Combinations of Adverse Childhood Events and Risk of Postpartum Depression Among Mothers Enrolled in a Home Visiting Program

Abstract

Purpose

The purpose of this study was to examine how combinations of adverse childhood events (ACEs) contribute to the risk of postpartum depression and the mediating role of prenatal social support.

Methods

The Adverse Childhood Experiences Scale Questionnaire and Edinburgh Postnatal Depression Scale Questionnaire were used to measure the study’s exposure and outcome. Among a cohort of 419 mothers enrolled in a home visiting program, latent class analyses were used to identify classes of ACEs exposure. General linear models assessed the risk of postpartum depression; prenatal social support was examined as a mediator.

Results

Four distinct classes of ACEs exposure were identified. On the Edinburgh Postnatal Depression scale, mothers who were classified in Classes 1-3 scored higher by 2.6-4.4 points compared to women in class 0. ACE class was found to be indirectly associated with postpartum depression scores through prenatal social support.

Conclusions

Identifying combinations of ACEs in a HV program has the potential to improve characterization of ACEs among low-income perinatal women in the United States. Elucidating how these combinations contribute to the risk of postpartum depression has the potential to identify women at increased risk which can help HV programs prioritize prevention efforts.

Introduction

Home visiting (HV) programs are implemented throughout the United States (US) to improve health outcomes for mothers and children from high risk populations.^1-3 Women enrolled in these programs are typically low income, young and unmarried. HV programs have demonstrated positive improvements in lowering the rates of preterm birth, reduction of depression symptoms, improving parental functioning and are associated with increased adherence to well-child care utilization.^4-6 However, these positive effects may be attenuated by high rates of traumatic exposures such as physical abuse.^7,8 The intergenerational effects of trauma have been increasingly identified and suggest an association between parental adverse childhood experiences (ACEs) and delayed child development .^9,10 ACEs are experienced by almost 60% of the US population and include household dysfunction, abuse and neglect.¹¹ Prior work has demonstrated mothers exposed to ACEs are at an increased risk of experiencing depression during the perinatal time period (pregnancy and postpartum).^8,12-19 In response to mounting evidence of the relationship between trauma and poor health outcomes, HV programs have begun to adopt trauma-informed models for interventions.²⁰ As programs expand to include trauma-informed approaches there is a need to improve characterization of ACEs experienced by low income women to better prepare HV to anticipate the needs of women enrolled in their programs.

The total number of ACEs experienced by an individual typically is dichotomized (Yes/No) at various cutoffs to examine the relationship between childhood trauma and poor health outcomes.^21,22 The limitations of characterizing ACEs as a total score has been described.²³ Briefly, the total score method ignores the heterogeneity of ACEs endorsed by individuals and assumes a linear relationship when specific types of abuse can have multiplicative effects on outcomes. ^23,24 Studies have also demonstrated that the type of abuse or trauma is predictive of specific outcomes. For example, studies have observed that child neglect and poverty are associated with an increased risk of cognitive deficits, whereas exposure to physical and sexual abuse have been associated with depression and anxiety risk.²⁵ In clinical practice, using the total score can also be problematic as it does not provide the clinician information about the type of ACEs experienced by their patient or help identify what interventions or resources may be required. ²⁶ Identification of combinations of ACEs experienced by women can improve the ability of clinicians, including home visitors, to provide trauma-informed care. Examining how the type or combinations of ACEs influences health outcomes, such as postpartum depression, has the potential to lead to better interventions informed by an individual’s trauma history. ²⁶ A recent study, using national survey data, identified distinct combinations of ACEs among children and demonstrated that the combination of ACEs, not the total score, predicted specific health outcomes.²⁶ Identifying distinct combinations of ACEs experienced by women enrolled in HV programs has the potential to detect the patterns of trauma experienced by low-income women. This may lead to earlier identification of groups of women and children at an increased risk of poor health outcomes, such as postpartum depression, based on trauma history. For example, recent studies have demonstrated that the type of adversity experienced in childhood differentially affects pregnancy outcomes, including antenatal depression and post-traumatic stress disorder (PTSD) symptoms. ^27,28

