Social support, stress, and maternal postpartum depression: A comparison of supportive relationships

Social Science Research 54 (2015) 246–262

Contents lists available at ScienceDirect

Social Science Research journal homepage: www.elsevier.com/locate/ssresearch

Social support, stress, and maternal postpartum depression: A comparison of supportive relationships Keshia M. Reid ⇑, Miles G. Taylor Pepper Institute for Aging and Public Policy, 636 West Call Street, Florida State University, Tallahassee, FL 32306, United States

a r t i c l e

i n f o

Article history: Received 9 August 2013 Revised 18 May 2015 Accepted 27 August 2015 Available online 31 August 2015 Keywords: Postpartum depression Social support Stress exposure Stress process Family type Fragile Families and Child Well-being Study

a b s t r a c t A large body of literature documents the link between social support, stress, and women’s mental health during pregnancy and the postpartum period; however, uncertainty remains as to whether a direct effect or stress mediating pathway best describes the relationship between these factors. Moreover, specific dimensions of social support that may be influential (family type, sources of support) have largely been neglected. Using data from the Fragile Families and Child Well-being Study (N = 4150), we examine the pathway between social support, stress exposure, and postpartum depression in greater detail. Findings reveal that social support is a significant, protective factor for postpartum depression, and the variety of support providers in a woman’s social network is important, especially in the context of family type. Findings also reveal the importance of considering social support and stress exposure as part of a larger causal pathway to postpartum mental health. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction Maternal postpartum depression is a serious and wide-spread mental health disorder. An estimated one in seven women experience postpartum depression following childbirth (Wisner et al., 2006), though prevalence may be even higher among socially disadvantaged women (Earls, 2010). Women with postpartum depression often have trouble adjusting to new social roles (Logsdon et al., 2006, 2009) and experience impairments that disrupt daily life, including sleep disturbance, feelings of irritability or anxiety, loss of appetite, and crying (Chaudron, 2003; Robertson et al., 2004). These features of postpartum depression not only have consequences for a woman’s own heath, but may also negatively impact the health and wellbeing of her children (Chaudron, 2003). For example, children of mothers with postpartum depression are more likely to experience problems in cognitive, social, and emotional development and have a higher risk of anxiety disorders and major depression (Goodman and Gotlib, 1999). Moreover, the adverse effects of postpartum depression on children have been observed throughout the life course, from infancy to adulthood (Ertel et al., 2011). Given the prevalence of postpartum depression, and the short and long term health implications it has for women and their children, this disorder represents a public health concern (Wisner et al., 2006; Almond, 2009). Understanding the causal factors and pathways leading to postpartum depression is necessary to establish prevention and intervention efforts that reduce negative maternal-child health outcomes. Evidence reveals that high levels of social support are associated with a reduced risk of postpartum depression (Robertson et al., 2003, 2004), while life stress is associated with an increased risk of postpartum depression (Swendsen and Mazure, 2000). Social support has been shown to be

⇑ Corresponding author at: 1767 Hermitage Blvd. #5203, Tallahassee, FL 32308, United States. E-mail addresses: [email protected] (K.M. Reid), [email protected] (M.G. Taylor). http://dx.doi.org/10.1016/j.ssresearch.2015.08.009 0049-089X/Ó 2015 Elsevier Inc. All rights reserved.

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

247

effective in diminishing the harmful effects of life stress elsewhere in the broader mental health literature (Paykel, 1994; Thoits, 2011), suggesting that it may also work to reduce the risk of postpartum depression by mediating the effects of causal mechanisms, namely stress. Yet uncertainty remains as to whether a direct effect or mediating pathway best describes the relationship between social support and postpartum depression. In addition, specific dimensions of social support that may be influential, including a woman’s family structure and different sources of support, have largely been neglected. In response to these gaps in the literature, we use data from the Fragile Families and Child Well-being Study to test whether social support directly impacts the risk for postpartum depression, or whether it acts as a mediator, fully or partially governing the harmful effects of specific stressors. We also compare the independent effects of social support from a woman’s friends and family to that from an intimate partner, with particular attention to differences across family type. 2. Background 2.1. Overview of social support Broadly defined, social support refers to the social resources that one perceives to be available to them, or that are actually provided to them, from those within his or her social network (Gottlieb and Bergen, 2009; Thoits, 2011). The two types of social support that receive the most attention in the mental health literature are emotional support and instrumental support (Hopkins and Campbell, 2008; Gottlieb and Bergen, 2009; Thoits, 2011). Emotional support refers to demonstrations of love, esteem, empathy, and encouragement (Thoits, 2011) and lets an individual know that he or she is valued. Examples of emotional support include giving or receiving positive feedback or talking over a concern. Instrumental support, on the other hand, is the offer and/or supply of assistance with responsibilities and problems, such as help with babysitting or household chores (Beck, 2002; Thoits, 2011). It is thought that instrumental support may be the most effective in alleviating stress because it not only reduces situational demands but also conveys the message that one matters to others and is valued (Thoits, 2011). It should also be noted that these two types of support are often broken down further into received and perceived support. Received support involves actual provisions of support (Thoits, 2011), whereas perceived support refers to an individual’s belief that support is available should he or she need it (Gottlieb and Bergen, 2009). The most commonly used measures of social support in the mental health literature are measures of perceived support (Cohen et al., 2000), primarily because the effects of perceived support are stronger and consistently beneficial for mental health (Thoits, 2011). 2.2. Social support, mental health, and major depression Over the last few decades, substantial evidence has accumulated demonstrating a positive and causal relationship between social support and mental health, particularly major depression (Paykel, 1994; Kawachi and Berkman, 2001; Thoits, 2011). Social support has been shown to be a consistent protective factor in the risk for major depression, with deficits in support leading to increased symptoms and levels of depression (Horenstein and Cohen, 2008). This association has been repeatedly observed within both community and inpatient populations, across a wide range of ages, and for both men and women (Horenstein and Cohen, 2008). Two pathways have been used to describe how social support works to reduce the risk of major depression, including a direct effect (protective) pathway and an indirect (mediating) pathway (Thoits, 2011). A direct effect pathway explains that social support has protective effects for mental health (and subsequently major depression) because it improves health behaviors, increases positive feelings, and enhances emotional regulation (Horenstein and Cohen, 2008). An indirect or mediating pathway explains that social support attenuates the deleterious effects of life stress (Pearlin, 1989), thereby allowing for better emotional adjustment to negative events and helping to prevent major depression (Horenstein and Cohen, 2008). The idea that social support has stress mediating effects is generally consistent with the stress process framework of mental health (see Pearlin et al., 1981 for a detailed overview). The stress process framework provides a conceptual outline for understanding how, and under which conditions, life stress and social support impact mental health. Central to this framework is the idea that social support is part of a larger causal pathway to mental health that is best understood by examining it both by itself and in combination with life stress (Pearlin, 1989; Thoits, 2011). Research has verified both the direct and mediating effects of social support for major depression drawing on a stress process framework (e.g., Ensel and Lin, 1991). 2.3. Social support and postpartum depression Similar to the literature on major depression, social support has been shown to play a beneficial role in reducing the risk of postpartum depression (e.g., Surkan et al., 2006; Xie et al., 2009; Webster et al., 2011). Specifically, studies demonstrate that women who report less social support present with higher levels of postpartum depression than women with more support. Moreover, social support has been found to be one of the strongest predictors of postpartum depression (Beck, 2001; Robertson et al., 2004), highlighting the importance of this resource for maternal mental health following childbirth. Although an abundance of literature has confirmed the benefits of social support for postpartum depression, uncertainty remains as to whether a direct effect or mediating pathway best describes the observed relationship. That is to say, very little is known about how social support works to improve (or sustain) a woman’s postpartum mental health (Haslam et al., 2006).

