–7.

Martin, J. A., Hamilton, B. E., Osterman, M. J., Driscoll, A. K., & Mathews, T. J. (2017). Births: Final Data for 2015. National vital statistics reports: From the Centers for Disease Control and

Prevention. National Center for Health Statistics, National Vital Statistics System, 66(1), 1.

McDowell, N. M., Wang, C. Y., & Kennedy-Stephenson, J. (2008). Breastfeeding in the United States: Findings from the national health and nutrition examination surveys, 1999–2006. NCHS Data Brief, 5, 1–8.

Mitchell-Herzfeld, S., Izzo, C., Greene, R., Lee, E., & Lowenfels, A. (2005). Evaluation of Healthy Families New York (HFNY): First year program impacts. New York State Albany: Office of Children and Family Services Bureau.

Nagahawatte, N. T., & Goldenberg, R. L. (2008). Poverty, maternal health, and adverse pregnancy outcomes. Annals of the New York Academy of Sciences, 1136, 80–85.

National Institue of Child Health and Human Development. (n.d.). Safe to Sleep 2017. Retrieved from https ://www.nichd .nih.gov/sts.

Office of Disease Prevention and Health Promotion. (2014). Healthy People 2020: Maternal, infant and child health. Retrieved 31 May 2017, from https ://www.healt hypeo ple.gov/2020/topic s-objec tives /topic /mater nal-infan t-and-child -healt h/objec tives .

Olds, D. L., Henderson, C. R. Jr., Tatelbaum, R., & Chamberlin, R. (1986). Improving the delivery of prenatal care and outcomes of pregnancy: A randomized trial of nurse home visitation. Pediat- rics, 77(1), 16–28.

Parents as Teachers. (2018). Retrieved May 28, 2018, from https :// paren tsast eache rs.org/.

Patient Protection and Affordable Care Act 42 U.S.C. § 18001 et seq. (2010).

Paulsell, D., Avellar, S., Martin, S., E., & Del Grosso, P. (2010). Home visiting evidence of effectiveness review: Executive Summary. Washington, DC: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services.

Radloff, L. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Journal of Applied Psychologi- cal Measure, 1, 385–401.

Rozga, M. R., Kerver, J. M., & Olson, B. H. (2015). Self-reported reasons for breastfeeding cessation among low-income women enrolled in a peer counseling breastfeeding support program. Journal of Human Lactation, 31(1), 129–137.

Stuebe, A. (2009). The risks of not breastfeeding for mothers and infants. Reviews in Obstetrics & Gynecology, 2(4), 222–231.

Task Force on Sudden Infant Death Syndrome. (2016). SIDS and other sleep-related infant deaths: Updated 2016 recommendations for a safe infant sleeping environment. Pediatrics, 138(5), e20162938. https ://doi.org/10.1542/peds.2016-2938.

Thompson, D. K., Clark, M. J., Howland, L. C., & Mueller, M.-R. (2011). The Patient Protection and Affordable Care Act of 2010 (PL 111–148): An analysis of maternal-child health home visita- tion. Policy Politics Nursing Practice, 12(3), 175–185.

Wardlaw, T. M. (Ed.). (2004). Low birthweight: Country, regional and global estimates. New York: UNICEF.

Wen, L. M., Baur, L. A., Simpson, J. M., Rissel, C., & Flood, V. M. (2011). Effectiveness of an early intervention on infant feeding practices and “tummy time”: A randomized controlled trial. Archives of Pediatrics & Adolescent Medicine, 165(8), 701–707.

Williams, C. M., Cprek, S., Asaolu, I., English, B., Jewell, T., Smith, K., & Robl, J. (2017). Kentucky health access nurturing develop- ment services home visiting program improves maternal and child health. Maternal and Child Health Journal, 21(5), 1166–1174.

Maternal & Child Health Journal is a copyright of Springer, 2018. All Rights Reserved.

ORIGINAL ARTICLE

An intervention to reduce postpartum depressive symptoms: a randomized controlled trial

Elizabeth A. Howell & Susan Bodnar-Deren & Amy Balbierz & Holly Loudon & Pablo A. Mora & Caron Zlotnick & Jason Wang & Howard Leventhal

Received: 1 May 2013 /Accepted: 27 August 2013 /Published online: 10 September 2013 # Springer-Verlag Wien 2013

Abstract Depressive symptoms and depression are a com- mon complication of childbirth, and a growing body of liter- ature suggests that there are modifiable factors associated with their occurrence. We developed a behavioral educational

intervention targeting these factors and successfully reduced postpartum depressive symptoms in a randomized trial among low-income black and Latina women. We now report results of 540 predominantly white, high-income mothers in a second randomized trial. Mothers in the intervention arm received a two-step intervention that prepared and educated mothers about modifiable factors associated with postpartum depres- sive symptoms (e.g., physical symptoms, low self-efficacy), bolstered social support, and enhanced management skills. The control arm received enhanced usual care. Participants were surveyed prior to randomization, 3 weeks, 3 months, and 6 months postpartum. Depressive symptoms were assessed using the Edinburgh Postnatal Depression Scale (EPDS of 10 or greater). Prevalence of depressive symptoms postpartum was unexpectedly low precluding detection of difference in rates of depressive symptoms among intervention versus en- hanced usual care posthospitalization: 3 weeks (6.0 vs. 5.6 %, p =0.83), 3 months (5.1 vs. 6.5%, p =0.53), and 6 months (3.6 vs. 4.6 %, p =0.53).

Keywords Postpartum depression . Randomized trial .

Behavioral intervention . Psychosocial

Introduction

Depressive symptoms and depression are one of the most common complications of childbirth and are the leading cause of disease-related disability among women (Gaynes et al. 2005; Kessler 2003). Estimates of prevalence rates for these symp- toms range from 10 to over 50 % depending on the screening instrument used and the population studied (Beeghly et al. 2003; Vera et al. 1991; Zayas et al. 2002). Symptoms of distress and depression impact hundreds of thousands of women annu- ally in the USA and affect women from all racial/ethnic and socioeconomic backgrounds. Both major depressive disorder

Supported by the National Institute of Mental Health (5R01MH77683) and the National Institute on Minority Health and Health Disparities (5P60MD000270).

Clinical trial registration Clinicaltrials.gov, www.clinicaltrials.gov, NCT00951717

E. A. Howell (*) : S. Bodnar-Deren :A. Balbierz : J. Wang Department of Health Evidence and Policy, Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, Box 1077, New York, NY 10029-6574, USA e-mail: [email protected]

E. A. Howell :H. Loudon Department of Obstetrics, Gynecology, and Reproductive Science, Icahn School of Medicine at Mount Sinai, New York, NY, USA

E. A. Howell Department of Psychiatry, Icahn School of Medicine at Mount Sinai, New York, NY, USA

P. A. Mora PsychologyDepartment, University of Texas at Arlington, Arlington, USA

C. Zlotnick Department of Psychiatry and Human Behavior, Warren Alpert Brown Medical School, Women and Infants Hospital, Providence, RI, USA

H. Leventhal Institute for Health, Health Care Policy and Aging Research and Department of Psychology, Rutgers, the State University of New Jersey, New Brunswick, NJ, USA

S. Bodnar-Deren Department of Sociology, Virginia Commonwealth University, 923 W. Franklin Street, Office 509, Richmond, VA 23284-2040, USA

Arch Womens Ment Health (2014) 17:57–63 DOI 10.1007/s00737-013-0381-8

and moderate symptoms of depression negatively affect the quality of life and daily functioning of mothers (Josefsson et al. 2002). Whether these symptoms do or do not reach the level of diagnosed major depressive disorder, they are associ- ated with multiple, negative consequences for mothers and infants (Dennis and Ross 2005; Duer et al. 1988; Herrera et al. 2004; Kroenke et al. 2001, 2003; Matthey et al. 2000; McLearn et al. 2006;Minkovitz et al. 2005; Seguin et al. 1999).

A growing body of literature suggests that potentially modi- fiable factors are associated with depressive symptoms postpar- tum. Situational factors such as postpartum physical symptoms, infant colic, overload from daily demands, and poor social support are associated with and likely trigger depressive symp- toms (Howell et al. 2005, 2006, 2009, 2010). We hypothesized that one or more of the three subsets of depressive symptoms, e.g., somatic, mood, and self-critical component, could be trig- gered and moderated by situational stimuli and that an interven- tion designed to normalize the postpartum experience would lower rates of these symptoms. The intervention had three ob- jectives: (1) to clarify the immediate postpartum experience as “normal,” i.e., shared event (e.g., physical symptoms and infant and social demands); (2) to provide simple procedures for man- agement and enhancing social support; and (3) to define the time frame for recovery. A randomized trial testing this intervention showed a significant reduction in positive depression screens among self-identified black and Latina postpartum mothers for up to 6 months of follow-up (Howell et al. 2012).

In this report, we present the results of a second randomized controlled trial testing this intervention among a different sociodemographic cohort of women: white and Asian mothers.

Methods

Study sample

Participants included 540 self-identified white and Asian postpartum mothers who delivered between July 2009 and April 2010 at a large tertiary inner-city hospital located in East Harlem, New York City. An earlier trial had recruited black and Latina postpartum mothers, and therefore, only white and Asian (non-black and non-Latina) women were in- cluded in this trial. The Program for the Protection of Human Subjects (the Institutional Review Board) at Mount Sinai School of Medicine approved this study. Eligible partic- ipants were ≥18 years of age and had infants with birth weights ≥2,500 g and 5-min Apgar scores ≥7. Women were identified by an electronic delivery system which includes information on maternal race/ethnicity, maternal age, infant Apgar scores, and infant birth weight.

Mothers whomet initial requirements were deemed eligible if they then self-identified as white, Asian, or other (non-black and non-Latina) minority women when questioned by a clinical

research coordinator. The clinical research coordinator approached mothers in hospital between their delivery day and postpartumday 2; approximately 12 to 17 patients were recruited Mondays through Fridays each week. Mothers who consented completed a 20-min survey and were assigned to intervention or enhanced usual care. Assignment involved giving each patient a number between 1 and 20, based on the order of delivery date, and using a list that had randomly assigned these numbers to intervention and control; the computer randomized list was prepared by the project statistician. The methods were similar to but completely separate from the trial with black and Latina mothers initiated 4 months earlier (Howell et al. 2012).

Intervention

Patients randomized to the intervention arm were given a two- step behavioral educational intervention. The theoretical un- derpinning of the intervention was based on the Common- Sense Model (CSM), a model that describes how patients automatically match and consciously interpret changes from their normal physical self to beliefs that theymay be acutely or chronically ill or merely tired or out of sorts (Leventhal et al. 2011;McAndrew et al. 2008). The interpretive process creates expectations about the likely duration of the current state and its consequences for daily activities and motivates action to return to normal, e.g., take a home remedy, seek care. The dramatic changes (e.g., in shape and weight, etc.) that occur postpartum can encourage a mother to see herself in her normal prepregnant state, a perception that will create unreal- istic expectations as to how she should feel and how quickly she should fully recover. These perceptions can encourage premature efforts to return to prepregnancy levels of activity, and failure to meet expectations can create depressive symp- toms. Our behavioral educational intervention was designed to address these potentially modifiable interpretations. It did so by preparing mothers to interpret and normalize triggers of depressive symptoms (e.g., vaginal bleeding, breast pain, etc.), and specified the need to bolster social support and conserve personal resources. Specific, simple actions were suggested to address each of these issues, e.g., symptoms and social needs; realistic time frames were provided for return to normal; and resources were listed in the event prob- lems exceeded specified time lines.

The in-hospital component of the two-step intervention involved a 15-min, in-hospital review of a patient education pamphlet and partner summary sheet by the mother with a masters-trained social worker. The patient education pamphlet described common postpartum physical symptoms, depres- sion, infant colic, and the importance of social support. The partner summary sheet described symptoms of depression and danger signs and gave partner suggestions of ways to help new moms. The second component of the intervention was a 2-week postdelivery call in which the social worker assessed

58 E.A. Howell et al.

patients' symptoms, skills in symptommanagement, and other needs. Fidelity of the intervention was maintained by repeated training and review of scripts for both the in-hospital and telephone components of the intervention. Approximately 5 % of both in-hospital sessions and 2-week telephone needs assessments were observed by a physician or project manager on the team. A coinvestigator or project manager shadowed each social worker, completed an intervention monitoring instrument, and shared the results of the assessment with the social worker and investigative team.

Enhanced usual care patients received routine postpartum hospital education (i.e., discharge materials, television educa- tional programs on infant care, breastfeeding, and peripartum care). To insure equivalent contact, patients assigned to en- hanced usual care received a 2-week postdelivery call to inform them of future surveys, and a list of health-related and community resources was mailed to them. The interven- tion materials were modified to represent white and Asian mothers but were identical in content and based on the same theoretical model (Howell et al. 2012). All materials were written at a sixth grade reading level and included pictorials. Interviewers were blinded to study arm assignment. All study participants were interviewed by phone at 3 weeks (mean days=27.2, SD 5.6; median days=26, range of 17 to 52), 3 months (mean days=84.3, SD 7.9; median days=81, range of 76 to 124), and 6 months (mean days=175.5, SD 8.7; median days=174, range of 162 to 218) to assess depressive symptoms and contributing and buffering factors.

Outcome variable

Postnatal depressive symptoms were assessed using the ten- item Edinburgh Postnatal Depression Scale (EPDS). The EPDS is a commonly used postpartum depression screening instrument and has been validated in many postpartum popu- lations and different languages (Areias et al. 1996; Boyce et al. 1993; Carpiniello et al. 1997; Cox et al. 1987; Eberhard-Gran et al. 2001a, b; Ghubash et al. 1997; Harris et al. 1989; Jadresic et al. 1995; Lawrie et al. 1998; Lundh and Gyllang 1993; Murray and Carothers 1990; Wickberg and Hwang 1996; Zelkowitz and Milet 1995). The recommended cutoff score of ≥10 has sensitivities of 0.59–0.81, and specificities ranged from 0.77 to 0.88 for major and minor depression (Gaynes et al. 2005). The Patient Health Questionnaire-9 (PHQ-9) was also administered, and a comparison was made between the EPDS and PHQ-9 scores over time in a secondary analysis. Women reporting high levels of depressive symp- toms (EPDS ≥13 or PHQ-9 ≥20 or suicidal ideation) at any of the four assessments were referred for psychiatric assessment and possible treatment. They were retained in the study.

Survey items also included questions on sociodemographics, clinical characteristics such as antepartum complications, co- morbid conditions (e.g., diabetes, hypertension, asthma, thyroid

disease, heart disease), past depression history, anxiety, social support, and health-care factors. Medical charts were reviewed for parity, delivery type, insurance, past medical history, mater- nal complications, delivery complications, and infant outcomes.

Sample size was powered to detect a clinically meaningful difference in reduction of symptoms of postpartum depression 3 weeks postrandomization for women in the intervention arm in comparison with women in enhanced usual care. Assuming that 34 % of white (non-black and non-Latina) women would report depressive symptoms (EPDS≥10) in enhanced usual care (based on our prior research) (Howell et al. 2006), our recruitment target of 492 patients (246 per arm) met 90 % power based on a two-sided 0.05-level chi-squared test to detect a clinically meaningful 13 % reduction in depressive symptoms (from 34 to 21 %, a relative risk reduction of about 33 %). To allow for decreased power due to patient loss at follow-up, we planned to enroll 540 participants.

Data analysis

Data were collected in person at baseline and by telephone during follow-up interviews by clinical research coordinators blinded to intervention status. Participants in the intervention arm (N =270) and enhanced usual care (N =270) were com- pared at baseline on demographic and clinical characteristics using t tests and their nonparametric equivalents, chi-square tests, as appropriate. Group differences were summarized by 95 % confidence interval estimates. Overall study attrition rate was low and equivalent across treatment groups. The amount of missing data for our primary outcome measure, EPDS, was low at 3 weeks (7 %; 40/540), 3 months (14 %; 73/540), and 6 months (13 %; 72/540) and similar between groups at each time point. The primary analysis examined the efficacy of the intervention in reducing the likelihood of positive postpartum depression screens. Logistic regression analyses tested the effect of the intervention on positive depression screens at 3 weeks, 3months, and 6months postpartum. To assess change over time across groups for the primary outcome, generalized estimating equations (GEEs), a widely used method for analyzing corre- lated, longitudinal outcome data with statistical accuracy, were employed (Zeger and Liang 1986). Significance level of 0.05 was used for the primary outcome of postpartum depression, and the final model was adjusted for baseline depression screen.

Results

There were 4,448 deliveries over the study period, and 2,628 met the initial eligibility criteria (i.e., were ≥18, etc.). A random sample of 734 (28 %) of the 2,628 eligible mothers were approached in hospital and reviewed the study consent form. Of the 734mothers, 540 (73.6%) completed the consent process and were enrolled in the trial; 194 (26.4 %) had

An intervention to reduce postpartum depressive symptoms 59

declined participation (Fig. 1). Mothers who declined to par- ticipate were slightly younger than mothers who consented (mean age of 32 vs. 33; p <0.05). Of the 270 mothers ran- domized to the intervention arm, 262 received the intervention in hospital. Eight mothers were discharged before the social worker could meet with them. These eight patients were sent the education materials, and the social worker reviewed the educational materials over the phone. Eighty-nine percent (241/270) of the intervention group and 93 % (252/270) of the control group were successfully reached for the 2-week calls. Of the 540 enrolled patients, 16 patients withdrew over the 6-month study period. Completion rates for the follow-up interviews were 93% (500/540) at 3 weeks, 86% (467/540) at 3 months, and 86 % (467/540) at 6 months. Follow-up was equivalent for intervention and control at 3 weeks (92.2 vs. 93.0 %, p =0.74), at 3 months (87.0 vs. 85.9 %, p =0.71), and at 6 months postpartum (84.8 vs. 88.2 %, p =0.26). There were baseline differences in rates of positive depression screens between women lost to follow-up versus those includ- ed in the analyses at 3 weeks: 30% (12/40) lost to follow-up at 3 weeks screened positive versus 11 % (57/500) who com- pleted the 3-week survey.

The overall mean age of enrolled participants was 33 (range 18–48); 89 % were white, 9 % were Asian, 2.4 % had Medicaid insurance, 11 % earned ≤$30,000 annually, and 24 % were foreign born. Baseline characteristics of the inter- vention and control groups are described in Table 1. There were no clinically important differences between the two groups at baseline. The mean EPDS scores at baseline were 4.0 (SD 4.0; range 0–22) in the intervention group and 5.0 (SD 4.0; range 0–19) in the control group.

In the intention-to-treat analysis (N =540), there was no difference in positive depression screens among mothers in the intervention arm versus those in the control arm: at 3 weeks, 6.0 % (15/249) versus 5.6 % (14/251), p =0.83, respectively; at 3 months, 5.1 % (12/235) versus 6.5 % (15/232), p =0.53, respectively; and at 6 months, 3.5 % (8/230) versus 4.6 % (11/238), p =0.53, respectively.

To assess change over time across groups for the primary outcome, we employed generalized estimating equations. An intention-to-treat generalized estimated equation model, for up to 6 months of follow-up, found no differences in rates of depressive symptoms among intervention versus control (OR of 0.97; 95 % CI 0.59–1.61). The time effect and interaction

Assessed for eligibility N=4,448

Excluded (did not meet eligibility criteria: infant birth weight, Apgar score, maternal age, race,

language, lack of telephone) n=1,820

Eligible to be approached for study n=2,628

Nonparticipants: n=1,894 Could not be approached (e.g. delivered on

a weekend, or weekly enrollment target already met): 1,782

Were approached in hospital but discharged before review of consent: 112

Refused: 194

Randomized n=540

Assigned to Enhanced usual care: n=270 Routine postpartum care: 270 2-week call: 252; 93.3%

Assigned to Behavioral educational intervention: n=270 Step 1: Educational session: 270 In-hospital session: 262; 97.0% Session by phone: 8; 3.0%

Step 2: 2-week telephone needs assessment: 241; 89.3%

Interview at 3 weeks completed n=251; 93.0%

Interview at 3 weeks completed n=249; 92.2%

Interview at 3 months completed n=232; 85.9%

Interview at 3 months completed n=235; 87.0%

Interview at 6 months completed n=238; 88.2%

Interview at 6 months completed n=229; 84.8%

Fig. 1 Participants' flow through study

60 E.A. Howell et al.

were not significant. Post hoc subgroup analyses examined whether the intervention was more effective among the sub- group of mothers with a high school education or less. There was no difference in positive depression screens among mothers in the intervention arm versus those in the control arm in this subgroup analysis.

Discussion

Our behavioral educational intervention aimed at addressing modifiable factors associated with postpartum depressive symptoms did not and essentially could not reduce depressive symptoms among 540 postpartum mothers as rates of depres- sive symptoms were low at baseline and subsequent time points among this population of highly educated mostly white women. Our results contrast with the results of our trial recruiting black and Latina mothers (which used the same recruitment strategy and methodologies), in which the odds of a positive depression screen were reduced by 33% for up to 6 months of follow-up for mothers randomized to the inter- vention arm as compared with the control arm (Howell et al. 2012).

The results of both of our trials suggest that this intervention may be successful in reducing racial/ethnic disparities in rates of depressive symptoms postpartum. In the trial with black and Latina mothers, the behavioral education intervention success- fully lowered the risk of a positive depression screen to roughly the same frequency as the overall rates for white and Asian mothers in the current trial. Rates of depressive symptoms in the intervention group of black and Latina mothers compared with combined rates in the current trial were similar: at 3 weeks, 8.8 versus 5.8 %, respectively, p =0.43; at 3 months, 8.4 versus 5.8 %, respectively, p =0.44; and at 6 months, 8.9 versus 4.0%, respectively, p =0.53. By targeting modifiable factors, the in- tervention helped black and Latina mothers buffer symptoms of depression.

An expanding body of research has documented the asso- ciation between potentially modifiable factors and postpartum depressive symptoms. For example, postpartum depressive symptoms are associated with physical symptoms (vaginal bleeding, breast pain, etc.), physical functioning, infant colic, and low levels of social support (Howell et al. 2005, 2006, 2009). Our intervention aimed to prepare and educate mothers about the postpartum experience, to normalize symptoms and other experiences, to provide specific, concrete suggestions for coping with realistic time frames for observing benefits, to bolster personal and social resources, and to better manage postpartum demands. We hypothesize that the intervention likely worked among low-income black and Latina women because the prevalence of their depressive symptoms was

Table 1 Demographic and clinical characteristics of trial participants

Behavioral educational intervention (N=270)

Enhanced usual care (N=270)

p value

Demographic characteristics

Age, mean (SD), years 33 (6) 32 (5) 0.12

Race, no. (%)

Non-Hispanic white or Caucasian 233 (86) 246 (91) 0.17

Asian 30 (11) 21 (8)

Other 7 (3) 3 (1)

Birthplace, no. (%) 0.27

US born 199 (74) 210 (78)

Foreign born 71 (26) 60 (22)

Language, no. (%)

English 270 (100) 270 (100)

Education, no. (%) 0.23

High school or less 34 (13) 44 (16)

Some college or more 235 (87) 226 (84)

Insurance, no. (%) 0.78

Medicaid/Medicaid-managed care 8 (3) 6 (2)

Private/other 262 (97) 263 (98)

Marital status, no. (%) 0.13

Single/separated/divorced/widowed 8 (3) 3 (1)

Married/living as if married 262 (97) 267 (99)

Parity, no. (%) 0.46

Primiparous 129 (49) 122 (46)

Multiparous 135 (51) 145 (54)

Breastfeeding, no. (%) 0.01

Yes 257 (95) 241 (89)

No 13 (5) 29 (11)

Clinical characteristics

Delivery type, no. (%) 0.02

C-section 90 (34) 65 (24)

Vaginal delivery 178 (66) 203 (76)

Comorbid condition, no. (%) 0.71

Yes 33 (12) 36 (13)

No 234 (88) 232 (87)

Antepartum complication, no. (%) 0.36

Yes 58 (21) 49 (22)

No 212 (79) 218 (82)

Past history of depression, no. (%) 0.25

Yes 65 (24) 54 (20)

No 205 (76) 216 (80)

Treatment for depression (medication or therapy/counseling) this pregnancy, no. (%) Yes 14 (5) 21 (8) 0.22

No 256 (95) 249 (92)

Positive baseline depression screen (EPDS≥10), no. (%)

0.37

Yes 31 (11) 38 (14)

No 239 (89) 232 (86)

An intervention to reduce postpartum depressive symptoms 61

higher at baseline, and many may have lacked the tools and support necessary to manage postpartum stressors. Providing tools and resources to low-income minority mothers, who are at highest risk of postpartum depressive symptoms, has the potential to lower the burden of depressive symptoms and place them at similar risk as higher-income majority women.

Studies do suggest that interventions selectively targeting women at elevated risk for postpartum depression may be more effective than universal interventions aimed at preventing post- partum depression (Dennis and Doswell 2013). However, there is little consistency in the identification of women “at risk,” and a review of 16 antenatal screening tools suggests that there are no measures with acceptable predictive validity to accurately identify asymptomatic women who will later develop postpar- tum depression (Austin and Lumley 2003). For this trial, we chose to include women regardless of their baseline depressive symptom score, a universal approach that is simple, because we expected much higher rates of depressive symptoms in our participants. The rates of depressive symptoms in our cohort were lower than previously published literature including those assessed in our prior longitudinal studies (Howell et al. 2005). Over the last decade, public awareness for postpartum depres- sion has risen, and numerous education campaigns on this topic may have impacted prevalence rates of depressive symptoms at our hospital and in the region. The base rates of depressive symptoms were roughly 20 % lower both in the current study and in the successful trial with black and Latina mothers. As behaviors, both symptomatic and nonsymptomatic, e.g., adher- ence, are triggered and moderated by many factors, there are multiple pathways for changing behavior outside the frame- work of a clinical trial. A number of high-profile public aware- ness campaigns and legislative initiatives aimed at educating women about postpartum depression were put into place in the period between 2002 and 2009 (Postpartum Support International 2010). It is possible that these initiatives and the longitudinal studies conducted in our setting interacted to gen- erate alternative pathways that normalized and destigmatized postpartum depression. Research on public service mental health campaigns has been found to be effective in engender- ing greater awareness, positive changes in attitudes regarding depression, attitudes toward antidepressants, and to a certain extent treatment-seeking behavior (Paykel et al. 1998).

There were a number of limitations with this trial. First, the rate of positive depressive symptom screens was much lower than previously published rates. It is possible that womenwith depressive symptoms were less likely to enroll in our random- ized controlled trial. However, our refusal rate was low. In addition, the low prevalence of depressive symptoms limited the power of this study to detect a difference in rates of depressive symptoms between the intervention and control arms in this trial. This study was powered to intervene on base rates of postpartum depressive symptoms detected in prior

longitudinal studies in this same catchment area. The results of this negative trial highlight the importance of monitoring ongoing rates of depressive symptoms, the focus of this be- havioral intervention. Second, women with positive depres- sion screens were more likely to be lost to follow-up at 3 weeks. However, the overall rate of positive depression screens was low, and only 12 women who screened positive for depression at baseline were lost to follow-up at 3 weeks. Third, we used a depression screening instrument rather than a formal structured interview to diagnose depression. Our inter- vention was implemented in an obstetrics settingwhere formal assessments are often too burdensome to perform. Further, positive screens for depression, whether or not they are asso- ciated with diagnosis of major depressive disorder, are an important measure to assess as they are associated with mul- tiple negative outcomes for mothers and infants (Zayas et al. 2002; Beeghly et al. 2003; Dennis and Ross 2005; McLearn et al. 2006). And finally, this trial was implemented at one institution, and the majority of the patients were educated and white which limits the generalizability of our results.

We found that a behavioral education intervention did not reduce postpartum depressive symptoms among a sample of highly educated and predominantly white mothers because their prevalence of depressive symp- toms was low. In contrast, the intervention was success- ful in reducing depressive symptoms among postpartum black and Latina mothers. Postpartum depressive symptoms affect women from all racial/ethnic and socioeconomic back- grounds but are more burdensome for low-income women of color who often have fewer resources available to them and less access to treatment. This simple, low-cost intervention appears to be of benefit for low-income women of color and may help address early modifiable factors associated with depressive symptoms. It remains to be demonstrated whether the intervention will prove equally beneficial for a sample of white, non-Latino women reporting higher rates of depressive symptoms.