The prevention and treatment of postpartum depression among women enrolled in HV is important for several reasons. Postpartum depression coincides during a critical window of attachment, bonding, development ²⁹ and is associated with increased parenting stress.^8,30 Additionally, women with depression have been found to benefit less from HV programs than non-depressed women.³¹ Due to the association between ACEs and depression ^8,12-19, identification of the mechanisms of risk has the potential to improve the effectiveness of HV interventions by helping to target the highest risk mothers. Studies are needed to examine how combination of ACEs are associated with depression risk, as prior work has used the total ACE score, or examined ACEs individually.^8,12-19

To inform interventions for HV programs, it is also important to identify how ACEs confer depression risk among postpartum women. Previous studies have identified individuals with a history of ACEs have smaller and less supportive social networks than their peers. ^8,32-34,35 The role of social support has been previously examined in HV populations. These showed the relationship between childhood maltreatment and parenting stress was partially mediated by depression and social support.^8,36 Social support has also been identified as a risk factor in numerous studies on postpartum depression risk.³⁷ A recent study found a woman’s total ACE score and social support were predictive of prenatal and postpartum depression symptoms. However, the study demonstrated that social support did not moderate the observed association of ACEs and depression score.³⁸ Results from this study along with overall knowledge related to social support and depression risk provides rationale to examine social support as a potential mediator. In another study among women enrolled in home visiting, intimate partner violence, antenatal depression and stress were identified as mediators, when examining the effect of ACE exposure on postpartum depression risk.¹⁹ Studies evaluating the mechanisms of risk are sparse, yet needed to inform interventions directed at reducing risk for depression postpartum.

Prior studies have operationalized ACE exposure as a total score or individual types of ACEs when examining the effect of ACEs on postpartum depression risk. ^19,39 In one study, authors examined the effect of the combination of physical and/or sexual abuse, however authors did not examine other patterns of ACEs. ³⁹ Therefore, the objective of this study was to identify specific combinations of ACEs experienced by women in a HV program using latent class analyses (LCA) and examine how these combinations of ACEs contribute to the risk of postpartum depression among women enrolled in a HV program. As a secondary analysis, we examined prenatal social support as a mediator in the relationship between ACEs and postpartum depression risk. We hypothesize that unique ACE classes will emerge, and that these classes will be associated with risk of postpartum depression. Our secondary hypothesis is that social support will partially mediate this relationship.

Methods

Study Population

Participants in this retrospective cohort study included a subset of first-time mothers who were enrolled from 2007-2016 in Every Child Succeeds (ECS), a regional HV program. ⁴⁰ To enroll in ECS, women residing within the greater Cincinnati area and Northern Kentucky had to meet at least one of the following criteria: low income (200% under federal poverty level), under the age of 18, unmarried and/or received suboptimal prenatal care (late prenatal care initiation). Among newly enrolled women in the HV program, three research studies drew randomly from the pool of available women. Two studies were clinical trials; one examined the efficacy of using multiple measures to reduce the risk of injury within the home and the other assessed the impact of motivational interviewing on retention in home visiting . The third was an observational study that examined the effect of neonatal DNA methylation on social-emotional development outcomes.^41-43 563 women participated in these studies. For the current study, these women were included if they completed a one-time screening of ACEs and were screened for depression up to three months postpartum (n = 419). Prenatal social support data was collected from a subset of participants enrolled in the injury prevention and motivational interviewing clinical trials (n=195). These data were used to address our second hypothesis. This study was approved by the Cincinnati Children's Hospital Institutional Review Board. Subjects provided informed consent for each individual study and the overall program evaluation.

Measures

Edinburgh Postnatal Depression Scale (EPDS) Questionnaire.

The EPDS questionnaire was administered to women during their participation in ECS. The EPDS is comprised of 10-items that measure depressive symptoms such as hopelessness, sleep disturbances and sadness, where higher scores are associated with increased depression risk. ⁴⁴ The scale was developed in 1987 and has become the most common instrument used to screen for prenatal and postpartum depression. Multiple studies have found high sensitivity, specificity and positive predictive value in identifying women with major depressive disorder during the perinatal period. ⁴⁵ This scale has excellent test-retest reliability scores (ICC=0.92).⁴⁶ For this study, EPDS scores collected up to 90 days postpartum were used to measure risk of postpartum depression based on a scoring 10 points or greater (score range 0-30).

Adverse Childhood Experiences Scale Questionnaire.

