Archives of Women's Mental Health

, Volume 17, Issue 1, pp 27–40

An examination of prenatal and postpartum depressive symptoms among women served by urban community health centers

Authors

    • Allina Health
  • Wendy L. Hellerstedt
    • Division of Epidemiology and Community HealthUniversity of Minnesota, School of Public Health
  • Patricia A. Harrison
    • Minneapolis Department of Health
  • Deborah Hennrikus
    • Division of Epidemiology and Community HealthUniversity of Minnesota, School of Public Health
Original Article

DOI: 10.1007/s00737-013-0378-3

Cite this article as:
Sidebottom, A.C., Hellerstedt, W.L., Harrison, P.A. et al. Arch Womens Ment Health (2014) 17: 27. doi:10.1007/s00737-013-0378-3

Abstract

We characterized depressive symptoms in the prenatal and/or postpartum periods and examined associated risk factors among 594 women who received care at community health care centers. Women were screened with comprehensive risk assessments, which included the Patient Health Questionnaire-9 depression screen, during pregnancy and at least 4 weeks after delivery. Fifteen percent had depressive symptoms in the prenatal period only; 6 % in the postpartum period only, and 8 % had depressive symptoms in both periods. Risk markers varied for women who reported depressive symptoms at one period only compared with those who reported persistent depressive symptoms. Age (25 years versus younger), having experienced abuse, not living with the infant’s father, and cigarette smoking were associated with depressive symptoms at both periods; being US-born, lacking social support, and experiencing food insecurity were associated with reporting symptoms only in the prenatal period, and lack of phone access was associated with risk only in the postpartum period. Our findings confirm the importance of repeated screenings for depressive symptoms during the perinatal period. The variability in risk markers associated with periods of reported depressive symptoms may reflect their varying associations with persistence, new onset, or recovery from depressive symptoms.

Keywords

Prenatal depressionPostpartum depressionDepression screening

Introduction

In recent years, the media and the public health community have focused much attention on postpartum depression. As a result, screening to identify postpartum depression has become more widely promoted (Lusskin et al. 2007; Wisner et al. 2002; Miller 2002). While there has been less attention focused on prenatal depression, there is a growing body of literature aimed at understanding how prenatal and postpartum depression are associated and identifying risk factors for depression during these periods. A better understanding of prenatal and postpartum depression prevalence, timing of onset, and the persistence or improvement of symptoms could inform health care and public health guidelines for screening and support services.

Prenatal and postpartum depression each has unique and serious risks. Prenatal depression is associated with pre-eclampsia, neonatal growth retardation, spontaneous abortion, and delivery of preterm and low birthweight infants (Bonari et al. 2004; Bennett et al. 2004; Lusskin et al. 2007; Davalos et al. 2012). Its association with poor fetal growth could explain the limited evidence that it is adversely associated with infant cognitive, behavioral, and psychomotor development (Kingston et al. 2012). Postpartum depression may limit a new mother’s capacity to fulfill her maternal role. Infants of women experiencing postpartum depression are at increased risk for poor bonding, cognitive and emotional developmental delays, and behavior problems as children (Lusskin et al. 2007; Kingston et al. 2012). These pregnancy, fetal, and infant development risks associated with prenatal and postpartum depression highlight the need to better understand depression during these time periods to inform screening, treatment, and prevention efforts.

Several prospective studies conducted in a variety of countries provide some evidence about the prevalence of depression during pregnancy and postpartum (Andersson et al. 2006; Barnet et al. 1996; Evans et al. 2001; Felice et al. 2004; Gotlib et al. 1989; Gotlib et al. 1991; Heron et al. 2004; Hobfoll et al. 1995; Milgrom et al. 2008; Mohammad et al. 2011; Rich-Edwards et al. 2006; Silva et al. 2012). These studies found a wide range in prevalence rates, and most reported higher prevalence in pregnancy than in the postpartum (Andersson et al. 2006; Barnet et al. 1996; Evans et al. 2001; Felice et al. 2004; Gotlib et al. 1989; Gotlib et al. 1991; Heron et al. 2004; Hobfoll et al. 1995; Silva et al. 2012) while others found similar prevalence rates in both time periods (Mohammad et al. 2011; Rich-Edwards et al. 2006; Milgrom et al. 2008).

To guide practice and public health efforts, it may be helpful to understand both the onset of symptoms and the persistence of those symptoms. Some research on the prevalence of depression has focused on postpartum-onset depression as compared with postpartum depression that began prenatally (Silva et al. 2012; Andersson et al. 2006; Gotlib et al. 1989; Gotlib et al. 1991). Prospective studies have generally found that prenatal-onset depression is more common than postpartum-onset (Evans et al. 2001; Felice et al. 2004; Hobfoll et al. 1995; Andersson et al. 2006; Rich-Edwards et al. 2006; Heron et al. 2004; Gotlib et al. 1989; Gotlib et al. 1991; Silva et al. 2012; Sexton et al. 2012). Several studies have provided evidence that prenatal depression may persist into the postpartum period and that perhaps half of “postpartum depression” had its onset prenatally (Hobfoll et al. 1995; Felice et al. 2004). For example, the Avon Longitudinal Study reported that 8.9 % of women had elevated depression scores postpartum, but only 3.5 % were new cases (Heron et al. 2004). Similarly, in a Canadian study of 360 women, 6.8 % had a postpartum depression diagnosis, but only half (3.4 %) experienced the onset of depression during the postpartum period (Gotlib et al. 1989). A few studies have also examined not only the persistence of prenatal depressive symptoms into the postpartum but what they label “recovery.” A study of 41 women with elevated prenatal depressive symptoms found that, in the postpartum period, 39 % “recovered” to non-elevated levels and 61 % had persistent elevated symptoms (Sexton et al. 2012). A larger study (n = 1,019) of urban Brazilian women found that 47 % of the women who had depression during pregnancy also had depression postpartum (Silva et al. 2012).