References

Areias ME, Kumar R, Barros H, Figueiredo E (1996) Comparative incidence of depression in women and men, during pregnancy and after childbirth. Validation of the Edinburgh postnatal depression scale in Portuguese mothers. Br J Psychiatry 169(1):30–35

Austin M, Lumley J (2003) Antenatal screening for postnatal depression: a systematic review. Acta Psychiatr Scand 107:10–17

BeeghlyM, OlsonKL,WeinbergMK, Pierre SC, Downey N, Tronick EZ (2003) Prevalence, stability, and socio-demographic correlates of depressive symptoms in black mothers during the first 18 months postpartum. Matern Child Health J 7(3):157–168

Boyce P, Stubbs J, Todd A (1993) The Edinburgh postnatal depression scale: validation for an Australian sample. Aust N Z J Psychiatry 27(3):472–476

62 E.A. Howell et al.

Carpiniello B, Pariante CM, Serri F, Costa G, CartaMG (1997)Validation of the Edinburgh postnatal depression scale in Italy. J Psychosom Obstet Gynaecol 18(4):280–285

Cox JL, Holden JM, Sagovsky R (1987) Detection of postnatal depres- sion. Development of the 10-item Edinburgh postnatal depression scale. Br J Psychiatry 150:782–786

Dennis CL, Doswell T (2013) Psychosocial and psychological interven- tions for preventing postpartum depression. Cochrane Database Syst Rev 2, CD001134

Dennis CL, Ross L (2005) Relationships among infant sleep patterns, maternal fatigue, and development of depressive symptomatology. Birth 32(3):187–193

Duer S, Schwenk TL, Coyne JC (1988) Medical and psychosocial correlates of self-reported depressive symptoms in family practice. J Fam Pract 27(6):609–614

Eberhard-Gran M, Eskild A, Tambs K, Opjordsmoen S, Samuelsen SO (2001a) Review of validation studies of the Edinburgh postnatal depression scale. Acta Psychiatr Scand 104(4):243–249

Eberhard-Gran M, Eskild A, Tambs K, Schei B, Opjordsmoen S (2001b) The Edinburgh postnatal depression scale: validation in a Norwegian community sample. Nord J Psychiatry 55(2):113–117

Gaynes BN,Meltzer-Brody S, Lohr KN, Swinson T, Gartlehner G, Brody S, Miller WC (2005) Perinatal depression: prevalence, screening accuracy, and screening outcomes. Paper presented at the AHRQ Publication No. 05-E006-2, Rockville

Ghubash R, Abou-Saleh MT, Daradkeh TK (1997) The validity of the Arabic Edinburgh postnatal depression scale. Soc Psychiatry Psychiatr Epidemiol 32(8):474–476

Harris B, Huckle P, Thomas R, Johns S, Fung H (1989) The use of rating scales to identify post-natal depression. Br J Psychiatry 154:813– 817

Herrera E, Reissland N, Shepherd J (2004) Maternal touch and maternal child-directed speech: effects of depressed mood in the postnatal period. J Affect Disord 81(1):29–39

Howell EA, Mora PA, Horowitz CR, Leventhal H (2005) Racial and ethnic differences in factors associated with early postpartum de- pressive symptoms. Obstet Gynecol 105(6):1442–1450

Howell EA, Mora P, Leventhal H (2006) Correlates of early postpartum depressive symptoms. Matern Child Health J 10(2):149–157

Howell EA, Mora PA, DiBonaventura MD, Leventhal H (2009) Modifi- able factors associated with changes in postpartum depressive symp- toms. Arch Womens Ment Health 12(2):113–120

Howell EA, Mora PA, Chassin MR, Leventhal H (2010) Lack of prepa- ration, physical health after childbirth, and early postpartum depres- sive symptoms. JWomens Health (Larchmt) 19(4):703–708. doi:10. 1089/jwh.2008.1338

Howell EA, Balbierz A, Wang J, Parides M, Zlotnick C, Leventhal H (2012) Reducing postpartum depressive symptoms among black and Latina mothers: a randomized controlled trial. Obstet Gynecol 119(5):942–949. doi:10.1097/AOG.0b013e318250ba48

Jadresic E, Araya R, Jara C (1995) Validation of the Edinburgh postnatal depression scale (EPDS) in Chilean postpartum women. J Psychosom Obstet Gynaecol 16(4):187–191

JosefssonA,Angelsioo L, Berg G, EkstromCM,Gunnervik C, Nordin C, Sydsjo G (2002) Obstetric, somatic, and demographic risk factors for postpartum depressive symptoms. Obstet Gynecol 99(2):223–228

Kessler RC (2003) Epidemiology of women and depression. J Affect Disord 74(1):5–13

Kroenke K, Spitzer RL, Williams JB (2001) The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 16(9):606–613

Kroenke K, Spitzer RL, Williams JB (2003) The patient health questionnaire-2: validity of a two-item depression screener. Med Care 41(11):1284–1292

Lawrie TA, Hofmeyr GJ, de Jager M, Berk M (1998) Validation of the Edinburgh postnatal depression scale on a cohort of south African women. S Afr Med J 88(10):1340–1344

Leventhal H, Bodnar-Deren S, Breland JY, Hash-Converse J, Phillips LA, Leventhal EA, Cameron LD (2011) Modeling health and illness behavior: the approach of the commonsense model. In: Baum A, Revenson TA, Singer J (eds) Handbook of health psychology, 2nd edn. Psychology Press, New York, pp 3–35

Lundh W, Gyllang C (1993) Use of the Edinburgh postnatal depression scale in some Swedish child health care centres. Scand J Caring Sci 7(3):149–154

Matthey S, Barnett B, Ungerer J, Waters B (2000) Paternal and maternal depressed mood during the transition to parenthood. J Affect Disord 60(2):75–85

McAndrew LM, Musumeci-Szabo TJ, Mora PA, Vileikyte L, Burns E, Halm EA, Leventhal EA, Leventhal H (2008) Using the common sense model to design interventions for the prevention and manage- ment of chronic illness threats: from description to process. Br J Health Psychol 13:195–204

McLearn KT, Minkovitz CS, Strobino DM, Marks E, Hou W (2006) Maternal depressive symptoms at 2 to 4 months post partum and early parenting practices. Arch Pediatr Adolesc Med 160(3):279– 284

Minkovitz CS, Strobino D, Scharfstein D, Hou W, Miller T, Mistry KB, Swartz K (2005) Maternal depressive symptoms and children's receipt of health care in the first 3 years of life. Pediatrics 115(2): 306–314

Murray L, Carothers AD (1990) The validation of the Edinburgh post- natal depression scale on a community sample. Br J Psychiatry 157: 288–290

Paykel ES, Hart D, Priest RG (1998) Changes in public attitudes to depression during the defeat depression campaign. Br J Psychiatry 173:519–522

Postpartum Support International (2010) Postpartum Support Internation- al, U.S. State Legislation http://www.postpartum.net/News-and- Events/Legislation.aspx. Retrieved May 8, 2012

Seguin L, Potvin L, St-Denis M, Loiselle J (1999) Depressive symptoms in the late postpartum among low socioeconomic status women. Birth 26(3):157–163

Vera M, Alegria M, Freeman D, Robles RR, Rios R, Rios CF (1991) Depressive symptoms among Puerto Ricans: island poor compared with residents of the New York city area. Am J Epidemiol 134(5): 502–510

Wickberg B, Hwang CP (1996) The Edinburgh postnatal depression scale: validation on a Swedish community sample. Acta Psychiatr Scand 94(3):181–184

Zayas LH, Cunningham M, McKee MD, Jankowski KR (2002) Depres- sion and negative life events among pregnant African-American and Hispanic women. Women's Health Issues 12(1):16–21

Zeger SL, Liang KY (1986) Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42(1):121–130

Zelkowitz P, Milet TH (1995) Screening for post-partum depression in a community sample. Can J Psychiatry 40(2):80–86

An intervention to reduce postpartum depressive symptoms 63

Copyright of Archives of Women's Mental Health is the property of Springer Science & Business Media B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Risk Factors for Major Antenatal Depression among Low-Income African American Women

Sabrina Luke, M.P.H.,1,2 Hamisu M. Salihu, M.D., Ph.D.,1,2,3 Amina P. Alio, Ph.D.,4 Alfred K. Mbah, Ph.D.,1

Dee Jeffers, M.P.H.,1 Estrellita Lo Berry, M.A.,1 and Vanessa R. Mishkit, R.N., B.S.N.1

Abstract

Objectives: Data on risk factors for major antenatal depression among African American women are scant. In this study, we seek to determine the prevalence and risk factors for major antenatal depression among low- income African American women receiving prenatal services through the Central Hillsborough Healthy Start (CHHS). Methods: Women were screened using the Edinburgh Postnatal Depression Scale (EPDS) with a cutoff of �13 as positive for risk of major antenatal depression. In total, 546 African American women were included in the analysis. We used logistic regression to identify risk factors for major antenatal depression. Results: The prevalence of depressive symptomatology consistent with major antenatal depression was 25%. Maternal age was identified as the main risk factor for major antenatal depression. The association between maternal age and risk for major antenatal depression was biphasic, with a linear trend component lasting until age 30, at which point the slope changed markedly tracing a more pronounced likelihood for major depression with advancing age. Women aged �30 were about 5 times as likely to suffer from symptoms of major antenatal depression as teen mothers (OR¼ 4.62, 95% CI 2.23-9.95). Conclusions: The risk for major antenatal depression increases about 5-fold among low-income African American women from age 30 as compared to teen mothers. The results are consistent with the weathering effect resulting from years of cumulative stress burden due to socioeconomic marginalization and discrimination. Older African American mothers may benefit from routine antenatal depression screening for early diagnosis and intervention.

Introduction

Depression is the most common mental health problemidentified during pregnancy and following delivery.1,2 In light of increasing racial disparities in birth outcomes and infant mortality, understanding the risk factors associated with maternal health issues, such as antenatal and postpar- tum depression, among disadvantaged African American women becomes crucial in the fight to improve the health of black women and their infants. Currently, research on ante- natal depression among African American women is scarce. Most research addresses postpartum depression; however, there is increasing evidence that antenatal depression impacts maternal health and birth outcome. Accordingly, we under- took this study to identify risk factors for antenatal depression among African American women. In the general population,

the prevalence rates for both antenatal and postpartum de- pression vary considerably, ranging between 10% and 52%.1,3,4 The prevalence of pregnancy-related depression in disadvantaged populations has been consistently higher, however, with rates of antenatal depression as high as 41.7% reported for low-income women.4–6

Risk factors for antenatal and postpartum depression include young age, poverty, lack of education, history of de- pression, history of miscarriage or abortion, anxiety, low self- esteem, lack of social support, stressful life events, and history of abuse.7 There is also evidence that women diagnosed with postpartum depression in their fist pregnancy are 50% more likely to develop the disorder in subsequent pregnancies or to develop depression that is unrelated to pregnancy.8

African American and Hispanic women have the highest rates of antenatal and postpartum depression.3,5–11 Single

1The Chiles Center for Healthy Mothers and Babies, 2Department of Epidemiology and Biostatistics, 3Department of Obstetrics and Gynecology, and 4Department of Community and Family Health, University of South Florida, Tampa, Florida.

JOURNAL OF WOMEN’S HEALTH Volume 18, Number 11, 2009 ª Mary Ann Liebert, Inc. DOI: 10.1089=jwh.2008.1261

1841

marital status and low income have been reported as being predominant risk factors for postpartum depression among African American women.6 Disadvantaged populations have been found to be more vulnerable to depression, as they have less control over their surroundings, are more likely to suffer stressful life events,5,6 have less access to financial and emo- tional resources, and are more likely to have experienced discrimination.11

Stress is a significant contributing factor to the onset of postpartum depression in African American women.4,9

Mothers are often depressed because of medical conditions following delivery, including hypertension, infection, and recovery from surgery.9 Loneliness and abandonment, as well as such external factors as financial difficulties, employment concerns, lack of social support, lack of support from a part- ner, and sibling care, may contribute additional stress, which culminates in fatigue associated with the accumulation of stressful circumstances.4,9 Both antenatal and postpartum depression have been associated with an increased likelihood of developmental problems in infants of afflicted mothers. There is evidence that these infants are more likely to show signs of depression and stress and exhibit changes in brain activity that mimic the mother’s prenatal depressive state.12

The CDC estimates that as many as 50% of postpartum depression cases go undiagnosed.3 Identifying pregnancy- related depression early is critically important in preventing postpartum depression and in improving birth outcome. In this study, we seek to identify factors associated with the risk for major antenatal depression among a low-income popula- tion of African American women.

Materials and Methods

The Central Hillsborough Healthy Start Project (CHHS) is one of the community-based programs in the State of Florida funded by the federal government through the Maternal and Child Health Bureau’s Healthy Start Initiative. Implemented by the Lawton and Rhea Chiles Center for Healthy Mothers and Babies, the CHHS project functions as a community= university partnership to narrow racial disparities in mater- nal and infant health outcomes in urban Tampa neighbor- hoods (Florida), where the black infant mortality and morbidity rates are more than double those among whites (unpublished data from the Florida Department of Health). In these neighborhoods, most of the births are to mothers who are black, many of whom are young, unmarried, underedu- cated, and Medicaid eligible.13 Prenatal and postnatal risk reduction services are provided by the project within the local perinatal healthcare system framework, the Hillsborough CHHS Program, and its overarching state system, the Florida Healthy Start Program. In collaboration with the Florida Department of Health and the Healthy Start Coalition of Hillsborough County, local and state efforts are integrated in a complementary manner. Unique to Florida, all pregnant women and newborn infants are offered risk screens to identify those who would benefit most from risk reduction services. Mothers who voluntarily accept the screen and ex- press interest in services are referred to local Healthy Start Programs. In Hillsborough County, women living in select East Tampa ZIP codes (33602, 33603, 33605, and 33610) who need services are referred to the CHHS Program for those services.

Data collection

Data were collected through services provided by the CHHS program during the period 2002–2007. Florida’s uni- versal screening of pregnant women and infants was used to identify women at risk of poor outcomes.14 Florida’s universal screening of pregnant women and infants includes a series of questions that focus on medical, environmental, and psy- chosocial factors that identify a patient as at risk. The score is determined by summing the contributing items, each worth 1 point, except for race, which contributes 2 points. The fol- lowing 15 variables comprise the components of the screening score: black race, maternal age <18 or >39, unmarried, less than high school education, low maternal weight (<110 pounds), problems keeping appointments, moving �3 times in the past year, feeling unsafe, going to bed hungry, tobacco use in the past 2 months, use of drug or alcohol in the past 2 months, unwanted pregnancy, current maternal illness, seeking prenatal care in the second trimester, and history of poor outcomes or no previous pregnancy experience. Women who scored �4 were considered at risk for adverse birth outcomes, which has been demonstrated previously.15

Women were administered the Edinburgh Postnatal De- pression Scale (EPDS) during prenatal visits. Measurements were conducted at different times during pregnancy, and the aggregate sum of EPDS scores from participating mothers was used to determine the prevalence of depressive symp- tomatology consistent with risk for antenatal depression. We believe that this is a better approach to capturing prevalence than administering the EPDS at a single point in time. The EPDS is a screening tool used to identify women at risk for antenatal and postpartum depression. It is usually adminis- tered at the 6–8-week postpartum examination, with a re- commended readministration of the test after 2 weeks. The EPDS instrument has been validated for antenatal admin- istration.16 The EPDS has been validated and used cross- culturally throughout the world and has been administered numerous times to black women in the United States and England.17–20 The scale can only be used to indicate how the mother felt during the previous 7 days and is not intended to detect anxiety, phobias, or personality disorders.21 Questions evaluate symptoms of depression, such as the inability to laugh, the inability to look forward to things with enjoyment, self-blame, anxiety and excessive worry, fear and panic, dif- ficulty sleeping because of sadness, crying, and self-harm.22

The maximum possible score is 30, with a cutoff point of 12=13 indicating that the mother is potentially suffering from major depression and a cutoff point of 9=10 indicating possible minor depression.23 According to Cox et al.,21 the cutoff point of 12=13 results in a sensitivity of 86% and a specificity of 78%. A total of 724 women were administered the EPDS instrument prenatally. Of these women, 75.4% (n¼ 546) were African American. We restricted our analysis to African American women because of a paucity of numbers for the other racial=ethnic groups in the sample. The study is cross-sectional in design, as we did not follow participants over time.

Statistical analyses

We applied chi-square tests to determine differences in sociodemographic characteristics between mothers with EPDS scores compatible with risk of major depression vs. those with lower scores. We used chi-square tests for linear

1842 LUKE ET AL.

trend to assess dose-response. In order to determine predic- tors for risk of major depression, we constructed logistic re- gression models that contained the following covariates: maternal age (categorized as <20, 20–24, 25–29, and �30), body mass index (BMI) (<30 vs. �30), marital status (single vs. married), education (high school graduate vs. nongradu- ate), prenatal screening score (<4 vs. �4), and smoking (smoker vs. nonsmoker). Teens (<20) were used as a reference group to compare the risk of antenatal depression with ma- ternal age. Variables were chosen based on previous research indicating their potential to impact depression during preg- nancy.24–26

Because the analyses depicted a strong association between maternal age and risk for major antenatal depression, we constructed a receiver operator characteristics (ROC) curve to assess the predictive ability of maternal age in categorizing a woman in the study sample as being at risk for major ante- natal depression. We performed this procedure using values generated from the adjusted logistic regression model with simultaneous computation of the area under the curve, the measure of accuracy. Similarly, we computed adjusted probabilities for major antenatal depression across maternal age and plotted the values to assess the trajectory of risk for major antenatal depression during pregnancy.

All tests of hypothesis were two-tailed, with a type 1 error rate fixed at 5%, and SAS version 9.1 (SAS Institute, Cary, NC) was used to perform all analyses. This study was approved by the Office of the Institutional Review Board at the University of South Florida.

Results

Of the 546 women in the study, 137 had EPDS scores con- sistent with risk for major depression (25.0%). Table 1 presents

a summary of the results of crude frequency comparisons between women at risk for major depression vs. those not at risk with respect to selected sociodemographic characteristics. There were significant differences between women with risk of major antenatal depression vs. those not at risk with respect to maternal age. The risk for major antenatal depression cor- related positively with maternal age in a dose-effect fashion ( p for trend<0.01). However, the prevalence of marital status, education, BMI, prenatal screening score, and smoking dur- ing pregnancy did not differ between the two maternal groups.

The mean EPDS scores steadily increased by age group, with women <20 years having a mean EPDS score of 7.5, those aged 20–24 having a mean score of 7.7, those aged 25–29 having a mean score of 9.9, and women �30 having a mean score of 11.1. The mean EPDS score for the entire study sample was 8.4.

Table 2 shows the results of the logistic regression analysis used to generate the odds ratios (ORs) for maternal age and other risk factors associated with antenatal depression. Model I includes screening score in the analysis, whereas model II does not. Prenatal screening score and maternal age were significantly associated with elevated risk for major antenatal depression in model I. The strongest association was with maternal age, which demonstrated a monotonic increase in risk for major antenatal depression in a dose-effect pattern ( p for trend<0.01). The association observed for maternal age was not impacted by the inclusion=exclusion of screening score in the analysis. As compared with teenagers, the risk for major antenatal depression was more than doubled among women aged 25–29 years (OR¼ 2.25, 95% CI 1.19-4.27) and almost quintupled by the time African American women reached the age of 30 and beyond (OR¼ 4.62, 95% CI 2.23- 9.55). Results of further analysis using adjusted probabilities to assess this dose-response relationship are illustrated in Figure 1. The observed trend between maternal age and possible antenatal depression appeared to be biphasic, with age 30 being a critical point in time when the risk of antenatal depression became more pronounced and maintained a much greater level of risk there and beyond.

The ROC curve in Figure 2 highlights the predictive ability of age in determining a woman’s risk of developing major antenatal depression. An accuracy rate of 60.4% was estab- lished, indicating that maternal age is a fairly good predictor of risk for major antenatal depression.

Discussion

We found up to 25% of women in this study to be at risk for major antenatal depression. This result is consistent with previous research on antenatal depression in women of low socioeconomic status.4–6,10,27 However, as only women at risk for major depression were of interest or counted as cases (and this implies that minor and moderate depression were non- cases), the actual prevalence of depressive symptomatology in the study population must be much higher.

Our study contributes important new insights into the re- lationship between maternal age and depression. Our find- ings indicate that African American women’s risk for depression increases with age. There is a distinct biphasic dose-response relationship that notably increases in the 30s. A shocking 47.5% of women in the 30 and older age group

Table 1. Demographics of African American Women by EPDS Score (n¼ 546)

EPDS <13 (n¼ 409)

EPDS �13 (n¼ 137)

Major depression

Variable n % n % p value

Age, years <20 128 81.0 30 19.0 0.04 20–24 183 79.6 47 20.4 0.03 25–29 67 67.7 32 32.3 0.07 �30 31 52.5 28 47.5 <0.0001

Marital status Married 34 69.4 15 30.6 0.4 Single 375 75.5 122 24.5

High school Graduate 216 73.0 80 27.0 0.3 Nongraduate 192 77.1 57 22.9

Body mass index (BMI) <30 187 74.8 63 25.2 0.4 �30 82 78.9 22 21.1

Prenatal screening score <4 45 77.6 13 22.4 0.4 �4 248 71.9 97 28.1

Smoking Smoker 10 58.8 7 41.2 0.1 Nonsmoker 393 75.7 126 24.3

MATERNAL AGE AND MAJOR ANTENATAL DEPRESSION 1843

experienced depressive symptomatology consistent with major antenatal depression. Older women were about 5 times more likely to be at risk for major antenatal depression com- pared with teen mothers. Previous research has reported teenage mothers as being a high-risk group for depression,7

and to our knowledge, there is no known documented re- search indicating that older women are at a greater risk for antenatal depression. However, our findings are consistent with the weathering effect hypothesis, which posits that the long-term cumulative exposure to stressors adversely affects the mental and physical well-being of African American women.4,28–36 The Life Course Model also provides a rationale for the observed effect by postulating that experiences accu- mulated from birth and throughout the life span significantly alter the trajectory for health and well-being.37 Women in this study came from disadvantaged backgrounds and had lim- ited financial resources. Approximately 45% did not graduate

from high school, and close to 91% were single throughout their pregnancy.

The weathering effect theory and Life Course Model sug- gest that exposure to repeated negative life events with age might explain why older African American women are more susceptible to poor mental health, in this case, antenatal de- pression. There is evidence that chronic stress caused by poverty and perceived discrimination causes higher rates of respiratory illness, hypertension, chronic illness, anxiety, de- pression, and psychosis.28,33 Among African Americans, the odds of experiencing an adverse health event that limits work ability increase by 158% with each decade of life.28 Research has shown that after the age of 26, African Americans are 24% more likely than Caucasians to suffer a health-related event.28

By age 35, African Americans experience health conditions that are on average more commonly diagnosed in 55-year-old Caucasians,29,31 and by age 46, African American women

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

10 20 30 40 50

P ro

ba bi

lit y

Maternal age (in years)

FIG. 1. The probability of risk for antenatal depression across maternal age.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

S en

si tiv

ity

Area under the curve = 0.6

False positive rate

FIG. 2. ROC for maternal age as a predictor of major an- tenatal depression.

Table 2. Predictors of Risk for Major Antenatal Depression among African American Womena

Risk factors for antenatal depression Model I

OR (95% CI)b Model II

OR (95% CI)

Age, years <20 1.0 (Reference group) 1.0 (Reference group) 20–24 1.15 (0.68-1.95)

0.59 1.15 (0.68-1.94)

0.61 25–29 2.25 (1.19-4.27)

0.01 2.24 (1.19-4.24)

0.01 �30 4.62 (2.23-9.55)

<0.0001 4.49 (2.18-9.24)

<0.0001 Prenatal screening score >4 1.78 (1.13-2.78)

0.01 N=A

Did not graduate from high school 1.02 (0.65-1.61) 0.92

1.15 (0.733-1.79) 0.55

Body mass index <30 1.12 (0.75-1.69) 0.58

1.17 (0.78-1.75) 0.46

Smoker 2.08 (0.76-5.74) 0.16

2.19 (0.80-6.04) 0.13

Marital status (single) 1.03 (0.51-2.07) 0.93

1.12 (0.56-2.23) 0.75

ap for trend for age. bOR, odds ratio. Adjusted ORs were obtained by controlling for the confounding effects of the following characteristics: BMI, smoking,

marital status, education, prenatal screening score and age; CI, confidence interval.

1844 LUKE ET AL.

have the highest probability of health-related work limita- tions compared to any other group.28

Most studies have reported a positive association among perceived discrimination, psychological distress, and a diag- nosis of major depression.28,31–36 It has been found that 12% of African American women who had experienced a negative life event in the previous year and lived in high-stress neighborhoods went on to suffer a major depressive episode compared with 2% of women who had experienced a negative life event and lived in low-stress environments.38 High rates of homicide in disadvantaged areas result in many women grieving the loss of a loved one.4,31 They may be caring for a parent with a chronic illness or may be suffering ill health themselves. According to one study, being a caregiver, ex- periencing parenting strain, and not being married are most strongly associated with depressive symptoms.39 In general, African American women in their 30s may have more chil- dren and are less likely to have financial support from their parents than women in their teens and 20s.40,41 Older women may be raising preteen or teenage children without the pres- ence of a father figure, which may act as an additional source of psychological stress.40,42 Women who lack partners suffer greater financial hardships and struggle with the reality of raising additional children by themselves.40 Given this prem- ise, we hypothesize that the accumulated psychological stress from financial burdens, poor living conditions, poor health, negative life events, and perceived discrimination built up over the years cause the increased risk of major antenatal depression with age among disadvantaged African American women. Further research is needed to determine the cause of the observed effect, however, as many factors were not directly measured in this analysis. For example, African American women may not receive adequate healthcare or may have less access to positive social support. We recom- mend evaluating these measures in future research.

Study limitations include selection bias, which might have impacted the study results if unidentified differences (in ad- dition to those adjusted for in the study) existed between younger and older women participating in the Healthy Start Program. Other limitations of the study include the fact that the EPDS is only a screening tool and cannot be used to di- agnose antenatal or postpartum depression. As well, the EPDS can only measure depressive symptoms within the past week, which may miss the severity of depressive symptom- atology at another point during pregnancy. Because of the variable sensitivity and specificity of the instrument, there is always the chance of obtaining false negatives or false posi- tives, leading to misclassification bias that should reasonably be random across age groups. Consequently, the reported results in this study could have underestimated the observed association between maternal age and major antenatal de- pression. A history of depression is also an important risk factor for antenatal and postpartum depression and was not evaluated in this study.1 Additional information about the course of depressive symptomatology over time could not be obtained because this was a cross-sectional study. Further research needs to be conducted to identify and confirm the factors responsible for the observed trend in maternal age and major depression in African American women. We recom- mend conducting a prospective cohort study to verify that lifelong stress and discrimination impacts maternal depres- sion during pregnancy in older African American mothers.

Conclusions

As African American women age, they face an accumula- tion of adverse life events from poverty and discrimination that have a profound impact on mental well-being. Symptoms consistent with the weathering effect become apparent during pregnancy, as African American women aged �25 become significantly at higher risk of developing major antenatal de- pression compared with younger women. This finding has important implications for public health and social policies. Healthcare providers need to be aware of the increased risk of pregnancy-related depression in older African American women in order to better provide effective prenatal screenings and eventual optimal management for this population.

Acknowledgments

This work was supported in part by project H49MC00090 from the Maternal and Child Health Bureau (Title V, Social Security Act), Health Resources and Services Administration (HRSA), U.S. Department of Human and Health Services. The funding agency did not play any role in any aspect of the analyses.

Disclosure Statement

The authors have no conflicts of interest to report.

References

1. Robertson E, Grace S, Wallington T, Stewart DE. Antenatal risk factors for postpartum depression: A synthesis of recent literature. Gen Hosp Psychiatry 2004;26:289–295.

2. Wisner K, Parry B, Pointek C. Postpartum depression. N Engl J Med 2002;347:194–199.

3. CDC. Prevalence of self-reported postpartum depressive symptoms—17 states, 2004–2005. MMWR 2008;57(14):361– 366.

4. Zayas L, Cuningham M, McKee M, Jankowski K. Depression and negative life events among pregnant African-American and Hispanic women. Womens Health Issues 2002;12:16–22.

5. Hobfoll S, Ritter C, Lavin J, Hulsizer M, Cameron R. De- pression prevalence and incidence among inner-city preg- nant and postpartum women. J Consult Clin Psychol 1995;63:445–453.

6. Beeghly M, Olson K, Weinberg K, Pierre SC, Downey N, Tronick EZ. Prevalence, stability, and socio-demographic correlates of depressive symptoms in black mothers during the first 18 months postpartum. Matern Child Health J 2003;7:157–168.

7. Leigh B, Milgrom J. Risk factors for antenatal depression, postnatal depression and parenting stress. BMC Psychiatry 2008;8:24–34.

8. NIH. News In Health. December 2005. Available at newsinhealth.nih.gov=pdf=NIHNiH%20December05.pdf

9. Amankwaa LC. Postpartum depression among African Amer- ican women. Issues Ment Health Nurs 2003;24:297–316.

10. Rich-Edwards J, Kleinman K, Abrams A, et al. Socio- demographic predictors of antenatal and postpartum de- pressive symptoms among women in a medical group practice. J Epidemiol Community Health 2006;60:221–227.

11. Surkan PJ, Pterson KE, Hughes MD, Gottlieb BR. The role of social networks and support in postpartum women’s de- pression: A multiethnic urban sample. Matern Child Health J 2006;10:375–383.

MATERNAL AGE AND MAJOR ANTENATAL DEPRESSION 1845

12. Field T, Diego M, Hernandez-Reif M. Prenatal depression effects on the fetus and newborn: A review. Infant Behav Dev 2006;29:445–455.

13. Jevitt C, Zapata L, Harrington M, Berry E. Screening for perinatal depression with limited psychiatric resources. J Am Psychiatr Nurses Assoc 2006;11:359–363.

14. Florida Department of Health. Florida Healthy Start Pre- natal Risk Screen. Available at www.doh.state.fl.us=Family= mch=pdf=DH3134.pdf

15. Salihu HM, Mbah AK, Jeffers D, Alio AP, Berry L. Healthy Start Program and feto-infant morbidity outcomes: Evalua- tion of program effectiveness. Matern Child Health J 2008; Aug 9 [Epub ahead of print].

16. Murray D, Cox J. Screening for depression during pregnancy with the Edinburgh Depression Scale (EDDS). J Reprod In- fant Psychol 1990;8:99–107.

17. Templeton L, Velleman R, Persaud A, Milner P. The expe- riences of postnatal depression in women from black and minority ethnic communities in Wiltshire, U.K. Ethn Health 2003;8:207–221.

18. Mayberry LJ, Andrews Horowitz J, Declercq E. Depression symptom prevalence and demographic risk factors among U.S. women during the first 2 years post partum. J Obstet Gynecol Neonatal Nurs 2007;36:542–549.

19. Morris-Rush JK, Freda MC, Bernstein PS. Screening for postpartum depression in an inner-city population. Am J Obstet Gynecol 2003;188:1217–1219.

20. Affonso DD, Lovett S, Paul S, et al. Dysphoric distress in childbearing women. J Perinatol 1992;12:325–332.

21. Cox JL, Holden JM, Sagovsky R. Detection of postpartum depression: Development of the 10-item Edinburgh Post- natal Depression Scale. Br J Psychiatry 1987;150:782–786.

22. Beck CT, Gable R. Further validation of the Postpartum Depression Screening Scale. Nurs Res 2001;50:155–164.

23. Cox JL, Murray D, Chapman G. A controlled study of onset, duration and prevalence of postnatal depression. Br J Psy- chiatry 1993;163:27–31.

24. Dayan J, Creveuil C, Herlicoviez M, et al. Role of anxiety and depression in the onset of spontaneous preterm labour. Am J Epidemiol 2002;155:293–301.

25. Zuckerman B, Amaro H, Bauchner H, Cabral H. Depressive symptoms during pregnancy: Relationship to poor health behaviors. Am J Obstet Gynecol 1989;160:1107–1111.

26. Bonari L, Pinto N, Ahn E, Einarson A, Steiner M, Koren G. Perinatal risks of untreated depression during pregnancy. Can J Psychiatry 2004;49:726–735.

27. Seguin L, Potvin L, St-Denis M, Loiselle J. Chronic stressors, social support, and depression during pregnancy. Obstet Gynecol 1995;85:583–589.

28. Walsemann K, Geronimus A, Gee G. Accumulating disad- vantage over the life course: Evidence from a longitudinal study investigating the relationship between educational advantage in youth and health in middle age. Res Aging 2008;30:169–199.

29. Geronimus A, Hicken M, Keene D, Bound J. Age patterns of allostatic load score among blacks and whites in the United States: Might allostatic load algorithms measure weather- ing? Am J Public Health 2006;96:826–833.

30. Geronimus A, Bound J, Waidmann T, Colen C, Steffick D. Inequality in life expectancy, functional status, and active life expectancy across selected black and white populations in the United States. Demography 2001;38:227–251.

31. Williams D, Williams-Morris R. Racism and mental health: The African American experience. Ethn Health 2000;5:243– 268.

32. Williams D, Neighbors H, Jackson J. Racial=ethnic discrim- ination and health: Findings from community studies. Am J Public Health 2003;93:200–208.

33. McKenzie K. Racism and health: Antiracism is an important health issue. BMJ 2003;326:65–66.

34. Kennedy B, Kawachi I, Lochner K, Jones C, Prothrow-Stith D. (Dis)respect and black mortality. Ethn Dis 1997;7:207–214.

35. Janssen I, Hanssen M, Bak M, et al. Evidence that ethnic group effects on psychosis risk are confounded by experi- ence of discrimination. BMJ 2003;326:65–66.

36. Geronimus AT. The weathering hypothesis and the health of African American women and infants: Evidence and spec- ulations. Ethn Dis 1992;2:207–221.

37. Lu M, Halfon N. Racial and ethnic disparities in birth out- comes: A life course perspective. Matern Child Health J 2003;7:13–30.

38. Cutrona C, Wallace G, Wesner K. Neighborhood character- istics and depression: An examination of stress process. Curr Dir Psychol Sci 2006;15:188–192.

39. Nolen-Hoeksema, Ahrens C. Age differences and similarities in the correlates of depressive symptoms. Psychol Aging 2002;17:116–124.

40. Jayakody R, Chatters L, Taylor RJ. Family support to single and married African American mothers: The provision of financial, emotional, and child care assistance. J Marriage Fam 1993;55:261–276.

41. Chase-Lansdale PL, Brooks-Gunn J, Zamsky ES. Young African-American multigenerational families in poverty: Quality of mothering and grandmothering. Child Dev 1994; 65:373–393.

42. Zimmerman M, Salem D, Maton K. Family structure and psychosocial correlates among urban African-American ad- olescent males. Child Dev 1995;66:1598–1613.

Address correspondence to: Hamisu M. Salihu, M.D., Ph.D.

Director, Center for Research and Evaluation Lawton and Rhea Chiles Center for Healthy Mothers and Babies

University of South Florida 3111 E. Fletcher Avenue

Tampa, FL 33613

E-mail: [email protected]

1846 LUKE ET AL.

Copyright of Journal of Women's Health (15409996) is the property of Mary Ann Liebert, Inc. and its content

may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express

written permission. However, users may print, download, or email articles for individual use.

Journal of Child and Family Studies (2018) 27:3169–3175 https://doi.org/10.1007/s10826-018-1157-6

ORIGINAL PAPER

Screening for and Preventing Perinatal Depression

Bonnie D. Kerker1 ● Judy A. Greene1 ● Rachel Gerson1 ● Michele Pollock1 ● Kimberly E. Hoagwood1 ●

Sarah McCue Horwitz1

Published online: 20 June 2018 © Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract New York City (NYC) public hospitals recently mandated that all pregnant women be screened for depression, but no funds were allocated for screening or care coordination/treatment, and research suggests that unfunded mandates are not likely to be successful. To address this, we implemented an on-site depression prevention intervention (NYC ROSE) for positive depression screens among pregnant, mostly Black and Hispanic, lower-income women in one public hospital. In this paper, we used Aarons’ implementation model to describe the successes and challenges of screening and intervention. Patient tracking sheets and electronic medical records were abstracted. Key informant interviews and an informal focus group were conducted, and staff observations were reviewed; common implementation themes were identified and fit into Aarons’ model. We found that a lack of funding and staff training, which led to minimal psychoeducation for patients, were outer context factors that may have made depression screening difficult, screening results unreliable, and NYC ROSE enrollment challenging. Although leadership agreed to implement NYC ROSE, early involvement of all levels of staff and patients would have better informed important inner context factors, like workflow and logistical/practical challenges. There was also a mismatch between the treatment model and the population being served; patients often lived too far away to receive additional services on site, and economic issues were often a higher priority than mental health services. Screening and interventions for perinatal depression are essential for optimal family health, and a detailed, thoughtful and funded approach can help ensure effectiveness of such efforts.