248

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

Unlike the literature on major depression, most postpartum research treats social support as working through a direct effect or protective pathway (e.g., Dennis and Ross, 2006; Webster et al., 2011), without testing for potential mediating pathways as defined by the stress process framework. There is, however, evidence to suggest that social support should be considered from both a direct effect and mediating pathway. In previous research, life stress has been found to be a salient and robust risk factor for postpartum depression (Swendsen and Mazure, 2000; Herrick, 2000) and there appears to be a negative association between social support and postpartum depression that is more pronounced in the presence of stressors (Barnet et al., 1996; Glazier et al., 2004). This apparent association between social support and postpartum depression in the presence of life stress, coupled with robust evidence of mediating effects in the major depression literature, lends theoretical support for a potential mediating pathway linking social support to postpartum depression. Research is necessary to establish whether or not social support does in fact act as a mediator, fully or partially reducing the risk of postpartum depression by diminishing the effects of other causal mechanisms such as stress. 2.4. Differential effects by family structure and source of support Additional dimensions of social support, including personal and contextual factors, play a role in shaping the impact that support has on health outcomes (Pierce et al., 1996). With regard to postpartum depression, a woman’s family structure and the source(s) of support available in her social network may in large part shape the benefits she gains from social support. Family is a primary network of supportive relationships (Thoits, 2011), and may influence which sources of support are most relevant to a woman’s mental health after childbirth. Yet, few studies specifically consider the relevance of both family type and distinct sources of support in the risk for postpartum depression. In general, family members and a woman’s intimate partner are cited as both the most important sources of social support and the most available sources of support following childbirth (Logsdon et al., 1997). In many circumstances, intimate partner support appears to be uniquely beneficial to postpartum mental health. Support from an intimate partner has been found to be a consistent and significant protective factor for postpartum depression (Dennis and Ross, 2006; Dennis and Letourneau, 2007). Women who perceive stronger support from their partners mid-pregnancy report lower emotional distress postpartum (Stapleton et al., 2012), and those who have experienced postpartum depression indicate that it was helpful to have a supportive partner when coping with depressive symptoms (Letourneau et al., 2007). However, a vast majority of this literature is based on samples of married women, presumably because marriage provides a relatively homogeneous context within which to study social support (Beach et al., 1996). Subsequently, these findings may not translate to women in differing family types (e.g., single-mother families, cohabiting families). This is of particular importance given that intimate partner support appears to be particularly important for postpartum mental health, but women who are not married may not have access to stable support from an intimate partner. For example, women who give birth outside of marriage are more likely to experience partnership instability (Meadows et al., 2008) and perceive their partner as less supportive (Gallagher et al., 1997). It is also unclear whether live-in partners provide the same support as married partners for women in stable, cohabiting unions. It is possible that many of the benefits of marriage, including partner availability and supportiveness, are attributable to any live-in relationship (Meadows et al., 2008). Another possibility is that women who are not married rely on other sources of support (i.e., friends and family) in the absence of a stable intimate partner. Indeed, some women who have experienced postpartum depression report that family and friends, particularly female friends and relatives with whom they have trusting relationships, were particularly important sources of support (Letourneau et al., 2007). However, whether support from other sources can substitute for a lack of partner support remains in question. Though the lack of a stable marriage partner may result in increased contact with family members and friends (Marks and McLanahan, 1993), it is also possible that the demands of being a single mother limit a woman’s contact with family and friends altogether, decreasing her access to supportive exchanges (Marks and McLanahan, 1993). Cairney et al. (2003) found that single women report lower levels of perceived social support, social involvement, and frequency of contact with family and friends compared to their married counterparts. In short, differences between family types (married, single, single and cohabiting) may contribute to differences in the experience of social support, postpartum depression, and stress. Given increases in cohabitation and non-marital childbearing in the United States (DeKlyen et al., 2006), studies of postpartum depression that consider family type and distinct sources of support are increasingly important. 2.5. Type of social support As previously mentioned, the bulk of the evidence pointing to the significance of social support for mental health outcomes, including postpartum depression, comes from studies of perceived support (Glazier et al., 2004; Haslam et al., 2006; Horenstein and Cohen, 2008; Thoits, 2011). Higher levels of perceived support are associated with lower levels of postpartum depression (Glazier et al., 2004; Haslam et al., 2006), while the benefits of received support are less clear (Robertson et al., 2003). There is also specific evidence for the salience of perceived support in the major depression and stress process literature (Turner and Lloyd, 1999; Cohen et al., 2000), whereas received support has yielded null to weak and often contradictory effects (Horenstein and Cohen, 2008; Thoits, 2011). This aligns with the consensus that measures that tap into an individual’s perceptions of support are the most consistent and strongest predictors of personal adjustment (Pierce et al., 1996). These findings, coupled with evidence that social support works in a similar manner for both postpartum and major depression, suggest that a focus on perceived support is more relevant for investigations of postpartum depression.

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

249

3. Present study While high levels of social support are associated with a reduced risk of postpartum depression, and life stress is a salient risk factor for developing postpartum depression, uncertainty remains as to whether a direct effect or mediating pathway best describes the relationship between social support and postpartum depression. In the present study, we draw on a stress process framework to test whether social support directly impacts the risk for postpartum depression, or whether it acts as a mediator, governing the harmful effects of specific stressors. We also account for specific dimensions of social support that may play a role in shaping the impact of social support on postpartum depression, including a woman’s family type and different sources of support available to her. Based on the literature outlined above, we developed several hypotheses. Hypothesis 1. In general, social support and life stress will each be independently predictive of postpartum depression for all women.

Hypothesis 2. Social support will not only have a direct, positive effect on postpartum depression (independent of life stress), but will also fully or partially mediate the detrimental effects of stressors. Hypothesis 3. Findings will vary across family type, such that the impact of specific stressors and the source of social support that is most important (intimate partner or friend and family) will be unique to women in each family type (married, single and cohabiting, single).

4. Methods 4.1. Data The data for this study are from the Fragile Families and Child Well-being Study, an NIH funded longitudinal sample of approximately 4900 births in 75 randomly selected hospitals across 20 U.S. cities (in a total of 15 states). The Fragile Families relies on a stratified, multistage, probability sample of new, mostly unwed (3600 unwed, 1300 married) parents and their children (Reichman et al., 2011) and contains rich information relevant to social support, life stress, and postpartum depression. While the Fragile Families is not nationally representative of mothers at risk for postpartum depression, one substantial benefit of this data is its large sample size. Much of the research on postpartum depression has relied on small, clinical samples (e.g., Misri et al., 2000; Dennis and Ross, 2006; Gremigni et al., 2011). Furthermore, the Fragile Families provides an ideal sample for comparing women in diverse family types and oversamples women with the highest risk of postpartum depression, since the respondents are mainly young, economically disadvantaged, and have reduced social resources. Although the Fragile Families was not explicitly created to study postpartum depression, recent research has utilized this data source to examine this outcome (Mitchell et al., 2011; Reid and Taylor, in press). The Fragile Families data includes five waves of measurement. Baseline measurement occurred in-person with parents in the hospital shortly after childbirth. Follow-up measurements were conducted either in-person or by telephone one year after the birth of the focal child and again when the child was 3, 5, and 9 years of age. Medical record data abstracted from the birth hospitalization record is also available for 3684 mothers and focal children. The remainder of respondents are missing medical record data for one of three reasons: (1) the hospital did not permit researchers to abstract records or there were too few cases for it to be financially feasible to collect data at that hospital, (2) the mother refused consent, or (3) the records could not be located in the hospital (http://www.fragilefamilies.princeton.edu/medrecs.asp). Medical record data is based on information collected from a data abstraction tool; therefore, a majority of the variables in the medical record data file are constructed by the Fragile Families and some specific items (e.g., ICD codes, open-ended responses) are not available to researchers. 4.2. Sample At baseline, the Fragile Families data consisted of 4898 women. We limit analyses to include only those women who participated in the one year follow-up and had complete information on the postpartum depression measure (n = 4362). Women missing data on marital status (n = 3) and women who indicated that their current intimate partner was a female (n = 7) were excluded. Although we incorporate an estimation strategy that allows missingness on predictor variables (full information maximum likelihood (FIML)), this strategy is computationally intense. Therefore, on variables where missingness was 1% or lower we used listwise deletion (excluding 202 women, a sample reduction of 5%). The final analytic sample for this study includes 4150 women. Missing values for all other variables were handled through FIML estimation which estimates a likelihood for any individual given their available information. This allows all individuals to be included in analyses without complete data. Findings from sensitivity analysis using listwise deletion are similar to those reported here. We use unweighted data for analyses because weighting decreases the analytic sample by 30% and the weights are not appropriate when using the medical records. However, the population based sampling strategy used by the Fragile Families eliminates

250

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

many of the selection biases often associated with clinical and convenience samples used to study depression (Mitchell et al., 2011). In addition, the Fragile Families has more variable population characteristics in terms of education and race/ethnicity than previous studies (Mitchell et al., 2011). Additionally, including the variables used for sample selection (socioeconomic status, race, age, education) as controls in the models suggests that the coefficients should be unbiased, consistent, and robust (Winship and Radbill, 1994). For the purposes of this study, we split the sample into three mutually exclusive groups of women (single, single and cohabiting, and married).

4.3. Measures 4.3.1. Postpartum depression Our measure of postpartum depression is based on a probable diagnosis of major depression from the Composite International Diagnostic Interview – Short Form (CIDI-SF), Section A (Kessler et al., 1998) at the time of the one year follow-up. The CIDI-SF takes a portion of questions from the full CIDI (the full CIDI is a standardized instrument for assessment of mental disorders consistent with criteria from the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; APA, 1994) and estimates the probability that a respondent would be positively diagnosed with major depression. Specifically, women were asked if they had experienced feelings of depression or anhedonia in the past year (that is, the first year following childbirth), which lasted two weeks or more. If so, they were asked about seven additional symptoms: (1) losing interest, (2) feeling tired, (3) weight changes, (4) trouble with sleep, (5) trouble concentrating, (6) feeling worthless or (7) thinking about death. Women who answered affirmatively to having three or more of these symptoms were considered to have depression (1 = depressed, 0 = not depressed). Although the CIDI-SF was created to assess major depression broadly and not postpartum depression specifically, the patterns of symptoms in women with postpartum depression are similar to those in women who have major depression (Robertson et al., 2003). Postpartum depression is typically distinguished by onset within the year following childbirth (Gaynes et al., 2005); therefore, we argue that women who meet the criteria for a probable diagnosis of major depression in the first year following childbirth are likely experiencing postpartum depression. At least one other study has also used the CIDI-SF depression measure in the Fragile Families data to capture postpartum depression (Mitchell et al., 2011).