The ACEs questionnaire was administered by study staff to mothers during pregnancy or postpartum to identify history of child abuse and neglect and is comprised of 10-items that measure the following forms of abuse (physical, sexual and emotional), neglect (physical and emotional), and household dysfunction (parents divorced or separated, mother treated violently, household member substance abuse, mental illness, or incarceration). Higher scores indicate more ACEs endorsed (range 0-10), for example if a subject endorsed three items on the questionnaire their total score would equal three. This measure was developed by the seminal epidemiology study on ACEs and health outcomes ⁴⁷ and has high internal consistency (Cronbach’s α=0.88).^48,49

Interpersonal Support Evaluation List (ISEL-12).

The ISEL consists of 12 items that measure the respondent’s perception of social support using a four-point scale. This instrument measures the following domains of social support: appraisal, perception of belonging, availability of tangible items and self-esteem, with higher scores indicating greater perception of social support (range 0-36). ⁵⁰ This measure has good test-retest reliability with coefficients ranging from 0.81 to 0.90 .^51-53 The ISEL-12 was collected during pregnancy at the time of study enrollment

Demographic Characteristics.

Maternal race, marital status, ethnicity, age at delivery, and education were abstracted from the ECS program enrollment data to control for potential confounding. Prior studies have demonstrated the risk of depression and number of ACEs varies by race or perceived racism.^54-56 Additionally, studies have demonstrated that young maternal age and advanced maternal age, lower levels of educational attainment, and being unmarried have been associated with an increased risk of postpartum depression.^37,57

Analysis

To estimate the incidence of depression postpartum (delivery up to 3 months postpartum) in our sample, the percentage of women in our sample scoring 10 or greater on the EPDS measures was calculated. A cut-off score of 10 or greater to indicate probable depression is commonly used and reduces the number of missed depression cases to less than ten percent. ^44,58 Then Chi-square and Fisher Exact tests (as appropriate) were used to examine maternal and child characteristics associated with depression during the postpartum time period. Chi-square tests were also used to compare demographic characteristics of our study population to demographic characteristics of the overall ECS population.

LCA was used to identify patterns of maternal ACEs. Model fit statistics including Akaike Information Criterion (AIC), average posterior probability (entropy), bootstrap likelihood ratio test, and group membership probabilities were used to select the number of classes for the LCA, starting with two classes. Class assignment was based on the highest posterior probability for each subject. Missing data (<5%) for the LCA was handled using a maximum likelihood approach. The LCA was performed in Mplus 7.1. Chi-square and Fisher Exact tests were used to examine the association of maternal characteristics and individual ACEs endorsed with assigned ACE class. The mean number of ACEs endorsed from the questionnaire was also calculated for each class.

General linear models were then used to test if ACE class assignment was associated with postpartum EPDS scores. Relative risk (RR) of postpartum depression (EPDS ≥ 10 vs <10) by class assignment was estimated using the RELRISK9 macro.⁵⁹ Models were adjusted for race, ethnicity, maternal age, and education. The effect of total number of ACEs endorsed on the EPDS score was estimated in a separate model. The Akaike information criterion (AIC ) and the Bayesian information criterion (BIC) derived from the adjusted model examining ACE class was compared to the AIC and BIC derived from the adjusted model examining the total ACE score as the exposure variable to assess model fit. Analyses were completed using SAS software (version 9.4).

In a secondary analysis, prenatal social support (total ISEL score) was examined as a mediator of ACEs class assignment and depression risk during the postpartum time period, as measured by EPDS scores, treated continuously (Y) using the PROCESS macro and following the approach described by Hayes and Preacher. ^60,61 Ordinary least squares regression models and 5,000 bootstrapped re-samples were used to derive bias corrected 95% confidence intervals and standard errors for the relative indirect effects. This approach allowed us to examine prenatal social support as a mediator of ACEs class assignment by analyzing the entire sample at one time ⁶⁰. If at least one of the bias corrected 95% confidence intervals for the indirect effects did not contain zero, then we concluded mediation existed. This analysis was conducted on a subsample of our cohort (N=195), who had a prenatal social support (total ISEL score) and postpartum depression (EPDS) score. This model was adjusted for race, ethnicity, maternal age, and education.