Keywords Postpartum depression ● Perinatal depression ● Primary care ● Depression treatment ● Depression screening

Introduction

Depression affects approximately 20% to 40% of mothers with young children (Chaudron et al. 2004; Field 2010), with rates highest among women with low socioeconomic status (Goyal et al. 2010; Lorant et al. 2007). The negative consequences of postpartum depression are well estab- lished, and include insecure attachment, unsafe parenting practices, increased emergency care for infants, and poor cognitive and mental health outcomes in children (Field 2010; Murray et al. 2010). However, although treatments for both prenatal and postnatal depression have been found to be effective (Fitelson et al. 2011; Misri and Kendrick

2007), most women with perinatal depression go undetected (Evins et al. 2000; Goodman and Tyer-Viola 2010). Data show that screening improves the detection of depression in primary care settings (Chaudron et al. 2004; Evins et al. 2000); thus, numerous professional organizations have issued recommendations for depression screening both prenatally and postnatally, but all stop short of specifying clear guidelines for what constitutes appropriate follow-up and treatment (Committee on Obstetric Practice 2015; Earls 2010; Sui and the U.S. Preventive Services Task Force USPSTF 2016).

Several policy initiatives have responded to these recommendations by encouraging healthcare providers to screen pregnant and postpartum women, and either treat on- site or refer to services. New York City (NYC) THRIVE is a Mayoral initiative that aims to “screen and treat all pregnant women and new mothers for pregnancy-related depression” among participating hospitals and clinics, as a part of a city-wide mental health strategy (City of New York 2016). As of this writing, NYC THRIVE was working with

* Bonnie D. Kerker [email protected]

1 New York University School of Medicine, One Park Avenue, 7th Floor, New York, NY 10016, USA

12 34

56 78

90 () ;,:

12 34 56 78 90 (); ,:

29 hospitals, including all public hospitals in NYC, to screen and treat (or refer) all women in their care. As part of this initiative, Health+Hospitals, which runs NYC’s public hospitals, pledged that depression screening and connection to care would be universal in its hospitals (reaching 78% of all NYC births) by the end of 2017 (Bologna 2015). When this perinatal screening initiative began, it relied on hospi- tals’ existing resources for both screening and providing/ referring to effective treatment.

Implementing standardized screening and treatment guidelines across the city is an important effort. However, adding a new unfunded screening and treatment mandate to an already overburdened system can be challenging (Yoo et al. 2007), and the literature shows that screening without staff supports and follow-up is not effective at increasing access to care or improving outcomes (Byatt et al. 2015; Kozhimannil et al. 2011). In 2005, for example, New Jersey established the New Jersey Postpartum Wellness Initiative to raise awareness about postpartum depression and to increase access to clinical services. In 2006, it became the first state to require postpartum depression screening among women who had recently given birth. However, the new policies did not include financing for screening services, nor did they address patient-level barriers to treatment. As of 2007, New Jersey’s policies were not found to have improved the detection or treatment of postpartum depres- sion among Medicaid recipients (Kozhimannil et al. 2011).

The extant literature suggests that the ability of primary care providers to screen and refer patients to on-site services may be crucial for successful implementation of perinatal mental health services, as many depressed mothers will not otherwise follow-up with referrals (Marcus et al. 2003; Schulberg et al. 1993). Miller et al. (2009) described the implementation of a stepped-care model in which pregnant women were screened for depression and referred for mental health treatment within the prenatal care setting. Almost all (98.6%) women who screened positive accepted the co-located formal diagnostic assessment, which the authors attributed to lack of stigma, ease of co-located care, and minimization of financial barriers to mental health care.

Some initiatives implemented on-site have been shown to reduce depressive symptoms among women at risk for postpartum depression. For example, Reach Out and Stay Strong Essentials for new mothers (ROSE), an evidence- based treatment based on interpersonal therapy (IPT) prin- ciples, aims to prevent postpartum depression among low- income pregnant women through psychoeducation on depression, coping skills, and effective utilization of social supports. ROSE has been evaluated in three randomized controlled trials with low-income women receiving prenatal care, and was effective in preventing depression in women who had recently delivered (Zlotnick et al. 2001, 2006, 2016). In the most recently published study, 31% of control

participants developed postpartum depression at 6-months post-delivery, compared to 16% in the intervention group. This effect was maintained, with marginal statistical sign- ficance, at 12 months post-delivery (40% of controls vs. 26% of intervention participants) (Zlotnick et al. 2016).

In an attempt to maximize benefits from the new screening protocols, we worked with providers at one large NYC public hospital to implement an adapted version of ROSE among at-risk women who screened positive for depressive symptoms during pregnancy (NYC ROSE). In this paper, we use an implementation model described by Aarons et al. (2011) to outline the successes and challenges involved in both screening low-income pregnant women for depressive symptoms and implementing a preventive intervention at a NYC public hospital.

Method

Participants

The intervention took place at the women’s health clinic in a NYC public hospital that has more than 500,000 ambu- latory care visits per year. The participants in this imple- mentation were the 559 women who attended their first prenatal appointment at the hospital clinic between Sep- tember 2016 and February 2017. More than half of the women identified as Hispanic (55.5%), 11.6% as Black, 6.1% as White, 4.3% as Asian, and 22.5% as another race or ethnicity. The mean age of the participants was 30 (SD= 6.09); women under the age of 18 attend a separate, ado- lescent prenatal clinic. More than three-quarter (78.5%) of the women used Medicaid (government-funded, needs- based health care) to pay for services, and 14% paid for services themselves (self-pay) (Table 1).

Procedure

As part of NYC Thrive and a NYC Health+Hospitals initiative, depression screening with the PHQ-9 during the first prenatal visit at the clinic began in September 2016. Women were handed the instrument (in either English or Spanish) with other medical forms when they registered and asked to complete it alone in the waiting room. Patients were told to keep the form and give it to social work staff who would review it with them during the social services intake. Women who met eligibility criteria were referred to NYC ROSE.

The NYC ROSE implementation discussed here occur- red between September, 2016 and February, 2017. Eligible women endorsed depressive symptoms (PHQ-9 score > 4 and <19), and were at least 18 years old, English- or Spanish-speaking, and in their second trimester. Women

3170 Journal of Child and Family Studies (2018) 27:3169–3175

who indicated they were suicidal (PHQ-9 item 9 > 0) or severely depressed (PHQ-9 score > 18), as well as those who were currently experiencing domestic violence, had been previously diagnosed with a serious mental health condition, or had an active substance use disorder, were ineligible for NYC ROSE and triaged to a higher level of care.

NYC ROSE was implemented in an individual format in order to maximize scheduling flexibility and meet the needs of women seeking care at the hospital. Sessions were offered in either English or Spanish by prenatal educators to women at the time of their prenatal appointments to reduce the travel burden on participants. The five prenatal educators were volunteer students, professionals and peer- partners. They were trained with over 10 h of classroom and didactic sessions.

Measures

The PHQ-9 has been shown to be a valid and reliable measure of depression severity, and its brevity makes it a useful instrument (Kroenke et al. 2001). The instrument has also been shown to have high sensitivity (85%) and speci- ficity (84%) for depression diagnosis, as well as sub- diagnosis (75% and 88%, respectively) among pregnant

women (Sidebottom et al. 2012). In this study, both English and Spanish versions of the instrument were administered.

We chose the Aarons et al. implementation model (2011) because it is widely recognized that implementing new interventions is heavily influenced by both clinical practices and implementation processes (Fixsen et al. 2009; Palinkas et al. 2008). In fact, many programs designed to improve outcomes have not been widely adopted due to imple- mentation challenges (Aarons et al. 2011). The model focuses on outer and inner context variables in four implementation phases: exploration, adoption/preparation, active implementation and sustainment. In this paper, we focus on the “active implementation phase,” when the intervention is put into place.

Data Analyses

Data on the prevalence of PHQ-9 screening in the clinic and PHQ-9 scores, and characteristics of the population came from electronic medical records and were provided by NYC Health + Hospitals. Data on the contacts with participants came from study tracking sheets, which were updated each time staff were in the field. These data were compiled using Microsoft Excel.

Information on reasons for non-participation came from staff observations and notes taken after conversations with participants. In addition, staff took notes on the imple- mentation process throughout the study, and the PI held an informal focus group with staff to further clarify the positive factors and challenges that influenced the process. Thematic analysis was used as a method for identifying, analyzing and reporting patterns within data (Braun and Clarke 2006). All notes were reviewed by study staff with previous qua- litative experience. After several group discussions, the reviewers reached consensus and common themes were identified. Interviews with key informant clinic social work staff were also conducted and, similarly, common themes were documented.

Results

Outer Context

Outer context factors include items such as funding and inter-organizational networks (Aarons et al. 2011). In our case, important outer barriers included a lack of resources available to train staff on how to administer the PHQ-9 or discuss the benefits of NYC ROSE with patients. As a result, staff (both administrative staff and providers) did not consistently describe the instrument to patients, assure patients about confidentiality, explain the importance of the PHQ-9 to identify depressive symptoms, interpret the

Table 1 Description of women attending their first prenatal appointment at the hospital clinic, September 2016–February 2017

N %

Total 559

Race/ethnicity

Non-Hispanic White

34 6.1

Non-Hispanic Black

65 11.6

Hispanic 310 55.5

Asian 24 4.3

Other 126 22.5

Age

<18 0 0

18–24 109 19.5

25–34 310 55.5

35–44 137 24.5

45+ 3 0.54

Health Insurance

Private 37 6.6

Medicaid (Government- funded, needs- based health care)

439 78.5

Self-pay 79 14.1

Other 4 0.72

Journal of Child and Family Studies (2018) 27:3169–3175 3171

measure’s findings, or describe the help available through NYC ROSE. For example, social work staff members had not received training regarding how to interpret endorse- ments of poor sleep or appetite changes, which are nor- mative during pregnancy; this may have impacted the results of the screenings and referrals to NYC ROSE.

Although the external NYC ROSE collaborators worked extensively with clinic administrators, who saw great potential value in the program, NYC ROSE staff did not include front-line staff or patients in planning and, conse- quently, created inappropriate and inefficient processes. For example, women were meant to be referred to NYC ROSE after the PHQ-9 screen at their first prenatal visit, which tended to be in the first trimester. This resulted in serious workflow issues as at-risk women were told to “come back” when they reached their second trimester and would be NYC ROSE-eligible. Consequently, NYC ROSE staff were not able to capitalize on the immediacy of the on-site ser- vice, which has been identified as an important variable in implementation (Miller et al. 2009).

Inner Context

Inner context factors include items such as organizational characteristics (structures and processes that exist and/or take place in organizations) and individual adopter char- acteristics (Aarons et al. 2011). In our case, due to a lack of communication between clinic administrators and front-line staff, staff were not prepared for workflow and responsi- bility changes associated with the screening tool and intervention, resulting in poor adherence to the screening and referral protocol. In fact, throughout the intervention, the NYC ROSE implementation was revised several times in an attempt to more effectively fit into the clinic workflow.

Further, because of insufficient clinic staff, women completed the PHQ-9 on their own and did not receive psychoeducation about the tool or postpartum depression until meeting with the social worker, which often occurred hours into their prenatal visit. As one social worker explained, women were not likely to comply if they felt they were completing a form “so we can find out if you’re depressed.” Consequently, many women did not complete the PHQ-9 by the time they saw the social work staff, who were often too overburdened to administer the instrument during their sessions. From September 2016 to February 2017, only 31% of patients completed the PHQ-9 on their first prenatal visit according to hospital electronic medical records, although the percentage did increase over time (from 17.4% in September to 56% in February).

Organizational structure was also a barrier to NYC ROSE implementation. The program relied on student and unpaid volunteers as prenatal educators. Although the volunteers attended extensive training and were dedicated to

the intervention, they often faced scheduling conflicts. As a result, NYC ROSE staff did not have a consistent presence in the clinic, which impeded full integration and acceptance into the workflow.

An additional obstacle involved the long wait times to see providers, ranging from 20 min to 2 h. Many women were afraid of missing their appointments, and were reluc- tant to leave the waiting area to speak with an NYC ROSE educator. It was also difficult to schedule NYC ROSE sessions after appointments; many patients travel great distances to get to the hospital and were not open to staying after medical appointments for additional services. As a result, the way NYC ROSE was administered was an inappropriate structural fit for this clinic.

Individual patient characteristics also influenced the success of screening implementation. Interviews with clinic social work staff suggested that patients rarely acknowl- edged that stress and mental health were concerns in their lives on screening tools; many experienced extreme stigma and were not used to having the freedom to speak about such issues. In addition, according to one social worker, “many patients believe that the minute they disclose a mental health issue, child welfare will come and take their children away.” In fact, the social workers noted that women who originally did not endorse items on the PHQ-9 often would disclose symptoms of stress and concern after they built a level of trust with a provider. These factors may have contributed to the fact that only 15% scored 10 or above on the PHQ-9 over the entire time period (ranging by month from 6.3 to 23.8%).

Even women who disclosed multiple stressors often declined the NYC ROSE referral. According to the social work staff, many patients could not commit to coming back to the hospital for continued services. Others felt that they could “handle the stress,” or that “the church would help with that.” Still others were not comfortable admitting that their symptoms might lead to a mental health condition. Of the 30 eligible women who were referred to NYC ROSE in this time period, one woman agreed in person to participate and 11 women agreed after being contacted by phone. However, of those 12 women, only two ultimately began the intervention, and each only received two sessions (out of five).

There were also strengths in this environment that aided in the implementation of this intervention. At the individual adopter level, the providers in the Women’s Health Clinic were very committed to their patients (many of whom had high levels of psychosocial stressors), which greatly facili- tated implementation. The perinatal screening mandate and NYC ROSE were highly congruent with providers’ goals and culture, even if they were somewhat incongruent with hospital systems and procedures. Some of the providers worked in a receptive sub-context, meaning that they were

3172 Journal of Child and Family Studies (2018) 27:3169–3175

open to and ready for change, and supportive of new poli- cies. These providers were champions of both screening and NYC ROSE. Other providers were supportive of the inter- vention, but did not have the time or motivation to actively assist in implementation, perhaps because of a lack of communication from leadership or necessary shifts in workflow.

Discussion

Implementing standardized screening and treatment guide- lines across the city is an important effort. However, similar to previous research (Biggar 2001; Kozhimannil et al. 2011), we found that an unfunded mandate may be insuf- ficient to ensure staff capacity and adequate implementation in a busy urban women’s health clinic. Given the early stages of the screening initiative, the hospital had some notable successes, but funding may be needed to hire new staff to implement a new protocol, train and supervise all involved staff, and ensure a dedicated staff person to manage changes in workflow. Furthermore, champions at every level should be identified, so that all providers have the same understanding of the context in which the screening is conducted.

Providers in our setting screened an impressive 56% of new prenatal patients by February 2017, only 6 months after the inception of the initiative. In addition, the social work staff were committed to the process and open to learning from their experiences thus far. However, there are several issues that, if addressed, could increase the usefulness of this initiative. For example, it is unclear if other services were compromised in the screening process since the staffing levels were not increased; this is an important component that deserves further exploration. In order to enhance screening compliance without impacting other services, additional personnel and training resources should be allocated, and a modified workflow should accommodate the screening practice. In fact, Kaiser Permanente recently showed that it could achieve a 96% screening rate among pregnant and postpartum women using an approach that identified and used best practices, identified champions and educated clinicians, used data that drove performance, and streamlined the office workflow (Flanagan and Avalos 2016).

We found a similar rate of PHQ-9 positive screens as a recent study of low-income pregnant women seeking pre- natal care at Federally Qualified Health Centers (Side- bottom et al. 2012). However, there is evidence to suggest that economically disadvantaged and minority women are at greater risk for PPD compared to the rest of the population (Goyal et al. 2010; Lorant et al. 2007). Research, for example, has found the average PHQ-9 score among low-

income women to be 10.6 (SD= 6.5) (Kneipp et al. 2010), which is not consistent with our finding of 15% scoring 10 or above. It is worth exploring whether differential ethnic makeups of the samples, and the cultural adaptability of the instrument, could account for some of this difference.

As our social worker interviews suggested, the lower scores might be due to the stigma still attached to mental health issues, particularly in certain immigrant communities prevalent at this hospital (Derr 2015; Nadeem et al. 2007), as well as the lack of psychoeducation provided and trust garnered. Similar to findings in the literature (Canvin et al. 2007), clinic social work staff noted that low-income pregnant populations may be particularly sensitive to answering these types of screenings honestly as they fear the negative repercussions, such as having a child removed from their care. In addition, many patients at this clinic struggle to secure basic needs such as housing, food, and employment, so they may be more focused on solving tangible problems than identifying and addressing mental health issues. Efforts to both educate patients on the importance of addressing depressive symptoms and develop trusting relationships with patients before asking them to answer sensitive questions might lead to more accurate results; in addition, it is essential that patient education and relationship-building be both culturally- and linguistically- appropriate.

Even with accurate depression screening scores, the lit- erature suggests that screening without staff supports and follow-up does not improve outcomes (Byatt et al. 2015; Kozhimnnil et al. 2011), and that on-site services may be crucial for successful implementation of perinatal mental health services, as many depressed mothers will not other- wise follow-up with referrals (Marcus et al. 2003; Schulberg et al. 1993). We attempted to address these issues by offering on-site mental health services (NYC ROSE), and encountered several challenges to doing so. Having lea- dership agree to implement a new intervention is a neces- sary but not sufficient component of introducing a new program. The support and involvement of all levels of staff and patients were necessary to modify the workflow effectively, identify logistical and practical challenges, and ultimately ensure success. In addition, even though external staff were provided to implement NYC ROSE, they were volunteers. Ideally, paid professionals or para-professionals who are part of clinic staff should be trained to implement interventions in order to maximize flexibility, availability, and integration into workflow.

There were also a few notable obstacles encountered with regard to patient acceptability. Many women did not identify stress or mental health concerns as a priority; they focused more on tangible concerns, such as housing. Psy- choeducation should be part of any screening and inter- vention effort to help women assess whether addressing

Journal of Child and Family Studies (2018) 27:3169–3175 3173

these less tangible issues might help with their day-to-day challenges. Alternatively, it may be prudent to introduce this type of service only once other more pressing issues have been addressed.

Finally, even for women who were interested in NYC ROSE, participation was difficult. Most patients did not live near the clinic, and travelled far to receive their prenatal care. Implementation teams would benefit from meeting with patients prior to launching a new program to better understand their specific needs and situations. Other, more useful models in this type of hospital might be to form strong linkages with follow-up services in other NYC boroughs, pilot a version of ROSE through an on-line electronic app, or implement ROSE through home visits or by telephone sessions.

Limitations and Future Research

This analysis has several limitations. Our data were derived from non-standardized field notes and were not system- atically collected. Future studies would benefit from sys- tematic collection of data. Similarly, while we did conduct interviews to better understand different players’ perspec- tives, we did not speak to all parties and can therefore not generalize beyond our participants. In addition, we do not have pre-post data for women who started the program because post-intervention measures were only scheduled to be collected after the completion of the sessions, and no women reached that point. Further, adolescents and teens did not attend this clinic, and although all clinic patients were given the PHQ-9, those who were deemed to have a high level of clinical need were excluded from this study. Unfortunately, the clinic workflow did not include a mechanism to capture the number of women excluded for this reason, but previous, exploratory work at this clinic has shown that only a small percentage of women would be expected to be excluded due to high clinical need (about 6.5%). Nonetheless, we cannot speak to the implementation process among all women. Finally, the data on screening and PHQ-9 scores came from electronic medical records, which may be inaccurate due to documentation errors.

This study is limited in what conclusions can be drawn, and future research should be conducted using more com- prehensive and systematic data collection. Nonetheless, it is useful to consider whether perinatal screening mandates are enough. Our experience suggests that unfunded mandates are not likely to succeed as workflow inefficiencies may lead staff to feel overburdened, potentially resulting in inaccurate screening results, strained referral processes, less effective interventions, and compromises in existing clinical services. Further, treatment models need to match the population being served; alternative models should be examined to meet the needs of the needs of specific popu- lations. While our study focused on perinatal depression,

these findings may also apply to the identification and treatment of other mental health conditions among pregnant women. Screening and interventions for perinatal mental health conditions are essential for optimal family health; a detailed and thoughtful approach can help ensure effec- tiveness of such efforts.

Acknowledgements We would like to acknowledge and thank Caron Zlotnick, Lauren Kincal Veznedaroglu, Priscilla Shorter, Betzabet Giron, Anya Urcuyo, Hannah Ephraim, Michele Knobel, Omobolanle Oladokun, Kelly Fitzgerald and Ming Tsai for their contributions to this study.

Author Contributions B.D.K.: designed and executed the study and wrote the paper. J.A.G.: designed and executed the study and con- tributed to the writing of the paper. R.G.: collaborated with the design and execution of the study, and contributed to the writing of the paper. M.P.: collaborated with the design and execution of the study, and contributed to the writing of the paper. K.E.H.: collaborated with the design and execution of the study, and contributed to the writing of the paper. S.M.H.: designed the study, collaborated with the execution of the study, and contributed to the writing of the paper.

Funding This study was funded by the National Institutes of Mental Health P30MH090322.

Compliance with Ethical Standards

Conflict of Interest The authors declare that they have no conflict of interest.

Ethical Approval This article does not contain any studies with human participants or animals performed by any of the authors.

References

Aarons, G. A., Hurlburt, M, & Horwitz, S. M. (2011). Advancing a conceptual model of evidence-based practice implementation in public service sectors. Administration and Policy in Mental Health and Mental Health Services Research, 38(1), 4–23.

Biggar, H. (2001). Homeless children and education: An evaluation of the Stewart B. McKinney Homeless Assistance Act. Children and Youth Services Review, 23, 941–969.

Bologna, C. (2015). The important thing hospitals have pledged to do for new moms. http://www.huffingtonpost.com/entry/the-importa nt-thing-hospitals-have-pledged-to-do-fornew-moms_us_ 56533801e4b0d4093a585dd4.

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychol- ogy. Qualitative Research in Psychology, 3(2), 77–101.

Byatt, N., Levin, L. L., Ziedonis, D., Simas, T. A. M., & Allison, J. (2015). Enhancing participation in depression care in outpatient perinatal care settings: a systematic review. Obstetrics & Gyne- cology, 126, 1048–1058.

Canvin, K., Jones, C., Marttila, A., Burström, B., & Whitehead, M. (2007). Can I risk using public services? Perceived consequences of seeking help and health care among households living in poverty: qualitative study. Journal of Epidemiology and Com- munity Health, 61, 984–989.

Chaudron, L. H., Szilagyi, P. G., Kitzman, H. J., Wadkins, H. I., & Conwell, Y. (2004). Detection of postpartum depressive symp- toms by screening at well-child visits. Pediatrics, 113, 551–558.

3174 Journal of Child and Family Studies (2018) 27:3169–3175

City of New York. ThriveNYC: Year One Update. (2016). https:// thrivenyc.cityofnewyork.us/wp-content/uploads/2017/02/Thrive_ Year_End_Updated-1.pdf.

Committee on Obstetric Practice. (2015). Screening for perinatal depression. Committee Opinion No. 630. American College of Obstetricians and Gynecologists. Obstetrics & Gynecology, 125, 1268–71.

Derr, A. S. (2015). Mental health service use among immigrants in the United States: A systematic review. Psychiatric Services, 67, 265–274.

Earls, M. F. (2010). Committee on Psychosocial Aspects of Child and Family Health, American Academy of Pediatrics. Incorporating recognition and management of perinatal and postpartum depression into pediatric practice. Pediatrics, 126, 1032–1039.

Evins, G. G., Theofrastous, J. P., & Galvin, S. L. (2000). Postpartum depression: A comparison of screening and routine clinical eva- luation. American Journal of Obstetrics and Gynecology, 182, 1080–1082.

Field, T. (2010). Postpartum depression effects on early interactions, parenting, and safety practices: A review. Infant Behavior and Development, 33(1), 1–6.

Fitelson, E., Kim, S., Baker, A. S., & Leight, K. (2011). Treatment of postpartum depression: Clinical, psychological and pharmacolo- gical options. International Journal of Women’s Health, 3, 1–14.

Fixsen, D., Blasé, K., Naoom, S., & Wallace, F. (2009). Core implementation components. Research Ion Social Work Practice, 19, 531.

Flanagan, T., & Avalos, L. A. (2016). Perinatal obstetric office depression screening and treatment: Implementation in a health care system. Obstetrics & Gynecology, 127, 911–915.

Goodman, J. H., & Tyer-Viola, L. (2010). Detection, treatment and referral of perinatal depression and anxiety by obstetrical provi- ders. Journal of Women’s Health, 19(3), 477–490.

Goyal, D., Gay, C., & Lee, K. (2010). How much does low socio- economic status increase the risk of prenatal and postnatal depressive symptoms in first-time mothers? Women’s Health Issues, 20(2), 96–104.

Kneipp, S. M., Kairalla, J. A., Stacciarini, J. M. R., Pereira, D., & Miller, M. D. (2010). Comparison of depressive symptom severity scores in low-income women. Nursing Research, 59, 380.

Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9: Validity of a brief depression severity measure. Journal of Gen- eral Internal Medicine, 16, 606–613.

Kozhimannil, K. B., Trinacty, C. M., Busch, A. B., Huskamp, H. A., & Adams, A. S. (2011). Racial and ethnic disparities in post- partum depression care among low-income women. Psychiatric Services, 62, 619–625.

Lorant, V., Croux, C., Weich, S., Deliege, D., Mackenbach, J., & Ansseau, M. (2007). Depression and socio-economic risk factors: 7-year longitudinal population study. The British Journal of Psychiatry, 190, 293–298.

Marcus, S. M., Flynn, H. A., Blow, F. C., & Barry, K. L. (2003). Depressive symptoms among pregnant women screened in obstetric settings. Journal of Women’s Health, 12, 373–380.

Miller, L., Shade, M., & Vasireddy, V. (2009). Beyond screening: Assessment of perinatal depression in a perinatal care setting. Archives of Women’s Mental Health, 12, 329.

Misri, S., & Kendrick, K. (2007). Treatment of perinatal mood and anxiety disorders: A review. The Canadian Journal of Psy- chiatry, 52, 489–498.

Murray, L., Arteche, A., Fearon, P., Halligan, S., Croudace, T., & Cooper, P. (2010). The effects of maternal postnatal depression and child sex on academic performance at age 16 years: a developmental approach. Journal of Child Psychology and Psy- chiatry, 51, 1150–1159.

Nadeem, E., Lange, J. M., Edge, D., Fongwa, M., Belin, T., & Mir- anda, J. (2007). Does stigma keep poor young immigrant and US- born black and Latina women from seeking mental health care? Psychiatric Services, 58, 1547–1554.

Palinkas, L. A., Schoenwald, S. K., Hoagwood, K., Landsverk, J., Chorpita, B. F., & Weisz, J. R. (2008). An ethnographic study of implementation of evidence-based treatments in child mental health: First steps. Psychiatric Services, 59, 738–746.

Schulberg, H. C., Coulehan, J. L., Block, M. R., Lave, J., Rodriguez, W., Scott, C. P. …, & Perel, J. (1993). Clinical trials of primary care treatments for major depression: Issues in design, recruit- ment and treatment. The International Journal of Psychiatry in Medicine, 23(1), 29–42.

Sidebottom, A. C., Harrison, P. A., Godecker, A., & Kim, H. (2012). Validation of the Patient Health Questionnaire (PHQ)-9 for pre- natal depression screening. Archives of Women’s Mental Health, 15, 267–374.

Sui, A. L., the U.S. Preventive Services Task Force (USPSTF). (2016). Screening for depression in adults: US Preventive Services Task Force Recommendation Statement. Journal of the American Medical Association, 315, 380–7.

Yoo, J., Brooks, D., & Patti, R. (2007). Organizational constructs as predictors of effectiveness in child welfare interventions. Child Welfare, 86(1), 53–78.

Zlotnick, C., Johnson, S. L., Miller, I. W., Pearlstein, T., & Howard, M. (2001). Postpartum depression in women receiving public assistance: Pilot study of an interpersonal-therapy-oriented group intervention. American Journal of Psychiatry, 158, 638–640.

Zlotnick, C., Miller, I. W., Pearlstein, T., Howard, M., & Sweeney, P. (2006). A preventive intervention for pregnant women on public assistance at risk for postpartum depression. American Journal of Psychiatry, 163, 1443–1445.

Zlotnick, C., Tzilos, G., Miller, I., Seifer, R., & Stout, R. (2016). Randomized controlled trial to prevent postpartum depression in mothers on public assistance. Journal of Affective Disorders, 189, 263–268.

Journal of Child and Family Studies (2018) 27:3169–3175 3175

Journal of Child & Family Studies is a copyright of Springer, 2018. All Rights Reserved.

ORIGINAL ARTICLE

Impact of a preventive intervention for perinatal depression on mood regulation, social support, and coping

Tamar Mendelson & Julie A. Leis & Deborah F. Perry & Elizabeth A. Stuart & S. Darius Tandon

Received: 29 August 2012 /Accepted: 4 February 2013 /Published online: 2 March 2013 # Springer-Verlag Wien 2013

Abstract Perinatal depression prevention trials have rarely examined proximal outcomes that may be relevant for under- standing long-term risk for depression. The Mothers and Babies (MB) Course is a cognitive-behavioral depression prevention intervention, which has been shown to prevent depressive symptoms among at-risk perinatal women of color. This study examined intervention impact on three proximal outcomes that are theoretically linked with the intervention’s model of change and have been empirically linked with risk for depression: mood regulation expectancies, perceived so- cial support, and coping. The study used data from a random- ized intervention trial of the MB Course with 78 low-income, predominantly African-American perinatal women enrolled at one of four home visitation programs in Baltimore City. Mood regulation expectancies, perceived social support, and coping were assessed with self-report instruments at baseline, post- intervention, and 3- and 6-month follow-ups. The intervention group experienced 16 % greater growth in mood regulation from baseline to 6-month follow-up compared to the usual care group, suggesting a prevention effect. The pattern of findings was similar, although not statistically significant, for

social support. Contrary to prediction, the control group expe- rienced less growth in avoidant coping than the intervention group. Findings indicate the MB Course enhances mood reg- ulation, which may facilitate prevention of depression over time. Assessment of intervention effects on proximal outcomes is beneficial for understanding how interventions may enhance protective factors relevant to successful long-term outcomes.

Keywords Perinatal depression . Prevention . Mood regulation expectancy . Social support . Coping . Mothers and Babies Course

Introduction

Psychosocial interventions to reduce or prevent postpartum depression are becoming increasingly popular, and the evi- dence base supporting their efficacy is growing (Muñoz et al. 2012). Both cognitive behavioral and interpersonal strategies have been shown to reduce depressive symptoms and prevent incidence of new cases of major depression in the postpartum period (e.g., Crockett et al. 2008; Muñoz et al. 2012; Tandon et al. 2011; Zlotnick et al. 2006). Studies of perinatal depres- sion prevention interventions have focused primarily on changes in depressive symptoms or disorders. Several studies have also evaluated and reported intervention effects on other proximal outcomes relevant to risk for depression (Brugha et al. 2011; Crockett et al. 2008; Dennis et al. 2009; Elliott et al. 2000); however, most have not done so.

Prevention researchers argue for the importance of con- sidering risk and protective factors both in designing and evaluating interventions (Garber 2006; Ialongo et al. 2006; Sutton 2007). Assessment of intervention impact on proxi- mal factors related to depression risk may be relevant for understanding the likelihood of long-term maintenance of

T. Mendelson (*) : E. A. Stuart : S. D. Tandon Johns Hopkins Bloomberg School of Public Health, 624 N. Broadway/615 N. Wolfe Street, Baltimore, MD, USA e-mail: [email protected]

J. A. Leis : S. D. Tandon Johns Hopkins University School of Medicine, 200 N. Wolfe Street, Baltimore, MD 21287, USA

D. F. Perry Georgetown University Center for Child and Human Development, 3300 Whitehaven Streets, N.W, Washington, DC 20007, USA

Arch Womens Ment Health (2013) 16:211–218 DOI 10.1007/s00737-013-0332-4

intervention effects. Findings have potential to influence further refinements to intervention content or delivery and may also have implications for shaping theories on factors associated with postpartum depression.

The Mothers and Babies Course

The Mothers and Babies (MB) Course is a manualized, cognitive-behavioral intervention designed to reduce risk for perinatal depression among low-income, ethnically di- verse women (Muñoz et al. 2012). Delivered in a group format by trained facilitators, the course teaches women how to modulate negative thoughts, enhance supportive and enjoyable contacts with other people, and increase pleas- ant activities. The mother–baby relationship is emphasized throughout the course, including ways to enhance parenting practices using course skills. Previous research has sup- ported the feasibility of the MB Course and has shown promising effects on depressive symptoms and/or risk for disorder (Le et al. 2011; Muñoz et al. 2012; Tandon et al. 2011). Originally developed for use with Latinas, the course has also been adapted for urban African-American women served by home visitation (HV) programs. Findings from a recent randomized pilot trial indicate that depressive symp- toms in a sample of 78 low-income, predominantly African- American women across four HV programs declined at a significantly greater rate for intervention participants than usual care participants between baseline and 1-week post- intervention, 3-month follow-up, and 6-month follow-up (Tandon et al. under review). At the 6-month follow-up, 11 of 34 (32.4 %) women receiving usual care in this sample were assessed via clinical interview as having a depressive episode compared with 6 of 41 (14.6 %) women receiving the MB Course (Tandon et al. under review). Given these encouraging findings, evaluation of proximal risk and pro- tective factors targeted by the intervention is warranted and may enhance our understanding of intervention mechanisms.