4.3.2. Previous depression Previous depression or mood disorder is one of the strongest predictors of postpartum depression (Chaudron, 2003; Horowitz and Goodman, 2005). Moreover, women with pre-existing depression may be less likely to seek social support (Pierce et al., 1996) and more likely to accumulate and/or report stress over the life course (Wheaton, 1994). To account for these factors, we use a constructed measure of previous depression available in the medical record data as a control (1 = previous depression, 0 = no previous depression). Women who did not seek medical attention for depression, were misdiagnosed, or whose healthcare provider did not record a diagnosis of depression on the medical record will not be captured in this measure. In addition, no supplemental self-reported information (e.g., anti-depressant prescriptions, visits with a mental health professional) was available to improve the reliability of this measure. As such, this measure likely reflects an underreporting of previous of depression. Nevertheless, we argue that this measure is a valuable control given that diagnosed cases of depression are usually more severe while undiagnosed cases tend to be mild and associated with higher functioning (Coyne et al., 1995).

4.3.3. Social support As noted above, the bulk of the evidence pointing to the significance of social support for mental health outcomes (including postpartum depression) comes from studies of perceived support (Robertson et al., 2003; Horenstein and Cohen, 2008; Thoits, 2011) and there is specific evidence for the salience of perceived support over received support in the major depression literature rooted in the stress process (Turner and Lloyd, 1999; Cohen et al., 2000). Moreover, the Fragile Families focuses primarily on perceived support from family and friends. We therefore use two measures of perceived social support in this study, due to both its relevance to postpartum mental health and for the purposes of comparing support from family and friends to that from an intimate partner. In addition, both of these measures of social support focus on instrumental support. Instrumental support from primary group members (i.e., family, friends, and/or a significant other) is thought to be the most effective in alleviating the harmful effects of stress given that it lessens situational demands and conveys a message of caring and esteem (Thoits, 2011). Friend and family support is a summed scale of five ‘‘yes/no” items taken from the one-year follow up interview (see Appendix A), measuring women’s perceptions of help available to them from family and friends. For each item in this measure, women were specifically asked if they could they count on someone other than the focal child’s birth father. Intimate partner support is measured using two items taken from the one-year follow up (see Appendix A). This measure captures women’s perceptions of how often help is available to them from a current romantic partner or the focal child’s birth father (including single mothers or those still married, cohabiting, or romantically involved with the birth father). Responses range from 0 (rarely) to 2 (always). For each measure of social support, higher values are equivalent to greater levels of support.

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

251

4.3.4. Life stress Central to this study is the idea that social support may act as part of a larger causal pathway to postpartum depression involving stress. In studies rooted in a stress process framework, measures of stress are most often defined by exposure to an inventory of acute life events that have occurred within a specified time frame (major/lifetime or recent), or exposure to more frequent, recurring chronic stressors (Pearlin, 1989; Turner and Lloyd, 1999). Consistent with a stress process framework, we include several indicators of acute life event and chronic stress in this study. The Fragile Families does not include acute life event inventories like those traditionally found in the stress process literature; however, it does ask women to report on various individual items that correspond with these inventories (for examples, see Turner and Lloyd, 2004; Peirce et al., 2009). Thus, we created two acute life event indices based on relevant items to approximate acute life event inventories. Major life stress ranges from 0 to 6, and assesses undesirable or negative events occurring at any point within a woman’s lifetime. Seven ‘‘yes/no” items are used to measure major life stress (see Appendix B). All items are taken from the baseline interview with exception of one item, capturing whether or not a woman has ever been forced to perform sexual activities. Although this item is from the one year follow-up, we argue that it should be considered a major life stressor because women who have been forced to perform sexual activities were likely forced to do so prior to pregnancy and birth (Beydoun et al., 2010; Scribano et al., 2013). In prior analyses of the stress indices (Reid and Taylor, 2015), alternate models excluding this variable from the measure of major life stress were tested with substantively similar results, thus we chose to retain it for analyses. In contrast to major life stress, our indicator of recent life stress assesses undesirable or negative events occurring within the twelve month period following childbirth. Recent life stress ranges from 0 to 5, and is measured using five ‘‘yes/no” items from the one-year follow up (see Appendix B). We also include four indicators of exposure to more frequent, recurring chronic stressors resembling those found in the stress process literature (see Turner and Lloyd, 1999 for examples). Neighborhood safety is measured using the baseline item ‘‘How safe are the streets around your home at night?” Responses range from 0 (very safe) to 3 (very unsafe). Child-related stress is a summed scale of seven items from the one year follow-up (see Appendix B), measuring the impact of the focal child’s temperament on the risk for postpartum depression. Parenting-related stress is measured using a scale of four items from the one year follow-up (see Appendix B) and captures the stressors associated with raising a family. Responses range from 0 (strongly disagree) to 3 (strongly agree). Because supportive relationships are not necessarily free from conflict (Hagerty and Williams, 1999), a measure of relationship strain is also included. Relationship strain is a summed scale of six items from the baseline survey measuring how often women report disagreements with the focal child’s birth father (see Appendix B). Responses range from 0 (never) to 2 (often). For each of the life stress measures, higher values are equivalent to greater levels of stress. 4.3.5. Intimate partner violence Intimate partner violence may work as a major life event stressor occurring before or after birth (Turner and Avison, 2003) or a recurring chronic relationship stressor (Adkins and Kamp Dush, 2010). In addition, intimate partner violence often escalates around the time of pregnancy and birth and is particularly impactful on depressive symptoms (Gustafsson and Cox, 2012). We therefore include a separate indicator of intimate partner violence capturing how often women experienced some form of domestic violence during the course of their relationship with the focal child’s birth father. Intimate partner violence is measured using a series of questions asked at baseline and the one-year follow up (see Appendix B). Responses for each item range from 0 (never encountered violence) to 2 (often encountered violence). 4.3.6. Controls Our analyses include a number of controls that have been identified as correlates of social support, postpartum depression, and life stress. Sociodemographic controls include mother’s age (in years), race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other; Non-Hispanic white = reference category), nativity (1 = U.S. native, 0 = non-native), education (less than high school, high school, some college, and college degree or above; less than high school = reference category), and employment (dummy variable indicating receipt of income from earnings in the past year). Household income is measured in thousands of dollars. Due to low levels of reporting, missing values on income were imputed by regressing household income on age, race/ethnicity, education, marital status, poverty level, and presence of other biological children. Other biological children (1 = yes, 0 = no) and number of children in the household (range from 0 to 10) are also included as controls. Finally, characteristics of the focal child are controlled for, including child’s health at birth (defined as low birth weight and/or preterm; 1 = yes, 0 = no), and both child’s age (in months) and any physical disability (1 = yes, 0 = no) at the time of the one-year follow-up. 4.4. Analytic methods To address our study questions, we estimate multiple probit regression models using a structural equation modeling (SEM) framework. Structural equation modeling allows the testing of simultaneous direct and indirect pathways in analyses and has the added benefit of a full information maximum likelihood (FIML) estimator. FIML estimation allows individuals to contribute to analyses given any available information. This approach also has the advantage of the missing at random (MAR) assumption, an improvement over the assumption underlying listwise deletion.