Results

Study Population

There were 419 women who had ACEs measures and a postpartum depression score on the EPDS included in this study. Of the 419 women included in this study, a majority were enrolled in the Cincinnati Home Injury Prevention trial (n= 337, 80.4%), 43 from the motivational interviewing trial (10.26%) and 39 from the DNA methylation on social-emotional development outcomes study (9.3%).^41,43 A majority of the study population was low income, single and had completed high school at the time of enrollment in the HV program. Approximately 50% of the study population was white, 45% were black and the remaining 5% were American Indian /Alaskan Native, Asian or of multi-race. Among these women, almost 20% scored 10 or higher on the EPDS, indicating high risk for major depressive disorder. Demographic characteristics of our sample population (n=419) were similar to demographic characteristics of the larger cohort of 563 women. Our study population was similar to the overall ECS population in terms of race and marital status. However, mothers in our study population were older, less likely to be Hispanic and more likely to report low income (200% below the federal poverty level). The risk of depression during the postpartum period differed by several maternal characteristics including race, education, and ACE class assignment (Table 1). Of the 210 White women in the study, 23.8% scored 10 or greater on the EPDS compared to 13.8% of 188 Black women in the study.

Table 1:

Maternal Characteristics and Depression Risk

Maternal Characteristics		EPDS <10 N=336 n (Col %)	EPDS ≥10 N=83 n (Col %)	P-value*
Race	White	160 (47.62)	50 (60.24)	0.012
	Black	162 (48.21)	26 (31.33)
	Other	14 (4.17)	7 (8.43)
Hispanic/Latina	No	324 (96.43)	77 (92.77)	0.141
	Yes	12 (3.57)	6 (7.23)
Age	<18	12 (3.57)	2 (2.67)	0.900
	18-21	174 (51.79)	42 (50.60)
	>21	150 (44.64)	39 (46.99)
Single	No	34 (10.18)	13 (15.66)	0.157
	Yes	300 (89.82)	70 (84.34)
Completed High School	No	65 (20.9)	26 (31.71)	0.039
	Yes	246 (79.1)	56 (68.29)
Low Income	No	9 (2.72)	3 (3.61)	0.714
	Yes	322 (97.28)	80 (96.39)
ACE Class	Low Exposure (Class 0)	219 (65.18)	27 (32.53)	<.001
	High Exposure (Class 1)	28 (8.33)	16 (19.28)
	Emotional/Physical Abuse (Class 2)	39 (11.61)	16 (19.28)
	Household Dysfunction (Class 3)	50 (14.88)	24 (28.92)

^*N = 419, Column total may not equal 419 due to missing covariate data

ACE Classes

Based on model fit and clinical interpretability, a 4-class model was selected (Table 2). The posterior probability of being classified into each class was calculated for all the women in the study. Class assignment was based upon the class with the highest posterior probability. Within this 4-class model, latent classes were distributed as follows: class 1(10.5% of the cohort, n=44), class 2 (13.3%, n=55), class 3 (17.7%, n= 74), and class 0 (58.7%, n=246) (Figure 1). Women in class 1 were more likely to have a high ACE score on the questionnaire than women in the other classes. Women in class 2 were more likely to endorse items related to physical and emotional abuse than women in class 3. Women in class 3 were more likely to report exposure to parent separation, witnessing domestic violence, living with an adult with a substance abuse problem or an adult in the home who was incarcerated when compared to women in class 2 (Table 3). Women in group 0 were more likely to have a low ACE score. The mean (standard deviation) number of ACEs endorsed by women in class 0 was 0.93 (std deviation (std) 0.75), class 1 was 7.75 (1.24), class 2 was 3.62 (1.62) and class 3 was 3.99 (1.31). Marital status and maternal race were associated with ACE class. When compared to White women, Black women were more likely to be assigned to the low exposure group, class 0 (65% vs 54%), and emotional and physical abuse group, class 2 (15.43% vs 11.43%). A higher proportion of single women were assigned to the high exposure group, class 1 (26%) than women who were married at the time of study enrollment (9%). The mean EPDS score varied by class. Class 1 had the highest mean score (8.4, std dev 6.1), followed by class 3 (7.6, std dev 4.7), class 2 (6.8, std dev 5.6) and class 0 had the lowest score (4.3, std dev 4.2).

Table 2.