There are few studies about how the risk markers for depressive symptoms may differ for the prenatal or postpartum periods, or how they may be associated with the persistence or “recovery” from prenatal depressive symptoms (Andersson et al. 2006; Gotlib et al. 1989, 1991; Silva et al. 2012). Several personal, social, behavioral, and environmental factors may be related to experiencing depression during either or both the prenatal or postpartum periods. Previously identified independent risks for prenatal or postpartum depression include maternal anxiety, life stress, or history of depression (Lancaster et al. 2010; Beck 2001), abuse (Gross et al. 2002; Nielsen Forman et al. 2000), lack of social support, social isolation (Nielsen Forman et al. 2000; Lancaster et al. 2010; Brugha et al. 1998), lower education, being unmarried or poor relationship quality, unintended pregnancy, smoking (Lancaster et al. 2010; Beck 2001), and poverty (Hobfoll et al. 1995; Gross et al. 2002; Lancaster et al. 2010). Alcohol and drug use are associated with general depression (Jane-Llopis and Matytsina 2006) and postpartum depression (Ross and Dennis 2009) and have had inconsistent findings related to prenatal depression (Lancaster et al. 2010). Studies that have examined risk during both the prenatal and postpartum periods have found some risks may be associated with recovery from depression experienced prenatally (Andersson et al. 2006; Sexton et al. 2012) such as living with a partner and lower baseline depressive symptoms. Another found that some risks associated with prenatal depression (e.g., age and number of children) are not associated with postpartum-onset (Gotlib et al. 1989). While the risk markers for depressive symptoms experienced in the postpartum period and to a lesser degree, prenatally are clear, the characteristics of women who “recover” from prenatal depressive symptoms after giving birth are relatively unexplored. Previously conducted studies provide some evidence through small sample sizes or a limited set of risk factors. More detailed examination of the experience of depressive symptoms perinatally is needed because of its impact on maternal and infant health and because risk markers may be modifiable.

The goals of this study were to: (1) describe elevated depressive symptom levels in pregnancy and postpartum and characterize the timing of detection and persistence or recurrence of symptoms in a sample of women who received care at urban community health clinics and (2) examine personal, social, behavioral, and environmental correlates of elevated depressive symptom levels in pregnancy only, postpartum only, or during both periods. Our analyses were informed by the Theory of Triadic Influences (Flay 1999). The theory was designed to assist researchers in identifying intervention approaches and posits that broad categories of personal, social, and environmental factors influence health outcomes.

Materials and methods

Study overview and context

This study was conducted through the Twin Cities Healthy Start program, one of 105 programs funded through the United States’ Healthy Start Initiative by the Health Resources and Services Administration. The United States’ Healthy Start Initiative targets funding to communities with high rates of infant mortality. The Twin Cities Healthy Start Program covered specific geographic communities in Minneapolis and St. Paul in Minnesota, USA. It served primarily African Americans and American Indians, the two groups with the highest long-term rates of infant mortality locally and statewide in Minnesota (Minnesota Department of Health 2009). The local program, administered by the Minneapolis Department of Health, offered outreach and case management services to women who received prenatal care at Federally Qualified Health Centers (FQHCs). FQHCs are typically located in—and serve—medically underserved communities that are home to disproportionate numbers of families living in poverty.

As part of the local Healthy Start program protocol, all women who sought prenatal care at participating FQHCs were screened using a multidimensional screening instrument developed for the program (Harrison and Sidebottom 2008) which includes the Patient Health Questionnaire-9 (PHQ-9) depression screen. Women who were enrolled in the Twin Cities Healthy Start program were also screened during the postpartum period with a similar assessment that also included the PHQ-9.

The University of Minnesota Institutional Review Board approved the use of data from the Twin Cities Healthy Start program for the study analyses.

Study sample

The study sample was women who sought prenatal care at five FQHCs and who enrolled in the Twin Cities Healthy Start program between November 2005 and May 2009. Women were included in this study if they completed both the prenatal and postpartum screenings. Postpartum data were not collected for women who transferred care to another clinic during pregnancy.

Of the 1,822 women enrolled in the program during the study period, 728 had both prenatal and postpartum screening data available. The postpartum risk assessment protocol required waiting until at least 4 weeks post-delivery because earlier symptoms might represent the initial “baby blues” period (Seyfried and Marcus 2003). However, some postpartum screenings occurred earlier because case managers believed they would not be able to meet with or reach a woman later to conduct the interview because of a previous history of missed appointments. Women (n = 119) whose postpartum screening was conducted before 4 weeks postpartum were excluded from the study. An additional 15 women were excluded because they were missing information on the timing of their postpartum screening. The final sample consisted of 594 women (33 % of those enrolled).

Data collection

The prenatal assessment was conducted at the end of the prenatal intake appointment, which included a substantial discussion of medical history. The risk assessment interview, lasting 10–15 min, was usually conducted by a registered nurse but sometimes by a social worker or a community health worker. As the interviewer administered the assessment, she entered responses into the web-based Twin Cities Healthy Start Screening and Case Management System, which linked to other databases with individual demographic descriptors, service-related information, and birth outcomes (Harrison and Sidebottom 2008). The postpartum assessment was conducted when the woman was at the clinic for another purpose, with the goal of administering assessment between 4 and 12 weeks after delivery. In some cases, when clinic staff members were unable to meet with the woman in person, they conducted the postpartum interview by telephone.

Measures

The Prenatal Risk Overview (PRO), the prenatal risk assessment instrument developed for the program consists of 58 questions that address 13 domains: telephone access, transportation access, food security, housing stability, social support, partner violence, physical/sexual abuse by a non-partner, depressive symptoms (i.e., the PHQ-9), cigarette smoking, alcohol use, drug use, legal problems, and child protection involvement. This instrument has been described in detail previously (Harrison and Sidebottom 2008). A shortened version of the PRO, which also includes the full PHQ-9, was used to assess risks in the postpartum period (PPRO). The criterion validity of the alcohol, drug, and depression domains was established against the Structured Clinical Interview for DSM Disorders (SCID), a structured diagnostic interview (Harrison et al. 2011; Sidebottom et al. 2012; Harrison and Sidebottom 2012). All three domains were found to have high sensitivity and specificity. Re-screening and interviewer equivalence studies have also been conducted (Harrison and Sidebottom 2008; Harrison et al. 2011).