252

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

Our analyses proceed in several steps. First, we estimate the direct effects of stressors on postpartum depression for each group of women (presented as Model 1 for each group), controlling for previous depression, sociodemographic characteristics, and characteristics of the focal child. Next, we enter both measures of social support into the previous model simultaneously (presented as Model 2 for each group). This allows us to examine the effects of social support and note any subsequent changes in the relationships between stressors and postpartum depression from Model 1 to Model 2. Finally, we evaluate potential mediating effects of social support. Following Baron and Kenny (1986; see also Judd and Kenny, 1981; MacKinnon et al., 2002), we use the Clogg test for equality of regression coefficients to determine if either measure of support substantially and significantly influences the relationship between stressors and postpartum depression (Clogg et al., 1995; Paternoster et al., 1998; MacKinnon et al., 2002). Path analyses displaying the standardized coefficients for each of these pathways are presented to further illustrate each of the above relationships. Because of the complex sampling design of the Fragile Families, we adjusted all models to correct for clustering by hospital. Analyses are performed using Mplus version 6 (Muthén and Muthén, 2007).1 5. Results 5.1. Descriptive analyses Descriptive statistics for the full sample and each group of women (single, single and cohabiting, and married) are presented in Table 1. Means, standard deviations, and ranges are reported for each variable. T-tests for differences in mean values, also reported in Table 1, reveal significant differences in sociodemographic characteristics between women in varying family types. Married women are more likely to be white, older (28 years of age vs. 24 years of age), and college educated (i. e., some college or college degree/above) compared to cohabiting and single women. In addition, married women are more likely to be employed and have a higher mean level of household income (M = 42.2 thousand vs. 23.5 thousand or less). Married women are, however, less likely to be U.S. natives. Compared to single women, both married and cohabiting women are more likely to have other biological children in the household and have a focal child younger in age at the time of the oneyear follow-up. In contrast, cohabiting and single women resemble one another in regard to the health of the focal child at birth. However, single women are more likely to report that the focal child has a physical disability (compared to both married and cohabiting women). Despite some similarities, cohabiting and single women in this sample are overall more economically and socially disadvantaged than married women. Significant differences in depression, social support, and stressors are evident across family type in Table 1 as well. Overall, the prevalence rates of postpartum depression among women in this sample are consistent with national prevalence rates (e.g., Chaudron, 2003). However, married and cohabiting women have the lowest mean values of depression (M = 0.12 and 0.14 respectively), while single women have a significantly higher mean value (M = 0.20). Single women are also substantially more likely to have a documented history of previous depression (M = 0.13). In fact, the prevalence of previous depression among single women in this sample is slightly higher than national estimates (findings from the National Health and Nutrition Survey III estimate the prevalence rate among women to be about 12.6%; Riolo et al., 2005). In comparison, only a small number of married and cohabiting women have a documented history of previous depression (M = 0.06 and 0.08, respectively). In terms of social support, married and cohabiting women report similar and substantial mean levels of perceived support from an intimate partner (M = 3.76 and 3.79, respectively). Single women, on the other hand, report somewhat lower mean levels of perceived partner support (M = 2.74). Perceptions of support from friends and family vary as well. Married women report the highest mean level of friend and family support (M = 4.71), with cohabiting women reporting a slightly lower mean level (M = 3.94) and single women reporting the lowest mean level (M = 3.51). With regard to stressors, the average woman has experienced few major life and recent life eventful stressors and low to moderate levels of chronic stress. In addition, very few women report a history of intimate partner violence with the focal child’s father (M = 0.20 or less). Notably, however, meaningful differences in mean levels of stress exposure are evident across family type. Single women report significantly higher mean levels of exposure across all stress indicators (except for neighborhood safety where they resemble cohabiting women), while married women report the lowest mean levels of exposure across all stress indicators. 5.2. Multivariate analyses 5.2.1. Direct effects of stress The direct effects of specific stressors on postpartum depression are presented in Models 1 of Table 2. All relationships in Table 2 are presented as unstandardized probit coefficients. In line with previous research and consistent with our first hypothesis, we find that life stress overall has negative implications for maternal mental health. According to Models 1, life stress is positively associated with postpartum depression, net of controls and previous depression. In partial support of our third hypothesis, several differences across family type are observed. Major life event stress is predictive of postpartum 1 In preliminary analyses not shown here, stress by support interaction terms were introduced into the full model (Model 2) for each group of women. No evidence of moderating effects were found.

253

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262 Table 1 Sample descriptive statistics. Variable

Range

Full sample

N

Married

Cohabiting

Single

Differencea

Postpartum depression Previous depression Age

0–1 0–1 15–43

0.15 (0.36) 0.09 (0.29) 25.15 (6.02)

4150 3116 4150

0.12 (0.32) 0.06 (0.25) 28.21(5.95)

0.14 (0.34) 0.08 (0.28) 23.66 (5.33)

0.20 (0.40) 0.13 (0.33) 23.82 (5.70)

MS, CS MS, CS MC, MS

Race/ethnicity Non-Hispanic white Non–Hispanic black Hispanic Other

0–1 0–1 0–1 0–1

0.22 0.47 0.27 0.04

4150 4150 4150 4150

0.39 0.26 0.28 0.06

0.16 0.51 0.31 0.02

0.12 0.63 0.22 0.02

MC, MS, CS MC, MS, CS MS, CS MC, MS

Education Less than high school High school Some college College or more Household income (thousands) U.S. native Employed Child’s age (months) Other biological children No. children in household Child’s health at birth Child physical disability

0–1 0–1 0–1 0–1 1.73–75 0–1 0–1 9–30 0–1 0–10 0–1 0–1

0.34 (0.47) 0.30 (0.46) 0.25 (0.43) 0.11 (0.31) 29.05 (21.37) 0.84 (0.37) 0.69 (0.46) 14.99 (3.46) 0.62 (0.49) 2.31 (1.32) 0.14 (0.35) 0.24 (0.15)

4150 4150 4150 4150 4150 4150 4150 4150 4150 4150 3113 4150

0.19 (0.39) 0.23 (0.42) 0.29 (0.45) 0.29 (0.45) 42.15 (22.75) 0.75 (0.43) 0.73 (0.44) 14.69 (3.42) 0.64 (0.48) 2.20 (1.19) 0.11 (0.31) 0.01 (0.12)

0.40 (0.49) 0.35 (0.48) 0.22 (0.42) 0.02 (0.14) 23.50 (17.73) 0.86 (0.35) 0.69 (0.46) 14.79 (3.29) 0.64 (0.48) 2.31 (1.34) 0.15 (0.36) 0.02 (0.14)

0.41 (0.49) 0.32 (0.47) 0.24 (0.42) 0.03 (0.18) 22.53 (17.72) 0.91 (0.29) 0.66 (0.47) 15.49 (3.63) 0.57 (0.50) 2.42 (1.41) 0.16 (0.37) 0.04 (0.19)

MC, MS MC, MS MC, MS MC, MS MC, MS MC, MS, CS MC, MS MS, CS MS, CS MC, MS, CS MC, MS MC, MS, CS

Social support Intimate partner support Family and friend support

0–4 0–6

3.49 (0.98) 4.04 (1.83)

3298 4146

3.76 (0.58) 4.71 (1.70)

3.79 (0.57) 3.94 (1.76)

2.74 (1.38) 3.51 (1.83)

MS, CS MC, MS, CS

Stressors Major life event stress Recent life event stress Neighborhood safety Relationship strain Intimate partner violence Child-related stress Parenting-related stress

0–6 0–5 0–3 0–12 0–2 0–25 0–12

1.76 (1.19) 0.75 (0.77) 0.94 (0.71) 2.56 (2.29) 0.10 (0.39) 11.26 (3.98) 4.08 (2.96)

2983 3862 4150 4067 4150 4150 4150

1.16 (1.03) 0.55 (0.68) 0.77 (0.68) 1.90 (1.81) 0.03 (0.19) 10.54 (3.81) 3.85 (2.73)

1.87 (1.15) 0.81 (0.76) 1.02 (0.71) 2.60 (2.26) 0.08 (0.33) 11.42 (3.84) 3.97 (2.91)

2.16 (1.15) 0.88 (0.82) 1.02 (0.72) 3.12 (2.56) 0.20 (0.55) 11.78 (4.18) 4.44 (3.19)

MC, MS, MC, MS MC, MS, MC, MS, MC, MS, MC, MS, MS, CS

(0.41) (0.50) (0.44) (0.19)

(0.49) (0.44) (0.45) (0.25)

(0.36) (0.50) (0.46) (0.15)

(0.33) (0.48) (0.41) (0.16)

CS CS CS CS CS

MC = Married and Cohabiting; MS = Married and Single; CS = Cohabiting and Single. a T-tests for differences in mean values; significant at .05 or higher.

depression for women across all three family types, but the impact of major life event stress is greatest for married women and decreases in both magnitude and significance for cohabiting and single women respectively. In contrast, recent life event stress is only predictive of postpartum depression for cohabiting women. The finding that major life event stress is significant for women in all three family types while recent life event stress is not suggests that stressful life events occurring before and after childbirth work independently of one another and major life event stressors have lasting effects for all women, while recent life event stressors are only salient for cohabiting women. In terms of chronic stress, the stressors related to parenting and children are positively associated with postpartum depression for women across all three family types, though differences in the magnitude and significance of these associations are observed. For example, the magnitude of parenting-related stress is greatest for married women, and child-related stress is the least substantial in magnitude for cohabiting women. In contrast to the stressors related to parenting and children, the stress associated with living in an unsafe neighborhood is significant for cohabiting and single women (not married women) and being in a partnership marked by conflict is only significant for single women. That all chronic stressors are significantly associated with postpartum depression for single women is consistent with patterns noted in Table 1 (single women report the highest levels of chronic stress exposure). Of particular interest, cohabiting and single women fare at least as well as married women (if not better) in terms of the magnitude of significant chronic stressors. Nevertheless, these findings indicate that chronic stress (compared to life event stress) plays a substantial role in the risk for postpartum depression among single women. Models 1 also reveal that intimate partner violence has a substantial impact on postpartum depression for single women, yet is not predictive of postpartum depression for married or cohabiting women when other types of stress are included in the model. Moreover, intimate partner violence has the largest magnitude among the significant stressors affecting single women. These findings not only highlight the importance of this particular stressor for single women, but also suggest that intimate partner violence is more in line with a recurring (i.e., chronic) stressor in its overall impact on single women (Adkins and Kamp Dush, 2010), given its similarities in significance to the chronic stressors discussed above. Finally, Models 1 in Table 2 display the predictive significance of sociodemographic and family characteristics. Age is a non-significant net predictor of postpartum depression across family type. Thus, the risk for postpartum depression does not vary significantly by age. Likewise, the risk for postpartum depression does not vary according to U.S. nativity. Some

254

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

Table 2 Unstandardized coefficients (SE) of stressors and social support on postpartum depression by family structure. Married (N = 1305) Model 1

a b c * ** ***

0.40 0.01 0.04 0.19 0.16 0.40 0.26 0.43 0.01 0.17 0.24 0.00 0.00 0.02 0.00 0.25 0.17 0.01 0.03 0.03 0.04 0.10 0.27