Proximal outcomes

In keeping with the cognitive-behavioral theoretical framework guiding the MB intervention, course modules teach skills for enhancing thoughts, contact with others, and pleasant activities. These skills would be expected to produce improvements in related areas of functioning, including mood regulation expec- tancies, perceived social support, and coping abilities. Each of these proximal outcomes has been associated empirically with risk for—or protection against—depression, suggesting that changes in these factors may contribute to reductions in depres- sion (see Fig. 1).

Mood regulation expectancies refer to belief in one’s ability to modulate negative mood and return to a more positive emotional state. Response expectancy theory (Kirsch 1985)

suggests that mood regulation expectancies influence mood directly via perceived self-efficacy in the context of negative mood, as well as indirectly through the use of coping strategies (Catanzano and Mearns 1990). Data suggest that mood regu- lation expectancies are inversely associated with depressive symptoms (Catanzaro 1993; Catanzano and Mearns 1990; Catanzaro et al. 2000; Drwal 2008; Kassel et al. 2007) and predict future depressive symptoms (Catanzaro et al. 2000; Kassel et al. 2007). We are not aware of studies that have investigated the association of mood regulation expectancies with depression among perinatal samples.

Research has identified significant associations of per- ceived social support with stress, coping, and distress/de- pression; both the absence of positive support and the presence of negative interpersonal influences have been linked with depression (see Coyne and Downey, 1991, for a discussion). Perceived social support is also a significant predictor of mental health during the perinatal period. For instance, low perceived social support from spouses and family during pregnancy and the postpartum period was found to predict major depressive disorder and depressive symptoms (Sheng et al. 2010; Xie et al. 2009).

There is a large theoretical and empirical literature on coping strategies that individuals employ to manage stress in response to situations involving adversity, threat, or loss (Lazarus 1966; Lazarus and Folkman 1984). Research sug- gests that coping approaches aimed at actively addressing a stressful situation are generally effective for maintaining psy- chological well-being, whereas coping strategies involving disengagement or avoidance of the stressor often increase risk for negative outcomes, including depression and anxiety (Littleton et al. 2007; Moskowitz et al. 2009). Some evidence suggests that avoidant, or disengaged, coping strategies are associated with perinatal depressive symptoms, whereas more active coping strategies focused on problem solving may be related to lower perinatal depression scores (Blaney et al. 2004; Edge and Rogers 2005; de Tychey et al. 2005).

The current study

This study examined the impact of the MB Course on mood regulation expectancies, perceived social support, and cop- ing. The study was not designed or powered to test

MB Course

Thoughts

Contact with Others

Pleasant Activities

Proximal Outcomes

Mood Regulation Expectancy

Perceived Social Support

Coping

Distal Outcomes

Depressive Symptoms

Depressive Disorders

Fig. 1 Intervention theory of change

212 T. Mendelson et al.

mediational pathways by which these proximal factors may contribute to intervention impacts on depression. However, measurement of the three proximal outcomes permits a preliminary evaluation of the theory of change guiding the MB intervention. Data for the current study were drawn from the randomized trial of urban, low-income, predomi- nantly African American women in four HV programs described above (n=78). The three proximal outcomes were assessed at baseline, post-intervention, and 3- and 6-month follow-ups. We hypothesized that the intervention would be associated with improved mood regulation expectancies, with higher perceived social support, and with more active and fewer avoidant coping strategies.

Methods

Participants

Pregnant women and women with a child less than 6 months old who were enrolled in one of four Baltimore City home visiting programs were eligible for study participation. Women were contacted by phone and screened for depressive symp- toms using the Center for Epidemiological Studies Depression Scale (CES-D; Radloff 1977) and for depressive disorder using the Maternal Mood Screener (MMS; Le and Muñoz 1998). Women currently experiencing elevated depressive symptoms (CES-D score≥16) and/or who reported a lifetime depressive episode but did not meet criteria for a current depressive episode were invited to join the study. Women meeting criteria for a current depressive episode on the MMS were not eligible to participate; they were referred to the clinical supervisor at the home visiting program who initiated further assessment and referral to mental health treatment.

Recruitment took place between October 2009 and March 2010. Details regarding participant eligibility, ran- domization, and flow are reported elsewhere (Tandon et al. under review). One hundred and twenty women were eligi- ble for the study and randomized. Of the 61 women assigned to the intervention group, 41 enrolled in the study while among the 59 who were assigned to the control group, 37 enrolled. There were no statistically significant differ- ences in maternal age, pregnancy status, or race/ethnicity between randomized women who enrolled and did not en- roll in the study for either the intervention or control groups.

Procedures

Women were randomized to receive standard home visiting services plus a modified, 6-week version of the MB Course or standard home visiting services plus information on perinatal depression. Participants were surveyed at four time points: pre-intervention, post-intervention, and at 3 and 6 months

follow-up. Mood regulation expectancies, social support, and coping were assessed at each time point. Study partici- pants were compensated $20 cash for each completed survey. The Johns Hopkins University School ofMedicine Institutional Review Board approved all study procedures. Written consent was obtained from participants prior to completion of the baseline assessment.

Intervention

Our version of the MB intervention, adapted to ensure cultural and contextual appropriateness for low-income, ur- ban women, consisted of six 2-h intervention sessions de- livered weekly in a group format by either a licensed clinical social worker or clinical psychologist. The six sessions were divided into three two-session modules that mapped onto core cognitive behavioral concepts: pleasant activities, thoughts, and contact with others. Each session contained didactic instruction on core content, as well as activities and group discussion. Booster sessions, which reviewed the main course concepts, were held at 3 and 6 months post- intervention. In addition to the group sessions, intervention participants received one-on-one home visitor reinforcement of group material. Reinforcements took place after each of the first five intervention sessions during home visitors’ regularly scheduled visits with all clients who were inter- vention participants. Of the 41 women in the intervention group, 40 (98 %) completed the 1-week post-intervention as- sessment, and all 41 completed the 3- and 6-month post- intervention assessments. Women in the intervention group attended a mean of 4.5 sessions (SD=1.36, range=1–6), and 71 % received weekly reinforcements from their home visitor; 78 % attended one booster session, and 51 % attended both. Of the 37women in the control group, all 37 completed the 1-week post-intervention assessment, 35 (95%) completed the 3-month assessment, and 34 (92 %) completed the 6-month assessment. Additional details about the intervention content and study procedures can be found in Tandon et al. (2011).

Measures

Mood regulation was measured using the Negative Mood Regulation Scale (Catanzaro and Mearns 1990), which evalu- ates an individual’s expectancy that she will be able to regulate a negative mood. The scale consists of 30 items such as “When I’m upset, I believe that I can do something to feel better” and “When I’m upset, I believe that it won’t be long before I can calm myself down.” Items are rated on a 5-point scale ranging from “strong disagreement” to “strong agreement.” Scores can range from 30–150; higher scores indicate a stronger belief in one’s ability to improve a negative mood state. Internal consis- tency of the scale was good; Cronbach’s alpha ranged from 0.84–0.91.

Impact of a perinatal preventive intervention 213

Perceived availability of social support was measured using the Interpersonal Support Evaluation List (ISEL; Cohen and Hoberman 1983). We used a 30-item version of the ISEL, which evaluates perceived availability of ap- praisal support, belonging support, and tangible support with items such as “There are several people that I trust to help solve my problems” and “I often meet or talk with family or friends.” Items are rated on a 4-point scale ranging from “definitely false” to “definitely true.” Scores can range from 0–120; higher scores indicate greater perceived sup- port. Internal consistency in this sample was acceptable; Cronbach’s alpha ranged from 0.73–0.92.

Coping was assessed using the Brief COPE (Carver 1997), a 28-item measure that includes 14 subscales assess- ing different coping reactions. The measure is designed to assess coping reactions in response to a specific stressor. The instructions were adapted in this study so as to be relevant for pregnancy-related stress: “These items deal with ways you’ve been coping with the stress in your life since you found out you were pregnant. There are many ways to try to deal with things. These items ask what you’ve been doing to cope with your pregnancy.” Items such as “I’ve been taking action to try to make the situation better” and “I’ve been refusing to believe that it has happened” measure both adaptive and problematic reactions and are rated on a 4-point scale ranging from “I haven’t been doing this at all” to “I’ve been doing this a lot.” The Brief COPE was not designed to yield a single coping index. Following the author’s recommendations, we subjected the 14 subscales to factor analysis to create higher-order factors. Using prin- cipal components analysis with varimax rotation we identi- fied two factors that explained 57 % of the variance. Factor 1, active coping, included the following subscales: self- distraction, active coping, using emotional support, using instrumental support, positive reframing, planning, accep- tance, and religion. Possible scores on this factor ranged from 16 to 64 with higher scores indicating better, more active coping. Cronbach’s alpha for this factor ranged from 0.80–0.89. Factor 2, avoidant coping, included the subscales measuring denial, substance use, behavioral disengagement, venting, and self blame. Possible scores on this factor ranged from 10 to 40; lower scores indicated better, less avoidant coping. Cronbach’s alpha for this factor ranged from 0.72–0.77. One subscale, humor, was not retained. Additional detail regarding the factor analysis is available upon request.

Statistical analyses

Differences between participants in the intervention and usual care groups at baseline were determined using inde- pendent sample t tests for continuous variables, Fisher’s exact test for categorical variables with small cell sizes,

and χ2 tests for categorical variables without small cell sizes. The effect of the intervention on the proximal outcomes was estimated using multilevel linear regression. A separate re- gression model was fit for each outcome. Each model includ- ed fixed effects for condition (intervention versus usual care), time, the condition by time interaction, and site (i.e., home visiting program), and a random effect for participant to account for the non-independence of measures for the same participant over time. To allow for flexible trends over time, the time variable in the model was coded as a factor, with an indicator variable for each time point (baseline [omitted cate- gory], post-intervention, 3-month follow-up, and 6-month follow-up). The distributions of mood regulation and avoidant coping scores were skewed and were log transformed before running the outcome models. All analyses were conducted using Stata version 11 (StataCorp 2009), with the xtmixed command used to run the multilevel (mixed effects) models.

Results

Sample characteristics

At baseline, participants ranged in age from 14 to 41 with a mean age of 24.1 (SD=6.1) years. Most women in the sample were African American (83.1 %), single (78.2 %), and were not currently working (72.4 %). Over half had a high school degree or GED (59.7 %). Approximately one third (28.2 %) of the sample was pregnant; the remaining women (71.8 %) were postpartum. The current child was the first child for 27.6 % of women. There were no significant differences between partic- ipants in the intervention and usual care groups at baseline on any demographic characteristics (Table 1).

Intervention impact on proximal outcomes

Results of the multilevel linear regression models are shown in Table 2, and the means showing the trends in each outcome over time by group are displayed in Fig. 2.

Mood regulation No significant differences in mood regu- lation were found between the groups at baseline, and growth in mood regulation did not differ by group from baseline to post-intervention or 3 months post-intervention. However, the groups differed in growth in mood regulation from baseline to 6-month follow-up; the intervention group experienced 16 % greater growth in mood regulation from baseline to 6-month follow-up compared to the usual care group (β=0.16, SE=0.03, p<0.001).

Social support A similar pattern emerged for perceived social support although the finding was not statistically significant at the 0.05 level: the growth in perceived social

214 T. Mendelson et al.

support from baseline to 6-month follow-up was 6.67 points larger among participants in the intervention group than par- ticipants in the usual care group (β=6.67, SE=0.03, p<0.10).

Coping There was no statistically significant difference in active coping between the groups over time. Surprisingly, for avoidant coping, participants in the intervention group had a 14 % larger increase in avoidant coping strategies between baseline and the 6-month post-intervention follow- up compared to participants in the usual care group (β= 0.14, SE=0.07, p<0.05).

Discussion

This study assessed the impact of the 6-week MB Course on proximal outcomes of mood regulation expectancies, per- ceived social support, and coping among pregnant women and new mothers. As predicted, women in the intervention group displayed greater growth in mood regulation expec- tancies between baseline and the 6-month follow-up com- pared with women in the control group. This finding suggests that the intervention enhances a cognitive factor with potential to protect against depression and promote

positive mental health. The pattern was similar but not statistically significant for perceived social support. The MB Course did not significantly impact active coping, and

Table 1 Participant demographic characteristics at baseline by inter- vention group

Characteristic Intervention (n=41)

Usual care (n=37)

p value

Age, mean (SD) 24.4 (6.4) 23.8 (5.9) 0.65

Race, no. (%)

African American 33 (82.5) 31 (83.8) 0.99

Caucasian 5 (12.5) 4 (10.8)

Other 2 (5.0) 2 (5.4)

Marital status, no. (%)

Single 31 (75.6) 31 (83.8) 0.37

Married 10 (24.4) 6 (16.2)

Employment status, no. (%)

Unemployed 28 (70.0) 27 (75.0) 0.25

Working full-time 5 (12.5) 7 (19.4)

Working part-time 7 (17.5) 2 (5.6)

Educational attainment, no. (%)

<High school degree 15 (37.5) 16 (43.2) 0.10

High school degree/GED 9 (22.5) 14 (37.8)

>High school degree/GED 16 (40.0) 7 (18.9)

Pregnancy status

Prenatal, no. (%) 12 (29.3) 10 (27.0) 0.83

Postpartum, no. (%) 29 (70.7) 27 (73.0)

p values were calculated using independent sample t tests for contin- uous variables, Fisher’s exact test for categorical variables with small cell sizes, and χ2 tests for categorical variables without small cell sizes

Table 2 Results of the random intercept multilevel model for second- ary outcomes

Coefficient (SE) z value

Mood regulation expectanciesa

Condition (intervention vs usual care)

0.01 (0.03) 0.38

1 week post-intervention −0.02 (0.02) −0.76

3 months post-intervention 0.02 (0.02) 0.92

6 months post-intervention −0.11 (0.02) −4.55***

Condition*1 week post-intervention

0.04 (0.03) 1.28

Condition*3 months post-intervention

0.06 (0.03) 1.86

Condition*6 months post-intervention

0.16 (0.03) 4.83***

Social support

Condition (intervention vs usual care)

3.89 (3.13) 1.24

1 week post-intervention 0.85 (2.54) 0.33

3 months post-intervention 1.78 (2.57) 0.69

6 months post-intervention −4.70 (2.61) −1.80‡

Condition*1 week post-intervention

−0.62 (3.52) −0.18

Condition*3 months post-intervention

−0.76 (3.50) −0.22

Condition*6 months post-intervention

6.67 (3.53) 1.89‡

Active coping

Condition (intervention vs usual care)

−1.18 (2.32) −0.51

1 week post-intervention −0.05 (1.74) −0.03

3 months post-intervention −2.86 (1.77) −1.62

6 months post-intervention −1.82 (1.78) −1.02

Condition*1 week post-intervention

0.35 (2.43) 0.14

Condition*3 months post-intervention

3.41 (2.43) 1.40

Condition*6 months post-intervention

3.30 (2.45) 1.35

Avoidant copinga

Condition (intervention vs usual care)

−0.01 (0.07) −0.09

1 week post-intervention 0.05 (0.05) 1.01

3 months post-intervention −0.01 (0.05) −0.02

6 months post-intervention −0.14 (0.05) −2.63**

Condition*1 week post-intervention

−0.08 (0.07) −1.17

Condition*3 months post-intervention

−0.03 (0.07) −0.35

Condition*6 months post-intervention

0.14 (0.07) 2.01*

‡ p<0.10; *p<0.05; **p<0.01; ***p<0.001 a Outcome was transformed by taking the log

Impact of a perinatal preventive intervention 215

contrary to predictions, was associated with a trajectory of increased avoidant coping among the intervention as com- pared with control participants.

Our analyses indicated that significant group differences in mood regulation expectancies emerged only at the 6- month follow up and were not evident at the post- intervention or 3-month follow up. As shown in Fig. 2, mood regulation expectancies remained relatively constant among intervention participants, whereas control group par- ticipants reported a decrease in positive expectancies be- tween the 3 and 6 month follow ups. This pattern of findings suggests a preventive intervention effect, in which intervention participants were buffered against feelings of lowered self-efficacy in regulating mood over the postpar- tum period. Demands of caretaking for an infant may have had a more negative effect on women’s mood regulation expectancies in the control group because they had not been exposed to specialized skills training through the interven- tion. Research using qualitative methods to explore experi- ences among postpartum women would be beneficial for better understanding this trajectory.

The pattern of findings for perceived social support, al- though not significant, also suggests that the control group may have experienced a trend toward reduced perceptions of social support from the 3- to 6-month follow-up. Findings may also reflect the fact that women in the control group were less likely than intervention participants to have experienced con- sistent supportive contact and information sharing with other at-risk pregnant women and new mothers, including the two booster sessions at follow-ups. This study was not powered to detect differences in perceived social support, but the effect size obtained in this study can inform the design of future trials

in which social support is assessed. In addition, we will con- sider whether additional emphasis in the intervention and booster sessions on increasing sources of social support—and identifying existing supports—may be beneficial.

The unexpected finding for avoidant coping merits in- vestigation in future studies on the MB Course. Similar to the pattern for the mood regulation outcome, the finding was driven by changes among control group women between the 3- and 6-month follow-ups; intervention group women remained relatively constant in levels of avoidant coping, whereas control group women decreased. We view this finding with caution because we are not confident that the Brief COPE scale adequately assessed coping strategies relevant for our study sample. As noted above, the measure is intended to assess coping reactions in response to a specific stressor, rather than more general coping strategies used regularly for coping with daily stresses. Instructions were adapted so as to target “stress in your life since you found out you were pregnant;” however, these instructions were not altered for the post-intervention and follow-up assessments and thus may not have adequately captured the sorts of stressors occurring during the postpartum peri- od. In future studies on the efficacy of the MB Course, it may be beneficial to develop a coping scale that evaluates more explicitly the coping skills targeted in the intervention, such as cognitive restructuring and pleasant activities. Given the pervasive and chronic nature of stress in our target population, it may also be more appropriate to assess the use of day-to-day coping strategies, rather than strategies used in response to a specific stress domain.

Several other perinatal depression prevention trials have evaluated and reported on proximal outcomes related to

a

b

BASE POST 3 MO 6 MO

BASE POST 3 MO 6 MO BASE POST 3 MO 6 MO

BASE POST 3 MO 6 MO

c

d

Fig. 2 Means showing trends in each outcome over time by group. a Mood regulation, b social support, c active coping, d avoidant coping. The means were obtained from multilevel linear regression models accounting for correlations within participants over time but do not account for site (i.e., home visiting program). These means are almost identical to those obtained in analyses accounting for site

216 T. Mendelson et al.

depression risk. These outcomes include anxiety (Dennis et al. 2009; Elliott et al. 2000), social support/loneliness (Brugha et al. 2011; Dennis et al. 2009; Elliott et al. 2000), coping (Brugha et al. 2011), and parental stress and adjustment (Crockett et al. 2008). Among those studies reporting significant intervention effects (Crockett et al. 2008; Dennis et al. 2009; Elliott et al. 2000), some degree of improvement was also found in anxiety (Dennis et al. 2009), social support (Elliott et al. 2000), and postpartum adjustment (Crockett et al. 2008). Ours is the only study of which we are aware to evaluate and report improve- ments in mood regulation expectancies. Taken together, these findings suggest that effective perinatal depression prevention programs may have positive impacts on depression-related risks and protective factors. Elliott and colleagues reported that higher ratings of marital and close relationships among intervention as compared with control participants did not mediate the effect of the intervention on depression (Elliott et al. 2000), whereas other studies did not include evaluation of mediation.

Limitations of the present study include the fact that the intervention trial was powered to detect group differences in depressive symptoms, not in the proximal outcomes assessed in this study. The relatively small sample size may have limited our ability to detect small effects, for example in perceived social support, and did not permit us to evaluate whether the proximal factors mediated intervention impacts on depression. Our Brief COPE measure, as noted above, was also potentially not an adequate index of coping strategies targeted in the intervention. This study also has notable strengths, including a randomized controlled design and lon- gitudinal assessments with 3- and 6-month follow-ups. It is one of the few intervention trials aimed at prevention of postpartum depression that incorporated and reported on mul- tiple measures of risk and protective factors.

Conclusion

In sum, the MB Course shows promise in preventing post- partum declines in women’s perceived ability to regulate their negative mood states effectively. This effect may serve a protective function with respect to prevention of perinatal depression, given evidence for concurrent and prospective associations of mood regulation expectancies with depres- sive symptoms (e.g., Catanzaro et al. 2000; Kassel et al. 2007). Evaluation of risk and protective factors as secondary outcomes in depression prevention trials is a potentially valuable way to better understand long-term intervention benefits. Next steps in the field of perinatal depression prevention include the implementation of larger-scale ran- domized controlled trials that are adequately powered to assess mediational pathways.

Acknowledgments The study was funded by the Johns Hopkins Institute for Clinical and Translational Research (1U54RR023561- 01A1) and The Abell Foundation. We gratefully acknowledge Karen Edwards for her assistance with recruitment, assessments, and other aspects of the study. We would also like to thank the four home visiting programs and their program participants for their strong support of this project.

References

Blaney NT, Fernandez MI, Ethier KA, Wilson TE, Walter E, Koenig LJ, Perinatal Guidelines Evaluation Project Group (2004) Psychosocial and behavioral correlates of depression among HIV-infected pregnant women. AIDS Patient Care 18:405–415

Brugha TS, Morrell CJ, Slade P, Walters SJ (2011) Universal preven- tion of depression in women postnatally; cluster randomized trial evidence in primary care. Psychol Med 41:739–748

Carver CS (1997) You want to measure coping but your protocol’s too long: consider the brief COPE. Int J Behav Med 4:92–100

Catanzaro SJ (1993) Mood regulation expectancies, anxiety sensitivity, and emotional distress. J Abnorm Psychol 102:327–330

Catanzaro SJ, Mearns J (1990) Measuring generalized expectancies for negative mood regulation: initial scale development and implica- tions. J Pers Assess 54:546–563

Catanzaro SJ, Wasch HH, Kirsch I, Mearns J (2000) Coping-related expectancies and dispositions as prospective predictors of coping responses and symptoms. J Per 68:757–788

Cohen S, Hoberman HM (1983) Positive events and social supports as buffers of life change stress. J Appl Soc Psychol 13:99–125

Coyne JC, Downey G (1991) Social factors and psychopathology: stress, social support, and coping processes. Annu Rev Psychol 42:401–425

Crockett K, Zlotnick C, Davis M, Payne N, Washington R (2008) A depression preventive intervention for rural low-income African- American pregnant women at risk for postpartum depression. Arch Womens Ment Health 11:319–325

Dennis CL, Hodnett E, Reisman HM, Kenton L, Weston J, Zupancic J, Stewart DE, Kiss A (2009) Effect of peer support on prevention of postnatal depression among high risk women: multisite random- ized controlled trial. BMJ 338:a3064. doi:10.1136/bmj.a3064

de Tychey C, Spitz E, Briançon S, Lighezzolo J, Girvan F, Rosati A, Thockler A, Vincent S (2005) Pre- and postnatal depression and coping: a comparative approach. J Affect Disord 85:323–326

Drwal J (2008) The relationship of negative mood regulation expectan- cies with rumination and distraction. Psychol Rep 102:709–717

Edge D, Rogers A (2005) Dealing with it: Black Caribbean women’s response to adversity and psychological distress associated with pregnancy, childbirth, and earlymotherhood. Soc SciMed 61:15–25

Elliott SA, Leverton TJ, Sanjack M, Turner H, Cowmeadow P, Hopkins J, Bushnell D (2000) Promoting mental health after childbirth: a controlled trial of primary prevention of postnatal depression. Br J Clin Psychol 39:223–241

Garber J (2006) Depression in children and adolescents: linking risk research and prevention. Am J Prev Med 31:S104–S125

Ialongo NS, Rogosch FA, Cicchetti D, Toth SL, Buckley J, Petras H, Neiderhiser J (2006) A developmental psychopathology approach to the prevention of mental health disorders. In: Cicchietti D, Cohen DJ (eds) Developmental psychopathology, vol.1: Theory and method, 2nd edn. Wiley, New Jersey, pp 968–1018

Kassel JD, Bornovalova M, Mehta N (2007) Generalized expectancies for negative mood regulation predict change in anxiety and de- pression among college students. Behav Res Ther 45:939–950

Kirsch I (1985) Response expectancy as a determinant of experience and behavior. Am Psychol 40:1189–1202

Impact of a perinatal preventive intervention 217

Lazarus RS (1966) Psychological stress and the coping process. McGraw-Hill, New York

Lazarus RS, Folkman S (1984) Stress, appraisal, and coping. Springer, New York

Le HN, Muñoz RF (1998) The Maternal Mood Screener (MMS). Unpublished questionnaire, University of California, San Francisco

Le HN, Perry DF, Stuart EA (2011) Randomized controlled trial of a preventive intervention for perinatal depression in high-risk Latinas. J Consult Clin Psychol 79:135–141

Littleton H, Horsley S, John S, Nelson DV (2007) Trauma coping strategies and psychological distress: a meta-analysis. J Traum Stress 20:977–988

Moskowitz JT, Hult JR, Bussolari C, Acree M (2009) What works in coping with HIV? A meta-analysis with implications for coping with serious illness. Psych Bull 135:121–141

Muñoz RF, Beardslee WR, Leykin Y (2012) Major depression can be prevented. Am Psychol 67:285–295

Radloff LS (1977) The CES-D Scale: a self-report depression scale for research in the general population. Appl Psych Meas 1:385–401

Sheng X, Le HN, Perry D (2010) Perceived satisfaction with social support and depressive symptoms in perinatal Latinas. J Transcult Nurs 21:35–44

StataCorp (2009) Stata Statistical Software: Release 11. College Station, TX, StataCorp LP

Sutton JM (2007) Prevention of depression in youth: a qualitative review and future suggestions. Clin Psychol Rev 27:552–571

Tandon SD, Perry DF, Mendelson T, Kemp K, Leis JA (2011) Preventing perinatal depression in low-income home visiting clients: a random- ized controlled trial. J Consult Clin Psych 79:707–712

Xie RH, He G, Koszycki D, Walker M, Wen SW (2009) Prenatal social support, postnatal social support, and postpartum depression. Ann Epidemiol 19:637–643

Zlotnick C, Miller IW, Pearlstein T, Howard M, Sweeney P (2006) A preventive intervention for pregnant women on public assistance at risk for postpartum depression. Am J Psychiat 163:1443–1445

218 T. Mendelson et al.

Copyright of Archives of Women's Mental Health is the property of Springer Science & Business Media B.V.

and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright

holder's express written permission. However, users may print, download, or email articles for individual use.

Perinatal Depressive Symptoms in HIV-Infected Versus HIV-Uninfected Women:

A Prospective Study from Preconception to Postpartum

Leah H. Rubin, Ph.D.,1 Judith A. Cook, Ph.D.,2 Dennis D. Grey, B.A.,2 Kathleen Weber, R.N., B.S.N.,3

Christina Wells, M.D.,4 Elizabeth T. Golub, Ph.D., M.P.H.,5 Rodney L. Wright, M.D., M.S.,6

Rebecca M. Schwartz, Ph.D.,7 Lakshmi Goparaju, Ph.D.,8 Deborah Cohan, M.D., M.P.H.,9

Melissa L. Wilson, Ph.D., M.P.H.,10 and Pauline M. Maki, Ph.D.1

Abstract

Objective: Depression is common among HIV-infected women, predicts treatment nonadherence, and conse- quently may impact vertical transmission of HIV. We report findings from a study evaluating preconception, pregnancy, and postpartum depressive symptoms in HIV-infected vs. at-risk, HIV-uninfected women. Methods: We examined the prevalence and predictors of elevated perinatal (i.e., pregnancy and/or postpartum) depressive symptoms using a Center for Epidemiological Studies-Depression (CES-D) scale score of ‡ 16 in 139 HIV-infected and 105 HIV-uninfected women (62% African American) from the Women’s Interagency HIV Study (WIHS). Results: The prevalence of elevated perinatal depressive symptoms did not differ by HIV serostatus (HIV-infected 44%, HIV-uninfected 50%, p = 0.44). Among HIV-infected women, the strongest predictor of elevated symptoms was preconception depression (odds ratio [OR] 5.71, 95% confidence interval [CI] 2.67-12.19, p < 0.001); crack, cocaine, and/or heroin use during preconception was marginally significant (OR 3.10, 95% CI 0.96-10.01, p = 0.06). In the overall sample, additional significant predictors of perinatal depression included having multiple sex partners preconception (OR 2.20, 95% CI 1.12-4.32, p = 0.02), use of preconception mental health services (OR 2.51, 95% CI 1.03-6.13, p = 0.04), and not graduating from high school (OR 1.92, 95% CI 1.06-3.46, p = 0.03). Conclusions: Elevated perinatal depressive symptoms are common among HIV-infected and at-risk HIV- uninfected women. Depressive symptoms before pregnancy were the strongest predictor of perinatal symptoms. Findings underscore the importance of early and ongoing assessment and treatment to ensure low vertical transmission rates and improving postpregnancy outcomes for mothers and children.

Introduction

Psychiatric illnesses, in particular mood disorders,commonly co-occur with the human immunodeficiency virus (HIV).1–4 Numerous studies report high rates of de-

pression and depressive symptoms in HIV-infected individ- uals compared to uninfected individuals in the general population, and the rates have been reported to be signifi- cantly higher in women than in men.5 The relationship be- tween depression and HIV infection is clinically important

1Department of Psychiatry and 2Center on Mental Health Services Research and Policy, Department of Psychiatry, University of Illinois at Chicago, Illinois.

3The Core Center of the John H. Stroger Jr. Hospital of Cook County, Chicago, Illinois. 4University of Illinois School of Medicine, Chicago, Illinois. 5Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. 6Department of Obstetrics and Gynecology and Women’s Health Division of Maternal Fetal Medicine, Albert Einstein College of Med-

icine, Bronx, New York. 7Department of Community Health Sciences in the School of Public Health at SUNY Downstate Medical Center, Brooklyn, New York. 8Department of Medicine, Georgetown University Medical Center, Washington, District of Columbia. 9San Francisco General Hospital, University of California San Francisco Department of Obstetrics, Gynecology and Reproductive Sciences,

San Francisco, California. 10University of Southern California, Department of Obstetrics and Gynecology and Department of Preventive Medicine, Los Angeles,

California.

JOURNAL OF WOMEN’S HEALTH Volume 20, Number 9, 2011 ª Mary Ann Liebert, Inc. DOI: 10.1089/jwh.2010.2485

1287

because the co-occurrence of the two conditions provides an additional health burden and can contribute to poor treatment adherence in women6,7 and thereby affect disease progres- sion, viral suppression, and survival.8–11 For example, the Women’s Interagency HIV Study (WIHS), a large multicenter investigation of 3766 women, found that in analyses adjusting for possible confounding factors, HIV-infected women with high levels of depressive symptoms were significantly less likely to be on highly active antiretroviral therapy (HAART) regimens.6 Depressive symptoms in HIV-infected women are associated with substance use, including the use of crack, cocaine, heroin, and amphetamines,7 and injected drugs.12

Depression in HIV-infected women has also been shown to be related to AIDS-related mortality,13 social and economic ad- verse events, and risk-taking behaviors, such as an increased number of sexual partners.12 Other predictors of depressive symptoms in HIV-infected women include low income, less than a high school education, and Hispanic ethnicity.12 Taken together, these findings suggest that reducing depressive symptoms in HIV-infected women may contribute to in- creased medication adherence, improved health, and a re- duction in drug use and other maladaptive behaviors.

Despite a wealth of knowledge about factors related to depression and depressive symptoms in HIV-infected wo- men, little is known about the prevalence of and risk factors for perinatal depression symptoms in this high-risk popula- tion. Perinatal depression encompasses major and minor de- pressive episodes that occur during pregnancy or within 1 year after childbirth. Prevalence estimates of perinatal de- pression in the general population range from 8.5% to 11% during pregnancy and from 6.5% to 12.9% during the first year postpartum depending on the assessment method (i.e., self-report measures, clinical interview), timing of the as- sessment (i.e., first, second, or third trimester of pregnancy or number of weeks after delivery), and population character- istics.14 Depressive symptoms that occur during this specific reproductive time are a serious mental health problem in the general population, and the consequences have critical im- plications for the mother; mother-infant relationship15; the emotional, behavioral, and cognitive development of the child16,17; and marital and family relationships.18 Early diag- nosis and treatment interventions are critical to ensure the welfare of the mother, child, and family.

