MATERNAL SOCIOECONOMIC MOBILITY:
RELATIONS WITH PRETERM DELIVERY AND PRENATAL DEPRESSION
By
Yan Tian

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Epidemiology—Doctor of Philosophy
2017

ABSTRACT
MATERNAL SOCIOECONOMIC MOBILITY:
RELATIONS WITH PRETERM DELIVERY AND PRENATAL DEPRESSION
By
Yan Tian
Preterm delivery (PTD) and prenatal depression are major public health problems.
Previous studies suggest that socioeconomic position (SEP) has been negatively associated with
the risk of PTD and mental health problems. However, the association between maternal
socioeconomic mobility (SM) and the risk of PTD or prenatal depression is rarely investigated.
We hypothesized that women with low childhood SEP and low adulthood SEP will be at greatest
risk of PTD or prenatal depression; upward SM from childhood to adulthood will lower women’s
risk of PTD or prenatal depression; and downward SM from childhood to adulthood will increase
risk of PTD or prenatal depression. We aimed (1) to evaluate the relationship between maternal
SM and risk of PTD; and (2) to assess the association between maternal SM and risk of prenatal
depression in a Michigan, community-based cohort study that is racially and socioeconomically
diverse. Because other studies reported that the relation of SEP (or SM) with risk of PTD or
prenatal depression varied by maternal race/ethnicity, we also examine evidence for this effect
modification.
Data were collected from 3,019 pregnant women recruited from 5 Michigan communities
at mid-trimester (16-27 weeks’ gestation) in the Pregnancy Outcomes and Community Health
(POUCH) Study (1998-2004). Pregnant women provided their parents’ SEP indicators
(education, occupation, receipt of public assistance) and their own and child’s father’s SEP
indicators (education, occupation, Medicaid status, and household income) at enrollment. PTD

was defined as deliveries less than 37 completed weeks’ gestation. Prenatal depressive symptoms
were assessed by the Center for Epidemiologic Studies Depression Scale (CES-D) and CESD≥16 was defined as high CES-D. Latent class analysis was conducted to identify latent groups
of childhood SEP indicators, adulthood SEP indicators, and SM from childhood to adulthood,
respectively. A model-based approach to latent class analysis with distal outcome assessed
relations between latent class and PTD or prenatal depression, overall and within race/ethnicity
groups.
Three latent groups (low, middle, high) were identified for childhood SEP indicators and
adulthood SEP indicators, respectively; while four latent groups (static low, upward, downward,
and static high) best described SM. Women with upward SM had decreased odds of PTD (odds
ratio (OR)=0.60, 95% confidence interval (CI): 0.42, 0.87), compared to those with static low
SEP. Women with upward SM had decreased odds of high CES-D (OR=0.22, 95% CI: 0.17,
0.29), compared to those with static low SEP. Women in downward group had elevated odds of
high CES-D (OR=4.13, 95% CI: 2.75, 6.22), compared to those in static high group. This SM
advantage was true for white/other women for PTD and most pronounced in white/other women
for high CES-D.
Maternal experiences of upward SM or downward SM may be important considerations
when assessing risk of PTD or prenatal depression. Our results support the argument that policies
and programs aimed at improving women’s SEP could lower PTD or prenatal depression risk.
Our study suggests that policies or programs to increase opportunities for upward mobility may
play an important role in decreasing the risk of PTD or prenatal depression. Further research is
needed to understand the specific elements accompanying SM that are protective for PTD or
prenatal depression and the best strategies for increasing SM among all race/ethnic groups.

This dissertation is dedicated to Emily and Grace,
with love.

iv

ACKNOWLEDGEMENTS

I would like to acknowledge many people for help with my doctoral education. I am most
grateful to my dissertation advisor, Dr. Claudia Holzman, for her mentorship, constant guidance
and support throughout the time of my dissertation research. I am thankful for my committee
members’ continuous guidance and support. Dr. Slaughter and Dr. Margerison assisted me in
deep understanding of relative aspects of social epidemiology. Dr. Luo and Dr. Todem gave their
time and wisdom to help me with statistical methods. All the committee members provided me
challenging ideas, careful reading, inspiring suggestions and critical comments. Without their
ongoing support, I could not have finished my research.
Furthermore, I would like to express my gratitude to many faculties for their helpful
teaching and guidance in the Department of Epidemiology and Biostatistics. In addition, it has
been a great opportunity to work on the Pregnancy Outcomes and Community Health (POUCH)
Study. I am grateful for Dr. Bertha Bullen and Crista Valentine’s direct assistance and constant
encouragement.
I am extremely grateful to my family and friends for constant support and unfailing
encouragement. I especially thank my husband, Xiaodong, for sharing the housework, taking
care of kids, and providing plentiful support. I am also grateful to my daughters, Emily and
Grace, for their encouragement. This would not have been possible without them by my side.

v

TABLE OF CONTENTS

LIST OF TABLES…………………………………………………………………………….…vii
KEY TO ABBREVIATIONS …………………………………………………………….……..ix
CHAPTER 1
BACKGROUND LITERATURE.…………………………………………………………….… 1
1.1 Preterm delivery ……………………………...………………………………..……..….….. 1
1.2 Maternal adulthood socioeconomic position and preterm delivery ………...………….…… 3
1.3 Maternal childhood socioeconomic position and preterm delivery ………...……..….…..… 4
1.4 Maternal socioeconomic mobility and preterm delivery ………...….…...……..………....… 5
1.5 Prenatal depression …………………………………………...……………….………..…....13
1.6 Adulthood socioeconomic position and depression …………………………………………14
1.7 Childhood socioeconomic position and depression …….. ……………………………....….16
1.8 Socioeconomic mobility and depression ……………………………………...….………….18
1.9 Gap in the literature …....…………………………………………………...………………..31
1.10 Aims …………….…………………………………...………………………………..…....33
CHAPTER 2
MATERNAL SOCIOECONOMIC MOBILITY AND PRETERM DELIVERY:
A LATENT CLASS ANALYSIS ……………………………..……………………………..….35
2.1 Introduction ………………………………………..……………………………………..….35
2.2 Methods ……………………………………………………..…………………………….…37
2.3 Results ……………………………………………………………..……………………..….41
2.4 Discussion ………………………………………………………………..……………..…...45
CHAPTER 3
MATERNAL SOCIOECONOMIC MOBILITY AND PRENATAL DEPRESSION:
A LATENT CLASS ANALYSIS …………..………………………………………………..….62
3.1 Introduction ……………………………………………………..……………………….…..62
3.2 Methods …………………………………………………………..……………………..…...64
3.3 Results ……………………………………………………………..…………………….…..68
3.4 Discussion …………………………………………………………..…………………….....72
CHAPTER 4
SUMMARY ………………………………………………………………………………..……89
4.1 Major findings ……………………………………………………………………….…....…89
4.2 Public health significance………………………………………..…………………………...90
BIBLIOGRAPHY ………………………………………………………………………….……94

vi

LIST OF TABLES

Table 1.1. Studies reporting on associations between maternal adulthood socioeconomic position
and risk of preterm delivery……………………………………………………………………… 7
Table 1.2. Studies reporting on associations between maternal socioeconomic mobility and risk
of preterm delivery……………………………………………………………………………….10
Table 1.3. Studies reporting on associations between adulthood socioeconomic position and risk
of depression……………………………………………………………………………………. 20
Table 1.4. Studies reporting on associations between childhood socioeconomic position and risk
of depression……………………………………………………………………………………..24
Table 1.5. Studies reporting on associations between socioeconomic mobility and risk of
depression………………………………………………………………………………………..27
Table 2.1. Maternal Characteristics and Socioeconomic Position Indicators (n=3,019), Pregnancy
Outcomes and Community Health Study, Michigan, 1998-2004………………………………..49
Table 2.2. Fit statistics for latent class analysis (n=3,019), Pregnancy Outcomes and Community
Health Study, Michigan, 1998-2004……………………………………………………………..51
Table 2.3. Estimated item-response probabilities of the socioeconomic position indicators
conditional on latent class membership with distal outcome preterm delivery overall and by
race/ethnicity (n=3,019), Pregnancy Outcomes and Community Health Study, Michigan, 19982004………………………………………………………………………………………………52
Table 2.4. Odds ratios of preterm delivery conditional on latent class membership (n=3,019),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004……………………55
Table 2.5. Fit statistics for latent class analysis (n=3,019), Pregnancy Outcomes and Community
Health Study, Michigan, 1998-2004………………………………………………………….….56
Table 2.6. Estimated item-response probabilities of the socioeconomic position indicators
conditional on latent class membership with distal outcome preterm delivery overall and by
race/ethnicity (n=3,019), Pregnancy Outcomes and Community Health Study, Michigan, 19982004……………………………………………………………………………………………....57
Table 2.7. Odds ratios of preterm delivery conditional on latent class membership (n=3,019),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004………………..…..61
Table 3.1. Maternal Characteristics and Socioeconomic Position Indicators (n=3,010), Pregnancy
Outcomes and Community Health Study, Michigan, 1998-2004………………………………..76
vii

Table 3.2. Fit statistics for latent class analysis (n=3,010), Pregnancy Outcomes and Community
Health Study, Michigan, 1998-2004……………………………………………………………..78
Table 3.3. Estimated item-response probabilities of the socioeconomic position indicators
conditional on latent class membership with distal outcome prenatal depression overall and by
race/ethnicity (n=3,010), Pregnancy Outcomes and Community Health Study, Michigan, 19982004………………………………………………………………………………………………79
Table 3.4. Odds ratios of prenatal depression conditional on latent class membership (n=3,010),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004……………………82
Table 3.5. Fit statistics for latent class analysis (n=3,010), Pregnancy Outcomes and Community
Health Study, Michigan, 1998-2004……………………………………………………………..83
Table 3.6. Estimated item-response probabilities of the socioeconomic position indicators
conditional on latent class membership with distal outcome prenatal depression overall and by
race/ethnicity (n=3,010), Pregnancy Outcomes and Community Health Study, Michigan, 19982004……………………………………………………………………………………………....84
Table 3.7. Odds ratios of prenatal depression conditional on latent class membership (n=3,010),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004……………………88

viii

KEY TO ABBREVIATIONS

CES-D

Center for Epidemiologic Studies Depression Scale

CI

Confidence Interval

OR

Odds Ratio

POUCH

Pregnancy Outcomes and Community Health Study

PTD

Preterm Delivery

RR

Rate Ratio

SEP

Socioeconomic Position

SM

Socioeconomic Mobility

UK

United Kingdom

USA

United States of America

ix

CHAPTER 1
BACKGROUND LITERATURE

1.1 Preterm delivery
Preterm delivery (PTD), defined as delivery of a baby before 37 completed weeks of
gestation age, is a major public health problem. The worldwide rates of PTD are generally 5-18%
and about 15 million babies are born preterm annually. According to data from Centers for
Disease Control and Prevention, PTD occurred in about one of every ten deliveries in the United
States in 2014 (Hamilton, et al. 2015).
PTD is a leading cause of perinatal mortality and morbidity, accounting for 70% of
neonatal death and 75% of neonatal morbidity (Blencowe et al. 2013). The mortality rate for
preterm babies is 15 times greater than that of term babies (Mathews et al. 2004). In addition,
PTD may cause infants and children to have morbidity or severe health problems, e.g. breathing
problems, jaundice, respiratory distress syndrome, as well as such developmental difficulties as
blindness, sensory, learning, cognitive and language deficits (Allen 2008). Some evidence
demonstrated that the very PTD (28 to <32 weeks) children at school ages had more inattention,
hyperactivity, and peer problems, compared with term-born children (Delobel-Ayoub et al. 2009).
Furthermore, children born preterm may have an increased risk of cardiovascular disease in adult
life (Barker et al. 1993).
PTD is an increasingly heterogeneous condition and has substantial impacts (Allen 2008).
From the respect of economy, PTD has a most important direct and indirect cost. The PTD

1

babies often need extra hospitalization and additional care, which increases the general health
expenses. The 2007 statistics shows the annual cost for PTDs is on average 10-fold that of term
births. The costs increase exponentially as the gestational age declines (Gilbert et al. 2003). So
the direct cost may be seen from intensive and extended newborn care, as well as financial,
emotional, and psychological burden to the parents. The indirect cost to the society can be
caused from long-term care of disabled premature children and the care of the premature children
by the parents who may give up their jobs (Behrman & Butler, 2007). Therefore, prevention of
PTD can help save the major infant and pediatric costs.
The PTD rate varies by maternal race/ethnicity. From 2008 to 2010, in the United States,
PTD rates were highest for non-Hispanic African-American babies (17.4%). PTD rates for
Native Americans, Hispanics, non-Hispanic whites, and Asians were 13.7%, 12.0%, 10.9%, and
10.6%, respectively. In 2012, the PTD rate remains the highest among African-American infants
at 16.5%. The difference between African Americans and whites has been slowly narrowing, but
the PTD rate among African Americans is still more than 1.5 times the rate of whites. During
2011-2013, 16.5% of African-American women and 10.3% of non-Hispanic white women in the
U.S. delivered prematurely. The disparities in PTD rates between population subgroups defined
by socioeconomic status are striking (Blumenshine et al. 2010). Many social factors greatly
affect African Americans in terms of the PTD rates. For example, poor education, unfavorable
neighborhood or low economic status increases low self-esteem, poor nutrition and sanitation,
and even level of stress (O’Campo et al. 2008). In addition, a body of previous studies suggested
lifetime adverse socioeconomic experiences and racism-related stress should also be considered
to elucidate the racial disparity in PTD (Lu & Halfon 2003, Domingues 2008, Domingues et al.
2008, Kramer & Hogue 2009, Braveman 2011).
2

There are many risk factors for PTD. Previous PTD, multifetal pregnancy, and cervical
incompetence have been repeatedly indicated as risk factors for PTD. Being an African
American, at low socio-economic status, and getting pregnant at extreme maternal age also
increase the risk of PTD. Some other behavioral, psychosocial, and nutrition factors have also
been implicated as risk factors of PTD, such as maternal smoking during pregnancy, maternal
cocaine use, maternal psychosocial stress, anxiety, low pre-pregnancy BMI, and poor weight
gain during pregnancy. Additionally, previous literature provided some evidence about potential
mechanisms linking SEP and PTD, e.g. behaviors, immune function, stress axis hormones,
hypertensive disorders, sexually transmitted disease, etc. (Kramer et al. 2000).

1.2 Maternal adulthood socioeconomic position and preterm delivery
Maternal disadvantaged socioeconomic position (SEP) in adulthood has been associated
with a higher risk of PTD. Measures of SEP have included occupation, education and household
income, and the relationship with PTD often varied by race/ethnicity (Morgen et al. 2008, Auger,
Giraud, and Daniel 2009, Blumenshine et al. 2010, Blumenshine et al. 2011, Mortensen,
Helweg-Larsen, and Andersen 2011, Matijasevich et al. 2012, Whitehead 2012, Joseph et al.
2014, Shankardass et al. 2014).
Table 1.1 shows studies reporting on associations between maternal adulthood SEP and
risk of PTD. The Danish National Birth Cohort study (Morgen et al. 2008) reported that the risk
of PTD increased among women with low SEP measured by maternal education level. In a
systematic review of 106 studies from 1999 to 2007, Blumenshine et al. (2010) suggested that
disadvantaged SEP was consistently linked to greater risk of adverse birth outcomes and
3

observed racial/ethnic heterogeneity in the effect of SEP measures. The inverse associations
between maternal education and adverse birth outcomes were significant among non-Hispanic
white pregnant women, but not among African-American women. In a California study,
Blumenshine et al. (2011) found that infants whose father had not completed college were at
significantly elevated risk of being born preterm; the association with paternal education was
particularly strong among unmarried women. A study of UK and Brazil noted higher rates of
PTD among the poorest and least educated women (Matijasevich et al. 2012). Whitehead et al.
(2012) examined 37 U.S. states data from the Pregnancy Risk Assessment Monitoring System,
and reported that the risk of PTD was greatest among women who had not finished high school,
whose household income was in the lowest 5%, and who were enrolled in Medicaid Insurance.
A Nova Scotia study by Joseph et al. (2014) observed a decreased risk of spontaneous PTD
among women with high family income compared to those with low family income. After
adjusting for other maternal characteristics the relationship between family income and
spontaneous PTD became attenuated. Shankardass et al. (2014) found that the risk of
spontaneous PTD was elevated across all the indicators of lower SEP, and women in the adverse
neighborhood had greater risk of medically indicated and spontaneous PTD.

1.3 Maternal childhood socioeconomic position and preterm delivery
Fewer studies have examined mother’s childhood SEP and pregnancy outcomes, though
there is evidence that SEP in childhood has a significant effect on adult health (Cohen et al.
2010). A Brazilian intergenerational, population-based cohort study by Gigante et al. (2015)
observed that maternal childhood poverty, measured by family income, was strongly associated
with shorter gestations.
4

Three conceptual models have been proposed to explain how disadvantaged life course
SEP affects health later in life (Kuh et al. 2003, Kendzor, Caughy, and Owen 2012). The critical
period model postulates that SEP-related factors have their greatest effect on adult health during
specific developmental period. The accumulation model supposes that increasing intensity and
duration of exposure to socioeconomic disadvantage may elevate the risk of poor adult health.
Individuals experiencing adverse SEP during both childhood and adulthood have higher risk for
poor adult health compared to those experiencing that during only one life stage. The
accumulation model focuses on the effect of gradually accumulated risk on later health problem.
The social mobility model hypothesizes that upward socioeconomic mobility (SM) remediates
the negative effect of earlier disadvantage SEP, however, downward SM may result in poorer
adult health despite relative socioeconomic advantages at earlier stages (Kuh et al. 2003, Cohen
et al. 2010, Kendzor, Caughy, and Owen 2012). These three models are theoretically different,
but closely correlated, which makes it hard to practically distinguish differential effects.
However, all of three models work together to provide a better understanding of how life course
SEP can be related to later life (Hallqvist et al. 2004). In all these models, determining the
temporal order of SEP and health can be challenging.

1.4 Maternal socioeconomic mobility and preterm delivery
SM was defined as “situations where an individual’s increase or decrease in SEP depends
on parental SEP compared with the level of SEP as an adult” (Owens and Jackson 2015). SM has
been studied in relation to low birthweight (Basso et al. 1997, Spencer 2004, Colen et al. 2006,
Astone, Misra, and Lynch 2007, Love et al. 2010, Osypuk et al. 2016), small for gestational age
(Love et al. 2010, Osypuk et al. 2016, Slaughter-Acey et al. 2016), and newborn body
5

composition (Sletner et al. 2014). In our review of the literature, however, we found fewer
studies examining SM and PTD (Collins et al. 2007, Love et al. 2010, Collins, Rankin, and
David 2011, Kramer, Dunlop, and Hogue 2014, Collins, Rankin, and David 2015, Osypuk et al.
2016).
Recent studies on associations between maternal SM and risk of PTD are shown in Table
1.2. A Detroit study by Osypuk et al. (2016) observed no association between maternal SM and
overall PTD, though improved SM was related to lower risk of spontaneous PTD among women
who experienced upward SM or had any stable SEP. Using the Illinois transgenerational birth
file with US Census income data, Collins et al. (2011) found that upward SM from an
impoverished childhood was related to a decreased risk of PTD among African Americans.
Women who experienced low, modest, or high upward SM had lower PTD rates of 16.0%,
15.2%, and 12.4%, respectively, than those with lifelong lower class status; rate ratio (RR) and
95% confidence interval (CI) = 0.9 (0.8, 0.9), 0.8 (0.7, 0.9), 0.7 (0.6, 0.8), respectively. In
contrast, downward SM was linked to an elevated risk of PTD among white urban women
(Collins et al. 2015). Those who experienced slight, moderate, or extreme downward SM had a
higher risk of PTD rates than women with lifelong upper class status, with RR and 95% CI of 1.2
(1.0, 4.0), 1.6 (1.3, 1.9), and 1.9 (1.3, 2.6), respectively. Using the same data, Collins et al. (2007)
revealed a twofold greater risk of PTD among African-American women than among whites
(RR=2.2, 95% CI: 1.7, 2.9). All of these three studies used quartiles of neighborhood census tract
median family income at two time points of the mother’s residence, when she was born and
when she delivered her baby.

6

Table 1.1. Studies reporting on associations between maternal adulthood socioeconomic position and risk of preterm delivery
First Author
(Year)
Joseph
(2014)

Sample

Canada
Nova Scotia
(N=132,714)

Shankardass
(2014)

Dependent
Variable

Statistical
Analyses

Results


PTD:
spontaneous
family income and
iatrogenic
PTD.

Logistic
regression

PTD:
spontaneous
and
iatrogenic
PTD.