(0.17)* (0.01) (0.10) (0.15) (0.29) (0.18)* (0.20) (0.27)** (0.00) (0.13) (0.10)* (0.01) (0.11) (0.03) (0.15) (0.52) (0.05)*** (0.05) (0.06) (0.02) (0.01)*** (0.02)*** (0.33)

Previous depression Age Non-Hispanic black Hispanic Other race High School Some college College or more Household income U.S. native Employed Child’s age (months) Other biological children No. children in household Child’s health status Child physical disability Major life stress Recent life stress Neighborhood safety Relationship strain Child-related stress Parenting-related stress Intimate partner violence Intimate partner support Friend/family support

– –

Chi-square RMSEAa CFIb WRMRc

294.86*** 0.04 0.17 1.93

Cohabiting (N = 1481) Model 2

Model 1

0.33 0.01 0.04 0.19 0.16 0.40 0.26 0.44 0.01 0.17 0.24 0.00 0.00 0.02 0.03 0.26 0.15 0.03 0.03 0.03 0.04 0.10 0.27 0.16 0.11

0.77 0.00 0.22 0.39 0.17 0.08 0.04 0.12 0.00 0.01 0.01 0.00 0.05 0.04 0.12 0.20 0.13 0.12 0.11 0.00 0.02 0.05 0.24

(0.17) (0.01) (0.10) (0.15) (0.29) (0.18)* (0.20) (0.27) (0.00) (0.13) (0.10)* (0.01) (0.11) (0.03) (0.15) (0.52) (0.05)** (0.05) (0.06) (0.02) (0.01)*** (0.02)*** (0.33) (0.05)*** (0.02)***

390.65*** 0.04 0.19 1.98

(0.11)*** (0.01) (0.11)* (0.13)** (0.29) (0.08) (0.09) (0.24) (0.00) (0.10) (0.08) (0.01) (0.10) (0.03) (0.11) (0.28) (0.04)** (0.05)* (0.06)* (0.02) (0.01)* (0.01)*** (0.15)

– – 330.08*** 0.04 0.12 2.07

Single (N = 1364)

Model 2 0.73 0.00 0.22 0.39 0.17 0.08 0.04 0.12 0.00 0.01 0.01 0.00 0.05 0.04 0.11 0.20 0.11 0.11 0.11 0.01 0.02 0.05 0.24 0.22 0.07

(0.11)*** (0.01) (0.11)* (0.13)** (0.29) (0.08) (0.09) (0.24) (0.00) (0.10) (0.08) (0.01) (0.10) (0.03) (0.11) (0.28) (0.04)** (0.05)* (0.05)* (0.02) (0.01)* (0.01)*** (0.15) (0.07)*** (0.05)***

367.48*** 0.04 0.24 1.93

Model 1 0.32 0.00 0.24 0.10 0.15 0.14 0.35 0.22 0.00 0.04 0.14 0.00 0.24 0.05 0.05 0.29 0.08 0.06 0.17 0.04 0.04 0.06 0.32

(0.11)** (0.01) (0.13)** (0.17) (0.27) (0.11) (0.10)*** (0.20) (0.01) (0.17) (0.06)* (0.01) (0.09)** (0.03) (0.08) (0.16) (0.04)* (0.05) (0.07)* (0.01)*** (0.01)*** (0.01)*** (0.05)***

– – 247.11*** 0.04 0.29 1.71

Model 2 0.33 0.00 0.24 0.10 0.15 0.14 0.35 0.22 0.00 0.04 0.14 0.00 0.24 0.05 0.07 0.29 0.06 0.06 0.17 0.04 0.04 0.06 0.32 0.07 0.07

(0.11)** (0.01) (0.13) (0.17) (0.27) (0.11) (0.10)*** (0.20) (0.00) (0.17) (0.06)* (0.01) (0.09)** (0.03) (0.08) (0.16) (0.04) (0.05) (0.07)* (0.01)** (0.01)*** (0.01)*** (0.05)*** (0.03)* (0.02)**

340.12*** 0.03 0.26 1.79

Root mean square error of approximation. Comparative fit index. Weight root mean square residual. p < 0.05. p < 0.01. p < 0.001.

variation in risk was observed by racial/ethnic minority status. Non-Hispanic black women who are single or cohabiting are less likely to report postpartum depression, as are Hispanic cohabiting women. Although income is a non-significant predictor, employment and education are significant for married and single women. Employment is associated with an increase in the risk of postpartum depression for married women, but a decrease in risk for single women. In contrast, single women who report at least some college education experience an increase in the risk for postpartum depression, while married women who report a high school or college or more education experience a decrease in the risk for postpartum depression. The presence of other biological children is associated with an increase in risk for single women; however, the number of children in the household is a non-significant predictor across family type. Characteristics of the focal child (age, health status, physical disability) were also non-significant across family type. However, previous depression is associated with an increase in the risk for postpartum depression for all women.

5.2.2. Unique effects of social support In Models 2 of Table 2, both sources of social support are entered into Models 1 simultaneously to determine whether either source of social support has a unique effect on postpartum depressive symptoms. Models 2 provide additional support for our hypotheses. In line with our first hypothesis, findings indicate that supportive relationships are associated with lower levels of postpartum depression. Moreover, the importance of different sources of support (providers) varies by family type as predicted in our third hypothesis. Both intimate partner support and friend and family support are independently associated with a significantly lower risk of postpartum depression for all women, regardless of family type. For married and cohabiting women, these associations are highly significant. Of particular interest, intimate partner support has a greater impact on depression for both married and cohabiting women compared to friend and family support, but is equal in magnitude to friend and family support for single women. Additionally, although friend and family support is significant for all women, it has the greatest impact on married women and is equal in magnitude for cohabiting and single women. With the addition of these variables to Models 2, several changes in stress coefficients are observed providing partial support for our second hypothesis that social support works to mediate the effects of stress. The coefficients for major life event

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

255

stress are reduced from Model 1 for married and cohabiting women but remain highly significant. In comparison, the coefficient for major life event stress is reduced to non-significance for single women. The coefficient for recent life event stress is also slightly reduced from Model 1 for cohabiting women, though no change in significance is observed. Several sociodemographic characteristics change from Model 1 to Model 2 as well. Of particular importance, the coefficient for previous depression is reduced for cohabiting women, remains relatively unchanged for single women, and is no longer significant for married women. The effects of education and race/ethnicity also change slightly. A college or more education is reduced to non-significance for married women, as are the race effects for single non-Hispanic black women. 5.2.3. Mediating effects The addition of support variables into the regression equation reduces the coefficients for certain stressors from Models 1 to Models 2. In additional analyses not presented here, results revealed that women who are exposed to more stressors also tend to have lower levels of social support. These findings are suggestive of our second hypotheses that social support works to mediate the effects of stressors; therefore, we formally test for mediating effects following in line with Baron and Kenny (1986). Using the Clogg test for equality of regression coefficients (Clogg et al., 1995; Paternoster et al., 1998; MacKinnon et al., 2002), we calculate whether either source of social support significantly diminishes the effects of stressors from Model 1 to Model 2 (see Appendix C) for each group of women. Results from the Clogg test calculations reveal that none of the observed reductions in stressors are significant, suggesting that social support does not work to mediate the relationship between stress exposure and postpartum depression refuting our second hypothesis. 5.2.4. Path analyses To further illustrate the relationships between social support, stress, and postpartum depression, path analyses are presented for each group of women (Figs. 1–3). Path analyses allow for a simultaneously estimated re-examination of the findings reported in the previous steps and provide additional backing for the findings listed above. These path analyses do not have the benefit of a significance test for mediation, since the mediating and outcome variables are of different types (support variables are assumed to be continuous whereas postpartum depression is binary). Therefore, direct, indirect, and total effects could not be calculated as in traditional path analysis (where both mediating and outcome variables are continuous). However, the path analyses do have the benefit of providing additional support to the probit regression models presented above while estimating all relationships simultaneously. In addition, they are a straightforward way of presenting standardized coefficients in order to examine which factors matter most in predicting both social support and postpartum depression. In all, results from the path analyses support the findings presented above with some notable additions. With the exception of neighborhood safety and child-related stress for cohabiting women, the significant relationships between stressors and postpartum depression (presented in Models 1 of Table 2) remain consistent in the path analyses, suggesting that the findings from the first step of analyses are robust. However, when comparing standardized coefficients in the figures to the unstandardized coefficients in the table, several differences are apparent. Parenting stress has the largest impact on depression among married and cohabiting women compared to other life event and chronic stressors, and child and parenting-related stressors have the largest impact on depression among single women when compared to other stressors. The figures also replicate the substantive findings reported in the second step of analyses. Numerous stressors reduce social support, and these effects vary by both the source of support and a woman’s family structure. Lastly, the figures paint a similar picture to that reported in the third step of analyses. Both sources of social support significantly reduce depression for women in all three family types, independent of the effects of stress on depression or support. However, a closer look at the standardized coefficients reveals that both sources of support have a relatively similar impact on cohabiting women, but friend and family support has a larger impact on depression among married and single women. 6. Discussion and conclusion In an attempt to strengthen current understanding of the relationship between social support, stress, and postpartum depression, we used data from the Fragile Families and Child Well-being Study to test whether social support directly impacts the risk for postpartum depression, or whether it acts as a mediator, fully or partially governing the harmful effects of specific stressors. We also compared the independent effects of social support from a woman’s friends and family to that from an intimate partner, with particular attention to differences across family type. Consistent with previous research and our first hypothesis, we found that both social support and life stress were independently predictive of postpartum depression for all women. In general, social support was beneficial for women across family type, and life stress was a significant risk factor postpartum depression. The negative effects of stressors on postpartum depression were slightly reduced for women across family type after social support was added into the models; however, further testing revealed that neither source of social support (intimate partner or friend and family) substantially or significantly reduced the effects of stressors, refuting our second hypothesis that social support works to fully or partially mediate stress. This indicates that social support is more of a protective factor in and of itself (e.g., Dennis and Ross, 2006; Webster et al., 2011), but insufficient to govern the harmful effects of stress exposure. Non-significant tests of moderating effects in preliminary analyses further support