Depressive symptoms (PHQ-9)

The PHQ-9, which is included in the PRO and PPRO, was developed using diagnostic criteria to screen for depression in a general primary care patient population (Kroenke et al. 2001, 2010). The PHQ-9 has been shown to have acceptable validity in diverse patient populations (Adewuya et al. 2006; Gjerdingen et al. 2009; Huang et al. 2006; Kroenke et al. 2001; Kroenke et al. 2010; Lotrakul et al. 2008; Martin et al. 2006; Monahan et al. 2009; Wittkampf et al. 2007; Yeung et al. 2008; Sidebottom et al. 2012), including obstetrics–gynecology patients (Kroenke et al. 2001), postpartum (Gjerdingen et al. 2009; Hanusa et al. 2008; Weobong et al. 2009), and pregnant women (Sidebottom et al. 2012). A meta-analysis of PHQ-9 diagnostic accuracy studies estimated a sensitivity of 77 % for major depressive disorder, specificity of 94 %, and a positive predictive value of 59 % in primary care populations and higher (85–90 %) in populations with a high prevalence of depressive disorder (Wittkampf et al. 2007). A study of postpartum women (Gjerdingen et al. 2009) yielded 82 % sensitivity and 84 % specificity. A prenatal validation of the PHQ-9 was conducted in a subset of community health clinics participating in the Twin Cities Healthy Start program. This study, with a sample similar to that of the current study, compared the prenatal PHQ-9, administered as part of the PRO, with a structured diagnostic interview and found sensitivity of 85 % and specificity of 84 % for major depressive disorder (Sidebottom et al. 2012). In studies assessing the concordance of the PHQ-9 with the Edinburgh Postnatal Depression Scale, another widely used depression screening tool, the two instruments performed similarly in the postpartum and prenatal periods (Flynn et al. 2011; Yawn et al. 2009).

PHQ-9 questions address the previous 2 weeks and ask about physical and mood symptoms of depression: problems sleeping, being tired or having little energy, appetite, restlessness, speaking or moving slowly, feeling little interest or pleasure, feeling down or hopeless, feeling bad about yourself, and suicidal ideation. We made one modification: The item measuring psychomotor retardation (moving or speaking slowly) or agitation (being fidgety or restless) was split into two questions to examine these symptoms separately but were scored as one item. Response categories for all questions were: “not at all”, “several days”, “more than half the days”, and “every day or nearly every day”. Each item was scored from 0 (“not at all”) to 3 (“every day or nearly every day”), and the item scores were summed. We created the following categories to reflect depressive symptom severity (Kroenke et al. 2001): scores less than 10 as low, 10–14 as moderate, 15–27 as high. In this study, a dichotomous measure was also used for low (<10) or moderate/high (10+) to categorize postpartum depressive symptom levels as the dependent measure in regression models. The cut point of 10 corresponds to the score at which providers recommend the development of a treatment plan for counseling, follow up, and/or pharmacotherapy (Kroenke 2002). A measure combining pregnancy and postpartum PHQ-9 scores was created to identify when depressive symptoms were reported resulting in the following categories: low (prenatal and postpartum PHQ-9 scores both < 10), prenatal only (prenatal PHQ-9 ≥ 10 and postpartum PHQ-9 < 10), postpartum only (prenatal PHQ-9 < 10 and postpartum PHQ-9 ≥ 10), and prenatal and postpartum (PHQ-9 scores ≥10 at both time periods).

Social, Behavioral, and Environmental Factors

These measures came from the PRO and were scored as low, moderate, and high risk categories as described elsewhere (Harrison and Sidebottom 2008). For analyses, we dichotomized each domain as low or moderate /high levels risk; scoring definitions are described below.

Social factors

To assess lack of social support, six items from the 21-item Maternal Social Support Index (Pascoe et al. 1988) were used. The questions asked about how many people a woman could count on in times of need or to take care of children for several hours and relationship satisfaction with boyfriend/husband/partner and with other adults. Risk was scored as moderate/high if a woman indicated she had no one or only one person to count on in times of need or to take care of children for several hours; if she had no husband/ partner, or had a partner but had unsatisfactory communication with him; or if she had satisfactory communication with a husband/boyfriend or another adult but not both. Risk was scored low if a woman had more than one person to take care of her children or to count on in times of need and satisfactory communication with both a husband/partner and an adult.

Six violence items, scored yes/no, adapted from the Abuse Assessment Screen (Soeken et al. 1998), asked about two time periods: 12 months prior to pregnancy awareness and the interval since pregnancy awareness. The questions asked about experiencing abuse, forced sex, fear of being abused relative to an intimate partner, and repeated to apply to anyone else. Risk was scored moderate/high if a woman indicated abuse during pregnancy or prior to pregnancy, or current fear of abuse. Risk was scored low if she denied abuse for both time periods. We combined responses about the two sources of abuse (i.e., intimate partner or someone else) into a single measure of any abuse. This combined measure was coded as low for women who reported no abuse by anyone during both of the time periods and moderate/high if she reported abuse from anyone during pregnancy or prior to pregnancy, or if she feared current abuse.

Child protection involvement was assessed with a single question about current or previous involvement with the child protection system (as a parent). Risk was scored moderate/high if the woman reported current or previous involvement.

Behavioral factors

Smoking and alcohol and drug use quantity and frequency questions were from the National Household Survey on Drug Use and Health (Institute 2003). Two time frames were addressed for each question: before pregnancy awareness (1 month for smoking and 12 months for alcohol and drug use) and after pregnancy awareness. Smoking was assessed with two questions for each time period: frequency (i.e., daily, less than daily, not at all) and average number of cigarettes smoked per day. If she smoked during pregnancy or was a daily smoker prior to pregnancy awareness, she was coded as moderate/high risk. If she did not smoke during either period or smoked prior to pregnancy awareness less frequently than daily, then she was coded as No/Low risk.

