Intimate partner violence and postpartum emotional distress among South African women: Moderating effects of resilience and vulnerability factors

Abstract

In this study we aimed to identify factors that condition (i.e., buffer or exacerbate) the impact of exposure to intimate partner violence (IPV) on postpartum emotional distress among South African women. Hypothesized buffering factors included: socioeconomic status, family social support, and religiosity. Hypothesized exacerbating factors included: baseline distress, HIV status, and childhood abuse. Longitudinal analyses examined interactions between putative buffering and exacerbating factors and exposure to physical or sexual IPV, assessed during pregnancy (T1), as predictors of emotional distress, measured at 14 weeks (T2) and 9 months postpartum (T3). Consistent with hypotheses, at both T2 and T3 the impact of IPV exposure on emotional distress was significantly stronger among women who reported greater baseline distress and weaker among women of greater socioeconomic status. At T3, an interaction emerged with HIV status; the impact of IPV exposure on emotional distress was stronger for women who were diagnosed as HIV-positive during pregnancy. Findings support the need for targeted mental health promotion interventions for IPV-exposed women who are newly diagnosed with HIV and/or report high levels of emotional distress during pregnancy. Although more research is needed, findings also suggest that strengthening socioeconomic supports for IPV-exposed women may buffer impacts on postpartum mental health.

Introduction

Maternal mental health problems during the postpartum period are a pervasive public health problem with devastating consequences, particularly among women in low and middle-income countries (LMICs; Bennett et al., 2016; Gelaye et al., 2016; Stein et al., 2014). A systematic review found that the weighted mean postnatal prevalence of common perinatal disorders (e.g., anxiety and/or depression) was approximately 20% for women in LMICs as compared to 13% for women in high-income countries (Fisher et al., 2012; Rahman et al., 2013). A growing body of research in LMICs suggests these disorders have detrimental impacts for mothers and for the growth and development of children (Bennett et al., 2016; Stein et al., 2014). Further, compared to women in high-income countries, women in LMICs who experience postnatal mental health problems may be at increased risk for these negative consequences due to limited access to care and increased vulnerability stemming from comorbid health problems and socioeconomic disadvantage (Rahman et al., 2013; Stein et al., 2014).

One key risk factor that has been linked to poor postnatal mental health in LMICs is intimate partner violence (IPV) victimization. Exposure to IPV may lead to feelings of fear, anger, shame, and loss of control that, in turn, contribute to poor mental health during the transition to motherhood. Indeed, numerous studies have documented a consistent and robust link between IPV exposure and a range of perinatal mental disorders (for reviews, see, Bacchus et al., 2018; Halim et al., 2018; Howard et al., 2013). However, investigators have also suggested that the relationship between IPV exposure and mental health is not deterministic, but rather may be buffered and/or strengthened by individual and contextual factors (Fogarty et al., 2008; Grych et al., 2015; Illangasekare et al., 2013; LaFlair et al., 2015; Machisa et al., 2018). Many survivors of IPV function well (i.e., demonstrate resilience) despite having been exposed to severe forms of victimization, suggesting that protective factors may enable healthy adaptation (Grych et al., 2015; Machisa et al. 2018; Scott & Babcock, 2010). Conversely, the negative impacts of IPV on mental health may be exacerbated (i.e., strengthened) among women who have preexisting vulnerabilities or who experience co-occurring stressors, such as HIV infection. This notion is consistent with the diathesis-stress model (Monroe & Simons, 1991), which suggests that prior vulnerabilities (diatheses) can make it more challenging for individuals to deal with life stressors such as IPV, and with syndemics theory (Illangasekare et al., 2013; Singer, 2006), which posits that the synergistic interaction of co-occurring adverse conditions (e.g., IPV and HIV) can produce a more severe health outcome than if each were experienced alone.