256

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

0.12*** Major Life Stress

NS

NS

–0.10*** Recent Life Stress

NS

–0.08*** NS

Neighborhood Safety

NS

Partner Support

–0.10***

–0.12***

Depression

–0.07***

0.12*** Child Stress

–0.17***

–0.07*** –0.12***

Friend and Family Support

0.23*** Parenting Stress

–0.07*** –0.09***

NS

Relationship Strain

NS Violence

NS

–0.08***

NS

Note: * significant at .05; ** significant at .01; ***significant at .001 Fig. 1. Path analysis of the mediating effects of social support for married women (standardized coefficients).

this assertion. Although it is possible that social support is not enough to mediate the impact of harmful stressors in a woman’s life on the risk for postpartum depression, it should be noted that social support (here) seems to decrease the risk of postpartum depression in and of itself. In fact, when comparing the standardized magnitudes of support against other variables in the model, social support from all sources is generally similar or greater in the magnitude of its effects compared to both major and recent life event stressors and child and parenting related stress. This suggests that the direct protective effects of social support are, in and of themselves, as powerful as some of the most salient stressors in increasing risk of postpartum depression. It should also be noted that our path models suggest that stress can deteriorate a woman’s social support, thus programs fostering support among mothers of infants is still important for those most disadvantaged women, who are likely to have high levels of stress. In addition to shedding light on the pathway that best describes the relationship between social support and postpartum depression, our findings confirm our third hypothesis that the impact of specific stressors and the source of social support that is most influential (intimate partner or friend and family) differs somewhat for married, cohabiting, and single women. In line with previous research, support from an intimate partner (compared to support from friends and family) appears to be uniquely beneficial and highly significant for married women. The same was observed for cohabiting women, suggesting that partner availability and supportiveness are attributable to, and significantly protective in, any live-in relationship. In comparison, intimate partner support was only marginally significant for single women and was also minimal in magnitude. With regard to friend and family support, all women gained significant protection. Of interest, however, intimate partner support had a greater impact on postpartum depression for both married and cohabiting women compared to friend and family support, and was equal in magnitude to friend and family support for single women. Surprisingly, married women gained the most benefits from friend and family support compared to cohabiting and single women (who gained similar benefits). That single women also report the lowest mean levels of support overall suggests that single women may perceive

257

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

0.11*** Major Life Stress

–0.07***

–0.09***

0.07*** Recent Life Stress

NS NS

NS

Neighborhood Safety

NS

Partner Support

–0.14***

–0.10***

Depression

–0.07***

NS Child Stress

–0.13***

NS

–0.14***

Friend and Family Support

0.12*** Parenting Stress

–0.07*** –0.18***

NS

Relationship Strain

NS

–0.07*** NS Violence

NS

Note: * significant at .05; ** significant at .01; ***significant at .001 Fig. 2. Path analysis of the mediating effects of social support for cohabiting women (standardized coefficients).

lower levels of support and have reduced frequency of contact with family and friends and/or an intimate partner compared to their married and cohabiting counterparts. Finally, findings demonstrate that the types of stress that influence postpartum depressive symptoms vary by family type as well. Although stress in general is a significant risk factor for all women, exposure to chronic stressors appears to be of greater relevance to single women while major life event stressors (and in the case of cohabiting women, recent life event stressors) have important and lasting effects for women in other family types. Of particular interest, married women do not fare better than single or cohabiting women in terms of the magnitude of significant chronic stressors. While it may be that some characteristic unique to marital unions amplifies the impact of certain stressors, it is presumably more likely that single women are less reactive to life event stress because they are consumed by the current stresses and strains they endure daily. Notably, neither source of social support was significantly influential in reducing these stressors. Nevertheless, these findings illustrate that, by not considering family type, the postpartum depression literature likely neglects important aspects of risk and vulnerability. Several limitations in this paper merit discussion. First, the Fragile Families data set is a uniquely ‘‘fragile” data source. Although this is a highly appropriate data set for the current research, we caution generalizing findings to the entire population, as these women are disproportionately disadvantaged. In addition, much of the data from the Fragile Families is based on self-reports. As such, women’s reports of stress and supportiveness may be influenced by current depressive symptoms. Sensitivity analyses suggest that previous depression did not significantly predict parenting-related stress (assumed to begin or increase at the birth of a child), but we cannot disentangle the causal directions between the stressors occurring in the postnatal period and postpartum depression. Several measures may also be subject to recall bias. Another limitation of the data centers on our depression measures. The Fragile Families lacks a retrospective self-report measure for depression at baseline, thus we rely on medical record data to establish previous depression. Due to imperfections in medical records, access to care, and geographic biases this measure likely underestimates previous depression among women in the sample.

258

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

NS

Major Life Stress

NS

Recent Life Stress

–0.12***

–0.09*** 0.09*** NS

0.10*** Neighborhood Safety

NS

Partner Support

–0.09**

–0.14***

Depression

–0.08***

0.16*** Child Stress

–0.13**

NS NS 0.16***

Parenting Stress

Friend and Family Support

–0.09*** –0.10***

0.08***

Relationship Strain

–0.06*** –0.20***

0.14*** Violence

–0.09***

Note: * significant at .05; ** significant at .01; ***significant at .001 Fig. 3. Path analysis of the mediating effects of social support for single women (standardized coefficients).

In addition, we use a general depression inventory (the CIDI-SF) to assess postpartum depression. The CIDI-SF does not clearly distinguish features associated with pregnancy and childbirth (e.g., fatigue, eating habits) from symptoms of major depression. In addition, assessment of depressive symptoms may be biased in the CIDI-SF across race and ethnicity (see Williams et al., 2007). However, the prevalence of postpartum depression in this sample was within the range of national estimates suggesting that women in our sample who met CIDI-SF criteria for a probable diagnosis of major depression at the time of the one-year follow-up were likely experiencing postpartum depression. Finally, the Fragile Families does not include measures of emotional support from family and friends, and focuses primarily on perceived support from this source. We therefore use two measures of perceived instrumental support to compare supportive relationships. Because women were asked to report on someone other than the focal child’s biological father for the friend and family support measures, there is a possibility that some women responded with a current romantic partner in mind. However, few women had a partner other than the biological father at the time reports were taken (fewer than 10%), suggesting that most women did provide accurate information for the friend and family measure. Nevertheless, future research should include more comprehensive measures of support for comparisons. Net of these limitations, the findings from this paper have important implications for research and practice. Given the evidence that the variety of support providers in a woman’s social network is important, future research should assess support from individuals outside of the romantic partnership as well as support from the intimate partner. Continued research should also consider women in all family types, given that differences between the three family types examined in this study (married, single, single and cohabiting) appeared to contribute to differences in the experience of social support, postpartum depression, and stress. A more thorough examination of the selective nature of social support is also warranted. It may be that the inability of supportive relationships to fully protect the most vulnerable women in this sample (i.e., single women), is at least in part driven by the day to day stressors that single women encounter. Additionally, because supportive relationships appear to be more of a protective factor overall in terms of instrumental support, efforts that focus on