In addition to the frequency and quantity questions, the alcohol measure included consequence questions from the Rapid Alcohol Problem Screen (Cherpitel 2000). These questions asked about having a feeling of guilt after drinking, being told about things they did while drinking they could not remember, neglecting responsibilities because of alcohol use, or drinking in the morning. Alcohol scoring was based on frequency, quantity, and consequences. The moderate/high risk was defined as any alcohol drinking after pregnancy awareness, or experiencing any adverse consequences related to drinking prior to pregnancy awareness, or a typical pattern of drinking in the year before pregnancy awareness of two or more drinks daily, three drinks per occasion once per week or more frequently, or four drinks per occasion once per month or more frequently. No/Low risk was defined for women reporting no use after pregnancy awareness and no adverse consequences prior to pregnancy awareness, and either no alcohol use prior to pregnancy awareness or typical use in the year prior to pregnancy awareness of one drink at any frequency level, two drinks on a single occasion at a weekly or monthly frequency, or three drinks on a single occasion monthly or rarely. The PRO identified an alcohol use disorder as measured by the Structured Clinical Interview for DSM Disorders (SCID) with sensitivity of 84 % and specificity of 80 % (Harrison et al. 2011).

Drug use scoring was based on frequency (i.e., daily, weekly, monthly, rarely, never) and on the response to one drug use consequence question (i.e., neglected responsibilities because of drug use) prior to pregnancy awareness and on frequency of use since pregnant. Moderate/high risk was defined as any drug use before or since pregnancy awareness or a history of failed obligations associated with drug use before pregnancy awareness. Low risk was defined as no use reported in either time period and no history of failed obligations associated with drug use. The PRO identified a drug use disorder as measured by the SCID with sensitivity of 89 % and specificity of 74 % (Harrison and Sidebottom 2012).

Environmental factors

Lack of access to telephone and transportation questions were developed specifically for the PRO and were assessed with one question each: When you are at home, how often do you have access to a telephone? How often do you have access to transportation? Moderate/high risk for each question was defined as responses of “none”, “rarely”, and “some of the time”. Low risk was defined as a response of “all of the time”.

Food insecurity was measured with four items adapted from the 12-month Food Security Scale of the U.S. Census Current Population Survey (Nord and Carlson 2003). The four items, scored 0–2, asked (1) how often food ran out and the woman did not have money to buy more; whether the woman could afford to eat balanced meals; whether the woman cut the size of a meal or skipped meals; and whether the woman had been hungry but unable to buy food. Responses to each question were “often”, “sometimes,” or “never”. We created a summary score and classified a score between 0 and 2 as low risk and between 3 and 8 points as moderate/high risk.

Housing instability was assessed with four questions, one from the Homeless Supplement to the Diagnostic Interview Schedule (North et al. 2004) and the others generated by clinic staff to reflect situations their clients experienced. The questions asked how many months out of the prior year a woman lived with family or friends as a temporary situation, how many nights out of the prior year she stayed in a shelter, how stable she felt her current housing situation was, and how concerned she was she would not have a place to live when her baby was born. Women were classified as moderate/high if they lived with family or friends as a temporary situation for three or more months in the past year; stayed in a shelter for one or more nights; described their current situation as unstable, and/or were very or somewhat concerned about not having a place to live when their infant was born. Women who did not stay in a shelter, were not concerned about having a place to live, and who stayed with relatives or friends for less than 3 months or not at all were classified in the low category.

Personal and clinical characteristics

Race/ethnicity was recorded as mutually exclusive categories: African American, American Indian, Asian/Pacific Islander, Hispanic (any race), white, or bi/multiracial. Nativity was categorized as US-born or foreign-born. Marital status was categorized as unmarried or married. Whether women were living with the father of the infant at prenatal intake was also recorded. While measures were available for both marital status and whether the woman was living with the father of the infant, we used only “living with the father of the infant” in regression models because only 16 % of the women reported being married, while 32 % reported living with the father of the infant. Trimester of pregnancy at initial PRO administration was calculated based on interview date and due date (or last menstrual period if due date was not entered) and categorized into first (1–13 weeks), second (14–26 weeks), or third (27+ weeks) trimester. Age was calculated based on date of birth and date of prenatal intake.

Analysis

We conducted statistical analyses with SPSS 18.0 (2010). We used frequency distributions to describe the study sample and depressive symptom levels. We conducted chi-square tests to examine differences between women included and excluded from the study sample and to examine correlates of elevated depressive symptom levels in pregnancy only, pregnancy and postpartum, and postpartum only. We compared mean PHQ-9 scores for the prenatal and postpartum periods using a t test for paired samples. We conducted multinomial logistic regression to examine adjusted associations of social, behavioral, and environmental risk factors with categories of elevated depressive symptom level time periods, with low symptom levels in both time periods as the reference. The reference categories for risk factors used in regression models were selected because of their association with low risk for depression.

Results

Included and excluded women differed significantly on four of the 18 factors collected at the time of the initial PRO (Table 1). Excluded women were more likely to be African American, Hispanic, or white and less likely to be American Indian or Asian. They were more likely than included women to have had their initial PRO conducted in the first trimester, to be smokers, and to report housing instability.
Table 1

Comparison of study sample with women excluded due to missing postpartum data

  

Women

Final study

 
 

Total

Excluded

Sample

 
 

(n = 1,822)

(n = 1,228)

(n = 594)

 

Personal characteristics

%

%

%

P value

Age, years

 <20

38.3

37.6

39.6

 

 20–24

34.9

34.0

36.6

0.115

 25+

26.9

28.4

23.8

 

Race/ethnicity

 African American

52.5

53.4

50.5

 

 Asian/Pacific Islander

13.6

12.4

16.0

 

 Hispanic

9.2

9.9

7.7

0.005

 American Indian

17.1

15.7

20.2

 

 White

5.0

5.5

3.9

 

 Multiple

2.6

3.1

1.7

 