Despite a compelling empirical and theoretical rationale for examining factors that might buffer or exacerbate the association between IPV and postpartum mental health problems, few studies have done so and almost no research along these lines has been conducted in LMIC. In fact, a recent systematic review of studies in LMIC examining the link between IPV and perinatal mental disorders identified only one study that examined potential effect modifiers (Halim et al., 2018). This is an important gap in the literature; the identification of factors that buffer the impact of IPV can inform the development of interventions that strengthen resilience among pregnant IPV victims. Similarly, the identification of factors that exacerbate the negative mental health impacts of IPV during the perinatal period can inform our understanding of who is most at risk and determine whether multicomponent interventions are needed.

The current study

The primary aim of the current study was to identify factors that condition (i.e., moderate) the impact of IPV on postpartum emotional distress in a longitudinal cohort of South African women who were assessed during pregnancy and then again at 14 weeks and 9 months postpartum. Drawing from resilience perspectives, we examined three factors-socioeconomic status, family social support, and individual religiosity-that may enable women to positively cope with the negative impacts of IPV and thus attenuate effects on postpartum distress (Howell et al., 2018; Machisa et al., 2018). Further, drawing from diathesis-stress and syndemics perspectives, we examined three potential effect magnifiers-history of childhood abuse, new HIV diagnosis, and baseline emotional distress levels-that may work synergistically with IPV to exacerbate impacts on postpartum emotional distress (Illangesekare et al., 2013; Fogarty et al., 2008; Warshaw et al., 2009). Our overarching study hypothesis was that the relationship between current relationship IPV, assessed during pregnancy, and postpartum emotional distress would be: (1) weaker for women with greater socioeconomic status, family social support, and/or religiosity levels and (2) stronger for women who are newly diagnosed with HIV, have a history of childhood abuse, and/or who report greater baseline emotional distress.

Methods

Design and procedure

Secondary analyses were conducted using extant data from a randomized controlled trial that examined the efficacy of enhanced HIV counseling for risk reduction during pregnancy and postpartum. Women seeking antenatal care from a health clinic in KwaZulu-Natal, South Africa were recruited into the study. Eligible participants had to have had an intimate partner for at least six months and be unaware of their HIV status (never tested for HIV or tested negative at least 3 months prior to enrollment). Women who consented to participate completed a baseline (T1) interview and were then randomized to receive either enhanced counseling (treatment group) or standard of care (control group) and tested for HIV. All women in the study were provided with HIV counseling, with enhanced counseling provided to those assigned to the treatment group. Follow-up interviews were conducted at 14 weeks (T2) and 9 months postpartum (T3). Recruitment for the study began in 2008 and data collection ended in 2011. Women were reimbursed 70 South African Rand, equivalent to $8 USD at each assessment visit.

A total of 1480 women completed a baseline interview, were assigned to a counseling treatment condition, and completed HIV-testing. Follow-up retention rates at T2 and T3 were 78% (n=1154) and 75% (n=1104) respectively. The study was approved by the Institutional Review Board at the University of North Carolina and the University of KwaZulu-Natal. For additional details on recruitment, randomization, intervention, and data collection, see Maman et al. 2014.

Measures

Emotional distress.

Emotional distress was measured using the Hopkins Symptoms Checklist (HSCL-25; Hesbacher, Rickels, Morris, Newman, & Rosenfeld, 1980), which has been found to be a reliable and valid measure of mental health in studies in sub-Saharan Africa (Ashaba et al., 2018; Kaaya et al., 2002; Kagee et al., 2017). The measure includes 25 items assessing symptoms of emotional distress; respondents were asked how much discomfort each symptom (e.g., feeling lonely) had caused in the past week with response options ranging from “not distressed at all” (1) to “extremely distressed” (4). Scores were averaged across items to create a composite emotional distress score at baseline (α=.90), 14-week (α=.92), and 9-month (α=.93) postpartum follow-up.

Physical or sexual IPV victimization.