Indices of Model Fit for Latent Class Analysis

Classes for Latent Construct	Log- likelihood	Number of Parameters	Akaike Information Criteria (AIC)	Bayesian Information Criterion (BIC)	Sample size adjusted BIC	Bootstrapped Rubin Likelihood ratio test (LRT)	Vuong- Lo- Mendell- Rubin LRT	Entropyc
2	−2552	21	5146	5237	5171	<0.001	<0.001	0.85
3	−2503	32	5070	5209	5107	<0.001	<0.001	0.79
4	−2458	43	5002	5188	5051	<0.001	0.010	0.82
5	−2443	54	4993	5227	5056	0.012	0.003	0.87
6	−2431	65	4991	5273	5066	0.270	0.299	0.81

Figure 1:

ACE Classes

Table 3:

Association of ACE class with Maternal Characteristics

Level	High Exposure N (Col %)	Emotional/Physical Abuse N (Col %)	Household Dysfunction N (Col %)	Low Exposure N (Col %)	P Value
Race					0.003
White	26 (59.09)	24 (43.64)	47 (63.51)	113 (45.93)
Black	17 (38.64)	29 (52.73)	19 (25.68)	123 (50)
Other	1 (2.27)	2 (3.64)	8 (10.81)	10 (4.07)
Ethnicity					0.067
Hispanic/Latina	0 (0)	5 (9.09)	5 (6.76)	8 (3.25)
Age					0.287
<18	4(9.09)	0 (0)	1 (1.35)	9 (3.66)
18-21	20 (45.45)	31 (56.36)	36 (48.65)	129 (52.44)
>21	20 (45.45)	24 (43.64)	37 (50)	108 (43.9)
Marital Status
Single	32 (72.73)	53 (96.36)	62 (83.78)	223 (91.39)	0.001
Education
Completed High school	28 (70)	43 (79.63)	52 (76.47)	179 (77.49)	0.711
ACE Items Endorsed
Emotional Abuse	44 (100)	53 (96.36)	12 (16.22)	4(1.63)	<.001
Physical Abuse	39 (88.64)	31 (56.36)	3 (4.05)	1 (0.41)	<.001
Sexual Abuse	26 (59.09)	13 (24.07)	25 (33.78)	17 (6.91)	<.001
Emotional Neglect	36 (81.82)	29 (53.7)	40 (54.05)	13 (5.28)	<.001
Physical Neglect	26 (59.09)	0 (0)	20 (27.03)	2 (0.81)	<.001
Parent Separation	34 (77.27)	32 (58.18)	48 (64.86)	149 (60.57)	0.159
Domestic Violence	36 (81.82)	9 (16.36)	25 (33.78)	6 (2.44)	<.001
Substance Abuse	42 (95.45)	12 (22.22)	58 (78.38)	25 (10.16)	<.001
Mental Illness	32 (72.73)	11 (20)	37 (50)	10 (4.07)	<.001
Prison	21 (47.73)	9 (16.67)	27 (36.49)	3 (1.22)	<.001

Depression During the Postpartum Period

In the unadjusted analysis, women who were classified in Classes 1 thru 3 scored 2.5-4 points higher on the EPDS than women in class 0, which was the lowest exposure group (Table 4). This effect remained after adjusting for age, race, ethnicity, and education. On average, women in the highest exposure class (class 1) scored 4.4 (95%CI 2.75, 5.96) points higher on the EPDS compared to women in the lowest ACES exposure class (Class 0) (model AIC 1914.9, BIC 1960.1). Class 2 women scored 2.6 (95%CI 1.14, 4.01) points higher and women in Class 3 scored 3.4 (95%CI 2.09, 4.77) points higher on average compared to woman in the lowest ACES exposure class (Class 0) (Table 4) . In a separate model we examined the association between the total ACE score (0-10) and EPDS scores. After adjusting for covariates, the EPDS score increased by 0.66 points (95% CI: 0.46,0.85) for each ACE endorsed (model AIC 1919.2, BIC 1956.8). The relative risk of postpartum depression by ACE class, adjusting for the same covariates in the main model was estimated. Women in class 1 had an increased adjusted relative risk (RR) of postpartum depression by 3.32 (1.99, 5.53), class 2 RR 2.69 (1.57, 4.61) and class 3 RR 2.73 (1.66, 4.49) when compared to women in the lowest exposure group (class 0).

Table 4.