Only a few studies have examined the prevalence and risk factors of perinatal depression in high-risk populations, such as HIV-infected women and at-risk HIV-uninfected women. To our knowledge, only one study has examined the rate of perinatal depression among HIV-infected women. That study was a retrospective cohort design and involved 273 pre- dominantly minority HIV-infected women from Los Angeles between 1997 and 2006.19 Perinatal depression was based on medical records or multidisciplinary chart notes indicating an onset of depression during pregnancy and/or within 4 weeks after delivery. The overall prevalence of perinatal depression was 30.8%, of depression during pregnancy was 22%, and of depression within 4 weeks postpartum was 18.3%. Estimates of the rate of perinatal depressive symptoms in HIV-infected women were not presented in the study.

There is also limited research on potential risk factors for perinatal depression in HIV-infected women. Only one study in HIV-infected women specifically examined correlates of depression in both the pregnancy and postpartum periods.

Kapetanovic et al.19 reported that past history of a psychiatric illness, substance use during pregnancy, social stress during pregnancy, and lower CD4 + pregnancy nadir were associ- ated with an increased risk of perinatal depression as defined by medical records and multidisciplinary notes in their sam- ple of HIV-infected women (n = 273). Other studies have at- tempted to elucidate the predictors of depressive symptoms in HIV-infected women during either pregnancy or postpar- tum but not both time periods. One study (n = 307) examined correlates of depression during pregnancy as defined by the Center of Epidemiological Studies—Depression scale (CES-D) (interval measure with somatic items removed) in a sample of young, pregnant ( ‡ 24 weeks), HIV-infected women who were predominantly low-income, low-education ( < 12 years), and minority women (71% African American, 20% Hispan- ic).20 Depression during pregnancy was significantly associ- ated with ineffective coping styles, stress, social isolation, and drug and alcohol use. Another study (n = 245) examined postpartum depressive symptoms operationally defined by the CES-D (median split of ‡ 15 was used to define depres- sion) in a sample of young, low-income, low-education ( < 12 years), unmedicated, asymptomatic HIV-infected patients who had recently given birth (18–24 months postpartum) in Thailand.21 Postpartum depression was related to broken relationships, ineffective coping strategies, having an HIV- infected infant, and nondisclosure of HIV status to others. Taken together, these studies help to elucidate some of the determinants of depression during pregnancy and the post- partum period in HIV-infected women.

The primary objective of the present investigation was to assess the prevalence of elevated perinatal depressive symp- toms in a sample of HIV-infected vs. at-risk HIV-uninfected women. The study design involved prospective evaluations of depressive symptoms as measured by the CES-D scale during preconception, pregnancy, and postpartum. Based on previous findings, we hypothesized that perinatal depressive symptoms would be increased in HIV-infected compared to at-risk HIV- uninfected women. The secondary objective was to examine risk factors for elevated perinatal depressive symptoms in HIV- infected and at-risk HIV-uninfected women. Given demon- strations that risky health behaviors7,12 (e.g., drug use, multiple sexual partners), sociodemographic factors12 (e.g., ethnicity, education, income), and service features6 (e.g., insurance sta- tus, use of mental health services) related to depressive symptoms in past studies of HIV-infected women, we exam- ined these factors as risk factors of elevated perinatal depres- sive symptoms. We also examined unintended pregnancy as an additional risk factor, given that unintended pregnancy is associated with an increase in smoking, illicit drug use, and alcohol use compared with intended pregnancy.22

Materials and Methods

Study population

WIHS is a longitudinal, multicenter study of the natural and treated history of HIV in women and includes six clinical sites: Bronx/Manhattan and Brooklyn, NY; Chicago, IL; Washington, DC; and San Francisco and Los Angeles, CA.23,24

Starting in 1994–1995 and again in 2001–2002 and with insti- tutional review board approval, HIV-infected and at-risk uninfected women matched by age, race/ethnicity, level of education, recruitment site, and risk factors, including history

1288 RUBIN ET AL.

of injection drug use and number of sexual partners were enrolled into WIHS.23,24 The HIV-uninfected women are considered at-risk because they are low income and they en- gage in high rates of risky behaviors (e.g., drug use, multiple sex partners), which are risk factors for HIV.

Women eligible for WIHS were aged ‡ 13 years and were able to provide informed consent and complete an interview in either English or Spanish. Although enrollment has been closed since the end of 2002, participants continue to undergo a clinical examination and extensive interview and provide biologic specimens every 6 months. Study methodology, training, and quality assurance procedures and the cohort’s baseline characteristics have been reported previously.23,24

The study sample for this analysis was restricted to WIHS participants who experienced one or more pregnancies be- tween October 1994 and May 2005 (n = 531), experienced a live birth (n = 474), had a known delivery date on file (n = 447), and had CES-D scores across three reproductive stages: precon- ception ( > 10 months before delivery), pregnancy ( £ 10 months before delivery), and postpartum ( £ 12 months after delivery)(N = 244). The primary reason why only 244 of the 447 with known delivery dates had CES-D scores available for each of the three stages was that WIHS administered the CES-D annually from 1994 until October 1998 and then began more frequent administrations, every 6 months, from October 1998 until the present. In the Results section we compare the characteristics of women who were included in this analysis and those who were excluded. CES-D scores at all three time points were necessary because (1) the primary outcome measure (i.e., perinatal depression) was operationally defined as elevated depressive symptoms during pregnancy and/or within 1 year postpartum and (2) the predictor variables, in- cluding elevated depressive symptoms before pregnancy, were from the preconception study visit.

Measures

Depressive symptoms. Elevated perinatal depressive symptoms were assessed with the CES-D,25 a 20-item self- report measure. In accordance with standard definitions of perinatal depression,14 elevated perinatal depressive symp- toms were defined as a CES-D score ‡ 16 during pregnancy ( £ 10 months before delivery) and/or postpartum ( £ 12 months after delivery). In secondary analyses, we examined a more stringent cutoff of 23 and an interval-level version of the sub- scale that excluded somatic items (i.e., fatigue, poor appetite, lack of energy, restlessness, poor concentration) similar to HIV symptoms. The CES-D has excellent reliability, validity, and factor structure25 and is commonly used in studies of HIV,12

including women with HIV.26 Its sensitivity for the DSM diag- nosis of major depression is excellent, in the range of 80%–90%, with a somewhat lower specificity, in the range of 70%–80%.27– 29 The Black Women’s Health Study has demonstrated that the CES-D is a valid measure in African American women.30

Independent measures/potential risk factors of perinatal depressive symptoms. Potential predictors of perinatal de- pression were obtained from the preconception visit, which was defined as the most recent visit at least 10 months before delivery. WIHS participants are asked about health-related variables in the 6 months since their last visit, so these pre- conception variables represent status/exposure in the 6 months preceding the preconception visit (e.g., drug use

represents drug use within the 6 months preceding the pre- conception visit). In addition to HIV status, we examined risky health behaviors, including current smoking status, alcohol—abstainer, light ( < 3 drinks/week), moderate (3–13 drinks/week), heavy ( ‡ 14 drinks/week)—marijuana/hash use, crack, cocaine, and/or heroin use, and current number of sexual partners (male and/or female, < 2, ‡ 2). The socio- demographic factors examined included age, race (Hispanic, African American, Caucasian), level of education ( < high school education, ‡ high school education), average house- hold income ( £ $12,000, > $12,000), employment status, marital status (married/partner, all others). We also exam- ined intention of getting pregnant, depression during pre- conception (CES-D ‡ 16 greater than 10 months before delivery), and service features, including current insurance and use of mental health services. Interactions were also ex- amined between HIV status and risky health behaviors, so- ciodemographic factors, service features, and intention of getting pregnant. In analyses restricted to HIV-infected wo- men, additional variables included HAART use (yes, no), HAART adherence ( ‡ 95% of the time considered adherent, < 95% of the time considered nonadherent), CD4 + cell count (an indicator of the progress of HIV infection where lower counts indicate more severe disease) ( < 200, 200–500, > 500 cells/lL), and viral load (a measurement of the amount of HIV in the blood where higher counts indicate more severe disease) (HIV RNA using cutoffs of 500 and 10,000 copies/mL).

Statistical analysis

Differences between HIV-infected and uninfected women in characteristics measured at preconception were examined using independent t tests for continuous variables and chi- square tests for categorical variables. To compare the preva- lence of perinatal depression (i.e., ‡ 16 on the CES-D during pregnancy and/or postpartum) in HIV-infected vs. uninfected women, chi-square analyses were conducted. To compare the prevalence on the interval-level version of the subscale that excluded somatic items, an independent t test was conducted. Next, logistic regression models were used to identify signif- icant predictors (i.e., serostatus, risky health behaviors, so- ciodemographic factors, service features, intention to get pregnant as well as interactions with serostatus) of elevated perinatal depressive symptoms first in the entire sample and then separately for HIV-infected and at-risk uninfected wo- men. Forward and backward selection procedures were used to determine the best predictors of elevated perinatal depres- sive symptoms. Likelihood ratio tests were used to compare models, and the most parsimonious model was selected. All p values were two-sided. The statistical significance level was set at p = 0.10 in order to examine marginally significant results/ trends (standard criteria in the literature). Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for each of the significant predictor variables using maximum likelihood estimates from logistic regression models. All analyses were conducted using SAS (version 9.2 for Windows, Cary, NC).

Results

Sample characteristics at preconception

Of the 244 participants, 139 were HIV-infected women and 105 were at-risk HIV-uninfected women. Each woman was

PERINATAL DEPRESSIVE SYMPTOMS AND HIV 1289

included in the analysis only once and was included for the first birth in WIHS only. Overall, the 244 women included in the analysis were similar across most sociodemographic and clinical variables to the 230 women who experienced a live birth but did not have a known delivery date on file or CES-D scores for each of the three reproductive stages. The exception was that women included in the analysis more frequently reported crack, cocaine, and/or heroin use during precon- ception (15%) compared to women not included in the anal- ysis (5%); chi-square (1, n = 474) = 11.31, p < 0.001.

Table 1 provides demographic information at preconcep- tion for both HIV-infected (n = 139) and uninfected (n = 105) women and for the two groups combined (n = 244). The sample ranged in age from 17 to 44 years (mean = 29.38, standard deviation [SD] = 5.70). Notably, our sample was largely representative of HIV-infected women in the United States in terms of ethnicity (62% African American), educa- tion (55% high school graduates or equivalent), employment status (39% employed), and household income (54% with < $12,000/year).31 Among HIV-infected women, the median CD4 + lymphocyte count was 423 cells/lL (range 0–1608, median 423), and 8% had CD4 + cell counts < 200 cells/lL. Fifty percent had been prescribed antiretroviral (ARV) ther- apy, and 77% were treatment adherent.32 Plasma viral load was undetectable for 31%, above the lower limit of quantita- tion (LLQ) but < 10,000 copies/mL for 42% and > 10,000 copies/mL for 27% of HIV-infected women.

Overall, HIV-infected and uninfected women were similar across many sociodemographic, clinical, and behavioral var- iables, although there were significant differences between HIV-infected and uninfected women in mean age (30.37 vs. 28.08 years, p = 0.002), health insurance status (80% vs. 64% insured, p = 0.01), and current number of sex partners (17% vs. 38% with ‡ 2 partners, p = 0.001). There were also trends for HIV-infected and uninfected women to differ on marijuana/ hash use (20% vs. 31%, p = 0.07) and use of mental health services (17% vs. 10%, p = 0.095) during preconception.

Prevalence of perinatal depressive symptoms

Table 2 provides the frequency of clinically significant depressive symptoms during preconception, pregnancy, postpartum, and perinatal for HIV-infected women, HIV- uninfected at-risk women, and for the two groups combined. The overall prevalence of elevated perinatal depressive symptoms across the two groups combined was 46%. There were no significant differences in overall prevalence as a function of HIV serostatus ( p = 0.44). There were no signifi- cant differences between HIV-infected and HIV-uninfected women in the prevalence of depressive symptoms during preconception ( p = 0.44), pregnancy ( p = 0.50; first trimester, p = 0.51; second trimester, p = 0.40; third trimester, p = 0.98), or postpartum ( p = 0.49).

In some studies of HIV-infected women, probable depres- sion is defined with a more stringent cutoff of 23 or with an interval-level version of the subscale that excludes somatic items that overlap with symptoms of HIV.6,20 Notably, the prevalence of elevated perinatal depressive symptoms did not differ by HIV serostatus using the more stringent cutoff of 23 (HIV-infected 26%, HIV-uninfected 31%, p = 0.47). There were also no differences between HIV-infected and HIV-uninfected women on the interval level version of the scale excluding

somatic items during preconception ( p = 0.91), pregnancy ( p = 0.26; first trimester, p = 0.25; second trimester, p = 0.82; third trimester, p = 0.92), or postpartum ( p = 0.43).

Risk factors of perinatal depressive symptoms

Table 3 provides the significant predictors from the final logistic regression models examining predictors of perinatal depressive symptoms (CES-D ‡ 16) in the entire sample (n = 244) and in HIV-infected (n = 139) and HIV-uninfected (n = 105) women separately. In the overall sample, precon- ception depression ( p < 0.001), education (less than high school, p = 0.03), current number of sexual partners precon- ception ( ‡ 2 partners, p = 0.02), and mental health service use preconception ( p = 0.04) were significant predictors of peri- natal depressive symptoms (i.e., during pregnancy and/or postpartum); crack, cocaine, and/or heroin use preconception was a marginally significant predictor ( p = 0.08). Serostatus was not a significant predictor of perinatal depressive symptoms in the overall sample and, thus, was not included in the final model. Among HIV-infected women, preconcep- tion elevated perinatal depressive symptoms ( p < 0.001) was a significant predictor of perinatal depression, and crack, co- caine, and/or heroin use almost reached statistical signifi- cance ( p = 0.058). Among at-risk HIV-uninfected women, preconception depression ( p < 0.001) and current number of sexual partners preconception ( ‡ 2 partners, p = 0.01) were significant predictors of perinatal depressive symptoms, and having less than a high school education almost reached sta- tistical significance ( p = 0.08).

In models using a more stringent CES-D cutoff of 23, many of the same variables were significant. In the overall sample, serostatus was not a significant predictor and, thus, was not included in the final models. Significant predictors in the overall sample included preconception depression (OR 4.78, 95% CI 2.51-9.08, p < 0.001), current number of sexual partners preconception ( ‡ 2 partners, OR 2.05, 95% CI 1.06-3.96, p = 0.03), and average household income ( £ $12,000, OR 2.19, 95% CI 1.16-4.14, p = 0.02). Among HIV-infected women, the only significant predictors of elevated perinatal depressive symptoms were preconception depressive symptoms (OR 7.50, 95% CI 3.05-18.41, p < 0.001) and average household income ( £ $12,000, OR 2.79, 95% CI 1.13-6.89, p = 0.03). Among at-risk HIV-uninfected women, preconception depression (OR 3.38, 95% CI 1.33-8.56, p < 0.001), and current number of sexual partners preconception ( ‡ 2 partners, OR 3.18, 95% CI 1.29- 7.83, p = 0.01) were significant predictors.

In models using the interval-level version of the subscale, serostatus was not a significant predictor of elevated de- pressive symptoms during the perinatal period (pregnancy or postpartum) (Table 4). Significant predictors in the overall sample included preconception depression ( p < 0.001), cur- rent number of sexual partners preconception ( p < 0.01), having less than a high school education ( p = 0.02), and crack, cocaine, and/or heroin use ( p = 0.01). Among HIV-infected women, the only significant predictors of perinatal depressive symptoms were preconception depression ( p < 0.001) and being of Hispanic origin ( p = 0.02); mental health service use preconception ( p = 0.05) and average household income ( £ $12,000, p = 0.05) were marginally significant predictors. Among at-risk HIV-uninfected women, preconception de- pression ( p < 0.001) and current number of sexual partners

1290 RUBIN ET AL.

preconception ( p < 0.03) were significant predictors. Crack, cocaine, and/or heroin use was a marginally significant pre- dictor ( p = 0.09) during pregnancy and significant during postpartum ( p = 0.03).

Discussion

The aim of this article was twofold: first, to assess the prevalence of clinically significant perinatal depressive

symptoms in a sample of HIV-infected vs. uninfected women and, second, to examine risk factors for elevated perinatal depressive symptoms in both HIV-infected and uninfected women. Consistent with previous findings,19 the proportion of women with elevated perinatal depressive symptoms was high in our sample before conception (39%) and in the peri- natal period (46%). There were no differences between HIV- infected and uninfected women in likelihood of clinically significant depressive symptoms. HIV-infected women were

Table 1. Demographics at Preconception ( > 10 Months Before Delivery) for HIV-Infected Women, HIV-Uninfected Women, and the Two Groups Combined (n = 244)

HIV status

HIV-infected n = 139 HIV-uninfected n = 105 Overall sample n = 244 Variables n (%) n (%) n (%)

Age (mean, SD)** 30.37 (5.37) 28.08 (5.90) 29.38 (5.71) Race

African American 90 (65) 60 (57) 150 (62) Hispanic 37 (27) 38 (36) 75 (31) White 27 (19) 16 (15) 43 (18)

At least high school graduate or equivalent 75 (54) 58 (55) 133 (55) Currently employed 52 (37) 43 (41) 95 (39) Currently married 57 (41) 39 (37) 96 (39) Average household Income ( £ $12,000/year) 77 (55) 54 (51) 131 (54) Carrying insurance** 111 (80) 67 (64) 178 (73) Current number of sex partners ( ‡ 2)*** 24 (17) 40 (38) 64 (26) Trying to get pregnant 88 (63) 76 (72) 164 (67) Mental health services{ 24 (17) 10 (10) 34 (14) Recent usea

Smoking 66 (48) 46 (44) 112 (46) Alcohol

Abstainer 70 (50) 41 (39) 111 (46) Light ( < 3 drinks/week) 46 (33) 41 (39) 87 (36) Moderate (3–13 drinks/week) 18 (13) 13 (12) 31 (12) Heavy ( ‡ 14 drinks/week) 5 (4) 10 (10) 15 (6)

Marijuana/hash{ 28 (20) 32 (31) 60 (25) Crack/freebase cocaine 15 (11) 11 (11) 26 (11) Cocaine 7 (5) 8 (8) 15 (6) Heroin 3 (2) 7 (7) 10 (4) Composite crack, cocaine, and/or heroin 19 (14) 18 (17) 37 (15)

Disease CD4 number > 500 59 (42) - - ‡ 200 and £ 500 69 (50) - - < 200 11 (8) - - Viral load (HIV RNA, copies/mL)

Undetectable 43 (31) - - < 10,000 58 (42) - - ‡ 10,000 38 (27) - -

Medication use HAART 47 (34) - - Non-HAART 22 (16) - - ARV naive 70 (50) - -

Medication compliance ( ‡ 95%)b 36 (77) - - Intervals (days: Mean, SD)

Preconception ( > 300 days before delivery) 440.30 (88.49) 453.81 (123.66) 446.11 (105.05) Pregnancy ( £ 300 days before delivery) 116.76 (74.32) 111.66 (76.58) 114.57 (75.19) Postpartum ( £ 365 days after delivery) 146.92 (82.10) 143.29 (83.83) 145.36 (82.70) aRecent use refers to within 6 months of the preconception Women’s Interagency HIV Study (WIHS) visit. bData were only available on 47 of the 69 women taking medication because this information was not collected until October 1998 in the

WIHS. Therefore, the percentage of women who were medication compliant is based on 36 of 47 women. ***p < 0.001; **p < 0.01. {p > 0.05 and p < 0.10. ARV, antiretroviral; HAART, highly active antiretroviral therapy; SD, standard deviation.

PERINATAL DEPRESSIVE SYMPTOMS AND HIV 1291

at increased risk for perinatal depression if they had a history of elevated depressive symptoms preconception or used crack, cocaine, and/or heroin preconception, although the substance use effect just missed statistical significance (p = 0.06). Additional risk factors for perinatal depressive symptoms across the combined sample of HIV-infected and

uninfected women included lower education, current number of sexual partners before pregnancy, and mental health ser- vice use before pregnancy.

Our data suggest that the occurrence of clinically signifi- cant depressive symptoms before pregnancy is a strong pre- dictor of perinatal depression in both HIV-infected and

Table 2. Frequency of Depressive Symptoms ( ‡ 16) as Measured by Center for Epidemiologic Studies-Depression Scale as Function of Stage of Pregnancy for HIV-Infected Women,

At-Risk HIV-Uninfected Women, and the Two Groups Combined (n = 244)

HIV Status

Infected (n = 139) Uninfected (n = 105) Overall sample n = 244 Stage of pregnancy n (%) n (%) n (%)

Preconception 58 (42) 38 (36) 96 (39) Pregnancya 47 (34) 40 (38) 87 (36)

First trimester 7 (35) 7 (47) 14 (40) Second trimester 24 (44) 21 (54) 45 (48) Third trimester 30 (47) 24 (47) 54 (47)

Postpartum 43 (31) 37 (35) 80 (33) Perinatalb 61 (44) 52 (50) 113 (46)

aThere were no significant differences in elevated depressive symptoms across trimesters in the overall sample ( p = 0.72), HIV-infected women ( p = 0.65), or HIV-uninfected women ( p = 0.79).

bPerinatal depression was defined as ‡ 16 on the CES-D during pregnancy ( £ 10 months before delivery) and/or postpartum ( £ 12 months following delivery). Perinatal depression was the primary outcome variable in the regression analyses.

Table 3. Significant Predictors from the Final Logistic Regression Models Examining Predictors of Perinatal Depressive Symptoms ( ‡ 16 on Center for Epidemiologic Studies-Depression Scale

During Pregnancy and/or Postpartum) in the Overall Sample (n = 244) and in HIV-Infected (n = 139) and HIV-Uninfected (n = 105) Women Separately

Sample Significant predictors OR 95% CI p value

Overall sample (n = 244) Preconception depression <0.001 No, CES-D < 16 (n = 148) 1.00 Reference Yes, CES-D ‡ 16 (n = 96) 5.62 3.08-10.27

Education 0.03 At least high school or equivalent (n = 133) 1.00 Reference Less than high school (n = 111) 1.92 1.06-3.46

Crack, cocaine, and/or heroin use preconception 0.08 No use (n = 207) 1.00 Reference Any use (n = 37) 2.24 0.92-5.46

Current number of sexual partners 0.02 < 2 (n = 180) 1.00 Reference ‡ 2 (n = 64) 2.20 1.12-4.32

Mental health services used preconception 0.04 No use of services (n = 210) 1.00 Reference Use of services (n = 34) 2.51 1.03-6.13

HIV-infected (n = 139) Preconception depression <0.001 No, CES-D < 16 (n = 78) 1.00 Reference Yes, CES-D ‡ 16 (n = 61) 5.71 2.67-12.19

Crack, cocaine, and/or heroin use preconception 0.06 No use (n = 120) 1.00 Reference Any use (n = 19) 3.10 0.96-10.01

HIV-uninfected (n = 105) Preconception depression <0.001 No, CES-D < 16 (n = 67) 1.00 Reference Yes, CES-D ‡ 16 (n = 38) 8.96 3.26-24.65

Education 0.08 At least high school or equivalent (n = 58) 1.00 Reference Less than high school (n = 47) 2.31 0.91-5.84

Current number of sexual partners <0.01 < 2 (n = 65) 1.00 Reference ‡ 2 (n = 40) 3.54 1.38-9.10

CI, confidence interval; OR, odds ratio.

1292 RUBIN ET AL.

uninfected women. This logical finding stresses the impor- tance of screening and treatment of depression in the pre- conception period that may lead to depression in the perinatal period. Additionally, our findings stress the need for contin- ued monitoring and management of depression in HIV- infected women and in at-risk HIV-uninfected women during pregnancy and the postpartum period. Substance use (crack, cocaine, and/or heroin) was a marginally significant predictor of perinatal depressive symptoms, particularly among HIV- infected women ( p = 0.06). Substance abuse, like HIV disease, has been associated with depression12 and has been shown to affect treatment adherence.6,7 Thus, consideration of sub- stance use among HIV-infected women remains of particu- lar importance when studying depression in general. Having multiple sex partners was a significant predictor of peri- natal depressive symptoms, particularly among at-risk HIV-uninfected women. Risky sexual behaviors, specifically having multiple sex partners, has also been associated with depression in a number of samples including both African American and white women,33 impoverished women,34 and inner-city drug users.35 Continued reinforcement needs to be placed on the negative consequences of risky sexual behaviors in at-risk HIV-uninfected women, which may help to reduce depression as well as HIV transmission.

HIV-infected women in our study did not show elevated perinatal depressive symptoms when compared to at-risk HIV-uninfected women. Factors, such as having multiple sexual partners, low education, mental health service use, and drug use (crack, cocaine, and/or heroin), were more important than serostatus in predicting perinatal depression in WIHS women overall. Although HIV-infected women may have a presumed vulnerability to perinatal depression because of their medical status, there are four possible reasons why the prevalence rates did not differ. First, HIV-infected women in our study and in the community frequently have greater ac- cess to medical care and social services compared with at-risk HIV-uninfected women,36 and this may mitigate against ele-

vated depressive symptoms. Second, women are living longer with HIV; thus, there are fewer concerns about low life ex- pectancy in the context of raising a child. Third, the risk of mother-to-child transmission of HIV has been significantly reduced from rates of 25%–30%37 to < 1%38 given routine HIV screening of pregnant women, use of ARV drugs and zido- vudine therapy, avoidance of breastfeeding, and use of ce- sarean deliveries. Fourth, this sample, like the larger WIHS sample from which it was drawn, has a high background rate of depression6,7,28,39 in part because of prevalence of other factors, including high rates of substance abuse. Those factors appear to have a stronger link to depression than pregnancy. However, this finding may be unique to the WIHS cohort and, thus, our findings need to be replicated in other prospective cohorts of pregnant women with HIV.

The present study is limited by several factors. First, we focused on elevated depressive symptoms as a marker of clinically relevant perinatal depression because we did not have clinical diagnoses. A better approach would have been to use a diagnostic interview, such as the Structured Clinical Interview for the DSM (SCID), to measure clinical depres- sion.6 Second, we did not have data on obstetric factors that may be associated with depression (e.g., preeclampsia, ges- tational diabetes, prematurity, HIV transmission to the infant, neonatal intensive care unit admissions, or other maternal or infant complications). We also lacked data on potential psy- chosocial predictors, including coping style, social stress, and broken relationships, as well as antidepressant use. Although these variables were not available for all the women in this study, these variables are currently being collected in WIHS and can be examined in future studies. Third, we did not have complete data for all women with live births in the WIHS, and this may have introduced biases. The primary reason CES-D scores were missing was because the CES-D was only ad- ministered annually at the onset of WIHS compared to once every 6 months later on. There was an overrepresentation of crack, cocaine, and/or heroin use among HIV-infected women

Table 4. Significant Predictors from the Final Regression Models Examining Predictors of Perinatal Depressive Symptoms (Interval-Level Subscale on Center for Epidemiclogic Studies-Depression

Scale During Pregnancy and Postpartum) in the Overall Sample (n = 244) and in HIV-Infected (n = 139) and HIV-Uninfected (n = 105) Women Separately

Stage

Pregnancy Postpartum

Sample b p b p

Overall sample (n = 244) Preconception depression 0.60 <0.001 0.47 <0.001 Current number of sexual partners ‡ 2 0.14 0.01 - - Having less than a high school education 0.12 0.02 - - Crack, cocaine, and/or heroin use preconception - - 0.14 0.01

HIV-infected (n = 139) Preconception depression 0.65 <0.001 0.55 <0.001 Hispanic origin 0.16 0.02 - - Mental health service use preconception - - 0.14 0.05 Average household income £ $12,000 - - 0.14 0.05

HIV-uninfected (n = 105) Preconception depression 0.53 <0.001 0.36 <0.001 Current number of sexual partners ‡ 2 0.19 0.02 - - Crack, cocaine, and/or heroin use preconception 0.14 0.09 0.20 0.03

PERINATAL DEPRESSIVE SYMPTOMS AND HIV 1293

who become pregnant in WIHS but an underrepresentation of use among HIV-infected women regardless of pregnancy status.7,40 Thus, the present results may still not be general- izable to the broader population of HIV-infected women who become pregnant. Our findings with respect to illicit sub- stances as a predictor of perinatal depression particularly among HIV-infected women are important, however, as this factor seems to have a stronger link to perinatal depression than serostatus.

Future studies need to examine these variables as potential risk factors of perinatal depression in large samples of HIV- infected women with demographically matched comparison groups. Finally, as previously mentioned, women with HIV in WIHS receive social, medical, and other benefits by partici- pating in WIHS, and these factors might mitigate against depressive symptoms. Therefore, studies are needed to see if HIV influences perinatal symptoms in women who do not benefit from participation in prospective clinical studies.

Conclusions

This is the first study to prospectively compare the preva- lence of perinatal depressive symptoms in HIV-infected and at-risk HIV-uninfected women and to identify predictors of elevated perinatal depressive symptoms in these two groups. The present findings suggest that the risk of elevated peri- natal depressive symptoms (1) does not differ between HIV- infected and at-risk HIV-uninfected women, (2) is associated with a history of depression before conception in both HIV- infected and HIV-uninfected women, and (3) is associated with illicit substances (i.e., crack, cocaine, and/or heroin), particularly among HIV-infected women. Identification and treatment of depression when a woman is trying to conceive may lead to decreased perinatal depression. HIV-infected women who use illicit substances should be monitored for the presence of elevated perinatal depressive symptoms.

Acknowledgments

Data in this article were collected by the Women’s Inter- agency HIV Study (WIHS) Collaborative Study Group with centers (Principal Investigators) at New York City/Bronx Consortium (Kathryn Anastos); Brooklyn, NY (Howard Minkoff); Washington, DC, Metropolitan Consortium (Mary Young); The Connie Wofsy Study Consortium of Northern California (Ruth Greenblatt); Los Angeles County/Southern California Consortium (Alexandra Levine); Chicago Con- sortium (Mardge Cohen); Data Coordinating Center (Stephen Gange). The WIHS is funded by the National Institute of Al- lergy and Infectious Diseases (UO1-AI-35004, UO1-AI-31834, UO1-AI-34994, UO1-AI-34989, UO1-AI-34993, and UO1-AI- 42590) and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (UO1-HD-32632). The study is cofunded by the National Cancer Institute, the National Institute on Drug Abuse, and the National Institute on Deafness and Other Communication Disorders. Funding is also provided by the National Center for Research Resources (UCSF-CTSI grant UL1 RR024131). This publication was also made possible by Grant Number K12HD055892 from the National Institute of Child Health and Human Development (NICHD) and the National Institutes of Health Office of Re- search on Women’s Health (ORWH). The contents of this publication are solely the responsibility of the authors and do

not necessarily represent the official views of the National Institutes of Health.

The data were presented at the Society of Biological Psy- chiatry, Vancouver, Canada, May 14–16, 2009.

Disclosure Statement

No competing financial interests exist.

References

1. Gaynes BN, Pence BW, Eron JJ Jr., Miller WC. Prevalence and comorbidity of psychiatric diagnoses based on reference standard in an HIV + patient population. Psychosom Med 2008;70:505–511.

2. Pence BW, Miller WC, Whetten K, Eron JJ, Gaynes BN. Prevalence of DSM-IV-defined mood, anxiety, and substance use disorders in an HIV clinic in the Southeastern United States. J Acquir Immune Defic Syndr 2006;42:298–306.

3. Berger-Greenstein JA, Cuevas CA, Brady SM, Trezza G, Richardson MA, Keane TM. Major depression in patients with HIV/AIDS and substance abuse. AIDS Patient Care STDs 2007;21:942–955.

4. Ciesla JA, Roberts JE. Meta-analysis of the relationship be- tween HIV infection and risk for depressive disorders. Am J Psychiatry 2001;158:725–730.

5. Hartzell JD, Janke IE, Weintrob AC. Impact of depression on HIV outcomes in the HAART era. J Antimicrob Chemother 2008;62:246–255.

6. Cook JA, Cohen MH, Burke J, et al. Effects of depressive symptoms and mental health quality of life on use of highly active antiretroviral therapy among HIV-seropositive wo- men. J Acquir Immune Defic Syndr 2002;30:401–409.

7. Cook JA, Grey DD, Burke-Miller JK, et al. Illicit drug use, depression and their association with highly active anti- retroviral therapy in HIV-positive women. Drug Alcohol Depend 2007;89:74–81.

8. Bangsberg DR, Hecht FM, Charlebois ED, et al. Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population. AIDS 2000;14: 357–366.

9. Ledergerber B, Egger M, Opravil M, et al. Clinical progres- sion and virological failure on highly active antiretroviral therapy in HIV-1 patients: A prospective cohort study. Swiss HIV Cohort Study. Lancet 1999;353:863–868.