Multiple
logistic
regression



PTD, IUGR,

Logistic
regression



SEP:

Canada

SEP:

Nova Scotia

family
income,
neighborhood
material
deprivation
index

(N=117,734)

Matijasevich
(2012)

Explanatory
Variable

UK, Brazil

SEP:

(N=5,168,
N=4,147)

education,
breastfeeding
family income for <3
months







7

Women with lowest family income had a
significantly increased risk of spontaneous
PTD (RR=1.14, 95% CI 1.03, 1.25), but not
iatrogenic PTD (RR=0.95, 95% CI 0.75, 1.19).
The relationship between family income and
spontaneous PTD became attenuated in the
adjusted model.
The risk of spontaneous PTD was elevated
across all the indicators of adverse SEP.
Higher neighborhood material deprivation was
related to greater risk for iatrogenic and
spontaneous PTD.

Higher prevalence of PTD was observed
among the poorest and less educated.
In the Pelotas cohorts, greater educationrelated inequalities in PTD were found.

Table 1.1. (cont’d)
First Author
(Year)
Whitehead
(2012)

Blumenshine
(2011)

Sample

USA

SEP:

37 states
(N=107,926)

income,
education,
insurance

USA

SEP:

California

father’s
education

(N=21,712)

Mortensen
(2011)

Explanatory
Variable

Dependent
Variable

Statistical
Analyses

Results

Preterm
contractions
and PTD

Logistic
regression



The risk of PTD was greatest among women
who had not finished high school, whose
income was in the lowest 5%, and who were
enrolled in Medicaid Insurance.

PTD

Logistic
regression



Infants whose father had not completed college
were at significantly elevated risk of being
born preterm (adjusted OR=1.26, 95% CI 1.01,
1.58).
The association was particularly strong among
unmarried women.



Denmark

SEP:

Danish
register data

education,
income

Stillbirth,
PTD,
birthweight,
infant
mortality

Review

8



The risk of PTD was reversely associated with
mother’s education attainment.

Table 1.1. (cont’d)
First Author
(Year)

Sample

Blumenshine
(2010)

106 studies
from 1999 to
2007

SEP:

Auger
(2009)

Canada

SEP:

Québec
(N =
353,120)

household
income,
immigrant
density,
income
inequality

Denmark

SEP:

(N=75,890)

education,
occupation,
household
income.

Morgen
(2008)

Explanatory
Variable

education,
income,
occupational
class

Dependent
Variable

Statistical
Analyses

Results

Birth weight;
PTD; low
birth weight;
small for
gestational
age

Systematic
review



The inverse associations between maternal
education and adverse outcomes were
significant among non-Hispanic white
pregnant women.

PTD and
small-forgestational
age

Multi-level
logistic
regression



Among foreign-born mothers, low immigrant
density was associated with decreased risk of
PTD (OR=0.79, 95% CI 0.63, 1.00), compared
to the high immigrant density.
Among Canadian-born mothers, low
immigrant density was associated with an
increased risk of PTD (OR=1.14, 95% CI 1.07,
1.21).



PTD:
extremely
PTD, very
PTD and
moderately
PTD.

Cox
regression
analyses





9

Women with lower lever of education
attainment had an elevated risk of PTD
compared to those with higher level of
education.
The above association interacted with parity.

Table 1.2. Studies reporting on associations between maternal socioeconomic mobility and risk of preterm delivery
First Author
(Year)
Osypuk
(2016)

Sample

Explanatory
Variable

Dependent
Variable

USA

SM:

Detroit

education,
income

PTD,
small for
gestational
age

Poisson
regression




SM was not associated with overall PTD.
Among upwardly mobile or stable high women,
improved financial mobility was related to 16%
marginally lower risk of spontaneous PTD.

PTD

Logistic
regression



Compared to women without downward
mobility, higher risk of PTD was found among
those who experienced slight (RR (95 % CI) =
1.2 (1.0, 4.0), moderate (RR (95 % CI) = 1.6
(1.3, 1.9), or extreme (RR (95 % CI) = 1.9 (1.3,
2.6), downward mobility.
Compared to non-low birthweight (LBW)
women, those who had former LBW and
experienced moderate to extreme downward
mobility had greater risk of PTD (RR (95% CI)
=2.8 (1.4, 5.8), and 1.6 (1.3, 1.9), respectively).

(N=1,410)
Collins
(2015)

USA
Illinois
White

Downward
SM:

Statistical
Analyses

Results

neighborhood
income

(N = 12,498)



10

Table 1.2. (cont’d)
First Author
(Year)
Kramer
(2014)

Collins
(2011)

Sample

USA

Explanatory
Variable

Dependent
Variable

SM:

PTD

cumulative
neighborhood
(N=413,048) deprivation
index (NDI)

Georgia

USA

SM:

Illinois

neighborhood
income

African
American

Statistical
Analyses
Binomial
regression

Results



PTD

Logistic
regression





(N = 11,265)

11

On average, Georgia women experienced
upward mobility.
The association between cumulative NDI and
PTD was modified by the effect of history of
prior PTD and advancing age.

Compared to women without upward SM,
those who experienced low, modest, or high
upward SM had decreased risk of PTD (RR
(95 % CI) = 0.9 (0.8, 0.9), 0.8 (0.7, 0.9), and
0.7 (0.6, 0.8), respectively).
Among former low birthweight (LBW) and
non-LBW mothers aged 20 to 35 years, those
who experienced modest to high upward SM
had decreased risk of PTD (adjusted RR (95%
CI) = 0.9 (0.5, 1.6) and 0.7 (0.5, 0.9),
respectively).

Table 1.2. (cont’d)
First Author
(Year)
Love
(2010)

Sample

Explanatory
Variable

Dependent
Variable

USA

SM:

Illinois

neighborhood
income

PTD, small
for
gestational
age, low
birthweight

Race-specific 
multilevel
logistic
regression


Among African American women with static
low SEP, those aged 20–24 years had the
lowest risk of PTD.
Among white women with static low SEP and
static high SEP, those aged ≥20 years older had
significantly decreased risk of PTD than teens.

PTD, low
birthweight

Stratified
logistic
regression



Compared to whites, African Americans had a
twofold greater PTD rate (RR=2.2, 95%
CI=1.7, 2.9).
The association became attenuated in the
adjusted model (adjusted OR=1.2, 95% CI=0.4,
2.0).

(N=70,409)

Collins
(2007)

USA

SM:

Illinois

neighborhood
income

(N=3,104)

Statistical
Analyses

Results



12

1.5 Prenatal depression
Prenatal depression is an important worldwide public health problem and is a major cause
of disease-related disability among women (Kessler 2003). The prevalence of prenatal
depression is about 10% to 20% in high income countries, (Marcus et al. 2003, Gavin et al. 2005,
Evans et al. 2007, Melville et al. 2010), while it is over 20% in low and middle income countries
(Faisal-Cury et al. 2009, Golbasi et al. 2010, Husain et al. 2011). It is estimated by the American
Congress of Obstetricians and Gynecologists that about 14-23% of women have prenatal
depression symptoms. A systematic review by Gavin et al. (2005) found that about 18.4% of
women experiencing prenatal depression, and 12.7% having an episode of major depression.
Another systematic review of 21 studies by Bennett et al. (2004) reported an overall prevalence
of 10.7% and 7.4%, 12.8%, and 12.0% of pregnant women experienced depression during the
first, second, and third trimester, respectively. Prenatal depression is likely to be under diagnosed
by health care providers since it is highly ascribed to emotional changes of pregnancy hormones
(Brown and Solchany 2004).
Depression during pregnancy have often linked with some socioeconomic and lifestyle
factors, e.g. smoking, substance use, alcohol consumption, improper nutrition, and inadequate
weight gain (Marcus 2009). Women with a history of depression before pregnancy have as 22.4fold increased risk for major depression during pregnancy as those who do not have any prior
history of depression (Räisänen et al. 2014). In addition, prenatal depression is positively
associated with postpartum depression and recurrent depression, all of which have the potential
to adversely affect the general health and well-being of mother and her children (Bennett et al.
2004, Cents et al. 2013). Some studies report a link between prenatal depression and increases in
adverse pregnancy outcomes, e.g. PTD, low birthweight, and stillbirth (Field, Diego, and
13

Hernandez-Reif 2006, Evans et al. 2007, Rahman et al. 2007, Wisner et al. 2009, Grote et al.
2010, Grigoriadis et al. 2013, Ncube et al. 2017). Several studies observe that the prevalence of
perinatal depression is more common among African Americans than among white women (Dole
et al. 2004, Nicholson et al. 2006, Li et al. 2008, Gavin et al. 2011, Mukherjee et al. 2016).
Furthermore, previous studies provided some evidence about potential mechanisms linking SEP
and depression, e.g. health behaviors (substance use, preventive care, etc.), psychological factors
(stress/negative affect, mastery/control, etc), physiological factors (physiological dysregulation,
altered cortical development, etc.) (Cohen et al. 2010). Compared to women with higher SEP,
lower SEP women experienced higher levels of perceived stress and negative affect which
increased the risk of depression during pregnancy (Rutter & Quine 1990). Hudson (2005) found
that economic stress mediated the association between SEP and mental illness. Mastery was also
reported to be mediators for the association between SEP and psychological distress (Myer et al.
2008, Gadalla 2009, Alang 2014). A recent study by Elwell-Sutton et al. (2017) observed that the
association between unemployment and job loss and depression was mediated through income.

1.6 Adulthood socioeconomic position and depression
Maternal disadvantaged SEP in adulthood has been related to a greater risk of prenatal
depression (Holzman et al.2006, Miyake et al. 2012, Bahk et al. 2015); measures of SEP have
included occupation, education and household income, although the results with different SEP
measures are not always consistent in general population (Lorant et al. 2003). A Michigan study
by Holzman et al. (2006) observed particularly high risk of prenatal depression among women
with low SEP measured by maternal income. Perinatal depression was measured by the Center
for Epidemiologic Studies Depression Scale (CES-D), a screening test for depression during the
14

past week (CES_D≥16 vs CES_D<16). About 46% of teens and 47% of disadvantaged women
had a positive CES-D screen (≥16), more common than advantaged women (23%).
Socioeconomic inequity in depression varies with the definition and measurement of SEP.
Table 1.3 shows recent studies reporting on associations between adulthood SEP and risk of
depression. In a met-analysis of 56 studies from 1980 to 2002, Lorant et al. (2003) suggested that
disadvantaged SEP was linked to higher odds of being depressed. Based on data from the
National Survey of American Life, Hudson et al. (2012) reported a link between SEP and
depression which varied according to sex. Among African-American men, low level of education
attainment, unemployment and low household income predicted higher risk of 12-month major
depressive episode (MDE). For African-American women, household income was inversely
associated with 12-month MDE. A study by Stewart et al. (2007) found low SEP and material
deprivation were associated with the presence of depression among pregnant women. Miyake et
al. (2012) examined Japanese data and identified that a lower prevalence of prenatal depression
were more common among women who had full-time employment, especially, a professional or
technical job. Production, service, sales, and other occupations were not significantly related to
prenatal depression. However, no significant association between household income or education
and prenatal depression were observed. A recent Korean study (Bahk et al. 2015) showed that
maternal education, paternal education, and household income were inversely associated with
maternal postpartum depressive symptoms, while no significant association was observed for
depressive symptoms one month before childbirth.
Previous studies among non-pregnant population indicate a consistent inverse
relationship between SEP and depression among whites, however, evidence for African
Americans is mixed (Lorant et al. 2003, Williams et al, 2007, Gavin et al. 2010). For instance,
15

some studies observed a higher prevalence of depression among African Americans with lower
SEP (Banks & Kohn-Wood 2002, Bromberger et al. 2004, Roxburgh 2009); however, other
studies of African Americans find no significant association between SEP and depression (Gavin
et al. 2010, Williams et al. 2007). Even studies examining multiple SEP indicators among
African Americans find no significant relationship between SEP and depression (Rodriguez et al.
1999, Hudson et al. 2012). These inconsistent findings may be explicated by the differences in
study sample, study design and methodology, such as various SEP measures, and variable
depression screening timing and instruments. Another potential explanation for the racial/ethnic
disparity for the effect of SEP on depression is that African Americans may not experience the
same protection as whites against the development of depression with higher SEP (Farmer &
Ferraro 2005). African Americans are more probable than whites to have adverse experiences in
childhood, hence, those African Americans who experience upward SM may have residual
effects of early disadvantage. Moreover, African Americans are more likely to encounter racial
discrimination which may limit them to attain some level of education or achieve advanced
occupation status compared to their white peers (Cole & Omari 2003). Experiences of racial
discrimination may provoke stress responses which ultimately wear down the health of Africa
Americans (Watkins et al. 2010) and may weaken the protective effects of improved levels of
SEP among Africa Americans (Hudson et al. 2012).

1.7 Childhood socioeconomic position and depression
Fewer studies have examined mother’s childhood SEP and maternal mental health,
though there is evidence that SEP in childhood has an inconsistent effect on adult mental health

16

among general population (Ritsher et al. 2001, Gilman et al. 2002, Harper et al. 2002, Nicholson
et al. 2008, Gale et al. 2011, Stansfeld et al. 2011). Table 1.4 shows recent studies reporting on
associations between childhood SEP and risk of depression. Most studies found that childhood
SEP had a strong impact on depression in later life (Ritsher et al. 2001, Gilman et al. 2002,
Nicholson et al. 2008, Stansfeld et al. 2011). A study from US National Collaborative Perinatal
Project (1959 – 1966) observed that low childhood SEP, measured by parental occupation, had
been linked to an elevated risk of major depression among adults (Gilman et al. 2002). In the
same study, participants with socioeconomically disadvantaged childhood had as three-fold risk
of depression as those with advantaged childhood, independent of adulthood SEP and family
history of mental illness. In addition, a study of the 1958 British Birth Cohort reported that
participants from lower childhood SEP only marginally increased the risk of depressive and
anxiety disorder in the mid-life, and part of the effect of childhood SEP was mediated through
adulthood SEP (Stansfeld et al. 2011). However, the impact of childhood SEP on later
depression has not been found in some studies (Harper et al. 2002, Gale et al. 2011).
The relationships between childhood SEP, adulthood SEP and mental health are likely
varied and complex with a multitude of explanations (Stansfeld et al. 2011).

The ‘social

causation’ theory postulates that low SEP increases the risk of a mental health disorder, mainly
depression and anxiety. This may be explained by critical period model or by pathway model
(Ljung and Hallqvist 2006). In one critical period model, the effect of disadvantage childhood
SEP on biological systems remains latent until adulthood, and is independent of adverse
adulthood SEP. In one pathway model, low childhood SEP increases the likelihood of low
adulthood SEP and this life course trajectory escalates the risk of mental disorders. There is also

17

the ‘health selection’ theory which asserts that childhood mental health disorders increase the
probability of low adult SEP.

1.8 Socioeconomic mobility and depression
Although a rather large body of evidence linked childhood SEP or adulthood SEP to
depression in later life, limited studies have tested whether SM is associated with depression.
Moreover, to our knowledge, no previous study has directly investigated the association between
SM and depression in pregnancy. In a study among 101 well-educated, African-American
pregnant women, Owens et al. (2015) reported life-course SEP and mobility, defined by family
class, did not moderate the association between contextualized stress and depression, and
contextualized stress was the only predictor for depression.
Prior studies of non-pregnant populations found that upward SM and stable high SEP
were protective against depressive symptoms (Luo and Waite 2005, Nicklett and Burgard 2009,
Hudson et al. 2013, Ro 2014, Dal Grande et al. 2015, Ward et al. 2016, Kwon et al. 2017). Table
1.5 shows recent studies reporting on associations between SM and risk of depression. Using the
United States nationally representative data of late middle age individuals, Kwon et al. (2017)
showed that high and increasing depression was associated with low financial status in childhood
and adverse adulthood SEP in terms of education and income; those with upward SM were less
likely to have high and increasing depression, compared to those with persistently low SEP. In
an observational study of mothers of young children in California, Walsemann et al. (2017)
found that compared with longstanding advantaged neighborhood, long-term disadvantaged and
increasing poverty neighborhood were associated with higher odds of depressive symptoms,

18

decreasing poverty neighborhood were about half as likely to report depressive symptoms. An
intergenerational study in a population of Mexican origin observed the lowest prevalence of
depression symptoms occurred in participants with upward SM, compared to those with stable
low education attainment; and among adults, stable low or downward SM was related to
depression symptoms (Ward et al. 2016). A significant association was also found in participants
from South Australia between stable low SEP for both childhood and adulthood and downward
SM, measured by housing tenure and financial situation, and mental health conditions (Dal
Grande et al. 2015). Using data from young adults in the United States, Hudson (2013) observed
an association between life course SEP and depressive symptoms, but this relationship was only
statistically significant in African Americans. In an American longitudinal study, participants
with upward SM had fewer depressive symptoms, compared to those with stable low SEP, in
terms of a single index created from parental education and other childhood socioeconomic
measures (Luo and Waite 2005). Adverse childhood SEP may be partially or wholly attenuated
by upward SM. A prosperity cohort study in northeast England by Tiffin (2005) also showed that
a downward SM over the whole life course was related to poorer mental health in men (p<0.001),
instead of women (p=0.8). Among immigrants, Nicklett and Burgard (2009) observed that
downward SM was linked to poorer mental health; Ro (2014) observed that both upward and
downward SM, as measured by a change in occupational prestige and educational match, were
related to higher depression among foreign-born women in the United States. These previous
findings implied that effect of adulthood SEP on depressive symptoms also depends on
childhood SEP. These results also demonstrated that joint effects of SEP across multiple
generations played an essential role in depression.

19

Table 1.3. Studies reporting on associations between adulthood socioeconomic position and risk of depression
First Author
(Year)
Bahk
(2015)

Sample

Korea

SEP:

(N=2,065)

education,
occupation,
household
income

Nonpregnant
population
Lindström
(2013)

Explanatory
Variable

Sweden

SEP:

(N=28,198) occupation

Dependent
Variable

Statistical
Analyses

Depression
(Kessler 6-Item
Psychological
Distress Scale)

Generalized
estimating
equation

Psychological
health
(GHQ12).

Logistic
regression

Results



Nonpregnant
population





20

Parental education and household income
were inversely associated with maternal
depressive symptoms.
No significant association was observed
between paternal occupation and maternal
depressive symptoms over the study period.
This study confirmed the social mobility
hypothesis and the accumulation
hypothesis, but not the critical period
hypothesis.
Exposures to economic stress during both
childhood and adulthood were related to
poor psychological health in adulthood.

Table 1.3. (cont’d)
First Author
(Year)
Hudson
(2012)

Sample

USA

SEP:

African
Americans
(N=2,137)

household
income,
education,
employment
status, wealth

Nonpregnant
population

Miyake
(2012)

Explanatory
Variable

Japan

SEP:

(N=1,741)

employment,
household
income,
education

Pregnant
population

Dependent
Variable
Major
Depressive
Episode (MDE)

Statistical
Analyses
Logistic
regression

Results




Depression
(Center for
Epidemiologic
Studies
Depression
Scale) (CES-D)

Logistic
regression





21

Among African-American men, low level of
education attainment, unemployment, and
low household income predicted greater risk
of 12-month major depressive episode
(MDE); unemployment was statistically
related to lifetime MDE.
For African-American women, household
income was inversely associated with 12month MDE.

A lower prevalence of prenatal depression
was more common among women who had
full-time employment, especially, a
professional or technical job. Production,
service, sales, and other occupations were
not significantly related to prenatal
depression.
No significant association between
household income or education and prenatal
depression were observed.

Table 1.3. (cont’d)
First Author
(Year)
Stewart
(2007)

Sample

USA

SEP:

(N=1,802)

education,
income,
public
assistance,
material
deprivation,
subjective
social
standing

Pregnant
population

Holzman
(2006)

Explanatory
Variable

USA

SEP:

(N=1,321)

Medicaid,
housing,
employment,
telephone
service,
transportation

Pregnant
population

Dependent
Variable
Depression
(CES-D)

Statistical
Analyses
Multiple
linear
regression

Results




Depression
(CES-D)

Generalized
linear
regression




22

Low education and low level of education
attainment were associated with higher risk
of depression.
Public assistance and material deprivation
were significantly associated with
depression.
Those with higher scores of subjective
social standing were less depressed.

Compared to advantaged women (23%),
depression symptom was more common
among disadvantaged women (47%).
Among disadvantaged and advantaged
women, abuse and economic problems were
associated with higher risk of depression
symptoms when these problems occurred in
both childhood and adulthood.