Association of ACEs and Depression Risk

ACE Class	Unadjusted EPDS Score β (95% CI)	Adjusted EPDS Score β (95% CI)
Ace Classes
Overall High Exposure	4.09 (2.58, 5.60) ***	4.36 (2.75,5.96) ***
Emotional/Physical Abuse	2.46 (1.08, 3.84) ***	2.58 (1.14,4.01) ***
Household Dysfunction	3.23 (2.01, 4.45) ***	3.43 (2.09,4.77) ***
Overall Low Exposure	REF	REF
Total ACEs (continuous)	0.63(0.44, 0.81) ***	0.66 (0.46,0.85) ***

^***<.001, adjusted models include the following covariates: Race, Ethnicity, Age, and Education

Mediation

195 women (of the 419 original cohort) who completed the ISEL-12 during pregnancy were included in the mediation analysis. Women from the study on the effect of neonatal DNA methylation on social-emotional development outcomes did not have measures of prenatal social support. Therefore, only women from the two clinical trials were included in the mediation analysis. Demographic characteristics of women included in the mediation analysis were similar to the overall cohort. From the mediation analysis we observed ACEs class assignment indirectly influenced depression risk during the postpartum period through its effect on perceived social support during pregnancy. As illustrated in Table 5 and Figure 2, women in ACEs classes 1-3 had an average decrease of two to four points the on the prenatal social support measure (ISEL scale) when compared to women in class 0. The relative indirect effects of prenatal social support on the ACEs class-EPDS score relationship was positive for all classes. The path coefficients for class 1, class 2, and class 3 were 0.49 (95% CI 0.0005, 1.32), 0.78 (0.09, 1.87), 0.45 (0.01, 1.20), respectively, with class 0 as the reference group, which suggests prenatal social support is a mediator for all classes. (Table 5)

Table 5.

Mediation analysis of Prenatal Social Support on the ACES Class-EPDS Score Relationship

Model	Path coefficient	SE	95% CI
ACES Class on Prenatal Social Support (path a)
Class 1 (High)	−2.56	1.26	−5.04, −0.08*
Class 2 (Emo/Phy)	−4.08	1.28	−6.62, −1.55**
Class 3 (Household Dys.)	−2.35	1.09	−4.51, −0.21*
ACES Class on EPDS Score
Prenatal Social Support (path b)	−0.19	0.06	−0.31, −0.07**
Relative Total Effect of ACES Class on EPDS Score (path c)
Class 1	5.92	1.08	3.78, 8.05***
Class 2	2.22	1.10	0.05,4.40
Class 3	3.76	0.93	1.91,5.60***
Relative Direct Effect of ACES Class on EPDS (path c')
Class 1	5.43	1.07	3.32,7.53***
Class 2	1.44	1.11	−0.74,3.63
Class 3	3.31	0.93	1.48,5.13***
Relative Indirect Effect of ACES Class on EPDS (path ab) †
Class 1	0.49	0.35	0.0005,1.32
Class 2	0.78	0.48	0.09,1.87
Class 3	0.45	0.31	0.01,1.20

n=189

^†Bootstrapped SE & 95% CI, P values are not calculated (CI that do not contain 0 are considered significant, indicated in bold ⁶⁵)

^*P<.05

^**P<.01

^***P<.001, Covariates including in the model: Race, Ethnicity, Age, and Education

Figure 2.

Statistical Diagram of Mediation analysis of Prenatal Social Support on the ACES Class-EPDS Score Relationship (adjusted for maternal age, race, ethnicity, education

*<.05, ** <.01, ***<.001

Discussion

In this retrospective cohort study of women enrolled in HV, we identified four classes based on the combination of ACEs endorsed. The classes we identified were consistent with previous studies that have used LCA methods to identify clusters of co-occurring ACEs. ^62,63 For example, a study examining ACE profiles and impulsivity identified four distinct classes of ACE exposure, including a household dysfunction group and maltreatment group, similar to classes identified in our study population.⁶⁴ In this study we observed differences by race and marital status in ACE class assignment and postpartum depression risk. Black women were less likely to be assigned to the high exposure ACE class and had lower risk of postpartum depression, when compared to White women. Next, using these classes we observed an association between classes and risk for depression postpartum. Women in class 1 (high exposure) had higher EPDS scores on average compared to women in the other classes and had greater than 3 times the RR of depression postpartum than women in the lowest exposure group. Our results align with prior research using the total ACE score and depression risk, as women in the highest exposure group had higher total ACE scores than women in the other groups. A previous study examining the type of ACEs (maltreatment vs. family dysfunction) and risk of depression during pregnancy observed the total number of maltreatment ACEs significantly increased total EPDS scores, however the authors did not observe a significant effect of total family dysfunction scores on depression risk during pregnancy.²⁷ In our study we demonstrated that women assigned to class 3, who had a high probability of experiencing family dysfunction and women in class 2, who had a high probability of experiencing maltreatment had similar and significant increased risk of postpartum depression. In this study we had a larger sample size (499 vs. 101), which allowed us to utilize LCA methods to examine how patterns of ACEs influence the risk of postpartum depression, which may contribute to the differential effect of family dysfunction on postpartum depression risk observed the current study. Additionally, the prior study examined depression during pregnancy, and we examined risk of postpartum depression in the current study.