10. Paterson DL, Swindells S, Mohr J, et al. Adherence to pro- tease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med 2000;133:21–30.

11. Garcia de Olalla P, Knobel H, Carmona A, Guelar A, Lopez- Colomes JL, Cayla JA. Impact of adherence and highly active antiretroviral therapy on survival in HIV-infected patients. J Acquir Immune Defic Syndr 2002;30:105–110.

12. Moore J, Schuman P, Schoenbaum E, Boland B, Solomon L, Smith D. Severe adverse life events and depressive symp- toms among women with, or at risk for, HIV infection in four cities in the United States of America. AIDS 1999;13: 2459–2468.

13. Leserman J. Role of depression, stress, and trauma in HIV disease progression. Psychosom Med 2008;70:539–545.

14. Gaynes BN, Gavin N, Meltzer-Brody S, et al. Perinatal de- pression: Prevalence, screening accuracy, and screening outcomes. Evid Rep Technol Assess No. 119. 2005:1–8.

15. Beck CT. The effects of postpartum depression on maternal- infant interaction: A meta-analysis. Nurs Res 1995;44:298–304.

1294 RUBIN ET AL.

16. Beck CT. The effects of postpartum depression on child de- velopment: A meta-analysis. Arch Psychiatr Nurs 1998;12:12–20.

17. Sohr-Preston SL, Scaramella LV. Implications of timing of maternal depressive symptoms for early cognitive and lan- guage development. Clin Child Fam Psychol Rev 2006;9:65–83.

18. Manopaiboon C, Shaffer N, Clark L, et al. Impact of HIV on families of HIV-infected women who have recently given birth, Bangkok, Thailand. J Acquir Immune Defic Syndr Hum Retrovirol 1998;18:54–63.

19. Kapetanovic S, Christensen S, Karim R, et al. Correlates of perinatal depression in HIV-infected women. AIDS Patient Care STDs 2009;23:101–108.

20. Blaney NT, Fernandez MI, Ethier KA, Wilson TE, Walter E, Koenig LJ. Psychosocial and behavioral correlates of de- pression among HIV-infected pregnant women. AIDS Pa- tient Care STDs 2004;18:405–415.

21. Bennetts A, Shaffer N, Manopaiboon C, et al. Determinants of depression and HIV-related worry among HIV-positive women who have recently given birth, Bangkok, Thailand. Soc Sci Med 1999;49:737–749.

22. Than LC, Honein MA, Watkins ML, Yoon PW, Daniel KL, Correa A. Intent to become pregnant as a predictor of ex- posures during pregnancy: Is there a relation? J Reprod Med 2005;50:389–396.

23. Barkan SE, Melnick SL, Preston-Martin S, et al. The Wo- men’s Interagency HIV Study. WIHS Collaborative Study Group. Epidemiology 1998;9:117–125.

24. Bacon MC, von Wyl V, Alden C, et al. The Women’s Inter- agency HIV Study: An observational cohort brings clinical sciences to the bench. Clin Diagn Lab Immunol 2005;12: 1013–1019.

25. Radloff LS. The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385–401.

26. Ickovics JR, Hamburger ME, Vlahov D, et al. Mortality, CD4 cell count decline, and depressive symptoms among HIV- seropositive women: Longitudinal analysis from the HIV Epidemiology Research Study. JAMA 2001;285:1466–1474.

27. Low-Beer S, Chan K, Yip B, et al. Depressive symptoms decline among persons on HIV protease inhibitors. J Acquir Immune Defic Syndr 2000;23:295–301.

28. Richardson J, Barkan S, Cohen M, et al. Experience and covariates of depressive symptoms among a cohort of HIV infected women. Soc Work Health Care 2001;32:93–111.

29. Breslau N. Depressive symptoms, major depression, and generalized anxiety: A comparison of self-reports on CES-D and results from diagnostic interviews. Psychiatry Res 1985; 15:219–229.

30. Williams CD, Taylor TR, Makambi K, et al. CES-D four-factor structure is confirmed, but not invariant, in a large cohort of African American women. Psychiatry Res 2007;150:173–180.

31. Centers for Disease Control and Prevention. MMWR anal- ysis provides new details on HIV incidence in U.S. popula- tions, 2008.

32. Department of Health and Human Services/Henry J. Kaiser Family Foundation Panel on Clinical Practices for the Treatment of HIV infection. Guidelines for the use of anti- retroviral agents in HIV-infected adults and adolescents, 2004 revision. Available at aidsinfo.nih.gov/guidelines/ adult/AA_040705.pdfOctober

33. Khan MR, Kaufman JS, Pence BW, et al. Depression, sexu- ally transmitted infection, and sexual risk behavior among young adults in the United States. Arch Pediatr Adolesc Med 2009;163:644–652.

34. Nyamathi AM, Bennett C, Leake B. Predictors of maintained high-risk behaviors among impoverished women. Public Health Rep 1995;110:600–606.

35. Williams CT, Latkin CA. The role of depressive symptoms in predicting sex with multiple and high-risk partners. J Acquir Immune Defic Syndr 2005;38:69–73.

36. Palacio H, Li X, Wilson TE, et al. Healthcare use by varied highly active antiretroviral therapy (HAART) strata: HAART use, discontinuation, and naivety. AIDS 2004;18:621–630.

37. Lindegren ML, Byers RH Jr., Thomas P, et al. Trends in perinatal transmission of HIV/AIDS in the United States. JAMA 1999;282:531–538.

38. Center for Disease Control and Prevention. Achievements in public health: Reduction in perinatal transmission of HIV infection—United States, 1985–2005. MMWR 2006;55: 592–597.

39. Cook JA, Grey D, Burke J, et al. Depressive symptoms and AIDS-related mortality among a multisite cohort of HIV-positive women. Am J Public Health 2004;94:1133– 1140.

40. Wilson TE, Massad LS, Riester KA, et al. Sexual, contra- ceptive, and drug use behaviors of women with HIV and those at high risk for infection: Results from the Women’s Interagency HIV Study. AIDS 1999;13:591–598.

Address correspondence to: Leah H. Rubin, Ph.D.

Department of Psychiatry (MC 913) University of Illinois at Chicago

Chicago, IL 60612

E-mail: [email protected]

PERINATAL DEPRESSIVE SYMPTOMS AND HIV 1295

Copyright of Journal of Women's Health (15409996) is the property of Mary Ann Liebert, Inc. and its content

may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express

written permission. However, users may print, download, or email articles for individual use.

Relationships of Race and Socioeconomic Status to Postpartum Depressive Symptoms in Rural African American and Non-Hispanic White Women

Christyn L. Dolbier • Taylor E. Rush •

Latoya S. Sahadeo • Michele L. Shaffer •

John Thorp • The Community Child Health Network Investigators

Published online: 9 September 2012

� Springer Science+Business Media, LLC 2012

Abstract This study examines the potential racial dispar-

ity in postpartum depression (PPD) symptoms among a

cohort of non-Hispanic white and African American women

after taking into consideration the influence of socioeco-

nomic status (SES). Participants (N = 299) were recruited

from maternity clinics serving rural counties, with over-

sampling of low SES and African Americans. The Edin-

burgh Postnatal Depression Scale (EPDS) was administered

1 and 6 months postpartum, and subjective SES scale at

6 months postpartum. Demographic information was col-

lected during enrollment and 1 month postpartum, with

updates at 6 months postpartum. Separate logistic regres-

sions were conducted for 1 and 6 month time points

for minor-major PPD (EPDS C 10) and major PPD

(EPDS [ 12); with marital status, poverty, education, sub- jective SES, and race predictors entered in block sequence.

After including all other predictors, race was not a signifi-

cant predictor of minor-major or major PPD at 1 or 6 months

postpartum. Subjective SES was the most consistent pre-

dictor of PPD, being significantly associated with minor-

major PPD and major PPD at 6 months postpartum, with

higher subjective SES indicating lower odds of PPD, even

after accounting for all other predictors. This study shows

that significant racial disparities were not observed for

minor-major or major PPD criteria at 1 or 6 months post-

partum. The most consistent and significant predictor of

PPD was subjective SES. Implications of these findings for

future research, as well as PPD screening and intervention

are discussed.

Keywords Postpartum depression � Race � Subjective socioeconomic status � Health disparity � Objective socioeconomic status

Introduction

For women, the postnatal period is the most vulnerable time

for depression than any other time in their lives [1]. In

research on this topic, postpartum depression (PPD) is

commonly characterized as major and minor depressive

symptom levels occurring within the months following

childbirth, with major PPD referring to a diagnosis of or

symptom level related to a form of clinical depression and

minor PPD to a less severe yet still impairing form [1].

Estimates of the prevalence of PPD range from 5 to 25 % or

more depending on whether major and/or minor PPD are

assessed, the population studied, as well as the method and

timing of assessment [1]. Given there are approximately four

million live births annually in the United States (US) [2], this

equates to a minimum of roughly two hundred thousand

women suffering from PPD annually. This maternal suffer-

ing translates into an estimated economic burden of $44

billion annually in the US [3], and deleterious effects asso-

ciated with mother’s health [4, 5], infant health and devel-

opment [6], and mother-infant attachment [7].

While racial disparities have been documented in a

variety of physical and mental health conditions, studies on

C. L. Dolbier (&) � T. E. Rush East Carolina University, Greenville, NC, USA

e-mail: [email protected]

L. S. Sahadeo � J. Thorp University of North Carolina, Chapel Hill, Chapel Hill, NC,

USA

M. L. Shaffer

Penn State Hershey College of Medicine, Hershey, PA, USA

123

Matern Child Health J (2013) 17:1277–1287

DOI 10.1007/s10995-012-1123-7

the prevalence of racial disparity in PPD have provided

mixed results. Some studies report African American

women have higher rates of PPD than non-Hispanic whites

[8–10], while others have reported no racial differences

[11–13]. These conflicting results may be due to the diffi-

culty of differentiating the confounding effects of race

versus socioeconomic status (SES) since African Ameri-

cans are over-represented in low SES, and a lack of con-

sistency in the method and timing of assessing PPD.

Research on traditional objective indicators of SES

(income, education, occupational status) indicates these

inter-linking factors can influence the development of PPD

[14]. For instance, mothers with lower income, education,

and employment status have a greater likelihood of

developing PPD, perhaps because they commonly are

younger, have lower social support, and are more likely to

be single parents [15]. Given the strong relationship of SES

with physical and mental health, researchers have begun to

explore possible mechanisms for this relationship. For

instance, psychosocial processes related to feelings of rel-

ative deprivation and social anxiety may at least partly

explain the SES-health relationship [16]. One such process

is subjective SES, one’s perceived position in the social

hierarchy [17]. Subjective SES is associated with physical

and mental health, and in some cases, is a stronger pre-

dictor than objective indicators of SES [17–19]. Thus,

subjective SES seems to contribute something unique in

the prediction of health outcomes. However, subjective

SES has not been studied in relation to PPD.

An understudied factor often related to race and SES that

may also relate to PPD is the type of area in which people

live (e.g., rural, suburban, urban). Most PPD research has

focused on urban, suburban, and national (mixed) samples,

while the specific challenges of rural settings (e.g., low

community support, low access to appropriate services,

limited transportation, isolated conditions) may influence

PPD [20]. Thus, PPD may affect rural women to a greater

extent [20], a finding supported by a recent study of low

income rural women [12].

The purpose of the current study is to determine whether

disparities in PPD symptoms exist between African Ameri-

can and non-Hispanic white rural women, and whether these

differences are accounted for by objective and subjective

SES, as well as marital status (a noted PPD risk factor) [21].

To address inconsistencies in the method and timing of

PPD assessment, one of the most valid and widely tested

instruments for PPD assessment, the Edinburgh Postnatal

Depression Scale (EPDS), was used at 1 and 6 months

postpartum. The EPDS has been used with diverse racial and

SES populations, and has a significant level of sensitivity

(proportion of depressed women correctly identified) and

specificity (proportion of non-depressed women correctly

identified) based on cut-off scores [22, 23].

Methods

This study is part of a larger study, the first being conducted

by the Community Child Health Network (CCHN), a group

of community organizations and universities partnering

with the Eunice Kennedy Shriver National Institute of Child

Health and Human Development and the National Institute

of Nursing Research to gain new insights into reasons for

disparities in maternal health and child development. The

goals of the network’s first study are to examine the factors

associated with maternal allostatic load (a possible factor in

adverse pregnancy outcomes), and to evaluate the usefulness

of community-partnered participatory research for con-

ducting research on health disparities. These goals are being

achieved through a community-academic partnered, multi-

site observational study examining how stress and resilience

factors interact with biological factors to result in racial

disparities in birth outcomes The CCHN study sites include

three urban (Baltimore, Los Angeles, Washington, DC), one

mixed urban-suburban (Lake County, IL), and one rural

(Eastern North Carolina, ENC). The analyses included here

are based on the ENC site.

Participants

The sample was an availability sample obtained from a

seven-county geographical catchment area in ENC (Bertie,

Edgecombe, Greene, Martin, Pitt, Tyrrell and Washington

counties). Women were recruited prenatally from maternity

clinics and through perinatal community outreach by the

research team and Eastern Baby Love Plus Maternity Care

Coordinators and Community Health Advocates.

Participants met the following inclusion criteria: (1)

18–40 years old; (2) African American or non-Hispanic

white; (3) resided in the catchment area for at least

6 months at time of delivery; and (4) live birth of greater

than or equal to 20 weeks of gestation. Exclusion criteria

included: (1) unable to give informed consent; (2) unable to

fully understand requirements of the study in English; (3)

four or more children; (4) incarcerated or otherwise unable

to participate in the study in a home, community or clinical

setting; and (5) surgically sterile or desired to be surgically

sterilized following the birth. The ENC site oversampled

low SES and African American women to help ensure the

majority of the CCHN total sample was comprised of low

SES and minority women.

At the time of this analysis, 433 participants were

enrolled in the study. Only participants who had completed

both the 1 and 6 months postpartum interviews were

included in the analyses. This excluded 86 (20 %) women

who had missed the window for completing the 6 months

postpartum interview, and 48 (11 %) for whom the window

was still open but who had yet to complete the interview.

1278 Matern Child Health J (2013) 17:1277–1287

123

Overall demographics of the ENC sample (N = 299,

69 %) are shown in Table 1. The majority of the sample

were African American (69 %), categorized as having

household income at or below the federal poverty threshold

(60 %), and were not employed at 1 month (63 %) or

6 months (57 %) postpartum. The largest percentage had

more than a high school education (43 %), and was in a

relationship (but not married) at enrollment (54 %),

1 month (48 %) and 6 months (47 %) postpartum.

Procedures

This study was conducted in accordance with ethical treat-

ment of human research participants after approvals by the

Institutional Review Boards at the participating institutions

were obtained. Women were ‘‘pre-enrolled’’ prenatally or

enrolled postnatally after completing an eligibility interview

and contacted within 1 month postpartum to complete a birth

interview (T0). A 90-min face-to-face interview was con-

ducted during home visits at 1 month (T1, window

2–16 weeks) and 6 months (T2, window 24–39 weeks)

postpartum. Interviewers resided in the catchment area,

underwent extensive training, and were matched with par-

ticipants based on race. Gift cards for completion of the T0

($20), T1 ($25), and T2 ($25) interviews were provided.

Measures

Demographics

Race and ethnicity were determined using two self-identifi-

cation questions included in the T0 eligibility interview that

were recommended by the US Office of Management and

Budget. First, individuals were asked to identify their eth-

nicity as ‘‘Hispanic or Latino’’ or ‘‘Not Hispanic or Latino.’’

Individuals who identified as Hispanic or Latino were

excluded from the analyses (n = 2). Then they were asked to

select one or more racial designations. Individuals who

answered yes to at least one of the two races of focus for the

ENC site (African American, non-Hispanic white) were

eligible for the study. Participants who indicated they were

multi-racial (n = 7) were included using their primary race

designation (4 African American, 3 non-Hispanic white).

Marital status was categorized as being married, in a rela-

tionship, or not in a relationship, and was determined using

questions from the T0 interview about the participant’s

relationship with the father of the baby or other romantic

interest, with updates requested during the T1 and T2 inter-

views. The T1 interview included education questions,

specifically how many years of school completed and highest

degree earned. The T1 interview also included employment

questions, with updates requested during the T2 interview.

Poverty

The T1 interview also included questions regarding

household income and number of people in the household.

Using responses to these questions, the following three

poverty categories were derived based on the US Census

Bureau, Weighted Average Poverty Thresholds 2009 [24],

which vary according to the size of the household without

requiring information on the number of related children

under 18 years: (1) B100 % federal poverty level (FPL)

(indicating income at or less than poverty threshold); (2)

101–200 % FPL; and (3)[200 % FPL. When a participant did not know or refused to report household income

(n = 53), poverty status was imputed based on her receipt

of Medicaid and/or public assistance [food stamps;

Women, Infants, and Children’s Program (WIC); Tempo-

rary Assistance to Needy Families (TANF)]. If she did not

receive any of these, she was categorized as[200 % FPL. If she only received WIC or Medicaid, she was categorized

as 101–200 % FPL. If she received food stamps or TANF,

she was categorized as B100 % FPL.

Subjective SES

The T2 interview included the MacArthur Scale of Sub-

jective Social Status (SES version), designed to capture

one’s sense of relative social standing across the objective

SES indicators [20]. Respondents view a picture of a lad-

der, with each rung labeled with a number from ‘‘1’’ at the

bottom to ‘‘10’’ at the top. It is explained to them that the

ladder represents where the people in the US stand, with

those at the top being people who are the best off (with the

most money, education, and respected jobs), and people at

the bottom being people who are the worst off (with the

least money, education, and respected jobs). Respondents

indicate the number that corresponds to the rung where

they think they stand compared to all the other people in

the US. This measure has demonstrated adequate test–ret-

est reliability and predictive validity [25].

Postpartum Depression Symptoms

The T1 and T2 interviews included the EPDS, which

consists of 10 questions that ask about the experience of

various symptoms of depression (e.g., felt sad or miserable,

so unhappy that had difficulty sleeping) during the past

7 days [22]. Respondents answer each question on a

4-point scale indicating lower to higher levels of the par-

ticular symptom. Question 10 asks about thoughts of

harming oneself. Cronbach’s alpha for the T1 EPDS was

0.83 and for the T2 EPDS was 0.85. Cut-off scores on the

EPDS were used to categorize participants as: (1) negative

Matern Child Health J (2013) 17:1277–1287 1279

123

Table 1 Descriptive Statistics of Study Variables Overall and by Race

Categorical Variables Overall (N = 299) African American (n = 206) Non-Hispanic white (n = 93) p

Frequency (percentage) Frequency (percentage) Frequency (percentage)

PPD (1 month postpartum)a

Minor PPD (scores 10–12) 19 (6.9) 16 (8.5) 3 (3.4) 0.124

Major PPD (scores 13 ?) 29 (10.5) 23 (12.2) 6 (6.9) 0.180

Minor–major PPD (scores 10 ?) 48 (17.5) 39 (20.7) 9 (10.3) 0.035

PPD (6 months postpartum)

Minor PPD (scores 10–12) 25 (8.4) 17 (8.3) 8 (8.6) 0.540

Major PPD (scores 13 ?) 27 (9.0) 21 (10.2) 6 (6.5) 0.206

Minor–major PPD (scores 10 ?) 52 (17.4) 38 (18.4) 14 (15.1) 0.294

Race

African American 206 (69)

Non-Hispanic white 93 (31)

Marital status (enrollment)b

Not in a relationship 52 (17) 47 (23) 5 (5) \0.0001 In a relationship 161 (54) 124 (60) 37 (40)

Married 85 (29) 34 (17) 51 (55)

Marital status (1 month postpartum)c

Not in a relationship 71 (24) 65 (32) 6 (6) \0.0001 In a relationship 141 (48) 104 (51) 37 (40)

Married 84 (28) 34 (17) 50 (54)

Marital status (6 months postpartum)

Not in a relationship 75 (25) 66 (32) 9 (10) \0.0001 In a relationship 142 (47) 108 (52) 34 (37)

Married 82 (27) 32 (16) 50 (54)

Poverty status (1 month postpartum)

B100 % FPL 179 (60) 136 (66) 43 (46) \0.0001 101–200 % FPL 73 (24) 53 (26) 20 (22)

[200 % FPL 47 (16) 17 (8) 30 (32) Employment status (1 month postpartum)

Working 47 (16) 31 (15) 16 (17) 0.210

On leave 64 (21) 39 (19) 25 (27)

Unemployed 188 (63) 136 (66) 52 (56)

Employment status (6 months postpartum)

Working 127 (42) 85 (41) 42 (45) 0.253

On leave 1 (\ 1) 0 (0) 1 (1) Unemployed 171 (57) 121 (59) 50 (54)

Highest degree (1 month postpartum)d

Less than high school 47 (16) 35 (17) 12 (13) 0.007

High school 124 (42) 95 (46) 29 (31)

More than high school 127 (43) 75 (37) 52 (56)

Continuous variables Overall (N = 299) African American (n = 206) Non-Hispanic white (n = 93) p

Mean (standard deviation) Mean (standard deviation) Mean (standard deviation)

EPDS (1 month postpartum)a 5.50 (±4.84) 5.75 (±5.08) 4.97 (±4.23) 0.211

EPDS (6 months postpartum) 4.76 (±4.86) 4.77 (±4.92) 4.72 (±4.74) 0.933

Subjective SES (6 months postpartum) 5.1 (±1.7) 5.1 (±1.8) 5.2 (±1.4) 0.612

Years of school (1 month postpartum) 13.2 (±2.2) 13.0 (±2.0) 13.6 (±2.7) 0.052

1280 Matern Child Health J (2013) 17:1277–1287

123

screen for PPD or non-symptomatic (scores of 0–9); (2)

positive screen for minor PPD (scores of 10–12); or (3)

positive screen for major PPD (scores of 13–30) or EPDS

item 10 responded to affirmatively indicating any suicidal

thoughts regardless of EPDS total score [22, 23]. The

sensitivity and specificity of this measure at the 10-point

cut-off are 83.6 and 88.3 %, respectively. The sensitivity

and specificity at the 13-point cut-off are 58.5 and 97.5 %,

respectively [26].

Statistical Analysis

Descriptive statistics were prepared for all variables

including frequencies and percentages for categorical

variables and means and standard deviations for continuous

variables. Demographic and study variables were com-

pared between African American and non-Hispanic white

women using X2 or Fisher’s exact tests (when expected cell

counts were too sparse for X2 tests to be appropriate) for

categorical variables and two-sample t tests, for quantita-

tive variables. PPD was defined as a binary variable in two

ways: (1) combining minor and major PPD for comparison

with non-symptomatic (minor-major PPD), and (2) com-

bining non-symptomatic and minor PPD for comparison

with major PPD (major PPD). Logistic regression was used

to examine the association between PPD and race after

accounting for the relationships between PPD and poverty

status, education, subjective SES, and marital status. Sep-

arate models were constructed for the 1 and 6 months

postpartum time points for minor-major PPD and major

PPD. Twenty-two mothers were excluded in the analyses

for the 1 month postpartum time point, as the interview

was completed at less than 2 weeks postpartum. Models

were constructed in four steps, sequentially adding in the

variables of interest with race as the primary predictor of

interest being added in the final step: (1) current marital

status; (2) current marital status, poverty status, and edu-

cation; (3) current marital status, poverty status, education,

and subjective SES; and (4) current marital status, poverty

status, education, subjective SES, and race. Exact logistic

regression methods were used when the number of PPD

cases was too small for traditional logistic regression

methods. Findings were considered statistically significant

for p \ 0.05. Analyses were conducted using SPSS (IBM Corporation, Somers, NY) and SAS (SAS Institute Inc.,

Cary, NC).

Results

Descriptive Statistics and Univariate Race Comparisons

Demographics of the sample by race are shown in Table 1.

Compared to non-Hispanic white participants, African

American participants were significantly younger (t =

-4.67, p \ 0.0001), poorer (X2 = 28.15, p \ 0.0001), less educated (X2 = 9.85, p = 0.01), and less likely to be

married at enrollment or one or 6 months postpartum

(X2 = 49.19, 49.93, and 50.32, respectively, all p\ 0.0001). African American and non-Hispanic white participants did

not differ with respect to subjective SES at 6 months post-

partum (t = -0.51, p = 0.61) or employment status at 1 or

6 months postpartum (X2 = 3.12, p = 0.21 and Fisher’s

exact table probability = 0.02, p = 0.25, respectively).

Descriptive statistics for the EPDS and minor, major, and

minor-major PPD categories at 1 and 6 months postpartum

for the overall sample and by race are shown in Table 1,

along with univariate tests for differences by race. At

1 month postpartum the mean EPDS score for the overall

sample was 5.5 (±4.8), with 6.9 % of participants having a

positive screen for minor PPD, 10.5 % having a positive

screen for major PPD, and 17.5 % having a positive screen

for minor or major PPD. African American participants had

a higher mean EPDS score 1 month postpartum compared

to non-Hispanic white participants, but this was not a sig-

nificant difference (t = 1.25, p = 0.21). A significantly

greater percentage of African American participants fell in

the minor-major PPD category (20.7 %) at 1 month post-

partum compared to non-Hispanic white participants

(10.3 %) (X2 = 4.46, p = 0.03). A similar pattern was

observed for minor PPD (African American 8.5 %, non-

Hispanic white 3.4 %) and major PPD (African American

Table 1 continued

Continuous variables Overall (N = 299) African American (n = 206) Non-Hispanic white (n = 93) p

Mean (standard deviation) Mean (standard deviation) Mean (standard deviation)

Age in years (enrollment) 23.6 (±4.7) 22.7 (±4.2) 25.6 (±5.3) \0.0001

FPL federal poverty level, EPDS Edinburgh Postnatal Depression Scale a 2 missing, 22 excluded who completed T1 at \2 weeks postpartum b 1 missing c 3 missing d 1 missing

Matern Child Health J (2013) 17:1277–1287 1281

123

12.2 %, non-Hispanic white 6.9 %) at 1 month postpartum,

but were not significant (X2 = 2.37, p = 0.12; X2 = 1.80,

p = 0.18, respectively).

At 6 months postpartum, the mean EPDS score for the

overall sample decreased to 4.8 (±4.9) with 8.4 % of

participants having a positive screen for minor PPD, 9.0 %

having a positive screen for major PPD, and 17.4 % having

a positive screen for minor or major PPD. The mean EPDS

scores decreased at 6 months postpartum for both African

American and non-Hispanic white participants, with a

greater decrease observed for African American partici-

pants; there was no significant difference in 6 month

postpartum EPDS scores by race (t = 0.09, p = 0.93). The

percentage of African American participants in the minor

PPD category at 6 months postpartum decreased, while

the percentage of non-Hispanic white participants in this

category increased. This led to a change in the racial pat-

tern for minor PPD at this time point, with a greater per-

centage of non-Hispanic white participants having a

positive screen compared to African American participants,

however, this difference was not significant (Fisher’s exact

table probability = 1.00, p = 0.54). The percentage of

African American participants in the major PPD category

at 6 months postpartum decreased and the percentage of

non-Hispanic white participants in this category stayed the

same; the overall pattern remained the same (greater per-

centage of African Americans than non-Hispanic whites),

but was not significant (Fisher’s exact table probabil-

ity = 0.39, p = 0.21). For the minor-major PPD category

at 6 months postpartum, the percentage of African Amer-

ican participants decreased, while the percentage of non-

Hispanic whites increased; the overall pattern remained the

same (greater percentage of African Americans than non-

Hispanic whites), but was not significant (Fisher’s exact

table probability = 0.51, p = 0.29).

Multivariable Logistic Regressions

Results of the logistic regressions for minor-major PPD,

modeled separately at 1 and 6 months postpartum are

summarized in Table 2. At 1 month postpartum, education

was a significant predictor of minor-major PPD until the

inclusion of race in the model after which it became

marginal. At 1 month postpartum, current marital status

was a significant predictor of minor-major PPD until the

inclusion of poverty and education in the model after which

it became marginal, and then lost significance after the

inclusion of race in the model. At 6 months postpartum,

subjective SES was significantly associated with minor-

major PPD, even after including all of the other predictors

in the model, with higher subjective SES indicating lower

odds of PPD. At 6 months postpartum, current marital

status was significantly associated with minor-major PPD

until accounting for poverty and education where it became

marginal.

Results of the logistic regression modeling for major

PPD are summarized in Table 3. At 1 month postpartum,

current marital status approached significance as a predic-

tor of major PPD; however, the significance was not

maintained after adding poverty, education, subjective

SES, and race to the model. Current marital status

approached significance as a predictor of major PPD at

6 months postpartum, but lost significance after subjective

SES and race were included in the model. Education

approached significance as a predictor of major PPD at

6 months postpartum. At 6 months postpartum, only sub-

jective SES was significantly associated with major PPD

even with current marital status, poverty, education, and

race in the model, with higher subjective SES indicating

lower odds of PPD.

Discussion

In the current study of rural African American and non-

Hispanic white women, the prevalence rates at 1 and

6 months postpartum of a positive screen for major PPD

(11, 9 %, respectively) and minor-major PPD (18, 17 %,

respectively) are within the range of those reported in

previous research using varying assessment methods and

time points (5–25 % or more) [1] and those reported in

previous studies that assessed PPD using the EPDS at

similar postpartum time points (6.5–34 %) [12, 14, 26–28].

Of comparable studies, only one focused on a rural,

although low income, sample and reported higher rates of

major PPD (15 %) and minor-major PPD (33 %) at

6–8 weeks postpartum [12], The inclusion of high income

women in the current study may help to explain in part the

lower rates of PPD observed. Comparable studies of urban

samples reported slightly lower rates of major PPD (8 %)

6 months postpartum [28] and minor-major PPD (13 %) at

4–6 weeks postpartum [27]. It is possible that the oral

administration of the EPDS in the current study led to

under-reporting of depressive symptoms as has been

observed in previous research [27]. In comparing the cur-

rent study’s results to those of previous comparable studies,

it appears that rural women with varying levels of SES

experience PPD symptoms to a greater extent than women

in urban areas but not as high as those experienced by rural,

low SES women, and that these rates persist up to 6 months

postpartum.

Initial univariate analyses revealed that a significantly

greater percentage of African American participants had

scores that fell into the minor-major PPD category at

1 month postpartum compared to non-Hispanic white

participants. However, after taking marital status, poverty,

1282 Matern Child Health J (2013) 17:1277–1287

123

education, and subjective SES into consideration, there

were no significant racial differences observed in symp-

toms of PPD at 1 or 6 months postpartum. These results

indicate that when focusing on a rural sample such as this

with varying levels of objective and subjective SES, PPD

symptoms may not differ between African American and

non-Hispanic white women at 1 and 6 months postpartum.