Table 1.3. (cont’d)
First Author
(Year)
Lorant
(2003)

Sample

Metaanalysis;
51prevalen
ce studies,
5 incidence
studies, 4
persistence
studies.
(19792001)

Explanatory
Variable
SEP:
education,
income,
occupation,
social class,
assets

Dependent
Variable
Depression

Statistical
Analyses
A random
effects
model;
metaregression

Results





23

Low SEP was associated with higher risk of
depression (OR = 1.81, p < 0.001), no
matter a new episode (OR = 1.24, p =
0.004), or a persisting depression (OR=
2.06, p < 0.001).
Education and income had a dose-response
relation with risk of depression.
The association between SEP and risk of
depression varied by the way depression is
measured, the definition and measurement
of SEP, region and time.

Table 1.4. Studies reporting on associations between childhood socioeconomic position and risk of depression
First Author
(Year)
Gale
(2011)

Sample

UK
Five cohorts
(N=5,570)

Explanatory
Variable

Dependent
Variable

Statistical
Analyses

SEP:

Depression
and
Anxiety
(the
Hospital
Anxiety
and
Depression
Scale)

Logistic
regression

Father’s
occupation;
Own or
husband’s
occupation

Stansfeld
(2011)

UK

SEP:

(N=9,377)

occupation

Results




Depressive Logistic
and anxiety regression
disorders





24

Lower social class in childhood was associated with
an increased risk of depression (OR=1.20, 95% CI
1.07, 1.34).
Lower social class in adulthood was linked to an
elevated risk of depression OR=1.35, 95% CI 1.19,
1.58).
The associations between social class and depression
became attenuated in the adjusted model (during
childhood OR=1.10, 95% CI 0.95, 1.26; during
adulthood: OR=1.09, 95% CI 0.81, 1.46)
Cumulative disadvantaged childhood SEP was
associated with higher risk of mid-life disorder.
Psychological disorder in childhood was related to
low adulthood SEP (OR = 3.33, 95% CI 2.63, 4.21).
Social causation and health selection explained the
relation between disadvantage SEP in childhood and
depression in adulthood.

Table 1.4. (cont’d)
First Author
(Year)
Nicholson
(2008)

Gilman
(2002)

Sample

Explanatory
Variable

Dependent
Variable

Statistical
Analyses

Russia,
Poland, and
the Czech
Republic
(N=12,053
men and
N=13,582
women)

SEP:
father’s
education;
household
amenities;
own
education;
current
financial
difficulties;
possession of
household
items.

Depression
(CESD-20)

Logistic
regression

USA

SEP:

(N=1,132)

occupation

Major
depressive
episode

Results




Survival
analysis





25

Among men, there was significant relationship
between adulthood SEP and depression (OR= 3.16,
95% CI= 2.57, 3.89 in Russia; OR= 3.16, 95% CI=
2.74, 3.64 in Poland; and OR= 2.17, 95% CI= 1.80,
2.63 in the Czech Republic).
Among female, the relationship between SEP and
depression was not observed.

Compared to those with advantaged childhood SEP,
participants with disadvantaged childhood SEP had
an increased risk of major depression, independent
of adult SEP and family history of mental illness.

Table 1.4. (cont’d)
First Author
(Year)
Harper
(2002)

Ritsher
(2001)

Sample

Explanatory
Variable

Dependent
Variable

Statistical
Analyses

Finland

SEP:

(N=2,585)

education,
income,
occupation

Depression
(the
Human
Population
Laboratory
depression
index)

Multiple
linear
regression

Major
depressive
disorder

Logistic
regression

USA

SEP:

(N=756)

education,
occupation

Results








26

Low level of education attainment and unskilled
occupation were associated with higher risk of ageadjusted levels of depressive symptoms in
adulthood.
There was no association between childhood SEP
and adult depressive symptoms.
There was a significant association between low
parental education and higher risk for offspring
depression, even after adjusting for parental
depression, offspring age and gender.
Depression cannot predict offspring education,
income or occupation.

Table 1.5. Studies reporting on associations between socioeconomic mobility and risk of depression
First Author
(Year)
Kwon
(2017)

Walsemann
(2017)

Sample

USA
(N=8,532)

Explanatory
Variable
SM:
parental
education,
financial
status; own
education,
income

USA

SM:

California

Neighborhood-level
poverty

(N=2,726)

Dependent
Variable
Depression
(CESD-20)

Statistical
Analyses
Multinomial
logistic
regression

Results




Depressive
symptoms, sense
of control,
number of
stressors

Logistic
regression,
linear
regression,
zero-inflated
Poisson
regression







27

Those who had a disadvantaged childhood
and continued to experience limited or no
upward mobility tended to experience more
depressive symptoms.
Those who had a socioeconomically
disadvantaged childhood but experienced
upward mobility over their life course (lowhigh) still tended to be high and increasing
depression symptoms.
Long-term high-poverty and increasing
poverty neighborhoods were associated with
higher odds of depressive symptoms than
long-term low-poverty neighborhoods.
Decreasing poverty neighborhoods were
associated with lower odds of depressive
symptoms than long-term low-poverty
neighborhoods.
Duration of residence was unrelated to
depressive symptoms.

Table 1.5. (cont’d)
First Author
(Year)
Ward
(2016)

Sample

Explanatory
Variable

USA

SM:

Mexicanorigin

education

Dependent
Variable
Depression
(CESD-10)

Statistical
Analyses
Generalized
linear
regression

(N=603)

Owens
(2015)

USA
Atlanta
African
American
(N= 101)

Results




SM:
family class

Depression
(Beck
Depression
Inventory-II)

Linear
regression




28

Compared to those with stable low education
attainment, participants with upwardly SM
and stable high SM had lower risk of
depression (PR (95% CI) =0.55 (0.39, 0.78);
0.62 (0.44, 0.87), respectively).
Compared to those with stable low education
attainment, participants with downwardly SM
also had lower risk of depression (PR = 0.65;
95% CI = 0.38, 1.11).
There was significant association between
high contextual stress and depression.
The association between contextualized stress
and depression was not modified by the lifecourse SEP and mobility

Table 1.5. (cont’d)
First Author
(Year)

Sample

Explanatory
Variable

Dependent
Variable

Statistical
Analyses

Results

Logistic
regression



Participants who had downward SM and
stable low SM from childhood to adulthood
had greater risk of anxiety, stress or
depression and psychological distress.

SM:
occupational
Immigrants prestige and
educational
(N = 2,305)
match

Depression:
Logistic
the World Health regression
Organization
Composite
International
Diagnostic
Interview ShortForm (CIDI-SF)



Both upward and downward SM were related
to higher depression among foreign-born
women in the United States.

USA
(N=3,620)

Depression
(CESD-20)



Overall, greater levels of life course SEP were
significantly associated with fewer depressive
symptoms (β = -0.33, p = 0.001).
There was a relationship between life course
SEP and depressive symptoms, but this
relationship was only statistically significant
in blacks (β = -0.52, p =0.001).

Grande
(2015)

Australia

Ro (2014)

USA

Hudson
(2013)

SM:

(N=10,429) family
financial
situation and
housing
tenure

SM:
parental
education;
own
occupation

Anxiety, stress
or depression,
suicidal ideation,
psychological
distress.

Multivariate
regression



29

Table 1.5. (cont’d)
First Author
(Year)
Nicklett
(2009)

Sample

USA
immigrants
(N=3,056)

Explanatory
Variable
Downward
SM:

Dependent
Variable

Statistical
Analyses

Results


Participants who had downward SM had an
elevated risk of depression (OR= 3.0, 95% CI:
1.3, 6.6), compared to those with stable SM.

Depression
(CESD)

Ordinary

least squares
regression

Participants with upward SM had fewer
depressive symptoms, compared to those with
stable low SEP, in terms of a single index
created from parental education and other
childhood socioeconomic measures.

General health
questionnaire
(GHQ-28)

Logistic
regression



Compared to those with the most advantaged
SEP, participants who had the disadvantaged
SEP at birth reported a significant poor mental
health at age 50 (OR=5.5, 95% CI 1.2, 25.4).
The association between a downward SM and
poorer mental health was found among men
(p<0.001), instead of women (p =0.8).

Major depression Logistic
episode
regression

social status

Luo
(2005)

USA

Tiffin
(2005)

UK

SM:

(N=508)

occupational
social class

SM:

(N=19,949) education,
household
income,
occupation



30

1.9 Gap in the literature
Epidemiologic studies investigating the relationship between SM and perinatal health
have employed various measures of maternal SEP. Some studies used neighborhood census tract
median family income to measure childhood and adulthood SEP. Without measuring women’s
SEP at multiple time points during the life course, many SM studies in the United States used the
linked birth files across generations and ecological SEP measures from the United States Census
data (Collins et al. 2007, Love et al. 2010, Collins, Rankin, and David 2011, Collins, Rankin, and
David 2015). These United States-based studies showed that women who experienced upward
SM from childhood in a disadvantaged neighborhood to adulthood in an advantaged
neighborhood had better birth outcomes, compared to those with lifelong residence in a
disadvantaged neighborhood. However, these studies were unable to assess the impact of
individual level wealth over the life course on birth outcomes. On the other hand, some studies
used various combinations of household income, individual occupation, and education
attainment to estimate the maternal SEP, but lack of information about contextual factors of
maternal SEP.
Many prior studies measured SEP at a single time point and used a single indicator, e.g.
education, occupation or family income (Auger, Giraud, and Daniel 2009, Nicklett and Burgard
2009, Blumenshine et al. 2011, Joseph et al. 2014, Gigante et al. 2015, Owens and Jackson 2015,
Ward et al. 2016). However, the multidimensional nature of SEP and the complexity of SEP
transition from one time point to another suggest that SEP and SM cannot be measured
completely with one indicator. In pregnancy outcomes studies, methods for combining multiple
SEP indicators and for measuring SM from childhood to adulthood varied (Gigante et al. 2015,
Slaughter-Acey et al. 2016). In a mental health study with multiple SEP indicators, a single index
31

of SEP measure was defined as the average of standardized SEP indicators (Luo and Waite
2005).
In this study, we used a latent class analysis (LCA) for measuring SM in a dataset with
individual-level data of maternal childhood and adulthood SEP measures, a novel approach for
pregnancy studies. We also adopted the model-based approach to LCA with distal outcome to
elucidate complex relationships between SM and the risk of PTD or prenatal depression among
pregnant women. Typically, traditional classify-analyze strategies have been used to predict a
distal outcome conditioned on latent class membership. In the first classification step, individuals
were assigned to latent classes based on probabilities. In a second analysis step, the distal
outcome was predicted from the latent class membership which was treated as observed. For
these traditional strategies the results are biased due to classification error. However, the modelbased approach that we use models measurement error (Collins & Lanza, 2010) and produces
less biased estimates for the probability of prenatal depression conditional on latent class
membership.
Despite growing evidence examining the effect of SM on pregnancy outcomes and
mental health later in life, few studies have taken into account the multi-dimensional aspect of
SEP when examining associations between SM and PTD; and to our knowledge, there has been
no previous study to investigate the association between SM and depression among pregnant
women. Therefore, there remains a gap in the literature exploring the associations of SM with
both PTD and depression among pregnant women in the United States.

32

1.10 Aims
The main objective of this study is to investigate associations between maternal SM from
childhood to adulthood and risk of PTD and risk of prenatal depression in a Michigan,
community-based cohort study that is racially and socioeconomically diverse. The availability of
in-depth, individual-level data on mothers and their childhood SEP provides a unique
opportunity to study SM. The aims of this study are: 1) to identify and describe distinct groups of
the childhood SEP indicators and adulthood SEP indicators and SM from childhood SEP to
adulthood SEP; and to investigate the associations between latent groups and PTD using modelbased approach; and to evaluate whether maternal race/ethnicity modifies the above associations;
2) to examine the associations between latent groups and perinatal depression using model-based
approach; and to evaluate whether maternal race/ethnicity modifies the above associations.
In Chapter 2, based on the life course model of SEP trajectories, we hypothesize that: 1)
women with lowest childhood SEP and lowest adulthood SEP will be at greatest risk of PTD;
and 2) women who experience upward SM from childhood to adulthood will have a decreased
risk of PTD relative to women who maintain a lower SEP during childhood and adulthood; and 3)
women who experience downward SM from childhood to adulthood will have a higher risk of
PTD relative to women who maintain a higher SEP during childhood and adulthood. Because
other studies reported that the relation of SEP (or SM) with PTD risk varied by maternal
race/ethnicity (P. Blumenshine et al. 2010, Collins, Rankin, and David 2011, Collins, Rankin,
and David 2015), we also examine evidence for this effect modification.
In Chapter 3, we hypothesize that: 1) women with lowest childhood SEP and lowest
adulthood SEP will be at greatest risk of prenatal depression; 2) women who experience upward

33

SM from childhood to adulthood will have a decreased risk of prenatal depression relative to
women who maintain a lower SEP; and 3) women who experience downward SM from
childhood to adulthood will have a higher risk of prenatal depression relative to women who
maintain a higher SEP during childhood and adulthood. Because some studies reported that the
relation between SEP (or SM) and depression varied by race/ethnicity (Luo and Waite 2005,
Owens and Jackson 2015, Ward et al. 2016), we also examine evidence for this effect
modification.

34

CHAPTER 2
MATERNAL SOCIOECONOMIC MOBILITY AND PRETERM DELIVERY:
LATENT CLASS ANALYSIS

2.1 Introduction
Preterm delivery (PTD, delivery less than 37 completed weeks of gestation) is a major
cause of perinatal mortality and morbidity (Blencowe et al. 2013). The disparities in PTD rates
between population subgroups defined by race/ethnicity and socioeconomic status are striking
(Blumenshine et al. 2010). Maternal disadvantaged socioeconomic position (SEP) in adulthood
has been associated with higher PTD risk (Blumenshine et al. 2010, Morgen et al. 2008, Auger,
Giraud, and Daniel 2009, Blumenshine et al. 2011, Mortensen, Helweg-Larsen, and Andersen
2011, Matijasevich et al. 2012, Whitehead 2012, Joseph et al. 2014, Shankardass et al. 2014).
Measures of SEP have included occupation, education and household income, and the
relationship with PTD often varied by race/ethnicity (Blumenshine et al. 2010). Childhood SEP
has also been examined to have a significant effect on adult health (Cohen et al. 2010). However,
few studies have examined mother’s childhood SEP and the risk of PTD. A Brazilian
intergenerational study observed that maternal childhood poverty, measured by family income,
was strongly associated with shorter gestation (Gigante et al. 2015).
Socioeconomic mobility (SM) has been studied in relation to low birthweight (Basso et al.
1997, Spencer 2004, Colen et al. 2006, Astone, Misra, and Lynch 2007, Love et al. 2010,
Osypuk et al. 2016), small for gestational age (Love et al. 2010, Osypuk et al. 2016, SlaughterAcey et al. 2016), and newborn body composition (Sletner et al. 2014). In our review of the

35

literature, however, we found fewer studies examining SM and PTD (Love et al. 2010, Collins,
Rankin, and David 2011, Collins et al. 2007, Kramer, Dunlop, and Hogue 2014, Collins, Rankin,
and David 2015). Collins et al. (2011) found that upward SM was related to a decreased risk of
PTD among African-American women. Downward SM was linked to an elevated risk of PTD
among white urban women (Collins, Rankin, and David 2015). However, these studies measured
SM based on neighborhood census tract median family income rather than individual or
household data.
Three conceptual models have been proposed to explicate how disadvantaged life course
SEP affects health later in life and they may not be mutually exclusive (Kuh et al. 2003,
Kendzor, Caughy, and Owen 2012). One model focuses on critical developmental periods (e.g.
childhood) when SEP may exert its strongest effect on later health. Another model views
“accumulation” of socioeconomic disadvantage (intensity and duration) as posing the greatest
health risk. A third model considers SEP trajectories, i.e. SM; it hypothesizes that upward SM
remediates the negative effects of earlier disadvantage SEP, however, downward SM may result
in poorer adult health despite relative socioeconomic advantages at earlier stages (Kuh et al.
2003, Cohen et al. 2010, Kendzor, Caughy, and Owen 2012). In all these models, determining
the temporal order of SEP and health can be challenging.
Previous studies have typically measured SEP at a single time point, and used a single
indicator, e.g. education, occupation or family income (Auger, Giraud, and Daniel 2009,
Blumenshine et al. 2011, Joseph et al. 2014, Gigante et al. 2015). However, the multidimensional
nature of SEP and the complexity of SEP transition from one time point to another suggest that
SEP and SM cannot be measured completely with one indicator. In pregnancy outcomes studies,
methods for combining multiple SEP indicators and for measuring SM from childhood to
36

adulthood vary (Gigante et al. 2015, Slaughter-Acey et al. 2016). Our study uses a latent class
analysis (LCA) for measuring SM in a dataset with individual-level data of maternal childhood
and adulthood SEP measures, a novel approach for pregnancy studies.
The main objective of this study is to examine associations between maternal SM from
childhood to adulthood and risk of PTD in a Michigan, community-based cohort study that is
racially and socioeconomically diverse. The availability of in-depth, individual-level data
provides a unique opportunity to study SM. Based on the model of SEP trajectories, We
hypothesize that: 1) women with lowest childhood SEP and lowest adulthood SEP will be at
greatest risk of PTD; and 2) women who experience upward SM from childhood to adulthood
will have a decreased risk of PTD relative to women who maintain a lower SEP during
childhood and adulthood; and 3) women who experience downward SM from childhood to
adulthood will have a higher risk of PTD relative to women who maintain a higher SEP during
childhood and adulthood. Because other studies reported that the relation of SEP (or SM) with
PTD risk varied by maternal race/ethnicity (Blumenshine et al. 2010, Collins, Rankin, and David
2011, Collins, Rankin, and David 2015), we also examine evidence for this effect modification.

2.2 Methods
Study population
This study included 3,019 pregnant women enrolled at 16-27 weeks’ gestation in the
prospective Pregnancy Outcomes and Community Health (POUCH) study (1998-2004) from five
Michigan communities. Eligibility criteria included: English-speaking, age >14 years, no preexisting diabetes, singleton pregnancy with no known congenital anomalies, and prenatal
37

maternal serum alpha-fetoprotein screening at 15-22 weeks of gestation. The detailed study
protocol can be found elsewhere (Holzman et al. 2001). Institutional review board approval for
this study was received from Michigan State University and all nine participating delivery
hospitals.
At enrollment each pregnant woman provided informed written consent, met with a study
nurse for a structured in-person interview, and completed a self-administered questionnaire that
collected information regarding demographics, lifestyle, psychosocial factors, medical history
and health status. Pregnancy outcome data were abstracted from hospital delivery records. Selfreported maternal race/ethnicity was dichotomized as white/others or African American. The
numbers of “others” race/ethnicity were small and their PTD rates were similar to that of whites,
therefore these groups were combined.
Measures
Childhood SEP and adulthood SEP
Childhood SEP measures included a woman’s self-report of her parents’ socioeconomic
indicators during her childhood, i.e. maternal mother’s and father’s highest level of education
attainment, maternal mother’s and father’s usual occupation, and receipt of public assistance by
the family. Adulthood SEP measures included maternal socioeconomic indicators at enrollment,
i.e. maternal and child’s father’s highest level of education, maternal and child’s father’s usual
occupation, Medicaid Insurance status, and maternal annual household income. SEP indicators
were modeled as binary variables, i.e. education (≤high school, >high school); usual occupation
status based on the United States Census Bureau’s 1990 Occupational Classification System
(“low”= service/blue collar, unemployed, and unknown; “high”= professional, manager,

38

technical, clerical/sales, homemaker, military, retired, and student); public assistance (yes/no);
Medicaid Insurance (yes/no); maternal household income (<$50,000 and ≥$50,000). “Unknown”
paternal and grandfather’s occupation was assigned as “low” occupation status because it
correlated with other low SEP indicators (Slaughter-Acey et al. 2016). Missing data for SEP
indicators were handled primarily through statistical methods (see statistical methods section
below). In separate models, a maternal neighborhood deprivation index (NDI) also was included
in adulthood SEP. The NDI was calculated using information from the 1986 Census and
followed a previously validated approach (Messer et al. 2006). The NDI was dichotomized at the
medium of the study population.
Maternal SM
Maternal SM was defined by modeling both childhood and adulthood SEP indicators in
LCA. The SEP indicator variables were dichotomized the same as the childhood SEP and
adulthood SEP variables.
Preterm delivery
Gestational age at delivery was estimated based on date of last menstrual period unless
this estimate disagreed with early ultrasound (< 25 weeks’ gestation) dating by two or more
weeks, in which case ultrasound dating was used. PTD was defined as deliveries less than 37
completed weeks of gestation.
Statistical Analysis
Initially we assessed the distribution of maternal characteristics and SEP indicators
overall and stratified by delivery timing (preterm/term), and used chi-square tests to test

39

differences in proportions. We used latent class analysis (LCA) to identify and describe distinct
groups of childhood SEP indicators and adulthood SEP indicators and SM from childhood to
adulthood. LCA classifies individuals with similar SEP characteristics into groups based on
conditional probabilities (Collins and Lanza 2010). SM was defined by LCA according to a
multidimensional set of SEP indicators from childhood to adulthood. We determined the optimal
number of LCA groups by: 1) visualizing the plot of log-likelihoods of each model and
eliminating model where the log-likelihood did not show a substantial improvement (Nylund,
Asparouhov, and Muthén 2007); 2) using the Bayesian Information Criterion (BIC), with a
smaller BIC representing a better-fitted model (Schwarz 1978); 3) evaluating the entropy, an
indicator for class separation and precision, that measures how well an individual fits into a
specific class with values closer to 1 representing better fit (Ramaswamy et al. 1993); and 4)
considering the size of the latent classes, uniqueness of classes, and meaningful interpretation of
the response probability pattern for each class. As a next step, after the optimal number of groups
was determined we used the model-based approach to LCA with distal outcome to predict the
risk of PTD based on the latent class membership of exposure variables (i.e. childhood SEP,
adulthood SEP, SM) (Lanza, Tan, and Bray 2013).
Race/ethnicity was added as a grouping variable in LCA models to investigate
associations among SEP, SM and risk of PTD within race/ethnic groups (white/others, African
American). Because SEP indicators may have different ranges and occur in different contexts
across race/ethnic groups, we computed group-specific LCA following the same methodology
described above. In another set of models, we added the NDI to adulthood SEP to examine its
contribution to identifying SEP associations with PTD risk.