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

259

identifying, treating, and preventing postpartum depression should explore resources or programs that foster supportive ties. Resources or programs that reduce or mitigate the impact of stressors should also be explored, since the protective effects of social support are insufficient to govern the harmful effects of stress exposure, and stress exposure is a substantial risk factor for postpartum depression. Acknowledgements We would like to thank Isaac Eberstein, Mary Gerend, and Kathryn Tillman for their thoughtful comments and helpful suggestions on earlier drafts of this paper. We would also like to thank the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) through Grants R01HD36916, R01HD39135, and R01HD40421, as well as a consortium of private foundations for their support of the Fragile Families and Child Wellbeing Study. Appendix A Construction of social support variables. Friend and family support Could you count on someone to loan you $200 in the next year? Could you count on someone to loan you $1000 in the next year? Could you count on someone to provide you with a place to live in the next year? Could you count on someone to help you with emergency child care? Could you count on someone to co-sign for a loan for $1000? Kuder-Richardson coefficient = 0.81 Intimate partner support How often can you trust father or partner to take good care of child? How often can you count on father or partner to watch child for a few hours? Cronbach’s a = 0.65 Appendix B Construction of stress variables. Major life event stress Did you think about aborting this pregnancy? (proxy for undesired pregnancy) Did focal child’s father want you to abort this pregnancy? Were you living with both your parents as a child? (proxy for parental divorce/separation) Have you ever had a miscarriage or abortion? Have you ever had a stillbirth? Has focal child’s father ever been in jail? Have you ever been forced into having sex? (from one year follow up) Kuder-Richardson coefficient = 0.33 Recent life event stress Have you moved since child was born? Since focal child’s birth, have you had another pregnancy/are you pregnant now? Has focal child’s father been jailed since baseline interview? Since focal child’s birth, have you had any miscarriages, abortions, or stillbirth? Have you divorced or separated from focal child’s father since focal child’s birth? Kuder-Richardson coefficient = 0.16 Relationship strain How often (did/do) you How often (did/do) you How often (did/do) you How often (did/do) you How often (did/do) you How often (did/do) you Cronbach’s a = 0.62

and and and and and and

baby’s baby’s baby’s baby’s baby’s baby’s

father father father father father father

disagree disagree disagree disagree disagree disagree

about about about about about about

money? spending time together? sex? the focal pregnancy? alcohol/drug use? being faithful? (continued on next page)

260

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

Child-related stress Child often fusses and cries Child gets upset easily Child reacts strongly when upset Child is sociable Child is friendly with strangers Child Is shy Child had to be spanked in past month Cronbach’s a = 0.49 Parenting-related stress Being a parent is harder than I thought I often feel trapped by parental responsibilities Taking care of children is more work than pleasure I often feel tired and worn out from parenting Cronbach’s a = 0.69 Intimate partner violence Has the relationship ended because of violence or abuse? Are you not living together/not planning to live together because of violence or abuse? Are you not married/not planning to marry because of violence or abuse? How often (did/does) baby’s father hit or slap? How often (did/does) baby’s father slap or kick you? How often (did/does) baby’s father s hit you with fist or dangerous object? Cronbach’s a = 0.64

Appendix C See Table C1.

Table C1 Clogg test for mediating effects.

⁄

Model 1

Model 2

Married Major life stress Recent life stress Neighborhood safety Relationship strain Child-related stress Parenting-related stress Intimate partner violence

0.17 0.01 0.03 0.03 0.04 0.10 0.27

(0.05) (0.05) (0.06) (0.02) (0.01) (0.02) (0.33)

0.15 0.03 0.03 0.03 0.04 0.10 0.27

(0.05) (0.05) (0.06) (0.02) (0.01) (0.02) (0.33)

0.30 0.26 0.27 0.27 0.06 0.08 0.07

Cohabiting Major life stress Recent life stress Neighborhood safety Relationship strain Child-related stress Parenting-related stress Intimate partner violence

0.13 0.12 0.11 0.00 0.02 0.05 0.24

(0.04) (0.05) (0.05) (0.02) (0.01) (0.01) (0.15)

0.11 0.11 0.11 0.01 0.02 0.05 0.24

(0.04) (0.05) (0.05) (0.02) (0.01) (0.01) (0.15)

0.24 0.15 0.00 0.48 0.00 0.00 0.00

Single Major life stress Recent life stress Neighborhood safety Relationship strain Child-related stress Parenting-related stress Intimate partner violence

0.08 0.06 0.17 0.04 0.04 0.06 0.32

(0.04) (0.05) (0.07) (0.01) (0.01) (0.01) (0.05)

0.06 0.06 0.17 0.08 0.04 0.06 0.32

(0.04) (0.05) (0.07) (0.03) (0.01) (0.01) (0.05)

0.32 0.03 0.00 0.37 0.00 0.00 0.00

p < 0.05;

⁄⁄

p < 0.01;

⁄⁄⁄

p < 0.001.

Z

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

261

References Adkins, Kate S., Kamp Dush, Claire, 2010. The mental health of mothers in and after violent and controlling unions. Soc. Sci. Res. 39, 925–937. Almond, P., 2009. Postnatal depression: a global public health perspective. Perspect. Public Health 129, 221–227. American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, fourth ed. Author, Washington, DC. Barnet, Beth, Joffe, Alain, Duggan, Anne K., Wilson, Michele D., Repke, John T., 1996. Depressive symptoms, stress, and social support in pregnant and postpartum adolescents. Arch. Pediatr. Adolesc. Med. 150, 64–69. Baron, R.M., Kenny, D.A., 1986. The moderator–mediator variable distinction in social psychological research: conceptual, strategic and statistical considerations. J. Pers. Soc. Psychol. 51, 1173–1182. Beach, Steven R.H., Katz, Jennifer, Fincham, Frank D., Bradbury, Thomas N., 1996. Social support in marriage. In: Pierce, Gregory R., Sarason, Barbara R., Sarason, Irwin G. (Eds.), Handbook of Social Support and the Family. Plenum, New York, pp. 43–65. Beck, Cheryl Tatano, 2001. Predictors of postpartum depression: an update. Nurs. Res. 50, 275–285. Beck, Cheryl Tatano, 2002. Revision of the postpartum depression predictors inventory. J. Obstet. Gynecol. Neonatal. Nurs. 31, 394–402. Beydoun, Hind A., Al-Sahab, Ban, Beydoun, May A., Tamim, Hala, 2010. Intimate partner violence as a risk factor for postpartum depression among Canadian women in the maternity experience survey. Ann. Epidemiol. 20, 575–583. Cairney, John, Boyle, Michael, Offord, David R., Racine, Yvonne, 2003. Stress, social support and depression in single and married mothers. Soc. Psychiat. Psychiat. Epidemiol. 38, 442–449. Chaudron, Linda H., 2003. Postpartum depression: what pediatricians need to know. Pediatr. Rev. 24, 154–161. Clogg, Clifford C., Petkova, Eva, Haritou, Adamantios, 1995. Statistical methods for comparing regression coefficients between models. Am. J. Sociol. 100, 1261–1293. Cohen, Sheldon, Underwood, Lynn, Gottlieb, Benjamin H., 2000. Social Support Measurement and Intervention: A Guide for Health and Social Scientists. Oxford University Press, New York. Coyne, James C., Schwenk, Thomas L., Fechner-Bates, Suzanne, 1995. Non-detection of depression by primary care physicians reconsidered. Gen. Hosp. Psychiat. 17, 3–12. DeKlyen, Michelle, Brooks-Gunn, Jeanne, McLanahan, Sara, Knab, Hean, 2006. The mental health of married, cohabiting, and non-coresident parents with infants. Am. J. Public Health 96, 1836–1841. Dennis, Cindy-Lee, Ross, Lori, 2006. Women’s perceptions of partner support and conflict in the development of postpartum depressive symptoms. J. Adv. Nurs. 56, 588–599. Dennis, Cindy-Lee, Letourneau, Nicole, 2007. Global and relationship-specific perceptions of support and the development of postpartum depressive symptomatology. Soc. Psychiat. Psychiat. Epidemiol. 42, 389–395. Earls, Marian F., 2010. Clinical report incorporating recognition and management of perinatal and postpartum depression into pediatric practice. Pediatrics 126, 1032–1039. Ensel, Walter M., Lin, Nan, 1991. The life stress paradigm and psychological distress. J. Health Soc. Behav., 321–341 Ertel, Karen A., Rich-Edwards, Janet W., Koenen, Karestan C., 2011. Maternal depression in the United States: nationally representative rates and risks. J. Women’s Health 20, 1609–1617. Gallagher, Robert W., Hobfall, Steven E., Ritter, Christian, Lavin, Justin, 1997. Marriage, intimate support and depression during pregnancy: a study of innercity women. J. Health Psychol. 2, 457–469. Gaynes, B.N., Gavin, N., Meltzer-Brody, S., Lohr, K.N., Swinson, T., Gartlehner, G., Brody, S., Miller, W.C., 2005. Perinatal Depression: Prevalence, Screening, Accuracy, and Screening Outcomes. Evidence Report No. 119. Rockville, Md. Agency for Healthcare Research and Quality. Glazier, R.H., Elgar, F.J., Goel, V., Holzapfel, S., 2004. Stress, social support, and emotional distress in a sample of pregnant women. J. Psychosom. Obstet. Gynecol. 25, 247–255. Goodman, Sherryl H., Gotlib, Ian H., 1999. Risk for psychopathology in the children of depressed mothers: a developmental model for understanding mechanisms of transmission. Psychol. Rev. 106, 458–490. Gottlieb, Benjamin H., Bergen, Anne E., 2009. Social support concepts and measures. J. Psychosom. Res. 69, 511–520. Gremigni, Paola, Mariani, Laura, Marracino, Valentina, Tranquilli, Andrea Luigi, Turi, Angelo, 2011. Partner support and postpartum depressive symptoms. J. Psychosom. Obstet. Gynecol. 32, 135–140. Gustafsson, Hanna C., Cox, Martha J., 2012. Relations among intimate partner violence, maternal depressive symptoms, and maternal parenting behaviors. J. Marriage Family 74, 1005–1020. Hagerty, Bonnie M., Williams, Reg A., 1999. The effects of sense of belonging, social support, conflict, and loneliness on depression. Nurs. Res. 48, 215–219. Haslam, Divna M., Pakenham, Kenneth I., Smith, Amanda, 2006. Social support and postpartum depressive symptomatology: the mediating role of maternal self-efficacy. Infant Ment. Health J. 27, 276–291. Herrick, Harry, 2000. The Effect of Stressful Life Events on Postpartum Depression – Results from the 1997–1998 North Carolina Pregnancy Risk Assessment Monitoring System (PRAMS), vol. 121. State Center for Health Statistics Studies, pp. 1–9. Hopkins, J., Campbell, S.B., 2008. Development and validation of a scale to assess social support in the postpartum period. Arch. Women’s Mental Health 11, 57–65. Horenstein, Jeffrey, Cohen, Sheldon, 2008. Social support. In: Ingram, Rick E. (Ed.), The International Encyclopedia of Depression. Springer Publishing Company, New York, pp. 514–517. Horowitz, June Andrews, Goodman, Janice H., 2005. Identifying and treating postpartum depression. J. Gynecol. Neonatal Nurs. 34, 264–273. Judd, C.M., Kenny, D.A., 1981. Process analysis: estimating mediation in treatment evaluations. Eval. Rev. 5, 602–619. Kawachi, I., Berkman, L.F., 2001. Social ties and mental health. J. Urban Health 78, 458–467. Kessler, R.C., Andrews, G., Mroczek, D., Ustun, T.B., Wittchen, H.U., 1998. The world health organization composite international diagnostic interview shortform (CIDI-SF). Int. J. Methods Psychiat. Res. 7, 171–185. Letourneau, Nicole, Duffett-Leger, Linda, Stewart, Miriam, Hegadoren, Kathy, Dennis, Cindy-Lee, Rinaldi, Christina M., Stoppard, Janet, 2007. Canadian mothers’ perceived support needs during postpartum depression. J. Obstet. Gynecol. Neonatal Nurs. 36, 441–449. Logsdon, Cynthia M., Birkimer, John C., Barbee, Anita P., 1997. Social support providers for postpartum women. J. Soc. Behav. Pers. 12, 89–102. Logsdon, Cynthia M., Wisner, Katherine L., Pinto-Foltz, Melissa D., 2006. The impact of postpartum depression on mothering. J. Obstet. Gynecol. Neonatal Nurs. 35, 652–658. Logsdon, Cynthia M., Wisner, Katherine L., Hanusa, Barbara H., 2009. Does maternal role functioning improve with antidepressant treatment in women with postpartum depression? J. Women’s Health 18, 85–90. MacKinnon, D.P., Lockwood, C.M., Hoffman, J.M., West, S.G., Sheets, V., 2002. A comparison of methods to test the significance of the mediated effect. Psychol. Methods 7, 83–104. Marks, F. Nadine, McLanahan, Sara S., 1993. Gender, family structure, and social support among parents. J. Marriage Family 55, 481–493. Meadows, Sarah O., McLanahan, Sara S., Brooks-Gunn, Jeanne, 2008. Stability and change in family structure and maternal health trajectories. Am. Sociol. Rev. 73, 314–334. Misri, Shaila, Kostaras, Xanthoula, Fox, Don, Kostaras, Demetra, 2000. The impact of partner support in the treatment of postpartum depression. Can. J. Psychiat. 45, 554–558. Mitchell, C., Notterman, D., Brooks-Gunn, J., Hobcraft, J., Garfinkel, I., Jaeger, K., Kotenko, I., McLanahan, S., 2011. Role of mother’s genes and environment in postpartum depression. Proc. Natl. Acad. Sci. USA 108, 8189–8193. Muthén, L.K., Muthén, B.O., 2007. MPlus User’s Guide, fifth ed. Muthén & Muthén, Los Angeles, CA.