Foreign-born

 Yes

23.9

23.5

24.8

0.523

 No

76.1

76.5

75.2

 

Trimester PRO conducted

 1st

48.6

51.5

42.8

 

 2nd

39.7

38.1

43.0

0.001

 3rd

11.7

10.4

14.3

 

Prenatal Depressive Symptom Level (PHQ-9)

 Low (<10)

77.8

78.0

77.3

 

 Moderate (10–14)

12.6

12.9

12.0

0.473

 High(15–27)

9.6

9.1

10.8

 

Social factors

Lack of social support

 Low

22.5

23.2

20.9

0.260

 Moderate/high

77.5

76.8

79.1

 

Intimate partner violence

 No

89.5

89.1

90.2

0.461

 Abuse or fear of abuse

10.5

10.9

9.8

 

Abuse by someone else

 No

87.3

87.3

87.4

0.998

 Abuse or fear of abuse

12.7

12.7

12.6

 

Child protection involvement

 No

85.8

85.0

87.5

0.156

 Yes (prior or current)

14.2

15.0

12.5

 

Marital status

    

 Married

15.0

14.5

16.0

0.415

 Single

85.0

85.5

84.0

 

Living with father of the baby

 Yes

32.6

32.8

32.2

 

 No

67.4

67.2

67.8

0.777

Behavioral factors

Alcohol use

 Low or no pre-pregnancy use/no prenatal use

70.8

70.0

72.6

0.257

 Any prenatal or high pre-pregnancy use

29.2

30.0

27.4

 

Drug use

 No

65.0

63.8

67.5

0.127

 Yes (prenatal or pre-pregnancy)

35.0

36.2

32.5

 

Cigarette use

 No prenatal/no or < daily pre-pregnancy

65.4

63.4

69.4

0.013

 Any prenatal use or daily pre-pregnancy

34.6

36.6

30.6

 

Environmental factors

Lack of phone access

 Low

87.6

87.1

88.6

0.386

 Moderate/high

12.4

12.9

11.4

 

Lack of transportation access

 Low

42.2

42.9

40.7

0.389

 Moderate/high

57.8

57.1

59.3

 

Food insecurity

 Low

57.7

57.2

58.7

0.539

 Moderate/high

42.3

42.8

41.3

 

Housing instability

 Low

33.5

31.8

37.2

0.021

 Moderate/high

66.5

68.2

62.8

 

The study sample was young: 76 % were younger than 25 years of age (mean 21.9, SD 5.45 years). Half were African American, 20 % American Indian, and 16 % were Asian. Nearly one-quarter were foreign-born. The three largest immigrant groups were Hmong, Somali, and Hispanic. Interpreters were used for only one-quarter of foreign-born women (6 % of the study sample): 3.2 % of the interviews were conducted in Hmong, 2.4 % in Spanish and less than 1 % in Somali or Laotian. A majority of women (84 %) were unmarried, and one-third (32.6 %) were living with the father of the infant at the time of the prenatal intake appointment. The PRO was administered in the first trimester for 42 % of women and the second trimester for 43 %. Mean gestation at the time of the prenatal assessment was 17.2 weeks (SD 8.0). Eighty-four percent of postpartum assessments were completed within 4 to 12 weeks after the infant was born (67 % were done in 4–8 weeks, 17.3 % done in 9–12 weeks, and 15.7 % were done after 12 weeks postpartum).

More women had elevated depressive symptom levels prenatally than postpartum: 10.8 % versus 6.4 % at high levels and 12.0 % versus 7.6 % at moderate levels, respectively (Table 2). The average PHQ-9 scores were consistent with the findings from categorical risk classifications with a mean score of 6.06 (SD 5.86) prenatally and 3.78 (SD 5.57) postpartum, (t = 9.532, p < 0.001). Based on the combination of prenatal and postpartum PHQ-9 scores, 71.4 % had low depressive symptom levels at both points. Elevated levels were seen for 14.6 % during the prenatal period only, 8.1 % during both periods, and 5.9 % during the postpartum period only. Thirty-six percent of those with elevated symptom levels prenatally also had elevated levels postpartum. Of the women with postpartum elevated depressive symptom levels, 58 % had elevated levels in the prenatal period (Table 2).
Table 2

Prenatal and postpartum depressive symptom levels

 

Postpartum Depressive Symptom Level (PHQ-9)

Prenatal depressive symptom level

Low

Moderate

High

  

(PHQ-9 score)

(<10)

(10–14)

(15–27)

Total

 

Low (<10)

424

22

13

459

77.3%

Moderate (10–14)

51

12

8

71

12.0%

High (15–27)

36

11

17

64

10.8%

Total

511

45

38

594

 
 

86.0%

7.6%

6.4%

  

Combined classifications in prenatal and postpartum periods

 Low both periods

424

71.4%

   

 Prenatal only

87

14.6%

   

 Prenatal and postpartum

48

8.1%

   

 Postpartum only

35

5.9%

   

 Total

594

    
With the exception of transportation access, all of the social, behavioral, and environmental risk correlates and personal characteristics examined were significantly associated with elevated depressive symptom levels (Table 3). All risk factors were least prevalent among women with low depressive symptom levels both during pregnancy and the postpartum. With the exception of lack of social support, higher proportions of women with elevated symptom levels during pregnancy and postpartum had social and behavioral risk factors than women in the other groups. Multinomial regression models examined adjusted associations for the risk factors that were significant in the cross-tabulations, using low depressive symptom levels during both time periods as the reference group (Table 4). Experiencing abuse, not living with the father of the infant, and smoking were all associated with increased odds of being in the group with elevated depressive symptom levels during both pregnancy and postpartum. The odds of having elevated depressive symptoms in both periods compared with low symptoms in both periods was 2.2 times higher for women experiencing abuse than those not experiencing abuse, 2.8 times higher for those not living with the father of the infant compared with those who are, and 2.1 times higher for smokers compared with non-smokers. Women under the age of 25 years were less likely than older women to report prenatal and postpartum depressive symptoms or symptoms in the prenatal period only compared with the low symptom group. The odds of experiencing elevated depressive symptom levels in the prenatal period only compared with the low symptom group were increased for women who were US-born, lacked social support, and were experiencing prenatal food insecurity. Specifically, compared with those with low depressive symptom levels in both periods, the odds of having elevated depressive symptoms in the prenatal period only was 2.6 times higher for women with poor social support compared with those with better social support and 2.4 times higher for women with moderate or high levels of food insecurity compared with those with low levels of food insecurity. Only one risk factor, lack of phone access, remained significantly associated with increased risk of experiencing elevated depressive symptom levels in the postpartum period only compared with the low symptom group.
Table 3