40

All statistical analyses were conducted using PROC LCA and its corresponding distal
outcome macro in SAS 9.4 (Statistical Analysis Software, Cary, NC) (Lanza et al. 2015, Dziak et
al. 2016). Proc LCA program uses full information maximum likelihood estimation to cope with
missing data in the SEP indicators.

2.3 Results
Sample characteristics
In the total sample of 3019 women, 11.1% of pregnancies ended in PTD and about one
fourth of women were African American (Table 2.1). For the childhood SEP indictors, about 60%
of grandmothers and 55% of grandfathers did not graduate from high school. Approximately 32%
of grandmothers and 71% of grandfathers had a low-level occupation. About one third of
families received public assistance during the mother’s childhood period. For adulthood SEP
indicators, 47% of mothers and 55% of fathers did not graduate from high school; 39% of
mothers and 65% of fathers had a low-level occupation; 48% of pregnant women were insured
by Medicaid; and 66% of families had an annual income of <$50,000. In bivariate analyses,
increased risk of PTD was associated with two childhood SEP indicators, i.e., grandmother’s
low-level occupation and family history of public assistance; and with all adulthood indicators of
lower SEP, except NDI.
Latent Class Profile
In the LCA of childhood SEP, a three-group solution was considered optimal according
to criteria described in our statistical methods section (Table 2.2). The three-group model had the

41

lowest BIC and substantial improvement in log-likelihood difference when moving from a threegroup to two-group model, compared to that from a four-group model to a three-group model.
The subgroups characterized by particular SEP indicators were interpreted and labelled
according to the parameter estimates. For childhood SEP, compared to other groups, Group 1,
“low group”, (38% of women) grandmothers and grandfathers were less likely to graduate from
high school (17% and 11%, respectively), to have a high-level occupation (23% and 11%,
respectively), or to avoid public assistance during the mother’s childhood (43%) (Table 2.3). The
probabilities of positive childhood SEP indicators improved in childhood SEP Group 2, “middle
group”, (38% of women) with the exception of grandfathers’ education (11%) and occupation
(17%). Childhood SEP Group 3, “high group”, (24% of women) showed high probabilities of all
positive SEP indicators.
For adulthood SEP, a four-group model had the lowest BIC, however there was minimal
improvement when moving from a three-group to four-group model, compared to the substantial
log-likelihoods difference from a two-group to a three-group model. Furthermore, the entropy for
the four-group model was smaller than that for the three-group model, indicating poor latent
group separation for the four-group model (Ramaswamy et al. 1993). Based on additional
considerations of group prevalence and interpretability we chose the three-group model for
adulthood SEP (Table 2.2). For adulthood SEP, compared to other group, Group 1,“low group”,
(51% of women) mothers and fathers were less likely to graduate from high school (22% and
15%, respectively), to have a high-level occupation (40% and 16%, respectively), to have
medical insurance other than Medicaid (15%), or to a family income > $50,000 (4%) (Table 2.3).
The probabilities of SEP indicators in adulthood SEP Group 2, “middle group”, (18% of women)
varied; for example, fathers in this group had low probabilities of having a high-level education
42

(19%) and occupation (23%) but women were more likely to have high-level occupations (73%)
and medical insurance other than Medicaid (94%). About half of the families had an annual
household income exceeding $50,000 (54%). In adulthood SEP Group 3, “high group”, (31% of
women), all probabilities of positive SEP indicators were increased; about 64% of these families
had annual household incomes exceeding $50,000.
We used both childhood and adulthood SEP indicators for the SM LCA. The BIC values
decreased with increasing number of groups until 6 groups. After reviewing the plot of loglikelihoods of each model with specified group, improvements in model fit, size of the groups,
and interpretability we chose the four-group solution for modeling SM from childhood to
adulthood (Table 2.2). In SM from childhood to adulthood, Group 1, “static low group”, (45%
women) women had low-level childhood SEP and low-level adulthood SEP indicators (Table
2.3). SM Group 2, “upwardly mobile group”, (24% of women) showed some improvement in
two childhood indicators, i.e. probabilities of grandmother with high-level occupation (75%) and
avoiding public assistance (81%), and higher probabilities in all positive adulthood SEP
indicators. SM Group 3, “downwardly mobile group”, (16% of women) tended to have highlevel childhood SEP and moderate to low-level adulthood SEP. In this group the fathers’
education and occupation were unlikely to be high-level (39% and 27%, respectively) and the
probability of household income >$50,000 was low (15%). SM Group 4, “static high group”, (15%
of women) had high-level SEP indicators throughout childhood and adulthood.
Latent class models also were examined for white/others women and African-American
women separately using the methodology described above. For each race/ethnic group, based on
fit indices and interpretability, we chose a three-group model for childhood SEP and adulthood
SEP indicators and a four-group model for SM. The race/ethnicity-specific item-response
43

probabilities for each indicator conditional on latent class membership are presented in Table 2.3.
For the childhood SEP, the group percentages for white/others and African American,
respectively, were: low group 24%, 78%; middle group 46%, 13%; and high group 30%, 9%. For
the adulthood SEP, the group percentages for white/others and African American, respectively,
were: low group 44%, 48%; middle group 19%, 31%; and high group 37%, 21%. For SM, the
group percentages for white/others and African American, respectively, were: static low 38%,
57%; upwardly mobile 28%, 15%; downwardly mobile; 16%, 21%; and static high 18%, 7%.
LCA with Distal Outcome: Preterm Delivery
Overall, the frequency of PTD in the low, middle, and high childhood SEP groups was
13.2%, 9.6%, 10.1%, respectively. The frequency of PTD in the low, middle, and high adulthood
SEP groups was 12.9%, 10.3%, and 8.5%, respectively. In comparisons with their respective
high SEP groups, women in the low SEP groups were more likely to deliver preterm: low
childhood SEP group PTD odds ratio (OR) =1.36, 95% confidence interval (95% CI): 0.98, 1.90;
and low adulthood PTD OR=1.59, (95% CI: 1.18, 2.15) (Table 2.4). PTD occurred in13.1%,
8.3%, 11.6%, and 8.8% of women who had experienced static low SM, upward SM, downward
SM, and static high SM, respectively. Women in the upward SM group had decreased odds of
PTD (OR=0.60, 95% CI: 0.42, 0.87) compared to women in the static low SM group. The odds
of PTD among women with downward SM was higher (OR=1.36, 95% CI: 0.80, 2.30) than that
of women with static high SM, however, the association was not statistically significant.
In a separate set of analyses we used race/ethnic-specific LCA models. Among the
white/others women, low childhood SEP and low adulthood SEP were associated with increased
odds of PTD in comparisons with their respective high SEP groups, PTD ORs=1.21, (95% CI:

44

0.98, 1.90) and 1.44 (95% CI: 1.02, 2.04). Women who experienced upward SM had a lower
odds of PTD, OR=0.66, (95% CI: 0.43, 1.00) when compared to those in the static low SM group.
Among African-American women, PTD odds were greater in the low childhood SEP and low
adulthood SEP groups compared to their respective high SEP groups, however, the small
samples sizes for each SEP group resulted in more limited statistical power and no statistically
significant associations.
We incorporated the NDI in the adulthood SEP LCA to examine its added effect on SEP
associations with PTD. Table 2.5 summaries the LCA model fit statistics. In consideration of a
combination of fit statistics and interpretability, we selected a three-group model for adulthood
SEP with NDI and a four-group model for SM that included childhood SEP and adulthood SEP
with NDI. The item-response probabilities for each SEP indicator conditional on latent class
membership of adulthood SEP with NDI and SM are shown in Table 2.6. The latent class
profiles were similar to those in adulthood SEP without NDI. When we assessed the associations
between latent class membership and PTD (Table 2.7), NDI did not substantially change our
original findings.

2.4 Discussion
Using a latent class approach with relevant indicators, we identified constructs of
pregnant women’s childhood SEP, adulthood SEP, and SM from childhood to adulthood in
association with risk of PTD. We found that women with low childhood SEP and low adulthood
SEP were at greatest risk of PTD; and pregnant women who experienced upward SM had a
lower likelihood of delivering preterm than those who had static low SEP. This SM advantage
45

was true in white/others women. Our findings are consistent with Slaughter-Acey et al.’s
analysis of POUCH Study data showing decreased risk of delivering a small-for-gestational-age
infant among pregnant women with upward SM (Slaughter-Acey et al. 2016). In another study of
African-American pregnant women, Collins et al. reported that upward SM was related to a
decreased risk of PTD (Collins, Rankin, and David 2011).
We also observed that disadvantaged SEP at adulthood was related to a higher likelihood
of PTD, a finding that is echoed across many previous studies (Morgen et al. 2008, Blumenshine
et al. 2011, Matijasevich et al. 2012, Whitehead 2012, Joseph et al. 2014). In a systematic review
of 106 studies from 1999 to 2007, Blumenshine et al. (2010) suggested that disadvantaged SEP
was consistently linked to greater risk of adverse birth outcomes and observed racial/ethnic
heterogeneity in the effect of SEP measures (Blumenshine et al. 2011). The inverse associations
between maternal education and adverse birth outcomes were significant among non-Hispanic
white pregnant women. Shankardass et al. (2014) found that the risk of spontaneous PTD was
elevated across all the indicators of lower SEP, and women living in neighborhoods with higher
NDI had greater risks of medically indicated and spontaneous PTD (Shankardass et al. 2014). In
our study, the addition of NDI to the model did not substantively change the adulthood SEP
relation to PTD risk.
Our findings support the hypothesis that improving women’s SEP from childhood to
adulthood may improve pregnancy outcomes. According to the cumulative pathway life-course
model, upward SM might decrease the accumulation of adverse exposures that create “wear and
tear” to the body (allostatic load) (McEwen 1998), and thereby create a more optimal in-utero
environment for the growing fetus (Slaughter-Acey et al. 2016). Pregnant women who
experience upward SM may have: 1) greater access to high-quality food, reliable housing, a safe
46

living environment, and high-quality health care; 2) more support for healthy lifestyle choices
(diet, physical activity, no smoking, no alcohol use, etc.); and 3) greater opportunity for
psychological health (less stress, more social integration, etc.). Our results also imply that the
childhood environment (SEP) is not deterministic for women’s future risk of PTD. In our data,
the adulthood SEP had the strongest association with PTD risk.
To our knowledge, this study is the first to use the model-based approach to LCA with
distal outcome to elucidate complex relationships between SM and the risk of PTD among
pregnant women in the United States. Typically, traditional classify-analyze strategies have been
used to predict distal outcome from latent class membership. In the first classification step,
individuals were assigned to latent classes based on probabilities. In a second analysis step, the
distal outcome was predicted from the latent class membership which was treated as observed.
For these strategies the results were biased to the extent that there is classification error.
However, this model-based approach models measurement error and produces less biased
estimates for the probability of PTD conditional on latent class membership (Collins and Lanza
2010). Moreover, use of the full-information expectation-maximization algorithm in LCA may
include all participants who responded to at least one SEP indicator (Collins and Lanza 2010).
Most previous studies investigating the effect of SM on PTD used a single SEP indicator to
measure SM (Love et al. 2010, Collins, Rankin, and David 2011, Collins et al. 2007, Kramer,
Dunlop, and Hogue 2014, Collins, Rankin, and David 2015). Our study used multiple SEP
indicators for childhood SEP and adulthood SEP measures, which may minimize
misclassification bias. In addition, this study enrolled pregnant women with diverse SEP, which
allowed us to observe a wider range of SM.

47

An important limitation in most SM research pertains to the inferences. When we study
individuals we don’t randomly assign SM, therefore it is difficult to disentangle the extrinsic
benefits of an improved socioeconomic environment from the unique attributes of individuals
who are able to climb the SEP ladder (self-selection). In addition, all SEP indicators were selfreported and recall bias cannot be ruled out, but the recall bias should apply equally to those with
and without PTD because reporting occurred mid-pregnancy, well before delivery. Few AfricanAmerican women experienced SM in this study, making it difficult to calculate accurate
race/ethnicity-specific estimates of the SM-PTD association in this group. Our software and
modeling strategy, i.e. model-based approach to LCA with distal outcome, could not
accommodate adjustment for other covariates (potential confounders); this leaves open the
possibility of unmeasured confounding. While we stratified on race/ethnicity, we considered
maternal age, parity, and pre-pregnancy body mass index as important covariates. In sensitivity
analyses we also looked to see whether our results would vary by these covariates. The analyses
results suggested that our findings were robust even after the stratification.
In summary, our study suggests that policies or programs to reduce socioeconomic
inequalities may play an essential part in decreasing adverse pregnancy outcomes such as PTD.
Upward SM among pregnant women was associated with decreased risk of PTD. This
relationship was most evident among white/others women in our study. Further research is
needed to understand the specific elements accompanying SM that are protective for PTD and
the best strategies for increasing SM among all race/ethnic groups.

48

Table 2.1. Maternal Characteristics and Socioeconomic Position Indicators (n=3,019),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004

Race/ethnicity
White/others
African American
Childhood SEP
Grandmother’s education
Missing
≤ high school
>high school
Grandfather’s education
Missing
≤ high school
>high school
Grandmother’s usual occupationb
Low occupational status
High occupation status
Grandfather’s usual occupationb
Low occupational status
High occupation status
Family history of public
assistance
Missing
Yes
No

Overall
(n=3,019)

Preterm
(n=335)

Term
(n=2,684)

No.

%

No.

%

No.

2276
743

75.4
24.6

225
110

67.2
32.8

%

P
valuea
0.0002

2051 76.4
633 23.6

0.55
128
1810
1081

4.2
60.0
35.8

19
193
123

5.7
57.6
36.7

109 4.1
1617 60.3
958 35.7
0.20

477
1662
880

15.8
55.1
29.2

62
188
85

18.5
56.1
25.4

415 15.5
1474 54.9
795 29.6
0.03

959
2060

31.8
68.2

124
211

37.0
63.0

835 31.1
1849 68.9
0.25

2145
874

71.1
29.0

247
88

73.7
26.3

1898 70.7
786 29.3
0.03

112
1081
1826

3.7
35.8
60.5

49

11
138
186

3.3
41.2
55.5

101 3.8
943 35.1
1640 61.1

Table 2.1. (cont’d)
Overall
(n=3,019)

Preterm
(n=335)

Term
(n=2,684)

No.

No.

No.

%

%

%

Adulthood SEP
Maternal education
≤ High school
>High school

1406
1613

46.6
53.4

175
160

52.2
47.8

1231 45.9
1453 54.1

Father’s education
Missing
≤ High school
>High school

138
1655
1226

4.6
54.8
40.6

18
199
118

5.4
59.4
35.2

120 4.5
1456 54.3
1108 41.3

Maternal usual occupationb
Low occupational status
High occupation status

1180
1839

39.1
60.9

151
184

45.1
54.9

1029 38.3
1655 61.7

Father’s usual occupationb
Low occupational status
High occupation status
Mother’s Medicaid insured
Missing
Yes
No
Annual household income
Missing
< $50,000
≥ $50,000
Neighborhood deprivation index
Missing
≥ Medium
< Medium

P
valuea
0.03

0.04

0.02

0.0004
1952
1067

64.7
35.3

246
89

73.4
26.6

1706 63.6
978 36.4
0.002

3
1454
1562

0.1
48.2
51.7

1
188
146

0.3
56.1
43.6

2
0.1
1266 47.2
1416 52.8
0.01

92
2005
922

3.1
66.4
30.5

9
243
83

2.7
72.5
24.8

83
3.1
1762 65.6
839 31.3
0.09

5
1506
1508

0.2
49.9
50.0

0
182
153

0.0
54.3
45.7

5
0.2
1324 49.3
1355 50.5

Abbreviation: SEP, socioeconomic position.
a
P value for chi-square test between term and preterm groups.
b
Low occupation status includes service/blue collar, unemployed, and unknown; high
occupation status includes professional, manager, technical, clerical/sales, homemaker, and other
(military, retired, student).

50

Table 2.2. Fit statistics for latent class analysis (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Group

Log-likelihood

Bayesian Information Criterion

Entropy

Childhood SEP
2
3
4
5

-8431.3
-8380.5
-8366.2
-8365.0

244.6
191.2
210.7
256.4

0.74
0.70
0.82
0.81

Adulthood SEP
2
3
4
5
6

-10102.4
-9978.9
-9896.7
-9876.6
-9869.4

609.1
418.3
309.9
325.8
367.6

0.81
0.78
0.72
0.74
0.74

3160.4
2637.1
2342.7
2300.2
2263.3
2269.8

0.84
0.80
0.76
0.74
0.73
0.69

SM: childhood SEP & adulthood SEP
2
-18443.8
3
-18134.1
4
-17938.8
-17869.5
5
-17803.0
6
-17758.2
7

Abbreviation: SEP, socioeconomic position; SM, socioeconomic mobility.

51

Table 2.3. Estimated item-response probabilities of the socioeconomic position indicators conditional on latent class
membership with distal outcome preterm delivery overall and by race/ethnicity (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004

Low
Childhood SEP
Latent class membership probabilities
Indicator
Grandmother’s education: >High
school
Grandfather’s education: >High
school
Grandmother’s occupation: High
Grandfather’s occupation: High
Family history of public assistance:
No

No.

Overall
(n=3,019)
Middle High

Latent Group
White/others
(n=2,276)
Low Middle High

1159
0.38

1130
0.38

730
0.24

0.36

0.17

0.34

0.74

0.11

0.27

0.29
0.68
0.29

0.11
0.23
0.11

0.11
0.98
0.17

0.97
0.93
0.76

0.08
0.09
0.11

0.60

0.43

0.67

0.89

0.47

%

52

538
1059
679
0.24
0.46
0.30
Item-response probabilities

African American
(n=743)
Low Middle High
579
0.78

97
0.13

67
0.09

0.72

0.22

0.87

0.93

0.13
0.93
0.16

0.94
0.94
0.77

0.13
0.41
0.11

0.01
0.97
0.12

0.98
0.82
0.64

0.72

0.90

0.30

0.50

0.69

Table 2.3. (cont’d)

Low
Adulthood SEP
Latent class membership probabilities
Indicator
Mother’s education: >High school
Father’s education: >High school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status: No
Annual household income:
≥$50,000

No.