These results are consistent with most of the other studies

that assessed PPD using the EPDS at similar time points

[12, 14, 27–29]. No significant racial differences in major

or minor-major PPD were found in a rural, low income

sample at 6–8 weeks postpartum [12], or a national sample

within 6 months postpartum [14]. SES and marital status

were not taken into consideration in the analyses of racial

Table 2 Logistic Regression Modeling of Minor–Major PPD at 1 and 6 months Postpartum

Variable Step 1 Step 2 Step 3 Step 4

Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p

Minor–Major PPD 1 month postpartum

Current marital status 0.046 0.066 0.075 0.185

Not in a relationship versus

Married

3.272

(1.256–8.523)

0.015 3.500

(1.189–10.305)

0.023 3.322 (1.131–9.757) 0.029 2.601 (0.842–8.035) 0.097

In a relationship versus

Married

2.556

(1.049–6.226)

0.039 2.897

(1.050–7.994)

0.040 2.884 (1.051–7.914) 0.040 2.525 (0.906–7.039) 0.077

Poverty status 0.175 0.169 0.167

B100 % FPL versus

[200 % FPL 0.394

(0.118–1.318)

0.131 0.402 (0.121–1.338) 0.137 0.389 (0.116–1.303) 0.126

101–200 % FPL versus

[200 % FPL 0.722

(0.219–2.383)

0.593 0.756 (0.230–2.479) 0.644 0.725 (0.220–2.392) 0.597

Education 0.038 0.048 0.052

\High school versus [High school

3.397

(1.312–8.793)

0.012 3.239 (1.247–8.408) 0.016 3.230 (1.240–8.409) 0.016

High school versus

[High school 2.147

(0.947–4.865)

0.067 2.114 (0.931–4.798) 0.074 2.032 (0.892–4.633) 0.092

Subjective SES 0.907 (0.751–1.095) 0.309 0.901 (0.747–1.088) 0.279

Race (African American

versus non-Hispanic white)

1.736 (0.743–4.055) 0.202

Minor–Major PPD 6 months postpartum

Current marital status 0.011 0.070 0.103 0.088

Not in a relationship

versus Married

3.915

(1.346–11.385)

0.012 3.288

(1.048–10.316)

0.041 2.605 (0.823–8.249) 0.104 2.889 (0.869–9.604) 0.083

In a relationship

versus Married

4.480

(1.671–12.014)

0.003 3.509

(1.193–10.320)

0.023 3.179 (1.096–9.221) 0.033 3.407 (1.146–10.129) 0.027

Poverty status 0.362 0.455 0.455

B100 % FPL versus

[200 % FPL 1.257

(0.362–4.360)

0.719 1.383 (0.398–4.809) 0.610 1.418 (0.407–4.944) 0.583

101–200 % FPL versus

[200 % FPL 0.680

(0.175–2.638)

0.577 0.808 (0.209–3.125) 0.757 0.832 (0.214–3.229) 0.791

Education 0.425 0.537 0.506

\High school versus [High school

1.621

(0.640–4.106)

0.309 1.477 (0.577–3.784) 0.416 1.493 (0.582–3.831) 0.405

High school versus

[High school 1.626

(0.758–3.487)

0.212 1.534 (0.708–3.321) 0.278 1.572 (0.723–3.419) 0.253

Subjective SES 0.766 (0.632–0.928) 0.006 0.766 (0.632–0.928) 0.007

Race (African American

versus non-Hispanic white)

0.789 (0.373–1.670) 0.536

CI confidence interval, FPL federal poverty level; models were constructed in four steps, increasing the number of predictors at each step:

Step 1—marital status; Step 2—marital status, poverty status, and education; Step 3—marital status, poverty status, education, and subjective

SES; and Step 4—marital status, poverty status, education, subjective SES, and race

Matern Child Health J (2013) 17:1277–1287 1283

123

differences in PPD in either of these studies. In two studies

of urban samples, initial racial differences in major PPD at

6 months postpartum [28] and minor-major PPD at

4–6 weeks postpartum [27] whereby African American

women had higher rates than non-Hispanic white women

were either accounted for by financial hardship [28] or not

confirmed after a clinical interview confirmation of PPD

[27]. Other studies identified in the literature as examining

racial differences in PPD show mixed results [8–11, 13, 30,

31], however, are not comparable given different PPD

assessment methods and assessment time points. Similar to

other comparable studies, the results of the current study

suggest that any initial racial differences in PPD that are

observed do not appear to maintain significance once SES

or confirmation of a clinical diagnosis is taken into

account.

Table 3 Logistic Regression Modeling of Major PPD at 1 and 6 months Postpartum

Variable Step 1 Step 2 Step 3 Step 4

Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p Odds Ratio

(95 % CI)

p

Major PPD 1 month postpartum

Current marital status 0.057 0.265 0.221 0.245

Not in a relationship

versus Married

2.994

(0.743–12.070)

0.123 1.793

(0.374–11.552)

0.642 1.767

(0.367–11.408)

0.659 1.530

(0.297–10.436)

0.847

In a relationship

versus Married

4.515

(1.291–15.796)

0.018 2.664

(0.720–14.892)

0.184 2.709

(0.731–15.169)

0.174 2.521

(0.666–14.313)

0.231

Poverty status 0.172 0.173 0.173

B100 % FPL versus

[200 % FPL 4.108

(0.577–Infinity)

0.182 4.117

(0.582–Infinity)

0.179 4.007

(0.570–Infinity)

0.188

101–200 % FPL versus

[200 % FPL 5.266

(0.726–Infinity)

0.110 5.379

(0.747–Infinity)

0.104 5.166

(0.720–Infinity)

0.113

Education 0.879 0.878 0.912

\High school versus [High school

1.357

(0.352–4.913)

0.805 1.301

(0.335–4.739)

0.864 1.304

(0.336–4.755)

0.861

High school versus

[High school 1.189

(0.434–3.423)

0.897 1.177

(0.429–3.389)

0.916 1.156

(0.419–3.342)

0.948

Subjective SES 0.922

(0.731–1.154)

0.478 0.917

(0.727–1.146)

0.445

Race (African American

versus non-Hispanic white)

1.380

(0.484–4.453)

0.556

Major PPD 6 months postpartum

Current marital status 0.059 0.098 0.173 0.161

Not in a relationship

versus Married

12.461

(1.555–99.885)

0.018 10.678

(1.239–92.008)

0.031 7.698

(0.891–66.491)

0.064 8.411

(0.931–76.014)

0.058

In a relationship

versus Married

10.286

(1.338–79.061)

0.025 8.128

(0.977–67.637)

0.053 6.819

(0.836–55.614)

0.073 7.202

(0.867–59.809)

0.068

Poverty status 0.390 0.442 0.434

B100 % FPL versus

[200 % FPL 2.235

(0.253–19.736)

0.469 2.890

(0.324–25.797)

0.342 2.968

(0.333–26.493)

0.330

101–200 % FPL versus

[200 % FPL 1.078

(0.105–11.032)

0.949 1.622

(0.156–16.831)

0.686 1.668

(0.160–17.344)

0.668

Education 0.078 0.079 0.076

\High school versus [High school

0.186 (0.022–1.575) 0.123 0.157 (0.018–1.358) 0.093 0.158 (0.018–1.362) 0.093

High school versus

[High school 1.690 (0.658–4.341) 0.276 1.544 (0.592–4.025) 0.374 1.578 (0.602–4.133) 0.353

Subjective SES 0.726 (0.563–0.937) 0.014 0.725 (0.562–0.937) 0.014

Race (African American

versus non-Hispanic white)

0.809 (0.286–2.287) 0.689

CI confidence interval, FPL federal poverty level; models were constructed in four steps, increasing the number of predictors at each step: Step 1—marital

status; Step 2—marital status, poverty status, and education; Step 3—marital status, poverty status, education, and subjective SES; and Step 4—marital

status, poverty status, education, subjective SES, and race

1284 Matern Child Health J (2013) 17:1277–1287

123

Subjective SES was the most consistent predictor of

PPD symptoms, predicting major and minor-major PPD at

6 months postpartum. While one other identified study

found a negative relationship between subjective SES and

depression measured using items from the General Health

Questionnaire 30 in male and female London civil service

employees [32], the present study is the first we are aware

of to examine the relationship between subjective SES and

PPD symptoms. The present study’s results suggest that

women who see themselves as less well-off in terms of

income, education, and occupation in comparison to others

may be at a higher risk of developing PPD. This could be

due to these women experiencing greater distress as a result

of their perceived inferior circumstances. For instance, they

may perceive themselves as having lower self-worth and

self-efficacy than other women, feel unable to adequately

provide for their families as well as themselves, and/or

view their current life situations as unlikely to change,

leaving them feeling hopeless and helpless (hallmark

indicators of depression). That the relationship could be

bidirectional (e.g., a woman experiencing depressive

symptoms may be more likely to perceive her SES position

as worse than others) raises the issue of a confounding

effect of depressive symptoms on the appraisal of one’s

subjective SES. However, research has demonstrated that

the appraisal of one’s subjective SES is not significantly

impacted by psychological biases [32] including negative

affect [25], which has conceptual overlap with depressive

symptoms. The observed relationship between subjective

SES and negative affect is more likely the result of the

influence of low subjective SES on negative affect rather

than the reverse [25, 33]. This research supports the idea

that low subjective SES increases the risk for PPD symp-

toms, perhaps in part by increasing negative affect in the

ways described above.

The significance of subjective SES for positive screen

for minor-major and major PPD at 6 months postpartum,

and that its inclusion in the regression models often

reduced the influence of indicators of objective SES sug-

gests that one’s perceived social status may provide pre-

dictive value that is not accounted for by the more

commonly used objective indicators of SES when exam-

ining factors related to PPD. This finding is consistent with

prior research relating SES to other health outcomes [19,

20, 22], but is the first report of an examination of sub-

jective SES in relation to PPD. It would be prudent for

future researchers to include both objective and subjective

measures of SES when trying to understand relative con-

tributions of race and SES in examining racial disparities in

health, especially in rural populations where SES and race

can be easily entangled.

Marital status was a significant predictor of PPD in our

study, but only when entered by itself in the first step of the

minor-major PPD regressions. When poverty status and

education were included, marital status became non-signif-

icant, and in most cases became even more non-significant

with the addition of subjective SES and race. This pattern

suggests that the other predictors, particularly poverty status

and education, may help to account for the initial observed

relationship between marital status and PPD.

Limitations

There were several limitations inherent in this study.

Although the use of established cut-offs using the EPDS for

assessment of PPD is consistent with much PPD research,

it only enables the determination of the likelihood of a

clinical diagnosis of PPD, not an actual diagnosis. Another

limitation is that subjective SES was only assessed at

6 months postpartum, so its relationship with PPD at

1 month postpartum should be interpreted with caution.

However, the MacArthur Scale of Subjective Social Status

has demonstrated adequate test–retest reliability, suggest-

ing this may not be a major concern [25]. As indicated

earlier, depressive symptoms could confound the appraisal

of subjective SES, so assessing both variables prospec-

tively will help elucidate how these variables affect one

another. In addition, that the assessments of subjective SES

and PPD were both via questionnaires, the strength of the

relationship between these two variables may be overesti-

mated given same source bias. However, previous research

showing negative affect has similar relationships with both

objective and subjective SES suggests that same-source

bias may not be a large concern [25]. Also, a potential

confound that was not included in these analyses due to

over-fitting the regression models was the variable preterm

birth (PTB). Racial disparities in PTB are well established

in the literature, with African American women exhibiting

significantly higher rates than non-Hispanic White women

(one in five births and one in 8–9 births, respectively) [34].

PTB has been shown to be directly correlated with

increased risk of PPD [35]; therefore future studies should

take this variable into account. However, exploratory

analyses that included PTB in the first step of the multi-

variable modeling showed that PTB did not affect the

significance of the variables currently presented. A statis-

tical limitation was that the number of cases of positive

screens for major PPD was not large, and thus the level of

power to detect significant effects in the logistic regression

models may be limited. As evidenced by the large 95 %

confidence intervals, the estimated odds ratios have low

precision. It is important to note that poverty status was

derived from household size, which was not asked directly,

and household income, which was not always provided.

Household size was estimated from a series of questions

detailing if the participant lived with parents, children,

Matern Child Health J (2013) 17:1277–1287 1285

123

other family members, and non-family members, which

could lead to an under-reporting of household size. As

previously described, when a participant did not know or

refused to report household income, poverty status was

imputed based on her receipt of Medicaid and/or public

assistance. Lastly, a small percentage of the sample (4 %)

reported taking medication for depression during the

6 months postpartum interview. These participants were

more likely to be non-Hispanic white, married, and

unemployed.

Practical Implications

The prevalence rate of PPD up to 6 months postpartum in

this study’s sample of rural women being higher than that

of urban women highlights the need for routine screening

mechanisms for PPD detection in rural areas. This may be

especially applicable to rural and low SES women, given

their actual and/or perceived limited personal resources and

few opportunities to seek help. Increased screening leads to

increased diagnosis, referral, and treatment, signifying that

screening is a crucial first step toward PPD treatment [36].

The feasibility of PPD screening has been demonstrated

in pediatrician and obstetrician/gynecologist offices and

health departments [36–38]. Additionally, screening could

extend to community-based infant mortality prevention

programs in order to more effectively reach rural popula-

tions. Given its established psychometric properties and

clinical utility, we concur with others who recommend the

EPDS be used as the standard screening measure for PPD

[1, 27], which would further enable comparisons across

studies.

That the most consistent predictor of PPD in this study

was subjective SES focuses attention on it as a possible risk

factor that may be modified through intervention services.

However, before specific interventions can be developed,

findings from this study need replication and further

understanding of why given the same level of objective

SES, rural women who perceive their SES to be lower are

more likely to have PPD symptoms. With this being said,

future avenues for exploration after the problem is more

fully understood include facilitating women’s awareness of

potential resources at their disposal, so they may not feel as

helpless to change their current situation and may help

instill hope that their situations can change for the better;

and enhancing problem-solving skills to help women learn

how to access support and services as well as facilitate

active coping towards presenting problems they are expe-

riencing. Relatedly, given that subjective SES is considered

an average appraisal of the combination of one’s income,

occupation, and education [16], designing interventions to

target improvement on any of these three objective SES

factors should also help improve one’s subjective SES.

Acknowledgment The Community Child Health Network (CCHN) is a community-based participatory research network supported

through cooperative agreements with the Eunice Kennedy Shriver

National Institute of Child Health and Human Development (U

HD44207, U HD44219, U HD44226, U HD44245, U HD44253, U

HD54791, U HD54019, U HD44226-05S1, U HD44245-06S1, R03

HD59584) and the National Institute for Nursing Research (U

NR008929). CCHN reflects joint endeavors of five local sites: (1)

Baltimore: Baltimore City Healthy Start and Johns Hopkins Univer-

sity (Community PI Maxine Vance, Academic PI Cynthia S.

Minkovitz, Project Coordinator Nikia Sankofa, Co-Is Patricia

O’Campo, Peter Schafer); (2) Lake County, Illinois: Lake County

Health Department and Community Health Center and the Northshore

University Health System (Community PI Kim Wagenaar, Academic

PI Madeleine Shalowitz, Project Coordinator Beth Clark-Kauffman,

Co-Is Emma Adam, Greg Duncan*, Chelsea McKinney, Rachel

O’Connell, Alisu Schoua-Glusberg); (3) Los Angeles: Healthy Afri-

can American Families, Cedars-Sinai Medical Center, and University

of California, Los Angeles (Community PI Loretta Jones, Academic

PI Calvin J.Hobel, Co-PIs Christine Dunkel Schetter, Michael C. Lu;

Project Coordinators Mayra Lizzette Yñiguez, Dawnesha Beaver,

Felica Jones); (4) East Carolina University, NC Division of Public

Health, NC Eastern Baby Love Plus Consortium, and University of

North Carolina, Chapel Hill (Community PIs Sharon Evans, Scharina

Oliver*, Richard Woolard, Academic PI John Thorp, Project Coor-

dinators Suzanne Kelly, Latoya S. Sahadeo, Kathryn Salisbury, Co-Is

Julia DeClerque, Christyn Dolbier, Mary Glascoff*, Vijaya Hogan*,

Carol Lorenz, Edward Newton, Belinda Pettiford, Research Partners

Shelia Bunch, Sarah Maddox, Judy Ruffin); and (5) Washington, DC:

Georgetown Center on Health and Education, Washington Hospital

Center, and Developing Families Center (Community PI Loral Pat-

chen, Academic PI Sharon L. Ramey, Academic Co-PI Robin Lanzi,

Project Coordinator Nedaa Timraz, Co-Is Lorraine V. Klerman,

Menachem Miodovnik, Craig T. Ramey, Linda Randolph, Commu-

nity Coordinator Rosalind German). The following individuals also

made critical contributions to CCHN: the Data Coordination and

Analysis Center at the Pennsylvania State University (PI Vernon M.

Chinchilli, Project Coordinator Gail Snyder, Co-Is Rhonda Belue,

Georgia Brown Faulkner*, Marianne Hillemeier, Erik Lehman, Ian

Paul, Jim Schmidt, Michele L. Shaffer, Christy Stetter), Steering

Committee Chairs Mark Phillippe and Elena Fuentes-Afflick*, and

NIH Program Scientists (V. Jeffrey Evans, Tonse Raju) and Program

Officers (Yvonne Bryan*, Michael Spittel, Linda Weglicki, Marian

Willinger). We thank the hospitals and other facilities sponsoring

participant recruitment and the local community advisory boards at

each site. For a detailed overview of CCHN please see the CCHN

public website at http://www.communitychildhealthnetwork.com.

*Indicates those who participated in the planning phase of the CCHN.

References

1. Gaynes, B. N., Gavin, N., Meltzer-Brody, S., et al. (2005).Peri-

natal depression: Prevalence, screening accuracy, and screening

outcomes. Rockville, MDL Agency for Healthcare Research and

Quality. Evidence Report/Technology Assessment, 119, AHRQ

Publication 05-E006-2. http://www.ahrq.gov/clinic/epcsums/peridepsum.

htm.

2. Centers for Disease Control. (2001). Births, marriages, divorces,

and deaths: Provisional data for January–December 2000.

National Vital Statistics Report 49.

3. Mann, R., Gillbody, S., & Adamson, J. (2010). Prevalence and

incidence of postnatal depression: What can systematic reviews

tell us? Archives of Women’s Mental Health, 13, 295–305.

1286 Matern Child Health J (2013) 17:1277–1287

123

4. Norman, E., Sherburn, M., Osborne, R. H., & Galea, M. P.

(2010). An exercise and education program improves well-being

of new mothers: A randomized controlled trial. Physical Therapy,

90, 348–355.

5. Zubaran, C., Foresti, K., Schumacher, M. V., Amoretti, A. L.,

Thorell, M. R., & Muller, L. M. (2010). The correlation between

postpartum depression and health status. Maternal and Child

Health Journal, 14, 751–757.

6. Gavin, N. I., Bradley, N. G., Lohr, K. N., Meltzer-Brody, A.,

Gartlehner, S., & Swinson, T. (2005). Perinatal depression: A

systematic review of prevalence and incidence. Obstetrics and

Gynecology, 106, 1071–1083.

7. Krause, K. M., Ostbye, T., & Swamy, G. K. (2009). Occurrence

and correlates of postpartum depression in overweight and obese

women: Results from the active mothers postpartum (AMP)

study. Maternal and Child Health Journal, 13, 832–838.

8. Howell, E. A., Mora, P. A., Horowitz, C. R., & Leventhal, H.

(2005). Racial and ethnic differences in factors associated with

early postpartum depressive symptoms. Obstetrics and Gyne-

cology, 105, 1442–1450.

9. Logsdon, M. C., & Usui, W. (2001). Psychosocial predictors of

postpartum depression in diverse groups of women. Western

Journal of Nursing Research, 23, 563–574.

10. Segre, L. S., Losch, M. E., & O’Hara, M. W. (2006). Race/

ethnicity and the perinatal depressed mood. Journal of Repro-

ductive and Infant Psychology, 24, 99–106.

11. Gross, K. H., Wells, C. S., Radigan-Garcia, A., & Dietz, P. M.

(2002). Correlates of self-reports of being very depressed in the

months after delivery: Results from the pregnancy risk assess-

ment monitoring system. Maternal and Child Health Journal, 6,

247–253.

12. Hutto, H. F., Kim-Godwin, Y., Pollard, D., & Kemppainen, J.

(2011). Postpartum depression among White, African Ameri-

can, and Hispanic low-income mothers in rural Southeastern

North Carolina. Journal of Community Health Nursing, 28,

41–53.

13. Hobfoll, S. E., Ritter, C., Lavin, J., Hulsizer, M. R., & Cameron,

R. P. (1995). Depression prevalence and incidence among inner-

city pregnant and postpartum women. Journal of Consulting and

Clinical Psychology, 63, 445–453.

14. Mayberry, L., Horowitz, J., & Declercq, E. (2007). Depression

symptom prevalence and demographic risk factors among U.S.

women during the first 2 years postpartum. Journal of Obstetric,

Gynecologic, and Neonatal Nursing, 36, 542–549.

15. Abrams, L. S., Dornig, K., & Curran, L. (2009). Barriers to

service use for postpartum depression symptoms among low-

income ethnic minority mothers in the United States. Qualitative

Health Research, 19, 535–551.

16. Adler, N. E., & Snibbe, A. C. (2003). The role of psychosocial

processes in explaining the gradient between socioeconomic

status and health. Current Directions in Psychological Science,

12(119), 23.

17. Demakakos, P., Nazroo, J., Breeze, E., & Marmot, M. (2008).

Socioeconomic status and health: The role of subjective social

status. Social Science and Medicine, 67, 330–340.

18. Adler, N. E., Epel, E., Castellazzo, G., & Ickovics, J. (2000).

Relationship of subjective and objective social status with psy-

chological and physiological functioning: Preliminary data in

healthy white women. Health Psychology, 19, 586–592.

19. Ghead, S. G., & Gallo, L. C. (2007). Subjective social status,

objective socioeconomic status, and cardiovascular risk in

women. Health Psychology, 26, 668–674.

20. Crockett, K., Zlotnick, C., Davis, M., Payne, N., & Washington,

R. (2008). A depression preventive intervention for rural

low-income African American pregnant women at risk for

postpartum depression. Archives of Women’s Mental Health, 11,

319–325.

21. Beck, C. T. (2001). Predictors of postpartum depression: An

update. Nursing Research, 50, 275–285.

22. Cox, J. L., Holden, J. M., & Sagovsky, R. (1987). Detection of

postnatal depression: Development of the 10-item Edinburgh

Postnatal Depression Scale. British Journal of Psychiatry, 150,

782–786.

23. Matthey, S., Henshaw, C., & Barnett, B. (2006). Variability in

use of cut-off scores and formats on the Edinburgh postnatal

depression scale: Implications for clinical and research practice.

Archives of Women’s Mental Health, 9, 309–315.

24. http://www.census.gov/hhes/www/cpstables/032010/pov/new35_

000.htm (accessed May 19, 2011).

25. Operario, D., Adler, N. E., & Williams, D. R. (2004). Subjective

social status: Reliability and predictive utility for global health.

Psychology & Health, 19, 237–246.

26. Murray, L., & Carothers, A. D. (1990). The validation of the

Edinburgh post-natal depression scale on a community sample.

British Journal of Psychiatry, 157, 288–290.

27. Horowitz, J., Murphy, C., Gregory, K., & Wojcik, J. (2011). A

community-based screening initiative to identify mothers at risk

for postpartum depression. Journal of Obstetric, Gynecologic,

and Neonatal Nursing, 40, 52–61.

28. Rich-Edwards, J. W., Kleinman, K., Abrams, A., Harlow, B. L.,

McLaughlin, T. J., Joffe, H., et al. (2006). Sociodemographic

predictors of antenatal and postpartum depressive symptoms

among women in a medical group practice. Journal of Epide-

miology and Community Health, 60, 221–227.

29. Yonkers, K. A., Ramin, S. M., Rush, A. J., Navarrete, C. A.,

Carmody, T., March, D., et al. (2001). Onset and persistence of

postpartum depression in an inner-city maternal health clinic

system. The American Journal of Psychiatry, 158, 1856–1863.

30. Ritter, C., Hobfoll, S., Lavin, J., Cameron, R., & Hulsizer, M.

(2000). Stress, psychosocial resources, and depressive symp-

tomatology during pregnancy in low-income, inner-city women.

Health Psychology, 19, 576–585.

31. Surkan, P. J., Peterson, K., Hughes, M. D., & Gottlieb, B. R.

(2006). The role of social networks and support in postpartum

women’s depression: A multiethnic urban sample. Maternal and

Child Health Journal, 10, 375–383.

32. Singh-Manoux, A., Adler, N., & Marmot, M. G. (2003). Sub-

jective social status: Its determinants and its association with

measures of ill-health in the Whitehall II study. Social Science

and Medicine, 56, 1321.

33. http://www.macses.ucsf.edu/research/psychosocial/subjective.

php (accessed November 10, 2011).

34. Centers for Disease Control and Prevention. CDC Health Dis-

parities and Inequalities Report- United States, 2011. U.S.

Department of Health and Human Services; 2011.

35. Vigod, S. N., Villegas, L., & Ross, L. E. (2010). Prevalence and

risk factors for postpartum depression among women with pre-

term and low-birth-weight infants: A systematic review. BJOG,

117, 540–550.

36. Chaudron, L. H., Szilagyi, P. G., Kitzman, H. J., Wadkins, I. M., &

Conwell, Y. (2004). Detection of postpartum depressive symptoms

by screening at well-child visits. Pediatrics, 113, 551–558.

37. Jesse, D. E., Morrow, J., Herring, D., Dennis, T., & Laster, B. M.

(2009). Translating research to prevent antepartum depression in a

local health department prenatal clinic: A model approach. Journal

of Public Health Management and Practice, 15, 160–166.

38. Flynn, H. A., O’Mahen, H. A., Massey, L., & Marcus, S. (2006).

The impact of a brief obstetrics clinic-based intervention on

treatment use for perinatal depression. Journal of Women’s

Health, 15, 1195–1204.

Matern Child Health J (2013) 17:1277–1287 1287

123

Copyright of Maternal & Child Health Journal is the property of Springer Science & Business Media B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Health Care for Women International, 32:39–56, 2011 Copyright © Taylor & Francis Group, LLC ISSN: 0739-9332 print / 1096-4665 online DOI: 10.1080/07399332.2010.529353

Ethnic-Specific Perceptions of Altered Control Among American Women: Implications for

Health Promotion Programs After Pregnancy

BOBBIE STERLING, EILEEN FOWLES, SUNGHUN KIM, LARA LATIMER, and LORRAINE O. WALKER

School of Nursing, University of Texas at Austin, Austin, Texas, USA

This study describes some ethnically diverse psychosocial and be- havioral contexts that influence low-income postpartum women’s ability to focus on their health. Content analysis was conducted on data from ethnically concordant focus groups of low-income American Anglo, African American, and Hispanic women 12 to 24 months postpartum. All women described altered sense of “per- ceived control” as the context contributing to their postpartum health status, but sources and management of this perception var- ied by ethnicity. Effective health promotion interventions may in- clude self-image building activities, stress management strategies and interventions that include family members but should address unique ethnic-specific contexts of low-income mothers.

Although family support of new mothers is widely practiced throughout the world, mothers continue to experience psychosocial and physical health needs after the first 4–6 weeks of recovery from childbirth (Cheng, Fowles, & Walker, 2006; Mercer, 1995). New mothers experience emotional and psy- chological stressors throughout the first year postpartum that alter their physi- cal and emotional well-being and functioning. As a result they may be unable to engage in health promoting activities, such as preparing well-balanced meals or engaging in physical activity, as they strive to meet the demands of

Received 31 March 2009; accepted 24 August 2010. We would like to acknowledge our focus group co-facilitators: Drs. Alexandra Garcia,

Sandra Jenkins, and Susan Wilkinson. This study was supported in part by grant R01 NR 04679.

Address correspondence to Dr. Eileen Fowles, PhD, RNC-OB, School of Nursing, Univer- sity of Texas at Austin, 1700 Red River Street, Austin, TX 78701-1499, USA. E-mail: efowles@ mail.nur.utexas.edu

39

40 B. Sterling et al.

parenting (Walker & Wilging, 2000). Consequently, low-income new mothers in particular may experience elevated depressive symptoms (Petterson & Al- bers, 2001) and higher prevalence of postpartum weight retention (Olson, Strawderman, Hinton, & Pearson, 2003; Parker & Abrams, 1993; Shrewsbury, Robb, Power, & Wardle, 2009; Wolfe, Sobal, Olson, Frongillo, & Williamson, 1997).

In this study of low-income American women we sought to under- stand their ethnic-specific (African American, Hispanic, and Anglo/White) psychosocial and behavioral contexts during postpartum. Understanding these contexts may aid in tailoring health promotion programs to reduce postpartum depressive symptoms and retained weight and increase mater- nal well-being. We believe these findings will be of interest in the United States and in other countries where international migration has led to ethnic diversity, such as countries of Europe and Oceania and to a lesser extent in Africa, Asia, and Latin America (Castles, 2000).

BACKGROUND

Depression and obesity (usually defined as a body mass index > 30; BMI) are examples of health-related conditions to which women are more vulnerable. As noted by the World Health Organization (WHO, 2009a), “Unipolar de- pression, predicted to be the second leading cause of global disability burden by 2020, is twice as common in women.” Although reports of the prevalence of diagnosed postpartum depression vary by country, its prevalence is sub- stantial in one or more countries on virtually every continent (Halbreich & Karkun, 2006). Similarly, obesity is more prevalent in women than in men (WHO, 2000). Among 91 WHO member states reporting rates of obesity between 2000 and 2006 for female adults 15 or more years of age, more than half reported obesity rates of 10% or higher. These rates occurred in countries as diverse as Columbia, Canada, Egypt, Lithuania, Spain, Mongolia, and Zimbabwe (WHO, 2009b). Consequently, both postpartum depression (Halbreich & Karkun) and weight retention after pregnancy (Viswanathan et al., 2008) are important maternal health problems. Elevated depressive symptoms and depressed mood not only affect women’s quality of life (Da Costa, Dritsa, Rippen, Lownesteyn, & Khalife, 2006), but also that of their young children (Beck, 1998; Civic & Holt, 2000; Petterson & Albers, 2001). Similarly, postpartum retained weight may increase lifetime weight gain (Linne, Dye, Barkeling, & Rossner, 2004; Rooney, Schauberger, & Mathiason, 2005), which is worrisome for the health effects on women (Field et al., 2001; Must et al., 1999) and impact on future pregnancies (Cedergren, 2004; Cedergren, & Kallen, 2003). Postpartum interventions focused on physical activity, sound eating patterns, and improved coping (which reduce risk factors) in turn may reduce depressed mood and weight retention and pro- mote health among low-income women (Howell, Mora, DiBonaventura, &

Ethnic Context Affecting Weight Loss 41

Leventhal, 2009; Oken, Taveras, Popoola, Rich-Edwards, & Gillman, 2007; Olson et al., 2003).

Furthermore, customizing health promotion interventions is one tool to increase their suitability, appeal, effectiveness, and relevance to low-income women. Focus group interviews (Krueger & Casey, 2000) or other interview methods are commonly used to gather data that may guide the develop- ment of health promotion interventions for specific population subgroups. These approaches are particularly relevant when seeking information on commonalities and differences in individual experiences. Several authors have reported focus group or interview findings relevant to postpartum pro- grams for low-income or ethnic minority women (Amankwaa, 2003; Chang, Nitzke, Guilford, Adair, & Hazard, 2008; Ebbeling, Pearson, Sorensen, Levine, Hebert, et al., 2007; Kieffer, Willis, Arellano, & Guzman, 2002; Sterling et al., 2009; Thornton et al., 2006). One key finding for a sample of African Amer- ican and White women was that stress and negative emotions led to eating calorie dense foods and erratic meal eating patterns, but it was unclear if this affected one ethnic group more characteristically than another (Chang et al., 2008). In another study of Spanish-speaking Latinas, authors cited spousal preferences and support as important influences on timing of meals and composition of meals for women during pregnancy and the postpartum period (Thornton et al., 2006), but it remains unclear if these findings may apply to women of other ethnic groups.

In this study we extend an analysis of recently reported focus group data related to women’s perceptions of control in their lives after childbirth (Sterling et al., 2009). In that earlier report, a low-income, ethnically diverse sample of women recounted their experiences described within the overar- ching theme of altered perceived control related to weight and depressive symptoms after childbirth. Subsequent analyses revealed the presence of ethnically distinct themes that may be relevant to the development of post- partum health promotion programs. As a result, the purpose of this study is to report the ethnic-specific themes of Anglo (White), African American, and Hispanic women related to their psychosocial and behavioral contexts de- scribed within interviews more than 1 year postpartum. The ultimate goal was to incorporate knowledge of these ethnic-specific contexts into the design of an intervention program to reduce postpartum depressive symptoms and retained weight and to promote health among low-income American women.

METHODS

Focus Group Participants

A purposeful sample of 25 low-income women (6 Anglo, 9 African Amer- ican, and 10 Hispanic) participated in six ethnically concordant focus groups conducted in a university setting in the southwestern United States. These women had completed a year-long longitudinal study of pregnancy

42 B. Sterling et al.

associated weight change and psychosocial and behavioral correlates of weight status among low-income ethnic minority women during postpartum (Walker, Freeland-Graves, et al., 2004). Women were invited if, at 12 months postpartum, they had a BMI of equal to or greater than 25 (WHO, 2000), had a total score of 16 or greater on the Center for Epidemiologic Study-Depression (CES-D; Radloff, 1977), or both (n = 10, 3, and 12, respectively). The preced- ing criteria were used to increase the likelihood that women would be able to describe experiences related to increased postpartum weight or depressive symptoms, or both. The high prevalence of each condition is likely to re- sult in a large number of low-income women with each condition (Walker, Timmerman, Kim, & Sterling, 2002; Walker, Timmerman, Sterling, Kim, & Dickson, 2004). Additional participant criteria included that women were 18 years or older; self-declared as Hispanic, African American, or Anglo (White) ethnicity; were able to speak and read English; had delivered no more than three children; were healthy, without coexisting medical problems during the pregnancy; had delivered a singleton healthy term newborn; and had low incomes defined as receiving prenatal care funded by Medicaid, a gov- ernment program with an income threshold of 185% of the federal poverty level. All focus group participants had received perinatal care by private physicians in a large metropolitan area in the southwest United States.

Demographic characteristics of women participating in this study are displayed by ethnic group in Table 1. The mean age of focus-group partici- pants at enrollment to the original longitudinal study was 24.3 years (SD ± 4.3). Eight women reported having a partial high school education, while 10 women had completed high school, and seven attended some college or had completed college. Eleven women reported a family income (at enroll- ment into the longitudinal study) of less than $15,000, while seven women reported a family income between $15,000 and $29,999. Thirteen women were living with a spouse or partner, while 11 were single at the time of initial enrollment into the larger study.