Association of personal, social, behavioral, and environmental factors with elevated depressive symptom levels

 

Prenatal and postpartum (n = 48)

Prenatal only (n = 87)

Postpartum only (n = 35)

Low (both periods) (n = 424)

P value

Personal characteristics

Age, years

 25+

37.5

35.6

22.9

19.9

0.002

 <25

62.5

64.4

77.1

80.1

 

Race/ethnicity

 Asian/Pacific Islander

6.3

12.6

5.7

18.6

0.039

 African American

54.2

62.1

57.1

47.2

 

 American Indian

29.2

13.8

28.6

19.8

 

 Hispanic, White, or multiple

10.4

11.5

8.6

14.4

 

Foreign-born

 Yes

12.5

16.1

17.1

28.7

0.008

 No

87.5

83.9

82.9

71.3

 

Social factors

Lack of social support

 Low

10.4

8.0

20.0

24.8

0.001

 Moderate/high

89.6

92.0

80.0

75.2

 

Any abuse

 No

59.6

68.6

70.6

85.2

<0.001

 Abuse or fear of abuse

40.4

31.4

29.4

14.8

 

Child protection involvement

 No

75.0

85.1

80.0

90.0

0.009

 Yes (prior or current)

25.0

14.9

20.0

10.0

 

Living with father of the baby

 Yes

14.6

23.0

25.7

36.6

0.002

 No

85.4

77.0

74.3

63.4

 

Behavioral factors

Alcohol use

     

 Low or no pre-pregnancy use/no prenatal use

54.2

67.8

62.9

76.4

0.003

 Any prenatal or high pre-pregnancy use

45.8

32.2

37.1

23.6

 

Drug use

     

 No

52.1

60.9

64.7

70.8

0.028

 Yes (prenatal or pre-pregnancy)

47.9

39.1

35.3

29.2

 

Cigarette use

     

 No prenatal/no or < daily pre-pregnancy

45.8

66.3

61.8

73.3

0.001

 Any prenatal use or daily pre-pregnancy

54.2

33.7

38.2

26.7

 

Environmental factors

Lack of phone access

     

 Low

77.1

87.4

77.1

91.0

0.004

 Moderate/high

22.9

12.6

22.9

9.0

 

Lack of transportation access

     

 Low

31.3

34.5

31.4

43.9

0.103

 Moderate/high

68.8

65.5

68.6

56.1

 

Food insecurity

     

 Low

41.7

37.9

58.8

64.9

<0.001

 Moderate/high

58.3

62.1

41.2

35.1

 

Housing instability

     

 Low

27.1

25.3

45.7

40.1

0.019

 Moderate/high

72.9

74.7

54.3

59.9

 
Table 4

Multinomial regression models for prenatal and postpartum, prenatal only, and postpartum only elevated depressive symptom levels compared with low in both periods (n = 594)

 

Prenatal and postpartum (n = 48)

Postpartum only (n = 87)

Postpartum only (n = 35)

 

OR

CI

P value

OR

CI

P value

OR

CI

P value

Personal characteristics

Age, years

 25+

Ref

  

Ref

  

Ref

  

 <25

39

(0.19,0.79)

0.009

0.40

(0.23, 0.70)

0.001

0.73

(0.30, 1.79)

0.496

Race/ethnicity

 Asian/Pacific Islander

Ref

  

Ref

  

Ref

  

 African American

1.55

(0.30, 7.96)

0.598

1.63

(0.56, 4.70)

0.370

0.35

(0.03, 3.73)

0.390

 American Indian

1.68

(0.56, 5.02)

0.351

1.53

(0.68, 3.45)

0.301

2.91

(0.76, 11.06)

0.118

 Hispanic, White, or multiple

1.30

(0.39, 4.38)

0.670

0.69

(0.25, 1.90)

0.469

1.64

(0.36, 7.51)

0.521

Foreign-born

 Yes

Ref

  

Ref

  

Ref

  

 No

1.81

(0.57, 5.77)

0.314

2.36

(1.06, 5.29)

0.036

0.81

(0.27, 2.48)

0.717

Social factors

Lack of social support

 Low

Ref

  

Ref

  

Ref

  

 Moderate/high

1.87

(0.68, 5.17)

0.228

2.60

(1.11, 6.06)

0.027

1.22

(0.45, 3.26)

0.697

Any abuse

 No

Ref

  

Ref

  

Ref

  

 Abuse or fear of abuse

2.20

(1.07, 4.55)

0.032

1.75

(0.96, 3.19)

0.069

1.91

(0.80, 4.57)

0.147

Child protection involvement

 No

Ref

  

Ref

  

Ref

  

 Yes (prior or current)

1.46

(0.61, 3.49)

0.391

1.06

(0.49, 2.29)

0.886

1.55

(0.52, 4.66)

0.434

Living with father of the baby

 Yes

Ref

  

Ref

  

Ref

  

 No

2.76

(1.13, 6.73)

0.026

1.64

(0.89, 3.03)

0.112

1.33

(0.54, 3.27)

0.536

Behavioral factors

Alcohol use

 Low or no pre-pregnancy use/no prenatal use

Ref

  

Ref

  

Ref

  

 Any prenatal or high pre-pregnancy use

1.56

(0.74, 3.31)

0.247

1.18

(0.64, 2.19)

0.600

1.74

(0.71, 4.24)

0.223

Drug use

 No

Ref

  

Ref

  

Ref

  