Overall
(n=3,019)
Middle High

1554
0.51

529
0.18

936
0.31

%
0.53
0.41
0.61
0.35
0.52

0.22
0.15
0.40
0.16
0.15

0.68
0.19
0.73
0.23
0.94

0.97
0.99
0.89
0.74
0.89

0.31

0.04

0.54

0.64

53

Latent Group
White/others
(n=2,276)
Low Middle High
1000
443
833
0.44
0.19
0.37
Item-response probabilities
0.25
0.70
0.97
0.17
0.21
0.99
0.39
0.76
0.90
0.18
0.23
0.76
0.22
0.98
0.91
0.06

0.58

0.66

African American
(n=743)
Low Middle High
359
0.48

226
0.31

158
0.21

0.12
0.09
0.09
0.13
0.09

0.26
0.09
0.94
0.12
0.07

0.90
0.70
0.77
0.50
0.61

0.004

0.03

0.35

Table 2.3. (cont’d)

SL

Overall
(n=3,019)
UM DM

Latent Group
White/others
(n=2,276)
SL UM DM SH

SH

African American
(n=743)
SL UM DM SH

SM: childhood SEP &
No. 1372 715 475
456
855 639 365 418
424 114 152
54
adulthood SEP
Latent class membership
probabilities
0.45 0.24 0.16 0.15
0.38 0.28 0.16 0.18
0.57 0.15 0.21 0.07
Indicator
%
Item-response probabilities
Grandmother’s
education: >High school
0.36 0.20 0.24 0.67 0.76
0.15 0.23 0.66 0.76
0.20 0.44 0.59 0.89
Grandfather’s
education: >High school
0.29 0.07 0.15 0.64 0.95
0.06 0.16 0.72 0.95
0.08 0.00 0.39 0.87
Grandmother’s
occupation: High
0.68 0.49 0.75 0.86 0.97
0.52 0.77 0.91 0.97
0.45 0.60 0.52 0.89
Grandfather’s occupation:
High
0.29 0.09 0.18 0.51 0.82
0.10 0.19 0.57 0.82
0.08 0.15 0.21 0.67
Family history of public
assistance: No
0.60 0.40 0.81 0.70 0.96
0.49 0.83 0.75 0.97
0.24 0.47 0.52 0.68
Mother’s education: >High
school
0.53 0.16 0.86 0.68 1.00
0.17 0.89 0.67 1.00
0.08 0.79 0.45 1.00
Father’s education: >High
school
0.41 0.11 0.69 0.39 0.93
0.12 0.71 0.37 0.94
0.08 0.57 0.08 0.91
Mother’s occupation: High 0.61 0.39 0.85 0.55 0.94
0.40 0.86 0.52 0.95
0.36 0.78 0.53 0.83
Father’s occupation: High
0.35 0.15 0.52 0.27 0.80
0.16 0.53 0.27 0.82
0.14 0.56 0.00 0.55
Mother’s Medicaid status:
No
0.52 0.18 0.93 0.44 0.97
0.27 0.94 0.50 0.97
0.03 0.45 0.30 0.71
Annual household income:
≥$50,000
0.31 0.05 0.65 0.15 0.73
0.09 0.65 0.19 0.74
0.00 0.25 0.06 0.46
Abbreviation: DM, downward mobility; SEP, socioeconomic position; SH, stable high; SL, stable low; SM, socioeconomic mobility;
UM, upward mobility.
54

Table 2.4. Odds ratios of preterm delivery conditional on latent class membership (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Overall
(n=3,019)

White/others
(n=2,276)

African American
(n=743)

N

OR

95% CI

N

OR

95% CI

N

OR

95% CI

Childhood SEP
Low
Middle
High (referent)

1159
1130
730

1.36
0.95
1.00

0.98, 1.90
0.62, 1.45

538
1059
679

1.21
0.84
1.00

0.79, 1.86
0.56, 1.26

579
97
67

1.48
1.18
1.00

0.53, 4.14
0.28, 5.07

Adulthood SEP
Low
Middle
High (referent)

1554
529
936

1.59*
1.24
1.00

1.18, 2.15
0.78, 1.98

1000
443
833

1.44*
1.29
1.00

1.02, 2.04
0.77, 2.17

359
226
158

1.44
0.67
1.00

0.78, 2.68
0.24, 1.89

0.42, 0.87
0.57, 1.33
0.44, 0.95

855
639
365
418

1.00
0.66*
1.10
0.75

0.43, 1.00
0.68, 1.77
0.49, 1.17

424
114
152
54

1.00
1.03
0.82
0.49

0.51, 2.07
0.35, 1.92
0.17, 1.46

0.80, 2.30

365

1.45

0.83, 2.56

152

1.65

0.46, 5.93

SM: childhood SEP & adulthood SEP
Static low (referent) 1372 1.00
715 0.60*
Upward mobile
475
Downward mobile
0.87
456 0.64*
Static high
Downward mobile
vs. Static high
(referent)

475

1.36

Abbreviation: CI, confidence interval; OR, odds ratio; SEP, socioeconomic position; SM, socioeconomic mobility.
*
P < 0.05.

55

Table 2.5. Fit statistics for latent class analysis (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Group
Childhood SEP
2
3
4
5

Log-likelihood Bayesian Information Criterion

-8431.3
-8380.5
-8366.2
-8365.0

Entropy

244.6
191.2
210.7
256.4

0.74
0.70
0.82
0.81

Adulthood SEP & neighborhood deprivation index
2
-11971.3
805.4
3
-11865.7
658.3
4
-11713.6
418.3
5
-11703.3
461.8
6
-11677.2
473.8
7
-11667.3
518.0

0.82
0.76
0.73
0.75
0.71
0.75

SM: childhood SEP & adulthood SEP & neighborhood deprivation index
-20295.5
2
4243.5
3
-19969.8
3696.5
4
-19755.6
3372.2
5
-19656.5
3278.0
6
-19588.7
3246.6
7
-19537.1
3247.6

0.85
0.80
0.76
0.75
0.70
0.70

Abbreviation: SEP, socioeconomic position; SM, socioeconomic mobility.

56

Table 2.6. Estimated item-response probabilities of the socioeconomic position indicators conditional on latent class
membership with distal outcome preterm delivery overall and by race/ethnicity (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004

Low
Childhood SEP
Latent class membership probabilities
Indicator
Grandmother’s education: >High
school
Grandfather’s education: >High
school
Grandmother’s occupation: High
Grandfather’s occupation: High
Family history of public assistance:
No

No.

Overall
(n=3,019)
Middle High

Latent Group
White/others
(n=2,276)
Low Middle High

1159
0.38

1130
0.38

730
0.24

0.36

0.17

0.34

0.74

0.11

0.27

0.29
0.68
0.29

0.11
0.23
0.11

0.11
0.98
0.17

0.97
0.93
0.76

0.08
0.09
0.11

0.60

0.43

0.67

0.89

0.47

%

57

538
1059
679
0.24
0.46
0.30
Item-response probabilities

African American
(n=743)
Low Middle High
579
0.78

97
0.13

67
0.09

0.72

0.22

0.87

0.93

0.13
0.93
0.16

0.94
0.94
0.77

0.13
0.41
0.11

0.01
0.97
0.12

0.98
0.82
0.64

0.72

0.90

0.30

0.50

0.69

Table 2.6. (cont’d)

Low
Adulthood SEP & NDI
Latent class membership probabilities
Indicator
Mother’s education: >High school
Father’s education: >High school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status: No
Annual household income:
≥$50,000
Neighborhood deprivation index:
< Medium

No.

Overall
(n=3,019)
Middle High

Latent Group
White/others
(n=2,276)
Low Middle High

1567
0.52

413
0.14

1039
0.34

%
0.53
0.41
0.61
0.35
0.52

0.20
0.09
0.40
0.13
0.22

0.83
0.77
0.72
0.56
0.46

0.93
0.77
0.89
0.60
0.98

0.31

0.08

0.06

0.76

0.12

0.06

0.50

0.32

0.43

0.80

0.47

0.47

58

1109
240
927
0.49
0.10
0.41
Item-response probabilities
0.26
0.87
0.94
0.12
0.95
0.79
0.41
0.76
0.90
0.16
0.70
0.62
0.33
0.55
0.99

African-American
(n=743)
Low Middle High
499
0.67

176
0.24

67
0.09

0.08
0.07
0.37
0.12
0.07

0.80
0.43
0.71
0.32
0.27

0.90
0.80
0.84
0.58
0.91

0.78

0.01

0.07

0.67

0.82

0.09

0.21

0.44

Table 2.6. (cont’d)
Overall
(n=3,019)
UM DM

SH

Latent Group
White/others
(n=2,276)
SL UM DM SH

1375

710

477

456

845

0.46

0.24

0.16

0.15

0.36

0.21

0.23

0.68

0.75

0.16

0.23

0.65

0.76

0.20

0.44

0.61

0.89

0.29

0.08

0.15

0.64

0.93

0.06

0.15

0.71

0.94

0.08

0.00

0.39

0.87

0.68

0.49

0.74

0.87

0.96

0.52

0.76

0.91

0.97

0.45

0.61

0.51

0.89

0.29

0.09

0.17

0.53

0.81

0.10

0.18

0.57

0.82

0.09

0.15

0.19

0.67

0.60

0.39

0.81

0.73

0.96

0.48

0.83

0.77

0.97

0.24

0.47

0.50

0.69

0.53

0.17

0.84

0.70

1.00

0.17

0.87

0.68

1.00

0.09

0.78

0.44

1.00

0.41
0.61
0.35

0.11
0.40
0.15

0.66
0.83
0.50

0.44
0.57
0.32

0.93
0.95
0.79

0.12
0.39
0.15

0.70
0.86
0.52

0.41
0.54
0.31

0.93
0.95
0.81

0.08
0.36
0.14

0.58
0.79
0.56

0.09
0.54
0.00

0.88
0.81
0.53

0.52

0.17

0.94

0.45

0.99

0.26

0.94

0.50

0.99

0.03

0.45

0.30

0.69

0.31

0.05

0.64

0.15

0.75

0.09

0.65

0.18

0.76

0.00

0.26

0.05

0.45

0.50

0.27

0.72

0.53

0.82

0.42

0.75

0.59

0.84

0.08

0.26

0.16

0.44

SL
SM: childhood SEP &
adulthood SEP & NDI
Latent class membership
probabilities
Indicator
Grandmother’s
education: >High school
Grandfather’s
education: >High school
Grandmother’s
occupation: High
Grandfather’s occupation:
High
Family history of public
assistance: No
Mother’s education: >High
school
Father’s education: >High
school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status:
No
Annual household income:
≥$50,000
Neighborhood deprivation
index: < Medium

No.

%

59

African American
(n=743)
SL UM DM SH

417

431

112

145

55

0.37 0.28 0.17 0.18
Item-response probabilities

0.58

0.15

0.20

0.07

629

384

Table 2.6. (cont’d)
Abbreviation: DM, downward mobility; NDI, neighborhood deprivation index; SEP, socioeconomic position; SH, stable high; SL,
stable low; SM, socioeconomic mobility; UM, upward mobility.

60

Table 2.7. Odds ratios of preterm delivery conditional on latent class membership (n=3,019), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Overall
(n=3,019)

Childhood SEP
Low
Middle
High (referent)

White/others
(n=2,276)

N

OR

95% CI

N

OR

95% CI

N

OR

95% CI

1159
1130
730

1.36
0.95
1.00

0.98, 1.90
0.62, 1.45

538
1059
679

1.21
0.84
1.00

0.79, 1.86
0.56, 1.26

579
97
67

1.48
1.18
1.00

0.53, 4.14
0.28, 5.07

1109
240
927

1.52*
1.15
1.00

1.08, 2.14
0.53, 2.49

499
176
67

1.42
1.23
1.00

0.57, 3.56
0.39, 3.85

431
112
145
55

1.00
1.04
0.80
0.50

0.52, 2.08
0.33, 1.96
0.17, 1.46

145

1.59

0.43, 5.88

Adulthood SEP & neighborhood deprivation index
1567 1.70* 1.26, 2.30
Low
413
Middle
1.22
0.65, 2.29
1039 1.00
High (referent)

SM: childhood SEP & adulthood SEP & neighborhood deprivation index:
845
Static low (referent) 1375 1.00
1.00
710 0.61* 0.43, 0.88
629
Upward mobile
0.66*
0.43, 1.00
477
384
Downward mobile
0.87
0.58, 1.32
1.07
0.67, 1.70
*
456 0.62
417
Static high
0.42, 0.92
0.73
0.47, 1.13
Downward mobile
vs. Static high
(referent)

African American
(n=743)

477

1.39

0.83, 2.36

384

1.47

0.85, 2.57

Abbreviation: CI, confidence interval; OR, odds ratio; SEP, socioeconomic position; SM, socioeconomic mobility.
*
P < 0.05.

61

CHAPTER 3
MATERNAL SOCIOECONOMIC MOBILITY AND PRENATAL DEPRESSION:
LATENT CLASS ANALYSIS

3.1 Introduction
Prenatal depression is a worldwide public health problem (Kessler 2003); Prevalence
estimates range from 8% to 20% in high income countries (Marcus et al. 2003, Gavin et al. 2005,
Evans et al. 2007, Vesga-López et al. 2008, Melville et al. 2010), and over 20% in low and
middle income countries (Faisal-Cury et al. 2009, Golbasi et al. 2010, Husain et al. 2011).
Prenatal depression is associated with postpartum depression and recurrent depression, which
have the potential to adversely affect the general health and well-being of mother and her
children (Bennett et al. 2004, Cents et al. 2013). Some studies report a link between prenatal
depression and increases in adverse pregnancy outcomes, e.g. stillbirth, preterm birth, and low
birth weight (Field, Diego, and Hernandez-Reif 2006, Evans et al. 2007, Rahman et al. 2007,
Wisner et al. 2009, Grote et al. 2010, Grigoriadis et al. 2013, Ncube et al. 2017). The prevalence
of perinatal depression is higher among African-American women compared to white women
(Dole et al. 2004, Nicholson et al. 2006, Li et al. 2008, Gavin et al. 2011, Mukherjee et al. 2016).
Maternal disadvantaged socioeconomic positions (SEP) in adulthood has been linked to a
greater risk of prenatal depression (Holzman et al.2006, Miyake et al. 2012, Bahk et al. 2015).
Measures of SEP have included occupation, education and household income, although the
results with different SEP measures are not always consistent in the general population (Lorant et
al. 2003). Childhood SEP has also been shown to be important risk factor for depression during

62

adulthood in the general population (Ritsher et al. 2001, Gilman et al. 2002, Harper et al. 2002).
However, few studies have examined mother’s childhood SEP and prenatal depression. In
addition, prior studies of non-pregnant populations have considered not only SEP but also
changes in SEP, i.e. socioeconomic mobility (SM), and found that upward SM and stable high
SEP were protective against depressive symptoms (Luo and Waite 2005, Nicklett and Burgard
2009, Hudson et al. 2013, Ro 2014, Dal Grande et al. 2015, Ward et al. 2016, Walsemann et al.
2017, Kwon et al. 2017). However, to our knowledge, no previous study has directly investigated
the association between maternal SM and depression in pregnancy.
Studies of non-pregnant women have found poorer mental health in association with
socioeconomic disadvantage during childhood (Ritsher et al. 2001, Gilman et al. 2002, Harper et
al. 2002) and during adulthood (Lorant et al. 2003, Hudson et al. 2012). These investigations
typically measured SEP at a single time point and used a single indicator, e.g. education,
occupation or family income (Nicklett and Burgard 2009, Owens and Jackson 2015, Ward et al.
2016). However, the multidimensional nature of SEP and the complexity of SEP transition from
one time point to another suggest that SEP and SM cannot be measured completely with one
indicator. Even in a study with multiple SEP indicators, a single index of SEP measure was
defined as the average of standardized SEP indicators (Luo and Waite 2005). Our study uses a
latent class analysis (LCA) for measuring SM in a dataset with individual-level data of maternal
childhood and adulthood SEP indicators, a novel approach for pregnancy studies.
Thus our study investigates associations between maternal SM from childhood to
adulthood and risk of prenatal depression in a Michigan community-based cohort study that is
racially and socioeconomically diverse. The availability of in-depth, individual-level data on
mothers and their childhood SEP provides a unique opportunity to study SM. We hypothesized
63

that: 1) women with lowest childhood SEP and lowest adulthood SEP will be at greatest risk of
prenatal depression; 2) women who experience upward SM from childhood to adulthood will
have a decreased risk of prenatal depression relative to women who maintain a lower SEP; and 3)
women who experience downward SM from childhood to adulthood will have a higher risk of
prenatal depression relative to women who maintain a higher SEP during childhood and
adulthood. Because some studies reported that the relation between SEP (or SM) and depression
varied by race/ethnicity (Luo and Waite 2005, Owens and Jackson 2015, Ward et al. 2016), we
also examine evidence for this effect modification.

3.2 Methods
Study population
This study included 3,019 pregnant women enrolled at 16-27 weeks’ gestation in the
prospective Pregnancy Outcomes and Community Health (POUCH) study (1998-2004) from five
Michigan communities. Eligibility criteria included: English-speaking, age >14 years, no preexisting diabetes, singleton pregnancy with no known congenital anomalies, and prenatal
maternal serum alpha-fetoprotein screening at 15-22 weeks of gestation. The detailed study
protocol can be found elsewhere (Holzman et al. 2001). Institutional review board approval for
this study was received from Michigan State University and all nine participating delivery
hospitals.
At enrollment each pregnant woman provided informed written consent, met with a study
nurse for a structured in-person interview, and completed a self-administered questionnaire that
collected information regarding demographics, lifestyle, psychosocial factors, medical history
64

and health status. Self-reported maternal race/ethnicity was dichotomized as white/others or
African American. The numbers of “others” race/ethnicity were small and their scores on the
prenatal depression symptoms measure were similar to that of whites, therefore these groups
were combined.
Measures
Childhood SEP and adulthood SEP
Childhood SEP measures included a woman’s self-report of her parents’ socioeconomic
indicators during her childhood, i.e. maternal mother’s and father’s highest level of education
attainment, maternal mother’s and father’s usual occupation, and receipt of public assistance by
the family. Adulthood SEP measures included maternal socioeconomic indicators at enrollment,
i.e. maternal and child’s father’s highest level of education, maternal and child’s father’s usual
occupation, Medicaid Insurance status, and maternal annual household income. SEP indicators
were modeled as binary variables, i.e. education (≤high school, >high school); usual occupation
status based on the United States Census Bureau’s 1990 Occupational Classification System
(“low”= service/blue collar, unemployed, and unknown; “high”= professional, manager,
technical, clerical/sales, homemaker, military, retired, and student); public assistance (yes/no);
Medicaid Insurance (yes/no); maternal household income (<$50,000 and ≥$50,000). “Unknown”
paternal and grandfather’s occupation was assigned as “low” occupation status because it
correlated with other low SEP indicators (Slaughter-Acey et al. 2016). Missing data for SEP
indicators were handled primarily through statistical methods (see statistical methods section
below). In separate models, a maternal neighborhood deprivation index (NDI) also was included
in adulthood SEP. The NDI was calculated using information from the 1986 Census and

65

followed a previously validated approach (Messer et al. 2006). NDI was dichotomized at the
medium of the study population.
Maternal SM
Maternal SM was defined by modeling both childhood and adulthood SEP indicators in
LCA. The SEP indicator variables were dichotomized the same as the childhood SEP and
adulthood SEP variables.
Prenatal depression
Prenatal depression symptoms were measured by the Center for Epidemiological
Studies – Depression (CES-D) scale (Radloff 1977) in the self-administered interview. The 20item CES_D is widely used in the general population to assess depressive symptoms during the
prior week. In the analysis, CES-D was dichotomized with a cutoff, i.e. > 16, frequently used to
indicate presence of depressive symptoms. Nine women did not complete the CES-D, leaving a
final sample size of 3,010.
Statistical methods
Initially we assessed maternal characteristics and SEP indicators overall and stratified by
CES-D score (low/high), and used chi-square tests to test for differences in proportions. We used
LCA to identify and describe distinct groups of childhood SEP indicators, adulthood SEP
indicators and SM from childhood to adulthood. LCA classifies individuals with similar SEP
characteristics into groups based on conditional probabilities (Collins and Lanza 2010). We
determined the optimal number of LCA groups by: 1) visualizing the plot of log-likelihoods of
each model and eliminating model where the log-likelihood did not show a substantial
improvement (Nylund, Asparouhov, and Muthén 2007); 2) using the Bayesian Information
66

Criterion (BIC), with a smaller BIC representing a better-fitted model (Schwarz 1978); 3)
evaluating the entropy, an indicator for class separation and precision, that measures how well an
individual fits into a specific class with values closer to 1 representing better fit (Ramaswamy et
al. 1993); and 4) considering the size of the latent classes, uniqueness of classes, and meaningful
interpretation of the response probability pattern for each class. As a next step, after the optimal
number of group was determined we used the model-based approach to LCA with distal outcome
to predict the risk of prenatal depression based on the latent class membership of exposure
variables (i.e. childhood SEP, adulthood SEP, SM) (Lanza, Tan, and Bray 2013).
Race/ethnicity was added as a grouping variable in LCA models to investigate
associations among SEP, SM and risk of CES-D within race/ethnic groups (white/others, African
American). Because SEP indicators may have different ranges and occur in different contexts
across race/ethnic groups, we computed group-specific LCA following the same methodology
described above. In another set of models, we added the NDI to adulthood SEP to examine its
contribution to identifying SEP associations with CES-D risk.
All statistical analyses were conducted using PROC LCA (Lanza et al. 2015) and its
corresponding distal outcome macro (Dziak et al. 2016) in SAS 9.4 (Statistical Analysis
Software, Cary, NC). The Proc LCA program uses full information maximum likelihood
estimation to cope with missing data in the SEP indicators.