When comparing the model fit of examining total ACE score and ACE class, we did not observe a meaningful difference based on the calculated AIC and BIC, indicating similar model fit. Although there was not a statistical advantage to examining ACE class versus using the total ACE score, identifying classes of ACEs is potentially clinically relevant. To illustrate, women in class 2 and 3 have similar total ACE scores, however they have very different profiles of ACEs. Women in class 2 had a high probability of experiencing physical abuse whereas women in class 3 had a high probability of living with an adult with substance abuse or mental illness. These differences in risk profiles are not apparent when using the total ACE score and support the need to identify the combinations of ACEs experienced to tailor resources. When examining the mediating role of social support in the relationship of ACE class and depression risk we observed a differential effect of ACE class on perceived social support during pregnancy. The effect of ACE class on prenatal social support, when compared to women in class 0 (low exposure class) was similar for women in class 1, high exposure class (path coefficient −2.56) and women in class 3, (path coefficient −2.35). However, prenatal social support was much lower, when compared to women in class 0, among women in class 2(path coefficient −4.08). The observed differences in the effect of ACE class on social support should be interpreted with caution as the sample size was small. Due to limitations of data, we were not able to examine why there are differences in perceived social support and future studies are needed to examine this relationship. There are several strengths of this study. The prospective cohort design allowed us to collect perceived level of social support and history of ACEs during pregnancy and captured depressive symptoms during the postpartum period. The LCA method allowed us to identify combinations of ACEs experienced by our cohort and examine how each class contributes to risk. Examining prenatal social support as a mediator extends our understanding of how ACEs contribute to depression risk among postpartum women. ACE class was associated with reduced prenatal social support. Results suggests an increase in prenatal social support could dampen the effect of ACEs on postpartum depression risk. There are important limitations of this study that need to be considered. Although our study sample demographics were comparable to the larger ECS population, our results may not be generalizable to all pregnant and postpartum women, as women from this study represent low-income women enrolled in a HV program in the states of Ohio and Kentucky. Our main exposure, ACE class, was based on data from the ACEs questionnaire. The ACEs questionnaire was administered at various time points, depending on when women enrolled in one of the three studies that comprised the current study’s population. It is possible the timing (prenatally vs. postpartum) of ACEs data collection could bias results. The mediation analysis was completed in a subpopulation of our cohort as not all women completed social support questionnaires prenatally. Nonetheless, the sample size was small, and results need to be replicated using a larger sample.

In conclusion, we identified four distinct classes of ACEs exposure in our study population and identified prenatal social support as a mediator between ACEs and depression risk. We have shown that various combinations of ACEs may infer differential risk for the development of postpartum depression in women participating in HV programs. This study extends prior research in this area through identifying classes of ACE exposure among women enrolled in a home visiting program and the mediating role of prenatal social support. Future interventions to ameliorate postpartum depression and its deleterious effect on child nurturing and development may be better informed by examining upstream maternal exposure to various ACEs. Therefore, future studies are needed to measure both the type and combinations of ACEs experienced by women along with collecting data on social support and other factors that may mediate the relationship between ACEs and postpartum depression risk. Overall, improving national policies and expanding research aimed at preventing and mitigating the effects of ACEs is needed to improve overall maternal and child outcomes.

List of abbreviations and acronyms

HV

Home Visiting

ACEs

Adverse childhood experiences

LCA

Latent class analyses

ECS

Every Child Succeeds

EPDS

Edinburgh Postnatal Depression Scale

ISEL-12

Interpersonal Support Evaluation List

AIC

Akaike Information Criterion

BIC

Bayesian information criterion