 Yes (prenatal or pre-pregnancy)

0.97

(0.46, 2.08)

0.946

1.19

(0.65, 2.18)

0.572

0.61

(0.244, 1.54)

0.299

Cigarette use

 No prenatal/no or < daily pre-pregnancy

Ref

  

Ref

  

Ref

  

 Any prenatal use or daily pre-pregnancy

2.11

(1.01, 4.43)

0.048

1.01

(0.55, 1.84)

0.975

1.36

(0.56, 3.31)

0.504

Environmental factors

Lack of phone access

 Low

Ref

  

Ref

  

Ref

  

 Moderate/high

2.21

(0.89, 5.47)

0.086

1.23

(0.55, 2.76)

0.617

3.91

(1.46, 10.48)

0.007

Food insecurity

 Low

Ref

  

Ref

  

Ref

  

 Moderate/high

1.84

(0.93, 3.67)

0.082

2.44

(1.43, 4.16)

0.001

1.14

(0.51, 2.56)

0.749

Housing instability

 Low

Ref

  

Ref

  

Ref

  

 Moderate/high

1.25

(0.60, 2.61)

0.553

1.63

(0.91, 2.94)

0.103

0.56

(0.26, 1.23)

0.150

Discussion and conclusion

Our examination of the prevalence of elevated depression symptom levels found a higher rate in pregnancy (23 %) than in the postpartum period (14 %). These findings are consistent with other prospective studies of prenatal and postpartum depression (Andersson et al. 2006; Barnet et al. 1996; Evans et al. 2001; Felice et al. 2004; Gotlib et al. 1989; Gotlib et al. 1991; Rich-Edwards et al. 2006; Silva et al. 2012). For example, a study of over 9,000 women in Avon England found 13.5 % with probable depression at 32 weeks gestation and 9.1 % at 8 weeks postpartum (Evans et al. 2001), while a study of 192 low-income women found 41.7 % with depression during pregnancy and 23.4 % with postpartum depression (Hobfoll et al. 1995). Our finding that experiencing elevated symptom levels in the postpartum period only is the least likely scenario is also consistent with the Avon study (Heron et al. 2004).

Our findings that 29 % of the study sample had elevated depressive symptom levels either prenatally or postpartum, with the majority of these cases detectable during pregnancy, support the practice of systematically screening all women in the prenatal and postpartum periods. Fifty-eight percent of women who had a PHQ-9 score in the postpartum period high enough to trigger a clinical treatment plan (Kroenke 2002) also met that clinical cut off level during pregnancy. This finding is similar to other studies (Evans et al. 2001; Felice et al. 2004; Hobfoll et al. 1995; Barnet et al. 1996). In one study (Felice et al. 2004), 11 (55 %) of the 20 women with postpartum depression had been diagnosed during pregnancy. In a study of low-income US women, 53 % of the 45 women with postpartum depression had prenatal-onset (Hobfoll et al. 1995).

We found that 64 % of the women with elevated depressive symptom levels in the prenatal period exhibited symptom levels below the PHQ-9 score of 10 in the postpartum period. This rate of recovery or remission was higher than the 39 % reported in a study of 41 women screened as depressed in their third trimester (Sexton et al. 2012).

While we cannot assume a causal relationship, our findings of a higher prevalence of some behavioral, social, and environmental risks among women with depressive symptoms during pregnancy suggest that prenatal interventions to address abusive relationships, food insecurity, and social isolation might also decrease depressive symptom levels. Only one of the risk factors examined at prenatal intake, lack of phone access, was significantly associated with postpartum-onset of depressive symptoms. This measure is likely a proxy for extreme poverty and/or social isolation, which may become more problematic after childbirth. It should be noted that our regression model may have lacked sufficient power to detect some associations: The postpartum-onset sub-sample had only 35 women.

A few other studies have also examined correlates of depression or depressive symptoms in either the prenatal period, postnatal period or both periods, but they differ from our study in the selection of predictors and depressive symptom classifications. Some characteristics associated with postpartum recovery from prenatal depression included exercise, living with a partner (Sexton et al. 2012), the absence of a history of psychiatric disorder (Andersson et al. 2006), and lower prenatal depression scores (Gotlib et al. 1991). Persistent depressive symptoms through prenatal and postpartum were related to higher prenatal depression severity scores (Sexton et al. 2012). Postpartum-onset was associated with higher mean scores on their prenatal depression screening and reported higher perceived stress in pregnancy, lower marital satisfaction, greater use of “escape-avoidance” as a coping strategy, and more negative perceptions of the amount of caring they received from their own parents in one study (Gotlib et al. 1991). Another study (Andersson et al. 2006) found a history of psychiatric disorder, being single, and obesity was associated with postpartum-onset of depression.

Despite a variety of measures and different study samples, a common theme among prior studies and ours is the role of social support as measured by marital status, living with a partner, or marital/partner satisfaction. We found that not living with the father of the infant was associated with increased odds of elevated symptom levels in both prenatal and postpartum periods. In addition to these relationship measures, our social support composite measure was associated with increased depression in the prenatal-only group but not the other two depressive symptoms groups. Other studies have found associations for social support measures with postpartum depression (Barnet et al. 1996; Felice et al. 2004; Milgrom et al. 2008; Mohammad et al. 2011; Neter et al. 1995). Our study may differ from these because our measure was of prenatal social support; a measure of postpartum social support may have been more predictive of postpartum depression. Additionally, the sample may have been too homogeneous with regard to lack of social support to detect a meaningful role for this measure within the sample.

Prenatal and postpartum depression have serious risks (Bennett et al. 2004; Bonari et al. 2004; Davalos et al. 2012; Kingston et al. 2012; Lusskin et al. 2007). Identifying women with elevated depressive symptom levels through screening is the first step in preventing poor outcomes. The convergence of study findings indicate that screening women early in pregnancy will identify a majority of women who will experience elevated depressive symptoms in the perinatal period. Earlier identification will enable both individual clinicians to better address individual needs as well as the health care system to better address by understanding risk levels in populations served. Increased screening will also allow for the ability to study the effect of interventions to address prenatal depression on the reduction of postpartum depression and poor outcomes related to depression during either period.