67

3.3 Results
Sample characteristics
For the childhood SEP indictors, about 60% of grandmothers and 55% of grandfathers
did not graduate from high school. Approximately 32% of grandmothers and 71% of
grandfathers had a low-level occupation. About one third of families received public assistance
during the mother’s childhood period. For adulthood SEP indicators, 47% of mothers and 55%
of fathers did not graduate from high school; 39% of mothers and 65% of fathers had a low-level
occupation; 48% of pregnant women were insured by Medicaid; and 67% of families had an
annual income of <$50,000. In bivariate analyses, all childhood SEP indicators and all adulthood
SEP indicators were significantly related to CES-D in the direction hypothesized, i.e. lower SEP
linked to greater prevalence of a positive CES-D screen (Table 3.1).
Latent class profile
In the LCA of childhood SEP, a three-group solution was considered optimal according
to criteria described in our statistical methods section (Table 3.2). The three-group model had the
lowest BIC and substantial improvement in log-likelihood difference when moving from a threegroup to two-group model, compared to that from a four-group model to a three-group model.
The subgroups characterized by particular SEP indicators were interpreted and labelled
according to the parameter estimates. For childhood SEP, compared to other group, Group 1,
“low group”, (40% of women) grandmothers and grandfathers were less likely to graduate from
high school (22% and 11%, respectively), to have a high-level occupation (45% and 9%,
respectively), or to avoid public assistance during the mother’s childhood (28%) (Table 3). The
probabilities of positive childhood SEP indicators were greater in childhood SEP Group 2,

68

“middle group”, (35% of women), compared to other group, with the exception of grandfathers’
education (10%) and occupation (19%). Childhood SEP Group 3, “high group”, (25% of women)
showed high probabilities of all positive SEP indicators.
For adulthood SEP, a four-group model had the lowest BIC, however there was minimal
improvement when moving from a three-group to four-group model, compared to the substantial
log-likelihoods difference from a two-group to a three-group model. Additionally, the entropy
for the four-group model was smaller than that for the three-group model, indicating poor latent
group separation for the four-group model (Collins & Lanza, 2010). Based on additional
considerations of group prevalence and interpretability we chose the three-group model for
adulthood SEP (Table 2). For adulthood SEP, compared to other group, Group 1,“low group”,
(52% of women) mothers and fathers were less likely to graduate from high school (23% and
15%, respectively), to have a high-level occupation (40% and 16%, respectively), to have
medical insurance other than Medicaid (15%), or to have a family income > $50,000 (4%) (Table
3). The probabilities of SEP indicators in adulthood SEP Group 2, “middle group”, (18% of
women) varied; for example, fathers in this group had low probabilities of having a high-level
education (20%) and occupation (25%) but women were more likely to have high-level
occupations (73%) and medical insurance other than Medicaid (94%). About half of the families
had an annual household income exceeding $50,000 (54%). In adulthood SEP Group 3, “high
group”, (30% of women), all probabilities of positive SEP indicators were increased; about 64%
of these families had annual household incomes exceeding $50,000.
We used both childhood and adulthood SEP indicators for the SM LCA. The BIC values
decreased with increasing number of groups until 6 groups. After reviewing the plot of loglikelihoods of each model with specified group, improvements in model fit, size of the groups,
69

and interpretability we chose the four-group solution for modeling SM from childhood to
adulthood (Table 3.2). In SM from childhood to adulthood, Group 1, “static low group”, (45%
women) women had low-level childhood SEP and low-level adulthood SEP indicators (Table
3.3). SM Group 2, “upward mobile group”, (24% of women) showed some improvement in two
childhood indicators, i.e. probabilities of grandmother with high-level occupation (75%) and
avoiding public assistance (82%), and higher probabilities in all positive adulthood SEP
indicators. SM Group 3, “downward mobile group”, (16% of women) tended to have high-level
childhood SEP and moderate to low-level adulthood SEP. In this group the fathers’ education
and occupation were unlikely to be high-level (38% and 27%, respectively) and the probability
of household income >$50,000 was low (14%). SM Group 4, “static high group”, (15% of
women) had high-level SEP indicators throughout childhood and adulthood.
Latent class models also were examined for white/other women and African-American
women separately using the methodology described above. For each race/ethnic group, based on
fit indices and interpretability, we chose a three-group model for childhood SEP and adulthood
SEP indicators and a four-group model for SM. The race/ethnicity-specific item-response
probabilities for each indicator conditional on latent group membership are presented in Table
3.3. For the childhood SEP, the class percentages for white/others and African American,
respectively, were: low group 38%, 77%; middle group 31%, 13%; and high group 31%, 10%.
For the adulthood SEP, the group percentages for white/other and African American,
respectively, were: low group 44%, 69%; middle group 20%, 23%; and high group 36%, 8%. For
SM, the group percentages for white/other and African American, respectively, were: static low
37%, 59%; upward mobile 28%, 15%; downward mobile 16%, 19%; and static high 19%, 7%.

70

LCA with distal outcome: prenatal depression
Overall, the frequency of a positive prenatal CES-D screen in the low, middle, and high
childhood SEP groups was 49.8%, 25.2%, 21.1%, respectively. Similar percentages were
observed in the low, middle, and high adulthood SEP groups, i.e. 48.3%, 25.6%, and 15.1%,
respectively. Using the respective high SEP as the referent group, a positive prenatal CES-D
screen was more common in the low childhood SEP group (OR= 3.72, 95% 95% CI: 2.74, 5.04);
and the low adulthood SEP group, (OR=5.26, 95% CI: 4.17, 6.64) (Table 3.4).
A positive prenatal CES-D screen was observed in 48.4%, 17.4%, 38.7%, and 13.2% of
women who experienced static low SM, upward SM, downward SM, and static high SM,
respectively (Table 3.4). Women in the upward SM group were less likely to have a positive
prenatal CES-D screen compared to women in the static low SM group (OR=0.22, 95% CI: 0.17,
0.29). By contrast, women in the downward SM group were at greater risk of a positive prenatal
CES-D screen compared to women with static high SM, (OR=4.13, 95% CI: 2.75, 6.22). In a
separate set of analyses we used race/ethnic-specific LCA models. Among the white/others
women, patterns of SEP (childhood and adulthood) and SM in relation to a positive prenatal
CES-D screen mirrored those observed overall and described above (Table 3.4). Among AfricanAmerican women, the associations between low SEP, SM and a positive prenatal CES-D screen
were more modest than those found in white/other women. Perhaps due to smaller sample sizes,
only low and middle adulthood SEP groups had a significantly greater prevalence of a positive
prenatal CES-D screen.
We incorporated NDI in the adulthood SEP to examine its added effect on SEP
associations with prenatal CES-D screen. Table 3.5 summaries the LCA model fit statistics. In

71

consideration of a combination of fit statistics and interpretability, we selected a three-group
model for adulthood SEP with NDI and a four-group model for SM that included childhood SEP
and adulthood SEP with NDI. The item-response probabilities for each SEP indicator conditional
on latent class membership of adulthood SEP with NDI and SM are shown in Table 3.6. The
latent class profiles were similar to those in adulthood SEP without NDI. When we assessed the
associations between latent class membership and a positive prenatal CES-D screen (Table 3.7),
NDI did not add appreciably to our original findings.

3.4 Discussion
Overall, we found that a positive screen for depression (CES-D >16) in mid-pregnancy
was more common in women with low childhood and adulthood SEP, or a downward SM
trajectory, and was less common in women who experienced upward SM. These relations
between SEP, SM and depressive symptoms were stronger in white/other women as compared to
African-American women. Links between SM and depression have been reported for nonpregnant women (Tiffin, Pearce, and Parker 2005, Hudson et al. 2013, Dal Grande et al. 2015,
Walsemann et al. 2017, Kwon et al. 2017), but, to our knowledge, have not been assessed
previously in pregnant women.
The relationships between childhood SEP, adulthood SEP and mental health are likely
varied and complex with a multitude of explanations (Stansfeld et al. 2011).

The ‘social

causation’ theory postulates that low SEP increases the risk of a mental health disorder, mainly
depression and anxiety. This may be explained by critical period models or by pathway models
(Ljung and Hallqvist 2006). In a critical period model, the effect of low SEP in childhood on
72

biological systems remains latent until adulthood and is independent of adult exposure to
adversity. In a pathway model, low childhood SEP increases the likelihood of low adulthood
SEP and this life course trajectory escalates the risk of mental disorders. There is also the ‘health
selection’ theory which asserts that childhood mental health disorders increase the probability of
low adult SEP.
Our findings are consistent with these theories explaining links between SEP and mental
health. In support of the social mobility model, upward SM may decrease the accumulation of
adverse exposures that create ‘wear and tear’ to the body (allostatic load) (McEwen 1998,
McEwen 2000). Pregnant women who experienced upward SM (compared to women with stable
low childhood and adulthood SEP) may have: 1) greater access to high-quality food, reliable
housing, a safer living environment, and higher-quality health care; 2) more support for healthy
lifestyle choices (diet, physical activity, no smoking, no alcohol use); and 3) greater opportunity
for psychological health (less stress, more social integration). Thus, the lower levels of ‘wear and
tear’ might be protective against depression in pregnant women who experienced upward SM
(Field, Diego, and Hernandez-Reif 2006). Our observation that a positive depression screen is
less likely in women with upward SM and more likely in women with downward SM points to
adulthood SEP is salient and perhaps a ‘critical window.’ The SM pattern and maternal
depression link also fits with the ‘health selection’ theory, i.e. women with mental health
problems may be less likely to obtain or maintain high SEP, though, without measures of
depression in childhood, we can’t complete this theory test in this study.
Previous studies among non-pregnant population indicate a consistent inverse
relationship between SEP and depression among whites, however, evidence for African
Americans is mixed (Lorant et al. 2003, Williams et al, 2007, Gavin et al. 2010). For example,
73

some studies observed a higher prevalence of depression among African Americans with lower
SEP (Banks & Kohn-Wood 2002, Bromberger et al. 2004, Roxburgh 2009); however, other
studies find no significant relationship between SEP and depression among African Americans
(Gavin et al. 2010, Williams et al. 2007). Even studies examining multiple SEP indicators find
no significant association between SEP and depression among African Americans (Rodriguez et
al. 1999, Hudson et al. 2012). In our study, we observed that the relations between SEP, SM and
depressive symptoms were stronger in white/other women as compared to African-American
women. Fewer African-American women experienced SM in this study, making it difficult to
calculate accurate race/ethnicity-specific estimates of the SM-CES-D association in this group.
To the best of our knowledge, this study is the first to use the model-based approach to
LCA with distal outcome to elucidate complex relationships between SM and the risk of prenatal
depression among pregnant women in the United States. Typically, traditional classify-analyze
strategies have been used to predict distal outcome from latent class membership. In the first
classification step, individuals were assigned to latent classes based on probabilities. In a second
analysis step, the distal outcome was predicted from the latent class membership which was
treated as observed. For these traditional strategies the results are biased to the extent that there is
classification error. However, our model-based approach models measurement error (Collins &
Lanza, 2010) and produces less biased estimates for the probability of prenatal depression
conditional on latent class membership. Moreover, use of the full-information expectationmaximization algorithm in LCA may include all participants who responded to at least one SEP
indicator (Collins & Lanza, 2010). Most previous studies examining the impact of SM on
depression used a single SEP indicator to measure SM (Nicklett and Burgard 2009, Owens and
Jackson 2015, Ward et al. 2016). Our study used multiple SEP indicators for childhood SEP and
74

adulthood SEP measures, which may minimize misclassification bias. In addition, this study
enrolled pregnant women with diverse SEP, which allowed us to observe a full range of SM.
An important limitation in most SM research pertains to the inferences. When we study
individuals we don’t randomly assign SM, therefore it is difficult to disentangle unique attributes
of individuals who climb the SEP ladder (self-selection) versus the extrinsic benefits of a better
environment. In addition, all SEP indicators were self-reported and recall bias cannot be ruled
out. Our software and modeling strategy, i.e. model-based approach to LCA with distal outcome,
could not accommodate adjustment for other covariates (potential confounders), which leaves
open the possibility of unmeasured confounding. While we stratified on race/ethnicity, we
considered maternal age and parity as important covariates. In sensitivity analyses we also
looked to see whether our results would vary by maternal age and parity, which they did not.
Comprehensive modeling of childhood and adulthood SEP indicators sheds light on risks
for prenatal depression. Our study suggests that low SEP in childhood alone does not raise the
risk of a positive depression screen in pregnancy as evidenced by the decreased risk for women
with upward SM. The explanations for these observations may be multiple, but overall the
findings lead to an optimistic conclusion: increasing opportunities for upward SM could help
lower the prevalence of maternal depression. Women with low SEP in childhood and adulthood
are the group most likely to screen positive for depression in pregnancy. Clinics and providers
serving these pregnant women need adequate capacity to conduct depression screens and carry
out appropriate referrals and interventions.

75

Table 3.1. Maternal Characteristics and Socioeconomic Position Indicators (n=3,010),
Pregnancy Outcomes and Community Health Study, Michigan, 1998-2004
Overall

CES-D<16

CES-D≥16

(n=3,010)
No.
%

(n=1,984)
No.
%

(n=1,026)
No.
%

<0.0001

Race/ethnicity
White/others
African American

2268 75.3
742 24.7

1610 81.1
374 18.9

658
368

64.1
35.9

Childhood SEP
Grandmother’s education
Missing
≤ high school
>high school

125 4.2
1806 60.0
1079 35.8

53
2.7
1183 59.6
748 37.7

72
623
331

7.0
60.7
32.3

Grandfather’s education
Missing
≤ high school
>high school

475 15.8
1656 55.0
879 29.2

238 12.0
1066 53.7
680 34.3

237
590
199

23.1
57.5
19.4

Grandmother’s usual
occupationb
Low occupational status
High occupation status
Grandfather’s usual
occupationb
Low occupational status
High occupation status
Family history of public
assistance
Missing
Yes
No

P
valuea

0.03

<0.0001

<0.0001
954 31.7
2056 68.3

566 28.5
1418 71.5

388
638

37.8
62.2
<0.0001

2138 71.0
872 29.0

1340 67.5
644 32.5

798
228

77.8
22.2
<0.0001

109 3.6
1079 35.8
1822 60.5

77
3.9
584 29.4
1323 66.7

76

32
495
499

3.1
48.3
48.6

Table 3.1. (cont’d)
Overall

CES-D<16

CES-D≥16

(n=3,010)
No.
%

(n=1,984)
No.
%

(n=1,026)
No. %

Adulthood SEP
Maternal education
≤ High school
>High school

1402 46.6
1608 53.4

779 39.3
1205 60.7

623
403

60.7
39.3

Father’s education
Missing
≤ High school
>High school

136 4.5
1652 54.9
1222 40.6

55
963
966

2.8
48.5
48.7

81
689
256

7.9
67.2
25.0

Maternal usual occupationb
Low occupational status
High occupation status

1175 39.0
1835 61.0

675 34.0
1309 66.0

500
526

48.7
51.3

Father’s usual occupationb
Low occupational status
High occupation status
Mother’s Medicaid insured
Missing
Yes
No
Annual household income
Missing
< $50,000
≥ $50,000
Neighborhood deprivation
index
Missing
≥ Medium
< Medium

P
valuea
<0.0001

<0.0001

<0.0001

<0.0001
1945 64.6
1065 35.4

1175 59.2
809 40.8

770
256

75.1
25.0
<0.0001

2
0.1
1451 48.2
1557 51.7

1
0.1
780 39.3
1203 60.6

1
671
354

0.1
65.4
34.5
<0.0001

89
3.0
2002 66.5
919 30.5

45
2.3
1183 59.6
756 38.1

44
819
163

4.3
79.8
15.9
<0.0001

5
0.2
1503 49.9
1502 49.9

5
0.3
879 44.3
1100 55.4

0
624
402

0.0
60.8
39.2

Abbreviation: SEP, socioeconomic position.
a
P value for chi-square test between CES-D<16 and CES-D≥16 groups.
b
Low occupation status includes service/blue collar, unemployed, and unknown; high
occupation status includes professional, manager, technical, clerical/sales, homemaker, and other
(military, retired, student).
77

Table 3.2. Fit statistics for latent class analysis (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Group

Log-likelihood

Bayesian Information Criterion

Entropy

Childhood SEP
2
3
4
5

-8431.3
-8380.5
-8366.2
-8365.0

244.6
191.2
210.7
256.4

0.74
0.70
0.82
0.81

Adulthood SEP
2
3
4
5
6

-10102.4
-9978.9
-9896.7
-9876.6
-9869.4

609.1
418.3
309.9
325.8
367.6

0.81
0.78
0.72
0.74
0.74

3160.4
2637.1
2342.7
2300.2
2263.3
2269.8

0.84
0.80
0.76
0.74
0.73
0.69

SM: childhood SEP & adulthood SEP
2
-18443.8
3
-18134.1
4
-17938.8
-17869.5
5
-17803.0
6
-17758.2
7

Abbreviation: SEP, socioeconomic position; SM, socioeconomic mobility.

78

Table 3.3. Estimated item-response probabilities of the socioeconomic position indicators conditional on latent class
membership with distal outcome prenatal depression overall and by race/ethnicity (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004

Low
Childhood SEP
Latent class membership probabilities
Indicator
Grandmother’s education: >High
school
Grandfather’s education: >High
school
Grandmother’s occupation: High
Grandfather’s occupation: High
Family history of public assistance:
No

No.

Overall
(n=3,010)
Middle High

Latent Group
White/others
(n=2,268)
Low Middle High

1211
0.40

1039
0.35

760
0.25

0.36

0.22

0.27

0.74

0.19

0.24

0.29
0.68
0.29

0.11
0.45
0.09

0.10
0.77
0.19

0.95
0.93
0.75

0.11
0.51
0.09

0.60

0.28

0.85

0.88

0.35

%

79

866
690
712
0.38
0.31
0.31
Item-response probabilities

African American
(n=742)
Low Middle High
569
0.77

98
0.13

75
0.10

0.71

0.22

0.85

0.91

0.08
0.81
0.21

0.94
0.94
0.75

0.11
0.40
0.11

0.01
0.98
0.14

0.98
0.78
0.60

0.99

0.90

0.30

0.49

0.67

Table 3.3. (cont’d)

Low
Adulthood SEP
Latent class membership probabilities
Indicator
Mother’s education: >High school
Father’s education: >High school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status: No
Annual household income:
≥$50,000

No.