Study Protocol

University Institutional Review Board approval for this current study was received and participants signed informed consent forms. Each focus group consisted of 3–5 ethnically similar women and met for approximately 2 hours in a private room located in a women’s wellness center administered by a large university School of Nursing in the southwestern United States. Partic- ipants were informed that the purpose of the focus group was to receive women’s stories and experiences of stress (because of the stigma of mental health problems, we did not use the term “depression” or “depressive symp- toms” with participants) and weight changes during postpartum. All partic- ipants were familiar with the focus group cofacilitators and the geographic location because of their earlier participation in the longitudinal study. Two members of the research team served as focus group cofacilitators:

Ethnic Context Affecting Weight Loss 43

TABLE 1 Sample Demographics

Ethnicity

Variables Anglo African

American Hispanic Total

Number of participants (n) 6 9 10 25 Age (mean, year) 26.2 25.7 22.0 24.3 BMI at 12th month after delivery

(mean) 35.87 30.76 34.60 33.52

Depressive symptoms (CESD; mean)

18.50 21.18 26.13 22.52

Education (n) Partial high school 1 2 5 8 (32%) Completed high school 2 5 3 10 (40%) Some college 3 2 2 7 (28%)

Family income (n) <$15,000 1 3 7 11 (44%) $15,000–$29,999 2 3 2 7 (28%) >$30,000 2 3 0 5 (20%)

Married (n) Yes 2 4 7 13 (52%) No 4 4∗ 3 11 (44%)

Parity (n) 1 1 1 3 5 (20%) 2 4 4 3 11 (44%) 3 1 4 4 9 (36%)

Type of delivery (n) Vaginal 3 5 7 15 (60%) Cesarean 3 4 3 10 (40%)

Feeding method at delivery (n) Breastfeeding only 1 1 1 3 (12%) Combination 2 4 2 8 (32%) Formula only 3 4 7 14 (56%)

∗Data not available from one participant.

one was Anglo and the other was from the same ethnic background as the group members. Women received a small stipend at the completion of the study. Women in each ethnically concordant focus group were asked to respond to similar questions. Interview guide questions were developed through research team consensus and were designed to elicit the women’s experiences, perceptions, and interpretations of their ethnic-specific context of weight changes, stress, and depressive symptoms during postpartum. (See Figure 1 for examples of interview questions.)

Data Analysis

Participant responses to the focus group questions were analyzed us- ing established content analysis procedures (Miles & Huberman, 1994). Discussions were audio-taped and transcribed verbatim. One cofacilitator recorded field observations during focus group discussions that included descriptions of the setting and key points discussed. At the conclusion, the

44 B. Sterling et al.

• How has your weight changed since becoming pregnant and having your baby? • How have those weight changes affected you? • How do the weight changes in this recent pregnancy compare to the weight changes

in previous pregnancies? • What were some situations that caused you stress during your postpartum period? • Do you think that you are more or less healthy now than before pregnancy, and what

do you think led to your feeling of being more or less healthy?

FIGURE 1 Examples of focus group interview questions.

cofacilitator repeated these key points to the focus group participants for confirmation and further clarification. The other cofacilitator recorded obser- vations shortly after completion of the discussion. Cofacilitators compared their notes that also were used in the analysis. The analysis procedure in- volved two phases. In the first phase, selected members of the research staff (BS, SK, LL) analyzed transcripts independently. Participant statements were subjected to open coding so that recurring common themes and subthemes could be identified within the overarching theme of altered perceived con- trol. In the second phase, the major themes and subthemes unique to each ethnic group were identified and then compared by members of the staff (BS, SK, LL) for commonality and consistency. The concepts identified as ethnically specific then were compared again with those of the other groups and refined within the ethnic contexts. All members of the research team met on multiple occasions to identify the participant words and phrases that were the most representative textual examples describing the ethnic-specific context. This was done to ensure credibility of the findings and to identify and describe the ethnic-specific contexts accurately.

FINDINGS

Overarching Theme of Altered Perceived Control

The overarching theme of altered perceived personal control was descriptive of the lived context during postpartum in this sample of low-income ethni- cally diverse women (Sterling et al., 2009). For analysis purposes “perceived control” was defined as “the sense of managing some areas of their lives with their own knowledge and skills as well as preferences and choices.” Women in this sample described instances in which they lacked or had lost individual control of their lives and activities during the extended postpar- tum period, yet each ethnic group perceived this altered control differently and this perception had differing meanings for them. This overarching theme was manifested in statements about having no time to meet their own needs and struggling with heavy duties of child rearing. A subtheme of strategies to

Ethnic Context Affecting Weight Loss 45

manage or cope with the changes caused by the context also emerged. For instance, while all of the focus groups commented on the heavy responsibil- ities involved in childrearing, ethnically distinct coping strategies were used to manage the child-caring obligations and reflect an ethnic-specific context.

Anglo Mothers Context

For Anglo mothers, the context of “perceived control” was manifested in the theme of an altered sense of self. One characteristic of this theme was that Anglo mothers described how their postpartum health status, physical pain, and chronic fatigue differed from their prepregnancy state. Because of the postpartum weight retention, they felt “heavier” than they used to be and, in turn, were less physically active. Moreover, they recognized that their increase in TV watching and snacking contributed to the weight retention. The secondary analysis reported here reveals that this perception was unique to Anglo mothers:

[I feel] just heavier. . . . I don’t know, maybe a little bit lazier . . . not doing as much like watching more TV.

I don’t work, so I don’t do anything, so I’m sitting a lot, so I’m getting bigger and bigger . . . just mainly I didn’t realize how much sitting time I did and then since I sit at home, you know, I watch TV and I want to go get a snack, and I snack. I don’t really eat, I snack.

Many Anglo mothers complained about people’s preferences for body im- ages of “thinner” women and differences in expectations about physical size between women and men:

People tend to be nicer and kinder to “thinner” women.

Men and women both are obsessed with women’s size. Men’s size does not matter that much anywhere

Another characteristic of the Anglo mother’s theme of an altered sense of self revolved around changes in their self-image. One mother agonized about her conflicting feelings about being mother and a career woman, since she had been laid off during maternity leave from her prior workplace:

I went from a $40,000 a year job down to $8.50 an hour because my employer laid me off while I was on maternity leave. That was fun. And you sacrifice, you know. OK, am I going to go back to that and work 20 . . . be on call 24-7, possibly work six 8-hour days a week or am I going to be a Mom?”

46 B. Sterling et al.

Anglo women adopted different approaches to manage the contextual changes by reframing their sense of self. Women who had conflicting feel- ings about being a mother and career woman managed by reframing their life priorities. One Anglo mother lamented, “I have no self-identity; I am just his Mom,” while another reframed her personal priorities by concluding the “Number one thing is being a Mom.” Another mother reframed her self- image by insisting that the dignity of a woman has nothing to do with her physical size but reflected the changes in sense of self during postpartum by saying, “I just want to become comfortable with myself and my size first.”

African American Mothers’ Context

For African American mothers, the context of altered “perceived control” was manifested in the theme focusing on the demands of parenting. One characteristic of this theme was that these mothers frequently mentioned the hardships of child care and their wishes to be relieved from the heavy duties:

I enjoyed and I believe that most of us do really enjoy being with our children and doing the motherly things, but . . . it did get overwhelming.

African American mothers appeared to live life “preoccupied” with chil- dren. Their schedule at work and home, the kinds of foods they ate, the methods of preparing meals, and physical activities were all framed with regards to the children:

There’s some days when I do a schedule, and I’ll stick to it, but then there’s some days when she [the baby] just doesn’t feel like it. She wants to do a whole other thing.

I think I try to wake up a little bit earlier before she wakes up in the morning, cause like I have time by myself. . . . I try to wake up a little bit early, she wakes up about 15–20 min-minutes later . . . but, you know, just have some quiet minutes for myself . . . just enough time to get me a quick shower and then . . . little feet are running around the couch.

My eating habit changed after I had the baby, cause it, I-I really felt like I had no time to just eat, you know, I skip breakfast, lunch. Sometime I eat at 4:00, sometime I eat at 1:00, and don’t eat again. I eat once a day, sometimes twice, but it’s the kids, it’s like whooo you just don’t have time to eat, cause it’s, I don’t know, it’s always something. So I know my eating habit is not good, but, it don’t. . . . I can’t change it because it’s, . . . I have no control.

Particularly, the responsibilities with multiple children were viewed as much more stressful than parenting a single child:

Ethnic Context Affecting Weight Loss 47

Once you actually cook and get the kids there, and then one’s finished before the other, so you have to deal with that, and then it’s like, you know, you don’t actually find the time to sit down and eat, yourself, especially having the little one you have to feed . . . so it is kind of hard to get something for you.

I think it was just the stress of dealing with two children instead of one, you know, cause you got my son, but when you have a second kid [the baby], then you have to change the diaper, you got to make sure you have the bottles and the bibs and the this and the that, and then, I thought, it was just like adjusting to having another one.

African American mothers managed the contextual changes caused by de- mands of parenting in different ways. Some chose smoking, drinking, or both as one way of “escaping” from child-caring duties, being “relieved” from stressors, and having their own time. Although some mothers revealed mixed feelings because of the smoking-related diseases that other family members developed, many mothers agreed that smoking was a way of hav- ing their “quiet time” from their adverse circumstances:

Mine [smoking] started where I worked. Everybody else was taking a cigarette break, so, you know, I want to be like . . . (Laughter). I wanted my time to myself, and so now, I do it at home, but I don’t smoke in the house. I smoke outside, so that’s my time to myself, when I’m outside smoking a cigarette.

One of my friends smokes, she be like, “Go smoke a cigarette. It’d help.” I be like, “No.” But finally I did it and smoked a cigarette. And it relieved me from the kids, and I was like, “great feeling.”’

African American mothers also discussed an eagerness to have “adult con- versations” as another strategy to manage changes resulting from an altered sense of “perceived control” during postpartum. Specifically, participants who maintained their employment often considered the workplace as a good space for having adult conversations:

At least going to work and stuff you be able to, you know, laugh and talk.

That’s the main thing—talking to adults and not little kids. They can’t say anything.

Also, for some African American mothers, leaving the child to go to their job was a way of coping with child-rearing duties:

48 B. Sterling et al.

[Mother talking about her husband] . . . uh, I was like, “How about you stay home, and I’ll go to work?” And he . . . he works from 3:00 to 11:00. I was working from 9:00 to 1:00 part time, and so that was my break, and that helped me get a schedule, get the kids on a schedule, and get myself on a schedule, and become more organized, and so, and I got my break even though I was working.

The demands of child care as a significant contextual event resulting in altered sense of perceived control was unique to African American women in this study.

Hispanic Mothers’ Context

For Hispanic mothers, the context of an altered sense of “perceived control” was manifested by the theme of family. These women made decisions or pursued health-related practices within the context of family. One character- istic noted was that family served as either a source of support or a source of stress. Even though help from close family members may not have eased the demands of child care, some Hispanic participants said their mother or partner was somehow helpful by “just being there”:

But my mother, she was always there, you know, emotionally, to give me my answers, and things like that.

Because my husband, right now, you know, he’s helping me a lot. I leave home at 5:00, so I can go in the library and do whatever I got to do, and finish up my homework or whatever, and he’s there.

In contrast, some Hispanic women did not always experience feelings of support from family members. Participants discussed difficulties in coping with their mother, spouse, or partner who at times imposed heavier burdens on daily life. One participant discussed the need to care for her sister’s children in addition to her own. Although she complained about the sister’s attitude about not carrying for her own children, the participant continued to assume primary care for both sets of children:

Because she [the sister of this participant] was in the process of . . . getting out of her house, and her boyfriend kicked her out, and so I had to take her to my house, and she had two kids, and I was taking care of them 24-7 . . .. It was real stressful for me . . . Because I was worried about her kids, because I had my two kids, and I had hers, so I was always just after them, you know, feeding them, bath . . . and putting them to bed. I was always doing stuff to them, so I didn’t have time for me.

Ethnic Context Affecting Weight Loss 49

Hispanic women employed several strategies to manage contextual changes in perceived control that resulted from both nuclear and extended family re- sponsibilities. To the Hispanic mothers, the increased focus on meeting their family’s needs seemed natural and even mandatory. This belief, however, resulted in a sacrificing or disregard of private or personal needs and often a sense of feeling guilty when engaging in activities focused on meeting their own needs if this took time away from meeting the needs of their children or family. One mother felt guilt when she took her children to a child care service so she could do something for herself, such as exercise.

I feel bad, like I’m taking, like I’m going to stick them in a day care and go do something for me. That makes me feel bad, you know . . . feel guilty.

In addition, several Hispanic mothers did not go to the 6-week postpartum checkup or did not see a doctor for their personal health conditions to fully take care of their children:

But, truthfully, since I’ve had the baby, you know, after you have the baby you see you have to go for the 6-weeks checkup. I still haven’t even been to the doctor for myself since I’ve had the baby. I haven’t even been to the doctor once. And while my husband tells me, he’s like, “You’re so caught up into her, you don’t even care about yourself anymore.”

I just don’t make the time for it. . . . I do take my son to the doctor when he needs to go, but I don’t for me. It’s, if I’m sick, I can handle it, but him he’s different.

Conversely, one mother managed the stress caused by family care demands in the context of altered perceived control by establishing a schedule for her life that included time needed for child care. Developing a schedule for herself and her children resulted in a certain level of perceived control in her life.

I think it’s important for us women to have a time for ourselves, . . . because, I mean, I know, in my home, I am, I have rules for my children. I get them to bed at 7:30, and from that time, I’m going to have my bath, do whatever I need to do that night for the next day.

I send her to my sister’s, you know, go play . . . with your little cousin, and they’ll go off and play, you know, while I can have some free time or (at) least and breathe or even if it’s just to relax and watch TV, just to myself. . . . You know, that’s my job [caring for child], but still I deserve time for myself. If not, you’ll go crazy.

50 B. Sterling et al.

Summary of the Ethnic Specific Context

For Anglo participants, the context of altered control was the result of having lost the sense of self that they experienced before pregnancy. The prepreg- nancy perceptions of being thinner, having a better job, and having good relationships with friends and family members seemed to contribute to Anglo women perceiving an altered sense of self related to their weight, financial status, and relationships with significant others during the postpartum pe- riod. Contrary to their Anglo counterparts, African American women were not concerned about a lack of control over their past self-image. Rather, their context was related to a lack of control in response to child care demands. Managing contextual changes and stress by beginning or resuming smoking was a vivid indication that they did not have perceived control over their life. Different from their counterparts, Hispanic participants’ perception of control was situated within the context of the family. Hispanic mothers man- aged contextual changes during postpartum by attending to the needs of the family, which was unlike that of Anglo or African American mothers.

Summary of Managing Changes in Perceived Control

Anglo and Hispanic postpartum women seemed to manage changes in per- ceived control by reframing their perceptions. Anglo women reframed their life priorities, adapted to the altered body image, or sought positive rein- forcement for reporting healthy eating habits. Hispanic women reframed their priorities to coincide with their perceptions of the importance of meet- ing the needs of their family, which resulted in either a sense of control or, at times, guilt. When discussing the topics of smoking and drinking, we found that the purpose of smoking could be distinguishable for one ethnic group to the other. For African American mothers, smoking was an exit to their own time from the heavy requirements of child care and their life, whereas one Anglo American mother smoked as a strategy to manage or lose weight: “I’m smoking a lot more because I’m trying to lose weight.” Because the dis- cussion of smoking as a management strategy was not as prominent among Anglo American or Hispanic mothers as it was with their African American counterparts, however, the difference between these ethnic groups cannot be generalized.

DISCUSSION

Ethnic-Specific Themes

The transition to parenthood is a common experience that involves a dy- namic transformation when women’s perception of themselves solely as women change to include the perception of themselves as mothers in or- der to meet the needs of their infants (Mercer, 2004; Rubin, 1984). For the

Ethnic Context Affecting Weight Loss 51

low-income women in this study, who were selected because they were either overweight or at risk of elevated depressive symptoms (or both con- ditions), an overarching theme of altered perception of control was evident in their lives, a finding replicated from a prior analysis of this sample (Ster- ling et al., 2009). For Anglo, African American, and Hispanic women, the themes of altered sense of self, demands of parenting, and family, respec- tively, emerged as distinguishing features of their respective psychosocial and behavioral contexts during focus group interviews. Our findings related to Anglo women differ from those reported by Ugarriza, Brown, and Chang- Martinez (2007), who conducted qualitative interviews of Anglo women who had not experienced a self-identified postpartum depression. Among the pro- tective factors cited by researchers was a “special recognition” given to the mothers by spouses, family and community. Mothers in this study by Ugar- riza and colleagues commented on their new and enriched sense of purpose and of self. Most of these Anglo women, however, were older and more advantaged than those in our study. This contrast illustrates the potential moderating role that socioeconomic status may have within the psychoso- cial context of Anglo women.

The theme of energy and emotional demands involved with parenting found among African American mothers also was reported by Amankwaa (2003) in a qualitative study of African American women with diagnosed or self-reported postpartum depression. In that study, women reported that stressors related to caring for infants with colic and the care of other children in the family were among those that preceded a shift to a depressed mood during the postpartum period.

The pervasive context of family that was noted for Hispanic mothers is similar to others in that the extended and nuclear family is the primary source of support for foreign-born Hispanic mothers (Page, 2004). Hispanic mothers rely on their family members, particularly their own mothers, for emotional support and advice. The women in this study, however, were U.S.-born Hispanic mothers, who typically experience more stress and have a less healthy diet than their foreign-born Hispanic or Anglo counterparts (Page). The support that women in this study felt by just the presence of their family may indicate the psychological and relational tie some Hispanic women have to their family members.

POTENTIAL STRATEGIES FOR ETHNIC-SPECIFIC MATERNAL HEALTH PROMOTION PROGRAMS

The goal of this study was to describe ethnic differences in the psychoso- cial contexts in a group of low-income American new mothers to serve as a guide in developing ethnic-specific health promotion interventions, es- pecially related to reducing postpartum depressive symptoms and weight

52 B. Sterling et al.

retention from pregnancy. Customizing health promotion interventions that address ethnic-specific forces influencing low-income mothers’ context may improve program effectiveness. In a group of Anglo American women who are experiencing an altered sense of self, weight loss programs might in- clude management strategies that improve their self-image, such as getting dressed daily in properly fitting clothes and avoiding baggy pants and shirts. Strategies that promote introspection, such as journaling of feelings related to a sense of control also may be effective in reducing depressive mood (Gortner, Rude, & Pennebaker, 2006). Anglo mothers could use the journal as a “consciousness-raising” activity to record emotional difficulties they are experiencing, reflect on the meaning of these difficulties, and propose strate- gies to address them. Providing assertive training and establishing realistic goals for weight loss and processing perceptions of self-image also should be components of effective programs targeting Anglo women.

Health promotion programs for African American mothers who may be struggling with parenting demands may be more effective if held in a group setting, allowing mothers to talk with one another without interruption from children. African American mothers in our larger study were more likely not to be partnered than were Anglo and Hispanic mothers, thus locating health promotion programs along public transportation routes and providing on- site child care in a separate room would enhance the likelihood that these women would participate in programs. Discussion could focus on identifying and managing daily events that lead to negative emotions and unhealthy be- haviors, such as smoking, among African American women (Pletsch, 2006). Providing anticipatory guidance regarding realistic expectations for child be- havior and parenting skills are some stress management strategies that may be effective for these women. Facilitating the development of support groups and providing telephone follow-up calls also should be included in health promotion intervention programs targeting African American mothers.

The importance of inclusion of the immediate and extended family should frame health promotion programs targeting Hispanic mothers. Strate- gies for providing healthy meals and promoting physical activity for all fam- ily members could be discussed. Encouraging family members, particularly Hispanic women’s mothers, to attend the program meetings may decrease the hesitancy that Hispanic women feel when attending to their individual health needs and provides an opportunity to educate the family on healthy eating practices (Thornton et al., 2006). While meeting the needs of family members is an inherent practice for many Hispanic mothers, some may expe- rience increased stress when trying to meet family needs as well as their own health needs. Health promotion interventions that include time management strategies may be more effective. For example, Hispanic mothers could be encouraged to establish daily schedules that included time for attending to family and child care needs, as well as time for personal health-promoting activities.

Ethnic Context Affecting Weight Loss 53

Limitations

Participant selection was limited to those women who had completed all phases of a larger research study in the United States on thriving during postpartum and who had either retained weight or had experienced ele- vated depressive symptoms or both at 12 months postpartum. Thus, these women may not have been characteristic of low-income U.S.-born Anglo, African American, and Hispanic new mothers. Because of the eligibility cri- teria regarding BMI and scores on the CES-D, the ethnic-specific themes identified within this sample of low-income women may not be found in women with normal weight, who were not experiencing depressive symp- toms, or in an otherwise dissimilar sample. Furthermore, data collection was limited to participant responses and facilitators’ field notes obtained at the time of the focus group. No attempt was made to contact the participants after the completion of the focus groups; therefore, no further analysis to achieve data saturation was conducted.

Women’s Health Perspectives

New mothers may experience emotional and psychological stressors throughout the first year postpartum that alter their physical well-being and functioning and, thus, alter their sense of personal control. In our study the overarching theme of altered “perceived control” was manifested differently within ethnic groups and illustrates several implications. First among these is the varying burden that women carry related to child care and parenting. Parenting is more likely to be viewed as burdensome when women with low resources have pregnancies that are unintended (Ispa, Sable, Porter, & Csizmadia, 2007).

Second, this study supports the importance of exploring ethnic-specific themes in the formative stages of program development to increase the rele- vance of such programming. Although it is unclear to what extent the ethnic- specific themes found in this study are applicable in other countries and settings, the findings still serve to alert program planners to the importance of ethnic-specific considerations. In summary, the obesity epidemic (WHO, 2000) and concerns about mental health, especially in low-income countries (WHO, 2008), support the importance of programs to reduce occurrence of these two health conditions. The postpartum period is a transition interval during which women may be at risk of elevated depressive symptoms and weight gain. Consequently, study of ethnic differences in women’s postpar- tum psychosocial contexts that may affect the success of health promotion programs to reduce weight retention and postpartum depression is needed. Addressing the ethnic-specific forces that frame the context of these women’s lives during the postpartum period is critical for enhancing the effectiveness of interventions for low-income and ethnically diverse mothers.

54 B. Sterling et al.

REFERENCES

Amankwaa, L. C. (2003). Postpartum depression among African-American women. Issues in Mental Health Nursing, 24, 297–316. doi: 10.1080/01612840390160801

Beck, C. T. (1998). The effects of postpartum depression on child development: A meta-analysis. Archives of Psychiatric Nursing, 12, 12–20. doi:10.1016/S0883- 9417(98)80004-6

Castles, S. (2000). International migration at the beginning of the twenty-first century: Global trends and issues. International Social Science Journal, 52, 269–181. doi: 10.1111/1468-2451.00258

Cedergren, M. I. (2004). Maternal morbid obesity and the risk of adverse pregnancy outcome. Obstetrics & Gynecology, 103, 219–224. doi: 10.1097/01.AOG.0000 107291.46159.00

Cedergren, M. I., & Kallen, B.A.J. (2003). Maternal obesity and infant heart defects. Obesity Research, 11, 1065–1071. doi: 10.1038/oby.2003.146

Chang, M.-W., Nitzke, S., Guilford, E., Adair, C. H., & Hazard, D. L. (2008). Moti- vators and barriers to healthful eating and physical activity among low-income overweight and obese mothers. Journal of the American Dietetic Association, 108, 1023–1028. doi: 10.1016/j.jada.2008.03.004

Cheng, C. Y., Fowles, E., & Walker, L. (2006). Postpartum maternal health care in the United States: A critical review. Journal of Perinatal Education, 15(3), 34–42. doi: 10.1624/105812406×119002

Civic, D., & Holt, V. L. (2000). Maternal depressive symptoms and child behavior problems in a nationally representative normal birthweight sample. Maternal and Child Health Journal, 4, 215–221. doi: 10.1092-7875/00/1200-0125

Da Costa, D., Dritsa, M., Rippen, N., Lownesteyn, I., & Khalife, S. (2006). Health- related quality of life in postpartum depressed women. Archives of Women’s Mental Health, 9, 95–102. doi: 10.1007/s00737-005-0108-6

Ebbeling, C. B., Pearson, M. N., Sorensen, G., Levine, R. A., Hebert, J. R., Salkeld, J. A., & Peterson, K. E. (2007). Conceptualization and development of a theory-based healthful eating and physical activity intervention for postpar- tum women who are low income. Health Promotion Practice, 8, 50–59. doi: 10.1177/1524839905278930

Field, A. E., Coakley, E. H., Must, A., Spadano, J. L., Laird, N., Dietz, W. H., . . . Colditz, G. A. (2001). Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Archives of Internal Medicine, 161, 1581–1586. Retrieved from http://www.archinternmed.com

Gortner, E. M., Rude, S. S., & Pennebaker, J. M. (2006). Benefits of expressive writing in lowering rumination and depressive symptoms. Behavior Therapy, 37, 292–303.

Halbreich, U., & Karkun, S. (2006). Cross-cultural and social diversity of prevalence of postpartum depression and depressive symptoms. Journal of Affective Disorders, 91, 97–111. doi: 10.1016/J.JAD.205.12.051

Howell, E. A., Mora, P. A., DiBonaventura, M. D., & Leventhal, H. (2009). Modifiable factors associated with changes in postpartum depressive symptoms. Archives of Women’s Mental Health, 12, 113–120. doi: 10.1007/s00737-009-0056-7

Ethnic Context Affecting Weight Loss 55

Ispa, J. M., Sable, M. R., Porter, N., & Csizmadia, A. (2007). Pregnancy acceptance, parenting stress, and toddler attachment in low-income Black families. Journal of Marriage and the Family, 69, 1–13. doi: 10.1111/j.1741-3737.2005.00174.x-i1

Kieffer, E. C., Willis, S. K., Arellano, N., & Guzman, R. (2002). Perspectives of preg- nant and postpartum Latino women on diabetes, physical activity, and health. Health Education and Behavior, 29, 542–556. doi: 10/1177/10901802237023

Krueger, R. A., & Casey, M. A. (2000). Focus groups: A practical guide for applied research (3rd ed.). Thousand Oaks, CA: Sage.

Linne, Y., Dye, L., Barkeling, B., & Rossner, S. (2004). Long-term weight development in women: A 15-year follow-up of the effects of pregnancy. Obesity Research, 12, 1166–1178. doi:10.1038/oby.2004.146

Mercer, R. (1995). Becoming a mother: Research on maternal identity from Rubin to the present. New York, NY: Springer.

Mercer, R. (2004). Becoming a mother versus maternal role attainment. Journal of Nursing Scholarship, 36, 226–232. doi: 10.1111/j.1547-5069.2004.04042.x

Miles, M. B., & Huberman, A. M. (1994). Qualitative data analysis (2nd ed.). Thou- sand Oaks, CA: Sage Publications, Inc.

Must, A., Spadano, J., Coakley, E. H., Field, A. E., Colditz, G., & Dietz, W. H. (1999). The disease burden associated with overweight and obesity. Journal of the American Medical Association, 282, 1523–1529. doi: 10.1001/jama.282.16.1523

Oken, E., Taveras, E. M., Popoola, F. A., Rich-Edwards, J. W., & Gillman, M. W. (2007). Television, walking and diet: Associations with postpartum weight retention. American Journal of Preventive Medicine, 32, 305–311. doi: 10.1016/j.amepre.2006.11.012

Olson, C. M., Strawderman, M. S., Hinton, P. S., & Pearson, T. A. (2003). Gestational weight gain and postpartum behaviors associated with weight change from early pregnancy to 1 y postpartum. International Journal of Obesity, 27, 117–127. doi: 10.1038/sj.ijo.0802156

Page, R. (2004). Positive pregnancy outcomes in Mexican immigrants: What can we learn? Journal of Obstetric, Gynecologic, and Neonatal Nursing, 33, 783–790. doi: 10.1177/0884217504270595

Parker, J. D., & Abrams, B. (1993). Differences in postpartum weight retention be- tween black and white mothers. Obstetrics and Gynecology, 81, 768–774.

Petterson, S. M., & Albers, A. B. (2001). Effects of poverty and maternal depression on early child development. Child Development, 72, 1794–1813. doi: 10.1111/1467- 8624.00379

Pletsch, P. K. (2006). A model for postpartum smoking resumption prevention for women who stop smoking while pregnant. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 35, 215–222. doi: 10.1111/J.1552-6909.2006.00036.x

Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401.

Rooney, B. L., Schauberger, C. W., & Mathiason, M. A. (2005). Impact of perinatal weight change on long-term obesity and obesity-related illnesses. Obstetrics & Gynecology, 106, 1349–1356. doi: 10.1097/01.AOG.0000185480.09068.4a

Rubin, R. (1984). Maternal identity and the maternal experience. New York, NY: Springer.

Shrewsbury, V., Robb, K., Power, C., & Wardle, J. (2009). Socioeconomic differ- ences in weight retention, weight-related attitudes and practices in postpartum

56 B. Sterling et al.

women. Maternal and Child Health Journal, 13, 231–240. doi: 10.1007/s10995- 008-0342-4

Sterling, B. S., Fowles, E. R., Garcia, A. A., Jenkins, S. K., Wilkinson, S., Kim, M.. . ., Walker, L. O. (2009). Altered perceptions of personal control about retained weight and depressive symptoms in low-income postpartum women. Journal of Community Health Nursing, 26, 143–157. doi: 10.1080/07370010903034524

Thornton, P. L., Kieffer, E. C., Salabarria-Pena, Y., Odoms-Young, A., Willis, S. K., Kim, H., & Salinas, M. A. (2006). Weight, diet, and physical activity-related beliefs and practices among pregnancy and postpartum Latino women: The role of social support. Maternal and Child Health Journal, 10, 95–104. doi: 10.1007/s10995-005-0025-3

Ugarriza, D. N., Brown, S.E.D., & Chang-Martinez, C. (2007). Anglo American moth- ers and the prevention of postpartum depression. Issues in Mental Health Nurs- ing, 28, 781–798. doi: 10.1080/01612840701413624

Viswanathan, M., Siega-Riz, A. M., Moos, M. K., Deierlein, A., Mumford, S., Knaack, J., . . . Lohr, K. N. (2008). Outcomes of maternal weight gain. Evi- dence Report/Technology Assessment, 168, 1–223. Retrieved from http://www. ahrq.gov/downloads/pub/evidence/pdf/admaternal/admaternal.pdf

Walker, L. O., Freeland-Graves, J. H., Milani, T., Hanss-Nuss, H., George, G., Sterling, B. S., . . . Stuifbergen, A. (2004). Weight and behavioral and psychosocial factors among ethnically diverse, low-income women after childbirth. I. Methods and context. Women & Health, 40(2), 1–17. doi: 10.1300/J013v40n02 01

Walker, L. O., Timmerman, G. M., Sterling, B. S., Kim, M., & Dickson, P. (2004). Do low-income women attain their pre-pregnant weight by the 6th week of postpartum? Ethnicity & Disease, 14, 119–126.

Walker, L. O., Timmerman, G. T., Kim, M., & Sterling, B. (2002). Relationships between body image and depressive symptoms during postpartum in ethnically diverse, low income women. Women & Health, 36(3), 101–121. Retrieved from http://www.haworthpressinc.com/store/product.asp?sku=J013

Walker, L. O., & Wilging, S. (2000). Rediscovering the “M” in “MCH”: Maternal health promotion after childbirth. Journal of Obstetric, Gynecologic, and Neona- tal Nursing, 29, 229–236. doi: 10.1111/j.1552-6909.2000.tb02044.x

Wolfe, W. S., Sobal, J., Olson, C. M., Frongillo, E. A., & Williamson, D. F. (1997). Parity-associated weight gain and its modification by sociodemographic and behavioral factors: A prospective analysis of US women. International Journal of Obesity, 21, 802–810.

World Health Organization (WHO). (2000). Obesity: Preventing and managing the global epidemic. Geneva: Author. Retrieved from http://www.who.int/nutrition/ publications/obesity/WHO TRS 894/en/index.html

World Health Organization (WHO). (2008). mhGAP: Mental Health Gap Action Programme: Scaling up care for mental, neurological and substance use dis- orders. Geneva: Author. Retrieved from http://www.who.int/mental health/ mhgap final english.pdf

World Health Organization (WHO). (2009a). Gender and women’s mental health. Retrieved from http://www.who.int/mental health/prevention/genderwomen/ en/

World Health Organization (WHO). (2009b). Statistical information system. Retrieved from http://apps.who.int/whosis/data/Search.jsp?countries=%5bLocation%5d. Members

Copyright of Health Care for Women International is the property of Routledge and its content may not be

copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.