Limitations

Sample attrition was high (60 %) but not surprising, given other reports about attrition of low-income populations from prenatal and parenting public health programs and services (Roggman et al. 2008; Katz et al. 2001). Clinic-based samples, like ours, are valuable because they are likely to be more representative of the diversity of women served by public programs than are samples of women who are specifically enrolled in longitudinal research studies. Such women may be very distinct if they are recruited or enrolled because they have qualities that are associated with long-term compliance and adherence (e.g., expectation of low residential transience; literacy). Thus, attrition may be one of the costs of studying a socially vulnerable sample. Like others who have studied samples similar to ours (O'Brien et al. 2012; Roggman et al. 2008), we believe that much of the attrition in our sample was related to the high level of residential transience of our socially vulnerable sample. And, consistent with our program protocol, women with medically high-risk pregnancies were transferred from our clinic sites to hospital-based specialty clinics and were thus unavailable for postpartum follow-up. In addition, we excluded another 7 % of the original sample because of data quality—another issue not as common in samples designed for research studies. We found few differences on a range of variables when we compared women who were included with those who were excluded from our study (Table 1). Three of the four characteristics where the groups demonstrated differences (i.e., race, housing stability, and tobacco use) were small differences, and only one factor (tobacco use) was related to depressive symptoms. The fourth difference was that excluded women were more likely to have their prenatal screening done in the first trimester. This finding is consistent with attrition due to the transfer of care because women who know they are medically at risk from past experience may have been more motivated to initiate care early.

Because our sample was predominantly urban women of diverse ethnic backgrounds who were served by Federally Qualified Health Centers, our findings may not be generalizable to populations who are more homogenous in terms of race, are of higher income, and/or are from rural areas. However, we have no evidence that our data are not generalizable. Our main finding—that the presence of depressive symptoms was higher in the prenatal period than the postpartum—is consistent with findings from other studies with diverse populations (Andersson et al. 2006; Barnet et al. 1996; Evans et al. 2001; Felice et al. 2004; Gotlib et al. 1989, 1991; Hobfoll et al. 1995; Rich-Edwards et al. 2006; Silva et al. 2012). Many of the risk markers we identified for prenatal and/or postpartum depressive symptoms have also been reported by others (e.g., social support, experience of domestic violence), and our findings thus add to an already established body of literature. The high prevalence of markers of social vulnerability in our sample (e.g., identification with historically disenfranchised race groups, non-marital status) allowed us to examine a sample in which public health programs have a high investment in serving. For example, we found that food insecurity—a potentially modifiable factor—was a risk marker for depressive symptoms. To our knowledge, few studies in developed countries have examined food insecurity in perinatal samples (Bitler et al. 2005; Herman et al. 2004; Laraia et al. 2009, 2006; Borders et al. 2007; Carmichael et al. 2007). We are aware of two studies, by the same study team, that have examined its association with prenatal depressive symptoms (Hromi-Fiedler et al. 2011, 2009). One of these two studies, a quantitative study with 135 subjects, reported that food insecurity was associated with an increased risk for prenatal depressive symptoms (adjusted odds ratio, 2.6; 95 % confidence interval (CI), 1.0, 6.5). Our findings were similar (adjusted odds ratio, 2.4; 95 % confidence interval, 1.4, 4.2) after adjustment for many more risk markers. We are aware of at least one study that also examined the association of food insecurity with postpartum depressive symptoms (Laraia et al. 2009) and found a slight association (adjusted Relative Risk, 1.09; 95 % CI 1.02 to 1.16), consistent with our findings of no association with more comprehensive adjustment. Our study extended knowledge about food insecurity by showing that it was not associated with depressive symptoms at both time periods. Furthermore, our disproportionately socially vulnerable sample allowed us to identify a risk marker for postpartum depressive symptoms—lack of phone access—that may serve as a simple and valid marker of economic risk in future studies with similar samples.

The measures of depressive symptoms were collected only at two points, once at entry into prenatal care and again about 4 to 12 weeks after delivery. We used these two measurement time points to describe depressive symptom status for the prenatal and postpartum periods, without information on how those symptoms potentially fluctuated during the time between measures. For example, women with high scores in both periods could have been experiencing persistent symptoms throughout or may have had a recovery and recurrence postpartum of those symptoms. Similarly, women who had high scores in the prenatal period but not in the postpartum may have represented a recovery from depressive symptoms or a remission period followed by recurring symptoms. While imperfect, the methodology to use measures at these time points as a proxy for the prenatal or postpartum period is similar to methods used in other studies.

This study was limited to risks measured during the prenatal period. Future examination of simultaneous measures of risk and depressive symptoms postpartum may provide additional evidence of the importance of the timing of measures and may yield stronger associations with postpartum-onset of depressive symptoms. We did not include some measures previously identified as associated with postpartum depression, such as pre-pregnancy psychiatric diagnosis or pregnancy intention (Lancaster et al. 2010), which may be useful in future examinations.

Conclusion

Our study found elevated depressive symptoms more common prenatally than postpartum. We found nearly one-fourth of prenatal clients scored at 10 or higher on the PHQ-9 at either time period which warrants a mental health assessment or treatment plan. Improved prenatal screening efforts would not only identify women earlier but allow more time to intervene to improve their health and ability to parent as well as to protect against poor birth outcomes associated with depressive symptoms. The finding that some women experience depression only in the prenatal period and some experience it persistently, suggests that aggressive intervention prenatally has the potential to reduce risk for postpartum depression. This study identified several social, behavioral, and environmental risk markers associated with depressive symptoms in the prenatal period only and during both prenatal and postpartum periods. How these risks and related interventions may be related to “recovery” warrants further research. Additionally, further research is needed related to postpartum-specific changes or development of risk markers that could be related to postpartum-onset.

Conflicts of interest

The authors have no financial conflicts of interest.

Copyright information

© Springer-Verlag Wien 2013