Overall
(n=3,010)
Middle High

1565
0.52

542
0.18

903
0.30

%
0.53
0.41
0.61
0.35
0.52

0.23
0.15
0.40
0.16
0.15

0.68
0.20
0.73
0.25
0.94

0.97
0.99
0.89
0.74
0.90

0.31

0.04

0.54

0.64

80

Latent Group
White/others
(n=2,268)
Low Middle High
998
454
816
0.44
0.20
0.36
Item-response probabilities
0.25
0.70
0.97
0.17
0.22
0.99
0.39
0.75
0.90
0.18
0.24
0.76
0.22
0.98
0.92
0.06

0.58

0.66

African American
(n=742)
Low Middle High
511
0.69

168
0.23

63
0.08

0.08
0.07
0.37
0.12
0.08

0.87
0.44
0.71
0.32
0.28

0.91
0.80
0.87
0.58
0.92

0.010

0.06

0.70

Table 3.3. (cont’d)

SL

Overall
(n=3,010)
UM DM

Latent Group
White/others
(n=2,268)
SL UM DM SH

SH

African American
(n=742)
SL UM DM SH

SM: childhood SEP &
No. 1355 722 482
452
840 631 370 426
434 115 140
54
adulthood SEP
Latent class membership
probabilities
0.45 0.24 0.16 0.15
0.37 0.28 0.16 0.19
0.59 0.15 0.19 0.07
Indicator
%
Item-response probabilities
Grandmother’s
education: >High school
0.36 0.20 0.24 0.68 0.76
0.15 0.23 0.65 0.76
0.21 0.44 0.61 0.89
Grandfather’s
education: >High school
0.29 0.07 0.16 0.63 0.94
0.06 0.16 0.70 0.95
0.08 0.00 0.41 0.86
Grandmother’s
occupation: High
0.68 0.49 0.75 0.86 0.97
0.52 0.77 0.91 0.97
0.45 0.60 0.52 0.89
Grandfather’s occupation:
High
0.29 0.09 0.18 0.51 0.82
0.10 0.19 0.55 0.82
0.08 0.15 0.22 0.66
Family history of public
assistance: No
0.60 0.39 0.82 0.70 0.95
0.49 0.83 0.75 0.96
0.24 0.47 0.52 0.68
Mother’s education: >High
school
0.53 0.16 0.85 0.67 1.00
0.17 0.88 0.66 1.00
0.08 0.79 0.46 1.00
Father’s education: >High
school
0.41 0.11 0.68 0.38 0.93
0.12 0.71 0.36 0.93
0.09 0.57 0.08 0.92
Mother’s occupation: High 0.61 0.39 0.84 0.55 0.94
0.40 0.86 0.52 0.94
0.36 0.77 0.54 0.83
Father’s occupation: High
0.35 0.15 0.51 0.27 0.79
0.16 0.53 0.26 0.82
0.14 0.55 0.00 0.55
Mother’s Medicaid status:
No
0.52 0.18 0.93 0.43 0.97
0.26 0.94 0.50 0.96
0.03 0.44 0.30 0.71
Annual household income:
≥$50,000
0.31 0.05 0.65 0.14 0.73
0.09 0.66 0.18 0.73
0.00 0.25 0.06 0.46
Abbreviation: DM, downward mobility; SEP, socioeconomic position; SH, stable high; SL, stable low; SM, socioeconomic mobility;
UM, upward mobility.
81

Table 3.4. Odds ratios of prenatal depression conditional on latent class membership (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Overall
(n=3,010)

White/others
(n=2,268)

African American
(n=742)

N

OR

95% CI

N

OR

95% CI

N

OR

95% CI

Childhood SEP
Low
Middle
High (referent)

1211
1039
760

3.72*
1.27
1.00

2.74, 5.04
0.86, 1.87

866
690
712

2.90*
1.27
1.00

2.18, 3.87
0.84, 1.92

569
98
75

1.73
0.99
1.00

0.94, 3.19
0.41, 2.37

Adulthood SEP
Low
Middle
High (referent)

1565
542
903

5.26*
1.94*
1.00

4.17, 6.64
1.37, 2.74

998
454
816

5.15*
2.14*
1.00

3.94, 6.74
1.47, 3.13

511
168
63

3.54*
2.76*
1.00

1.65, 7.60
1.14, 6.67

0.17, 0.29
0.50, 0.91
0.12, 0.23

840
631
370
426

1.00
0.25*
0.74
0.18*

0.19, 0.34
0.53, 1.04
0.13, 0.26

434
115
140
54

1.00
0.65
0.52*
0.35*

0.38, 1.13
0.27, 0.98
0.18, 0.69

2.75, 6.22

370

4.08*

2.62, 6.36

140

1.47

0.79, 4.40

SM: childhood SEP & adulthood SEP
Static low (referent) 1355 1.00
722 0.22*
Upward mobile
482 0.67*
Downward mobile
452 0.16*
Static high
Downward mobile
vs. Static high
(referent)

482

4.13*

Abbreviation: CI, confidence interval; OR, odds ratio; SEP, socioeconomic position; SM, socioeconomic mobility.
*
P < 0.05.

82

Table 3.5. Fit statistics for latent class analysis (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Group
Childhood SEP
2
3
4
5

Log-likelihood Bayesian Information Criterion

-8431.3
-8380.5
-8366.2
-8365.0

Entropy

244.6
191.2
210.7
256.4

0.74
0.70
0.82
0.81

Adulthood SEP & neighborhood deprivation index
2
-11971.3
805.4
3
-11865.7
658.3
4
-11713.6
418.3
5
-11703.3
461.8
6
-11677.2
473.8
7
-11667.3
518.0

0.82
0.76
0.73
0.75
0.71
0.75

SM: childhood SEP & adulthood SEP & neighborhood deprivation index
-20295.5
2
4243.5
3
-19969.8
3696.5
4
-19755.6
3372.2
5
-19656.5
3278.0
6
-19588.7
3246.6
7
-19537.1
3247.6

0.85
0.80
0.76
0.75
0.70
0.70

Abbreviation: SEP, socioeconomic position; SM, socioeconomic mobility.

83

Table 3.6 Estimated item-response probabilities of the socioeconomic position indicators conditional on latent class
membership with distal outcome prenatal depression overall and by race/ethnicity (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004

Low
Childhood SEP
Latent class membership probabilities
Indicator
Grandmother’s education: >High
school
Grandfather’s education: >High
school
Grandmother’s occupation: High
Grandfather’s occupation: High
Family history of public assistance:
No

No.

Overall
(n=3,010)
Middle High

Latent Group
White/others
(n=2,268)
Low Middle High

1211
0.40

1039
0.35

760
0.25

0.36

0.22

0.27

0.74

0.19

0.24

0.29
0.68
0.29

0.11
0.45
0.09

0.10
0.77
0.19

0.95
0.93
0.75

0.11
0.51
0.09

0.60

0.28

0.85

0.88

0.35

%

84

866
690
712
0.38
0.31
0.31
Item-response probabilities

African American
(n=742)
Low Middle High
569
0.77

98
0.13

75
0.10

0.71

0.22

0.85

0.91

0.08
0.81
0.21

0.94
0.94
0.75

0.11
0.40
0.11

0.01
0.98
0.14

0.98
0.78
0.60

0.99

0.90

0.30

0.49

0.67

Table 3.6. (cont’d)

Low
Adulthood SEP & NDI
Latent class membership probabilities
Indicator
Mother’s education: >High school
Father’s education: >High school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status: No
Annual household income:
≥$50,000
Neighborhood deprivation index:
< Medium

No.

Overall
(n=3,010)
Middle High

Latent Group
White/others
(n=2,268)
Low Middle High

1505
0.50

602
0.20

903
0.30

%
0.53
0.41
0.61
0.35
0.52

0.23
0.15
0.41
0.17
0.12

0.64
0.23
0.69
0.24
0.92

0.98
0.98
0.90
0.75
0.91

0.31

0.03

0.48

0.66

0.12

0.06

0.50

0.23

0.64

0.98

0.47

0.47

85

1100
235
933
0.49
0.10
0.41
Item-response probabilities
0.26
0.87
0.94
0.12
0.96
0.78
0.41
0.76
0.90
0.16
0.70
0.62
0.33
0.55
0.99

African American
(n=742)
Low Middle High
505
0.68

173
0.23

64
0.09

0.09
0.07
0.37
0.12
0.07

0.82
0.45
0.70
0.33
0.28

0.90
0.79
0.86
0.57
0.92

0.78

0.01

0.07

0.68

0.82

0.09

0.21

0.44

Table 3.6. (cont’d)
Overall
(n=3,010)
UM DM

SH

Latent Group
White/others
(n=2,268)
SL UM DM SH

1355

722

482

452

830

0.45

0.24

0.16

0.15

0.36

0.21

0.22

0.68

0.75

0.15

0.22

0.65

0.75

0.22

0.44

0.63

0.89

0.29

0.08

0.15

0.63

0.92

0.06

0.15

0.69

0.93

0.08

0.00

0.44

0.87

0.68

0.49

0.74

0.88

0.96

0.52

0.76

0.91

0.96

0.45

0.62

0.51

0.89

0.29

0.09

0.16

0.52

0.81

0.10

0.17

0.56

0.81

0.09

0.15

0.22

0.67

0.60

0.39

0.81

0.73

0.95

0.48

0.83

0.77

0.96

0.25

0.48

0.52

0.69

0.53

0.17

0.83

0.69

1.00

0.17

0.86

0.68

1.00

0.10

0.78

0.47

1.00

0.41
0.61
0.35

0.11
0.40
0.15

0.65
0.83
0.49

0.43
0.57
0.31

0.93
0.94
0.79

0.12
0.39
0.16

0.70
0.85
0.51

0.40
0.54
0.30

0.93
0.95
0.81

0.08
0.36
0.14

0.58
0.78
0.55

0.09
0.55
0.00

0.89
0.82
0.54

0.52

0.17

0.93

0.45

0.98

0.25

0.94

0.50

0.98

0.04

0.45

0.31

0.71

0.31

0.04

0.64

0.15

0.75

0.08

0.66

0.18

0.76

0.00

0.26

0.06

0.45

0.50

0.26

0.72

0.53

0.82

0.41

0.75

0.59

0.83

0.08

0.26

0.17

0.44

SL
SM: childhood SEP &
adulthood SEP & NDI
Latent class membership
probabilities
Indicator
Grandmother’s
education: >High school
Grandfather’s
education: >High school
Grandmother’s
occupation: High
Grandfather’s occupation:
High
Family history of public
assistance: No
Mother’s education: >High
school
Father’s education: >High
school
Mother’s occupation: High
Father’s occupation: High
Mother’s Medicaid status:
No
Annual household income:
≥$50,000
Neighborhood deprivation
index: < Medium

No.

%

86

African American
(n=742)
SL UM DM SH

424

451

112

124

55

0.36 0.28 0.17 0.19
Item-response probabilities

0.61

0.15

0.17

0.07

625

389

Table 3.6. (cont’d)
Abbreviation: DM, downward mobility; NDI, neighborhood deprivation index; SEP, socioeconomic position; SH, stable high; SL,
stable low; SM, socioeconomic mobility; UM, upward mobility.

87

Table 3.7. Odds ratios of prenatal depression conditional on latent class membership (n=3,010), Pregnancy Outcomes and
Community Health Study, Michigan, 1998-2004
Overall
(n=3,010)

Childhood SEP
Low
Middle
High (referent)

White/others
(n=2,268)

N

OR

95% CI

N

OR

95% CI

N

OR

95% CI

1211
1039
760

3.72*
1.27
1.00

2.74, 5.04
0.86, 1.87

866
690
712

2.90*
1.27
1.00

2.18, 3.87
0.84, 1.92

569
98
75

1.73
0.99
1.00

0.94, 3.19
0.41, 2.37

1100
235
933

4.61*
2.14*
1.00

3.58, 5.93
1.30, 3.52

505
173
64

3.45*
2.73*
1.00

1.58, 7.53
1.10, 6.79

451
112
124
55

1.00
0.64
0.47*
0.37*

0.37, 1.10
0.23, 0.98
0.19, 0.71

124

1.29

0.50, 3.32

Adulthood SEP & neighborhood deprivation index
1505 5.65* 4.42, 7.22
Low
602 2.14* 1.52, 3.01
Middle
903
High (referent)
1.00

SM: childhood SEP & adulthood SEP & neighborhood deprivation index:
830
Static low (referent) 1355 1.00
1.00
722 0.22* 0.17, 0.29
625
Upward mobile
0.25*
0.18, 0.34
482 0.63* 0.47, 0.84
389
Downward mobile
0.70*
0.50, 0.97
452 0.15* 0.11, 0.21
424
Static high
0.17*
0.12, 0.25
Downward mobile
vs. Static high
(referent)

African American
(n=742)

482

4.10*

2.72, 6.17

389

3.99*

2.57, 6.21

Abbreviation: CI, confidence interval; OR, odds ratio; SEP, socioeconomic position; SM, socioeconomic mobility.
*
P < 0.05.

88

CHAPTER 4
SUMMARY

4.1 Major findings
The two studies in this dissertation have explored different outcomes of maternal
socioeconomic mobility using data from the POUCH Study. First, we investigated the
association between socioeconomic mobility and the risk of PTD. We used latent class approach
with relevant multiple indicators and identified constructs of pregnant women’s childhood SEP,
adulthood SEP, and SM from childhood to adulthood in association with risk of PTD. We found
that women with low childhood SEP and low adulthood SEP were at greatest risk of PTD
compared to women who experienced SM or had high childhood and adulthood SEP; and
upward socioeconomic mobility among pregnant women was associated with decreased risk of
PTD compared to women with low childhood and adulthood SEP. This SM advantage was true
in white/other women. In chapter 2, we also confirmed previous observations that disadvantaged
SEP at adulthood was associated with a greater likelihood of PTD. In a previous publication
from the POUCH Study, decreased risk of delivering a small for gestational age baby was
observed among pregnant women with upward SM (Slaughter-Acey et al. 2016).
Next, we assessed the relationship between SM from childhood to adulthood and the risk
of prenatal depression. A positive screen for depression (CES-D >16) in mid-pregnancy was
more common in women with low childhood and adulthood SEP or a downward SM trajectory
than women with upward SM or high childhood and adulthood SEP; and was less common in
women who experienced upward SM than women with low childhood and adulthood SEP. These

89

relations between SEP, SM and depressive symptoms were stronger in white/other women as
compared to African-American women.
Previous studies have used predominately ecological measures of SEP, and only a few of
these have considered SM in relation to PTD or prenatal depression. Our study is novel in
several ways: it includes in-depth, individual-level and ecological level data to measure SM and
it uses a latent class analysis with distal outcome, i.e. PTD or prenatal depression. In addition,
the sample is large for this type of in-depth data and is racially and socioeconomically diverse.
Our Michigan, community-based cohort was obtained by sampling from 52 prenatal care clinics,
thereby avoiding biases that can arise in samples from select inner-city clinics and
academic/teaching hospitals. While vital data, i.e. birth files, and other administrative data can
provide large numbers of pregnant women, these data sources lack the granularity of childhood
and adulthood SEP measures that is obtained in this study by carefully querying women during
pregnancy.

4.2 Public health significance
Our study supports some conceptual models that explicate how life course SEP affect
pregnancy outcomes and mental health. In chapter 2, our findings support the social mobility
model that upward SM remediates the negative effects of earlier disadvantage SEP and
improving women’s SEP from childhood to adulthood may improve pregnancy outcomes among
white/other women. Our results also imply that the childhood SEP is not deterministic for
women’s future risk of PTD. In chapter 3, our findings are consistent with these theories
explaining links between SEP and mental health. Our observation that a positive depression
90

screen is less likely in women with upward SM and more likely in women with downward SM
points to adulthood SEP is salient and perhaps a ‘critical window.’ The SM pattern and maternal
depression link also fits with the ‘health selection’ theory, i.e. women with mental health
problems may be less likely to obtain or maintain high SEP, though, without measures of
depression in childhood, we can’t complete this theory test in this study.
Our study suggests that programs or policies to increase opportunities for upward SM
may play an essential part in women’s reproductive and mental health. Our study provided
evidence that PTD and prenatal depression are shaped by socioeconomic exposures across the
life course and childhood disadvantages might be buffered by advantaged experiences in later
life. From a life course perspective, disparities in birth outcomes are the consequences of not
only differential exposures during pregnancy, but perhaps more important differential
accumulation of risk and protective factors across the life course. Therefore, in order to reduce
the risk of adverse pregnancy outcomes and mental illness, it is necessary to improve women’s
SEP not only during pregnancy, but before and between pregnancies and indeed, across their
entire life course. Previous literature indicated that education may be a more precise indictor of
SEP across the life course (Morgen et al. 2008). In addition, higher education has been effective
for poor people to find opportunities to transform their SEP. Therefore, education policies are
necessary to help promote equal opportunity for children, no matter their socioeconomic
background. Moreover, good education policies are critical for reducing health inequalities
during both childhood and adulthood. It is challenging to change health across the entire
distribution of SEP, particularly, when the current generation may be less socially mobile than
prior generations. Hence, a wide range of national education policies may promote to improve
the quality of and access to education and potentially reduce health inequalities (Heveman &
91

Smeeding, 2006). Accordingly, improving pregnant women’s socioeconomic circumstances
could be a vital intervention, particularly for those with adverse SEP in childhood and adulthood.
As a result, some policies may target low income populations to help achieve financial selfsufficiency and provide a financial safety net for basic needs. The United States has high rates of
PTD relative to other high income countries; this is true when we consider all births or when we
look only at births to non-Hispanic white women living in the United States. Our findings are
interpreted with caution but support suggestions that policies or programs to advance women’s
upward SM could help decrease rates of PTD in the United States.
In spite of growing evidence examining SM and depression later in life, no study has
investigated the association among pregnant women or taken into account the multi-dimensional
aspect of SEP. Consequently, the findings of this dissertation fill an essential gap in the literature.
Latent class analysis was used in this dissertation to examine the effect of SM across the life
course on pregnancy outcome and prenatal depression and the findings would benefit the
pregnant women in the United States. Findings from the second study in this dissertation suggest
that low SEP in childhood alone does not raise the risk of a positive depression screen in
pregnancy as evidenced by the decreased risk for women with upward SM. The explanations for
these observations may be multiple, but overall the findings lead to an optimistic conclusion;
increasing opportunities for upward SM could help lower the prevalence of maternal depression.
Women with low SEP in childhood and adulthood are the group most likely to screen positive
for depression in pregnancy. Clinics and providers serving these pregnant women need adequate
capacity to conduct depression screens and carry through with appropriate referrals and
interventions. In addition, inequalities in maternal depression may affect health disparities in
later life, so it is imperative to have early detection and treatment for depressive mothers with
92

low SEP (Bahk et al. 2015). Further research is needed to understand the specific elements
accompanying SM that are protective for PTD or prenatal depression and the best strategies for
increasing SM among all race/ethnic groups.

93

BIBLIOGRAPHY

94

BIBLOGRAPHY

Alang SM. Racial variations in the effects of structural and psychological factors on depressive
symptoms : a structural equation modeling approach. Mental Health & Prevention. 2014; 2: 2–10.
Allen, MC. Neurodevelopmental outcomes of preterm infants. Current Opinion in Neurology.
2008; 21(2):123-8.
Astone NM, Misra D, Lynch C. The effect of maternal socio-economic status throughout the
lifespan on infant birthweight. Paediatric and Perinatal Epidemiology. 2007; 21(4):310–318.
Auger N, Giraud J, Daniel M. The joint influence of area income, income inequality, and
immigrant density on adverse birth outcomes: A population-based study. BMC Public Health.
2009; 9(1).
Bahk J, Yun SC, Kim YM, Khang YH. Changes in the relationship between socioeconomic
position and maternal depressive symptoms: results from the Panel Study on Korean Children
(PSKC). Maternal and Child Health Journal. 2015; 19(9): 2057–65.
Banks KH, Kohn-Wood LP. Gender, ethnicity and depression: intersectionality in mental health
research with African American women. African American Research Perspectives. 2002; 8(1):
174-200.
Barker DJ, Osmond C, Simmonds SJ, & Weild A. The relationship of small head circumference
and thinness at birth to death from cardiovascular disease in adult life. BMJ. 1993; 306 (6875).
Basso O, Olsen J, Johansen AMT, Christensen K. Change in social status and risk of low birth
weight in Denmark: Population based cohort study. BMJ. 1997; 315(7121):1498–1502.
Behrman RE, Butler AS. Preterm birth: causes, consequences, and prevention. Institute of
Medicine (US) committee on understanding premature birth and assuring healthy outcomes;
Washington (DC): National Academies Press (US); 2007. 12, Societal Costs of Preterm Birth.
Available from: https://www.ncbi.nlm.nih.gov/books/NBK11358/
Bennett HA, Einarson A, Taddio A, Koren G, Einarson TR. Prevalence of depression during
pregnancy: systematic review. Obstetrics and Gynecology. 2004; 103(4): 698–709.
Blencowe H, Cousens S, Chou D, et al. Born too soon: the global epidemiology of 15 million
preterm births. Reproductive Health. 2013:10 (Suppl 1): S2.
Blumenshine P, Egerter S, Barclay CJ, et al. Socioeconomic disparities in adverse birth
outcomes: a systematic review.” American Journal of Preventive Medicine. 2010:39 (3): 263–72.
95