262

K.M. Reid, M.G. Taylor / Social Science Research 54 (2015) 246–262

Paternoster, Raymond, Brame, Robert, Mazerolle, Paul, Piquero, Alex, 1998. Using the correct statistical test for the equality of regression coefficients. Criminology 36, 859–866. Paykel, E.S., 1994. Life events, social support and depression. Acta Psychiatr. Scand. 89, 50–58. Pearlin, Leonard I., Lieberman, Morton A., Menaghan, Elizabeth G., Mullan, Joseph T., 1981. The stress process. J. Health Soc. Behav. 22, 337–356. Pearlin, Leonard I., 1989. The sociological study of stress. J. Health Soc. Behav. 30, 241–256. Peirce, J.M., Burke, C.K., Stoller, K.B., Neufeld, K.J., Brooner, R.K., 2009. Assessing traumatic event exposure: comparing the traumatic life events questionnaire to the structured clinical interview for DSM-IV. Psychol. Assess. 21, 210–218. Pierce, Gregory R., Joseph, Helene J., Henderson, Ciarda A., Sarason, Barbara R., Sarason, Irwin G., 1996. Conceptualizing and assessing social support in the context of the family. In: Pierce, Gregory R., Sarason, Barbara R., Sarason, Irwin G. (Eds.), Handbook of Social Support and the Family. Plenum, New York, pp. 3–23. Reichman, Nancy E., Teitler, Julien O., Garfinkel, Irwin, McLanahan, Sara S., 2011. Fragile families: sample and design. Child. Youth Serv. Rev. 23, 303–326. Reid, Keshia M., Taylor, Miles G., 2015. Examining postpartum depression from a stress-process perspective: the importance of stress type and timing. Popul. Res. Policy Rev. http://dx.doi.org/10.1007/s11113-015-9368-7. Riolo, Stephanie A., Nguyen, Tuan Anh, Greden, John F., King, Cheryl A., 2005. Prevalence of depression by race/ethnicity: findings from the national health and nutrition examination survey III. Am. J. Public Health 95, 998–1000. Robertson, Emma, Grace, Sherry, Wallington, Tamara, Stewart, Donna E., 2004. Antenatal risk factors for postpartum depression: a synthesis of recent literature. Gen. Hosp. Psychiat. 26, 289–295. Robertson, E., Celasun, N., Stewart, D.E., 2003. Risk factors for postpartum depression. In: Stewart, D.E., Robertson, E., Dennis, C.-L., Grace, S.L., Wallington, T. (Eds.), Postpartum Depression: Literature Review of Risk Factors and Interventions. Scribano, Philip V., Stevens, Jack, Kaizar, Elly, 2013. The effects of intimate partner violence before, during, and after pregnancy in nurse visited first time mothers. Matern. Child Health J. 17, 307–318. Stapleton, Lynlee R., Tanner, Schetter, Dunkel, Christine, Westling, Erika, Rini, Christine, Glynn, Laura M., Hobel, Calvin J., Sandman, Curt A., 2012. Perceived partner support in pregnancy predicts lower maternal and infant distress. J. Fam. Psychol. 26, 453–463. Surkan, Pamela J., Peterson, Karen E., Hughes, Michael D., Gottlieb, Barbara R., 2006. The role of social networks and support in postpartum women’s depression: a multiethnic urban sample. Matern. Child Health J. 10, 375–383. Swendsen, Joel D., Mazure, Carolyn M., 2000. Life stress as a risk factor for postpartum depression: current research and methodological issues. Clin. Psychol.: Sci. Pract. 7, 17–31. Thoits, Peggy A., 2011. Mechanisms linking social ties and support to physical and mental health. J. Health Soc. Behav. 52, 145–161. Turner, R. Jay, Lloyd, Donald A., 1999. The stress process and the social distribution of depression. J. Health Soc. Behav. 40, 374–404. Turner, R. Jay, Lloyd, Donald A., 2004. Stress burden and the lifetime incidence of psychiatric disorder in young adults. Arch. Gen. Psychiat. 61, 481–488. Turner, R. Jay, Avison, William R., 2003. Status variations in stress exposure: implications for the interpretation of research on race, socioeconomic status, and gender. J. Health Soc. Behav. 44, 488–505. Webster, Joan, Nicholas, Catherine, Velacott, Catherine, Cridland, Noelle, Fawcett, Lisa, 2011. Quality of life and depression following childbirth: impact of social support. Midwifery 27, 745–749. Wheaton, Blair, 1994. Sampling the stress universe. In: Avison, W., Gotlib, I. (Eds.), Contemporary Issues and Prospectus for the Future. Plenum, New York, pp. 77–114. Williams, David R., Gonzalez, Hector M., Neighbors, H., Nesse, Randolph, Abelson, Jamie M., Sweetman, Julie, Jackson, James S., 2007. Prevalence and distribution of major depressive disorder in African Americans, Caribbean blacks, and non-Hispanic whites: results from the National Survey of American Life. Arch. Gen. Psychiat. 64, 305–315. Winship, Christopher, Radbill, Larry, 1994. Sampling weights and regression analysis. Sociol. Methods Res. 23, 230–257. Wisner, Katherine L., Chambers, Christina, Sit, Dorothy K.Y., 2006. Postpartum depression: a major public health problem. J. Am. Med. Assoc. 296, 2616– 2618. Xie, Ri-Hua, He, Guoping, Koszycki, Diana, Walker, Mark, Wen, Shi Wu, 2009. Prenatal social support, postnatal social support, and postpartum depression. Ann. Epidemiol. 19, 637–643.