Blumenshine PM, Egerter SA, Libet ML, Braveman PA. Father’s education: An independent
marker of risk for Preterm birth. Maternal and Child Health Journal. 2011; 15(1):60–67.
Braveman P. Black–White disparities in birth outcomes: is racism-related stress a missing piece
of the puzzle? In: Lemelle AJ, Reed W, Taylor S, editors. Handbook of African American Health.
New York, NY: Springer; 2011. pp. 155–163.
Bromberger JT, Harlow S, Avis N, Kravitz HM, Cordal A. Racial/ethnic differences in the
prevalence of depressive symptoms among middle aged women: the study of women’s health
across the nation (SWAN). American Journal of Public Health. 2004; 94(8): 1378-1385.
Brown MA, Solchany JE. Two overlooked mood disorders in women: subsyndromal depression
and prenatal depression. Nursing Clinics of North America. 2004; 39:83–95.
Cents RM, Diamantopoulou S, Hudziak JJ, Jaddoe VW, Hofman A, Verhulst FC, et al.
Trajectories of maternal depressive symptoms predict child problem behaviour: the Generation R
Study. Psychological Medicine. 2013; 43(1): 13–25.
Cole ER, Omari SR. Race, class and the dilemmas of upward mobility for African Americans.
Journal of Social Issues. 2003; 59(4), 785-802.
Cohen S, Janicki-Deverts D, Chen E, Matthews KA. Childhood socioeconomic status and adult
health. Annals of the New York Academy of Sciences. 2010; 1186(1):37–55.
Colen CG, Geronimus AT, Bound J, James SA. Maternal upward socioeconomic mobility and
Black–White disparities in infant Birthweight. American Journal of Public Health. 2006;
96(11):2032–2039.
Collins JW, David RJ, Simon DM, Prachand NG. Preterm birth among African American and
white women with a lifelong residence in high-income Chicago neighborhoods: an exploratory
study. Ethnicity & disease. 2007:17(1): 113–17.
Collins JW, Rankin KM, David RJ. African American women’s lifetime upward economic
mobility and Preterm birth: The effect of fetal programming. American Journal of Public Health.
2011; 101(4):714–719.
Collins JW, Rankin KM, David RJ. Downward economic mobility and Preterm birth: An
exploratory study of Chicago-born upper class white mothers. Maternal and Child Health Journal.
2015; 19(7):1601–1607.
Collins LM, Lanza ST. Latent class and latent transition analysis: with applications in the social,
behavioral, and health sciences. New York: Wiley. 2010.
Dal Grande E, Chittleborough CR, Wu J, Shi Z, Goldney RD, Taylor AW. Effect of social
mobility in family financial situation and housing tenure on mental health conditions among
96

south Australian adults: results from a population health surveillance system, 2009 to 2011.
BMC Public Health. 2015; 17(15): 675.
Delobel-Ayoub M, Arnaud C, White-Koning M, Casper C, Pierrat V, Garel M, Burguet A, Roze
JC, Matis J, Picaud JC, Kaminski M, Larroque B; EPIPAGE Study Group. Behavioral problems
and cognitive performance at 5 years of age after very preterm birth: the EPIPAGE Study.
Pediatrics. 2009; 123(6):1485-92.
Dole N, Savitz DA, Siega-Riz AM, Hertz-Picciotto I, McMahon MJ, Buekens P. Psychosocial
factors and preterm birth among African American and White women in Central North Carolina.
American Journal of Public Health. 2004; 94:1358–65.
Dominguez TP. Race, racism, and racial disparities in adverse birth outcomes. Clinical
Obstetrics and Gynecology. 2008; 51(2):360–370.
Dominguez TP, Dunkel-Schetter C, Glynn LM, Hobel C, Sandman CA. Racial differences in
birth outcomes: the role of general, pregnancy, and racism stress. Health Psychology. 2008;
27(2):194–203.
Dziak JJ, Yang J, Tan X, Bray BC, Wagner AT, Lanza ST. LCA distal SAS macro users’ guide
(Version 3.0.2). 2016. University Park: The Methodology Center, Penn State. Retrieved from
http://methodology.psu.edu.
Elwell-Sutton T, Folb N, Clark A, Fairall L, Lund C, Bachmann M. Socioeconomic position and
depression in South African adults with long-term health conditions: A longitudinal study of
causal pathways. Epidemiology and Psychiatric Sciences. 2017; 1-11
Evans J, Heron J, Patel RR, Wiles N. Depressive symptoms during pregnancy and low birth
weight at term: longitudinal study. The British Journal of Psychiatry. 2007; 191: 84–85.
Faisal-Cury A, Menezes P, Araya R, Zugaib M. Common mental disorders during pregnancy:
prevalence and associated factors among low-income women in São Paulo, Brazil: depression
and anxiety during pregnancy. Archives of Women’s Mental Health. 2009; 12(5): 335–43.
Farmer MM, Ferraro KF. Are racial disparities in health conditional on socioeconomic status?
Social Science & Medicine. 2005; 60: 191-204.
Field T, Diego M, Hernandez-Reif M. Prenatal depression effects on the fetus and newborn: a
review. Infant Behavior & Development. 2006; 29(3): 445–55.
Gale CR, Sayer AA, Cooper C, Dennison EM, Starr JM, Whalley LJ, . . .the HALCyon Study
Team. Factors associated with symptoms of anxiety and depression in five cohorts of
community-based older people: The HALCyon (Healthy Ageing across the Life Course)
Programme. Psychological Medicine. 2011; 41: 2057-2073.

97

Gavin NI, Gaynes BN, Lohr KN, Meltzer-Brody S, Gartlehner G, Swinson T. Perinatal
depression: a systematic review of prevalence and incidence. Obstetrics and Gynecology. 2005;
106(5 Pt 1): 1071–83.
Gavin AR, Walton E, Chae DH, Alegria M, Jackson JS, Takeuchi D. The associations between
socioeconomic status and major depressive disorder among Blacks, Latinos, Asians, and NonHispanic Whites: findings from the collaborative psychiatric epidemiology studies.
Psychological medicine. 2010; 40(1):51-61.
Gadalla TM. Determinants, correlates and mediators of psychological distress: a longitudinal
study. Social Science & Medicine. 2009; 68:2199–2205.
Gavin, AR, Melville JL, Rue T, Guo Y, Dina KT, Katon WJ. (2011). Racial differences in the
prevalence of antenatal depression. General Hospital Psychiatry. 2011; 33(2), 87–93.
Gigante DP, Horta BL, Matijasevich A, et al. Gestational age and newborn size according to
parental social mobility: An intergenerational cohort study. Journal of Epidemiology and
Community Health. 2015; 69(10):944–949.
Gilbert WM, Nesbitt TS, Danielsen B). The cost of prematurity: quantification by gestational age
and birth weight. Obstetrics & Gynecology. 2003; 102 (3): 488–492.
Gilman SE, Kawachi I, Fitzmaurice GM, Buka SL. Socioeconomic status in childhood and the
lifetime risk of major depression. International Journal of Epidemiology. 2002; 31(2): 359–67.
Golbasi Z, Kelleci M, Kisacik G, Cetin A. Prevalence and correlates of depression in pregnancy
among Turkish women. Maternal and Child Health Journal. 2010; 14(4): 485–91.
Goodman JH, Tyer-Viola L. Detection, treatment, and referral of perinatal depression and
anxiety by obstetrical providers. Journal of Women’s Health. 2010; 19(3): 477–90.
Grigoriadis S, VonderPorten EH, Mamisashvili L, Tomlinson G, Dennis CL, Koren G, et al. The
impact of maternal depression during pregnancy on perinatal outcomes: a systematic review and
meta-analysis. The Journal of Clinical Psychiatry. 2013; 74(4): e321-341.
Grote NK, Bridge JA, Gavin AR, Melville JL, Iyengar S, Katon WJ. A meta-analysis of
depression during pregnancy and the risk of preterm birth, low birth weight, and intrauterine
growth restriction. Archives of General Psychiatry. 2010; 67(10): 1012–24.
Hallqvist J, Lynch J, Bartley M, Lang T, Blane D. Can we disentangle life course processes of
accumulation, critical period and social mobility? An analysis of disadvantaged socio-economic
positions and myocardial infarction in the Stockholm Heart Epidemiology program. Social
Science and Medicine. 2004; 58(8): 1555-1562.

98

Hamilton BE, Martin JA, Osterman MJK, et al. Births: Final data for 2014. National vital
statistics reports; vol 64 no 12. Hyattsville, MD: National Center for Health Statistics. 2015.
Harper S, Lynch J, Hsu WL, Everson SA, Hillemeier MM, Raghunathan TE, et al. Life course
socioeconomic conditions and adult psychosocial functioning. International Journal of
Epidemiology. 2002; 31(2): 395–403.
Haveman R, Smeeding T. The Role of Higher Education in Social Mobility. The Future of
Children. 2006; 16(2), 125-150.
Holzman C, Bullen B, Fisher R, Paneth N, Reuss L, and Prematurity Study Group. Pregnancy
outcomes and community health: The POUCH study of preterm delivery. Paediatric and
Perinatal Epidemiology. 2001; 15 Suppl 2: 136–58.
Holzman C, Eyster J, Tiedje LB, Roman LA, Seagull E, Rahbar MH. A life course perspective
on depressive symptoms in mid-pregnancy. Maternal and Child Health Journal. 2006; 10(2):
127–38.
Howard LM, Piot P, Stein A. No health without perinatal mental health. The Lancet. 2014;
384(9956): 1723–24.
Hudson CG. Socioeconomic status and mental illness: tests of the social causation and selection
hypotheses. American Journal of Orthopsychiatry. 2005; 75:3–18.
Hudson DL, Neighbors HW, Geronimus AT, Jackson JS. The relationship between
socioeconomic position and depression among a US nationally representative sample of African
Americans. Social Psychiatry and Psychiatric Epidemiology. 2012; 47(3):373-381.
Hudson DL, Puterman E, Bibbins-Domingo K, Matthews KA, Adler NE. Race, life course
socioeconomic position, racial discrimination, depressive symptoms and self-rated health. Social
Science & Medicine. 2013; 97:7–14.
Husain N, Parveen A, Husain M, Saeed Q, Jafri F, Rahman R, et al. Prevalence and psychosocial
correlates of perinatal depression: a cohort study from urban Pakistan. Archives of Women’s
Mental Health. 2011; 14(5): 395–403.
Joseph K, Fahey J, Shankardass K, et al. Effects of socioeconomic position and clinical risk
factors on spontaneous and iatrogenic preterm birth. BMC Pregnancy and Childbirth. 2014;
14(1).
Kendzor DE, Caughy MO, Owen MT. Family income trajectory during childhood is associated
with adiposity in adolescence: A latent class growth analysis. BMC Public Health. 2012; 12(1).
Kessler RC. Epidemiology of women and depression. Journal of Affective Disorders. 2003;
74(1): 5–13.

99

Kramer MR, Dunlop AL, Hogue CJR. Measuring women’s cumulative neighborhood
deprivation exposure using longitudinally linked vital records: A method for life course MCH
research. Maternal and Child Health Journal. 2013; 18(2):478–487.
Kramer MR, Hogue CR. What causes racial disparities in very preterm birth? A biosocial
perspective. Epidemiologic Reviews. 2009; 31:84–98.
Kramer MS, Seguin L, Lydon J, Goulet L. Socio-economic disparities in pregnancy outcome:
why do the poor fare so poorly? Paediatric and Perinatal Epidemiology. 2000;14:194–210.
Kuh D, Ben-Shlomo Y, Lynch J, Hallqvist J, Chen WJ. Life Course Epidemiology. Journal of
Epidemiology & Community Health. 2003; 57: 778-783.
Kwon E, Park S. Heterogeneous trajectories of physical and mental health in late middle age:
importance of life-course socioeconomic positions. International Journal of Environmental
Research and Public Health. 2017; 14:582. DOI: 10.3390/ijerph14060582.
Lanza ST, Dziak JJ, Huang L, Wagner A, Collins LM. PROC LCA & PROC LTA users' guide
(Version 1.3.2). 2015. University Park: The Methodology Center, Penn State. Retrieved from
http://methodology.psu.edu
Lanza ST, Tan X, Bray BC. Latent class analysis with distal outcomes: a flexible model-based
approach. Structural Equation Modeling. 2013; 20(1): 1–26.
Li D, Liu L, Odouli R. Presence of depressive symptoms during early pregnancy and the risk of
preterm delivery: a prospective cohort study. Human Reproduction. 2008; 24(1):146–53.
Ljung R, Hallqvist J. Accumulation of adverse socioeconomic position over the entire life course
and the risk of myocardial infarction among men and women: results from the Stockholm Heart
Epidemiology Program (SHEEP). Journal of Epidemiology and Community Health. 2006;
60(12): 1080–84.
Lorant V, Deliège D, Eaton W, Robert A, Philippot P, Ansseau M. Socioeconomic Inequalities
in depression: a meta-analysis. American Journal of Epidemiology. 2003; 157(2): 98–112.
Love C, David RJ, Rankin KM, Collins JW. Exploring weathering: Effects of lifelong economic
environment and maternal age on low birth weight, small for gestational age, and Preterm birth
in African-American and white women. American Journal of Epidemiology. 2010; 172(2):127–
134.
Lu MC, Halfon N. Racial and ethnic disparities in birth outcomes: a life-course perspective.
Maternal & Child Health Journal. 2003; 7(1):13–30.
Luo Y, Waite LJ. The impact of childhood and adult SES on physical, mental, and cognitive
well-being in later life. The Journals of Gerontology. 2005; 60(2): S93–101.
100

Marcus SM, Flynn HA, Blow FC, Barry KL. Depressive symptoms among pregnant women
screened in obstetrics settings. Journal of Women’s Health. 2003; 12(4): 373–80.
Marcus SM. Depression during pregnancy: rates, risks and consequences--Motherisk Update
2008. Canadian Journal of Clinical Pharmacology. 2009; 16(1):15-22.
Mathews TJ, Menacker F, & MacDorman MF. Infant mortality statistics from the 2002 period
linked birth/infant death data set. Hyattsville, Maryland. National Center for Health Statistics.
National Vital Statistics Reports. 2004; 53(10):1-30.
Matijasevich A, Victora CG, Lawlor DA, et al. Association of socioeconomic position with
maternal pregnancy and infant health outcomes in birth cohort studies from Brazil and the UK.
Journal of Epidemiology and Community Health. 2012; 66(2):127–135.
McEwen BS. Allostasis and allostatic load: implications for neuropsychopharmacology.
Neuropsychopharmacology. 2000; 22(2): 108–24.
McEwen BS. Stress, adaptation, and disease: allostasis and allostatic load. Annals of the New
York Academy of Sciences. 1998; 840(1): 33–44.
Melville JL, Gavin A, Guo Y, Fan MY, Katon WJ. Depressive disorders during pregnancy:
prevalence and risk factors in a large urban sample. Obstetrics and Gynecology. 2010; 116(5):
1064–70.
Messer LC, Laraia BA, Kaufman JS, et al. The development of a standardized neighborhood
deprivation index. Journal of Urban Health. 2006; 83(6):1041–1062.
Miyake Y, Tanaka K, Arakawa M. Employment, income, and education and prevalence of
depressive symptoms during pregnancy: the Kyushu Okinawa Maternal and Child Health Study.
BMC Psychiatry. 2012; 12: 117.
Morgen CS, Bjørk C, Andersen PK, et al. Socioeconomic position and the risk of preterm birth-a study within the Danish National Birth Cohort. International Journal of Epidemiology. 2008;
37(5):1109-20.
Mortensen LH, Helweg-Larsen K, Andersen AM. Socioeconomic differences in perinatal health
and disease. Scandinavian Journal of Public Health. 2011; 39(7 Suppl):110–114.
Myer L, Stein DJ, Grimsrud A, Seedat S, Williams DR. Social determinants of psychological
distress in a nationally-representative sample of South African adults. Social Science and
Medicine. 2008; 66:1828–1840.
Nicholson A, Pikhart H, Pajak A, Malyutina S, Kubinova R, Peasey A, Bobak M. Socioeconomic status over the life-course and depressive symptoms in men and women in Eastern
Europe. Journal of Affective Disorders. 2008; 105: 125-136.
101

Nicholson WK, Setse R, Hill-Briggs F, Cooper LA, Strobino D, Powe NR. Depressive
symptoms and health-related quality of life in early pregnancy. Obstetrics & Gynecology. 2006;
107(4):798–806.
Nicklett EJ, Burgard SA. Downward social mobility and major depressive episodes among
Latino and Asian-American immigrants to the United States. American Journal of Epidemiology.
2009; 170(6): 793–801.
Nylund KL, Asparouhov T, Muthén BO. Deciding on the number of classes in latent class
analysis and growth mixture modeling: a Monte Carlo simulation study. Structural Equation
Modeling. 2007; 14(4): 535–69.
O’Campo P, Burke JG, Culhane J, et al. Neighborhood deprivation and preterm birth among
non-hispanic black and white women in eight geographic areas in the united states. American
Journal of Epidemiology. 2008; 167:155-163.
Osypuk TL, Slaughter-Acey JC, Kehm RD, et al. Life-course social mobility and reduced risk of
adverse birth outcomes. American Journal of Preventive Medicine. 2016; 51(6):975–982.
Owens CT, Jackson FM. Examining life-course socioeconomic position, contextualized stress,
and depression among well-educated African-American pregnant women. Women’s Health
Issues. 2015; 25(4): 382–89.
Radloff LS. The CES-D scale: a self-report depression scale for research in the general
population. Applied Psychological Measurement. 1977; 1(3): 385–401.
Rahman A, Bunn J, Lovel H, Creed F. Association between antenatal depression and low
birthweight in a developing country. Acta Psychiatrica Scandinavica. 2007; 115(6): 481–86.
Räisänen S, Lehto SM, Nielsen HS, Gissler M, Kramer MR, Heinonen S. Risk factors for and
perinatal outcomes of major depression during pregnancy: a population-based analysis during
2002-2010 in Finland. BMJ Open 2014; 4(11): e004883.
Ramaswamy V, Desarbo WS, Reibstein DJ, Robinson WT. An empirical pooling approach for
estimating marketing mix elasticities with PIMS data. Marketing Science. 1993; 12(1): 103–24.
Ritsher JE, Warner V, Johnson JG, Dohrenwend BP. Inter-generational longitudinal study of
social class and depression: a test of social causation and social selection models. The British
Journal of Psychiatry. 2001; Supplement 40: s84-90.
Ro A. occupational mobility and depression among the foreign-born in the United States. Journal
of Immigrant and Minority Health. 2014; 16(6): 1149–56.

102

Rodriguez E, Allen JA, Frongillo EA Jr., Chandra P. Unemployment, depression, and health: a
look at the African-American community. Journal of Epidemiology Community Health. 1999;
53(6):335–42.
Roxburgh S. Untangling inequalities: gender, race, and socioeconomic differences in depression.
Sociological Forum. 2009; 24: 357-381.
Rutter DR, Quine L. Inequalities in pregnancy outcome: a review of psychosocial and
behavioural mediators. Social Science & Medicine. 1990; 30(5):553-68.
Schwarz G. Estimating the dimension of a model. The Annals of Statistics. 1978; 6(2): 461–64.
Shankardass K, O’Campo P, Dodds L, et al. Magnitude of income-related disparities in adverse
perinatal outcomes. BMC Pregnancy and Childbirth. 2014; 14(1).
Slaughter-Acey JC, Holzman C, Calloway D, Tian Y. Movin’ on up: Socioeconomic mobility
and the risk of delivering a small-for-gestational age infant. Maternal and Child Health Journal.
2015; 20(3):613–622.
Sletner L, Jenum AK, Mørkrid K, et al. Maternal life course Socio-Economic position and
offspring body composition at birth in a multi-ethnic population. Paediatric and Perinatal
Epidemiology. 2014; 28(5):445–454.
Spencer N. Accounting for the social disparity in birth weight: Results from an intergenerational
cohort. Journal of Epidemiology & Community Health. 2004; 58(5):418–419.
Stansfeld SA, Clark C, Rodgers B, Caldwell T, Power C. Repeated exposure to socioeconomic
disadvantage and health selection as life course pathways to mid-life depressive and anxiety
disorders. Social Psychiatry and Psychiatric Epidemiology. 2011; 46(7): 549–58.
Stewart AL, Dean ML, Gregorich SE, Brawarsky P, Haas JS. Race/ethnicity, socioeconomic
status and the health of pregnant women. Journal of Health Psychology. 2007; 12(2): 285–300.
Tiffin PA, Pearce MS, Parker L. Social mobility over the lifecourse and self-reported mental
health at age 50: prospective cohort study. Journal of Epidemiology and Community Health.
2005; 59(10): 870–72.
Tiffin Y, Waite LJ. The impact of childhood and adult SES on physical, mental, and cognitive
well-being in later life. The Journals of Gerontology. 2005; 60(2): S93–101.
Turrell G, Lynch JW, Kaplan GA, Everson SA, Helkala EL, Kauhanen J, et al. Socioeconomic
position across the lifecourse and cognitive function in late middle Age. The Journals of
Gerontology. 2002; 57(1): S43-51.
Walsemann KM, Child S, Heck K, Margerison-Zilko C, Braveman P, Marchi K, Cubbin C. Are
the poverty histories of neighbourhoods associated with psychosocial well-being among a
103

representative sample of California mothers? An observational study. Journal of Epidemiology
and Community Health. 2017;71(6):558-564. DOI: 10.1136/jech-2016-207866.
Ward JB, Haan MN, Garcia ME, Lee A, To TM, Aiello AE. Intergenerational education mobility
and depressive symptoms in a population of Mexican origin. Annals of Epidemiology. 2016;
26(7): 461–66.
Watkins DC, Hudson DL, Howard CC, Siefert K, Jackson JS. Discrimination, mastery, and
depressive symptoms among African American men. Research on Social Work Practice. 2010;
21(3), 269-277.
Whitehead NS. The relationship of socioeconomic status to Preterm contractions and Preterm
delivery. Maternal and Child Health Journal. 2012; 16(8):1645–1656.
Wisner KL, Sit DKY, Hanusa BH, Moses-Kolko EL, Bogen DL, Hunker DF, et al. Major
depression and antidepressant treatment: impact on pregnancy and neonatal outcomes. The
American Journal of Psychiatry. 2009; 166(5): 557–66.

104