Seven items from a modified version of the Violence Against Women Instrument (VAWI; Garcia-Moreno et al., 2006) were used to assess severe physical and sexual IPV in the current relationship at baseline. Women were asked how many times their current partner had perpetrated the following acts of IPV against them prior to or during pregnancy: hit them with a fist or something else that could hurt; kicked, dragged, or beat them up; choked or burnt them on purpose; threatened to use or actually used a gun, knife or other weapon that could hurt them; physically forced them to have sex; used threats to make them have sex; or forced them to do something sexual that was degrading or humiliating. Response options ranged from “never” to “ten or more times.” Because item distributions were highly skewed and zero-inflated, scores on the items were summed and dichotomized to create a binary indicator that denoted whether the woman had or had not ever experienced any of these acts of IPV in their current relationship.

Socioeconomic status.

Following previous research in South Africa (Bärnighausen et al., 2007), we created a measure of socioeconomic status (SES) by using principal components analysis to derive a linear index from a series of asset ownership indicators and then categorized participants as belonging to the poorest 40%, middle 40%, or wealthiest 20% on the asset index scale.

Family social support.

The Norbeck Social Support Questionnaire (Norbeck et al., 1981) was used to assess family social support. Respondents were asked to list up to six individuals who provide them with support in their lives and then were asked six questions assessing the extent to which each individual provided them with instrumental and/or emotional support. A sample item is “how much does this person make you feel liked or loved?” Response options ranged from “not at all” (0) to “a great deal” (4). To create a composite support score, for each individual who the respondent identified as a “family member or relative,” we summed scores on the social support questions and then averaged across all family members named as support providers (range 0 to 24). (Cronbach’s α=.70). Due to sparseness and skew in the distribution of this score (47% of the sample were scored at the minimum or maximum on the scale) we classified individuals into four categories as follows: scores of 0 (no support; 20% of the sample) were coded as “0”; scores of 1–18 (low support; 12% of the sample) were scored as “1”; scores of 19–23 were scored as “2” (moderate support; 40% of the sample); and a score of 24 (high support; 27% of the sample) was scored as “3.”

Religiosity.

Religiosity was assessed by two items. Religious participation was assessed by the question: “how often do you participate in church/faith-related activities?” Response options ranged from never (0) to one or more times per week (4). Religious beliefs were assessed by the question: “how strong would you say your religious beliefs are?” Response options ranged from not at all religious (0) to very religious (4). Items were standardized and averaged to create a composite religiosity score (r=.55, p<.0001).

HIV status.

HIV status, coded as “1” for HIV-positive and “0” for HIV-negative, was determined at baseline by two rapid HIV tests with a fourth-generation assay used as a tiebreaker in cases of discordant findings. To be eligible for the study women had to report never having tested for HIV or having tested negative at least three months prior to enrollment; thus, an HIV-positive status represented a new HIV diagnosis for all participants.

Experience of childhood abuse.

Childhood abuse exposure was coded as “1” if the participant reported “unwanted sexual experiences” (defined as inappropriate touching or unwanted sexual intercourse) and/or “serious physical violence” (defined as being hit, punched, kicked, or beaten up in a way that resulted in serious harm) prior to age 12 and as “0” otherwise.

Covariates.

We included several covariates that could plausibly be associated both with emotional distress and IPV exposure and thus could potentially confound associations including age, education, relationship length, cohabitation status, previous pregnancy, and gestational age at baseline. Age was coded as number of years. Education was coded as “0” for those who reported that the highest standard passed was 5 or less (grade 7 or lower); those who reported reaching standards 6–9 (grades 8–11) were coded as “1”; and those who reported matriculation from high school or higher were coded as “2”. Relationship length was coded as number of years in current relationship. Lives with current partner was coded as “1” for those who reported currently living with their partner (regardless of marital status) and as “0” otherwise. Previous pregnancy was coded as “0” if the women reported never having been pregnant prior to the index pregnancy; “1” if she reported one previous pregnancy; and “2” if she reported more than one previous pregnancy. Estimated gestational age was coded as number of weeks gestation. Treatment condition and baseline to follow-up time interval were also included as covariates in all analyses.

Analytic strategy

A small number of cases (2% of those retained at T2 and 1% of those retained at T3) were missing on study covariates and were not included in analyses yielding analytic samples of n=1129 at 14 weeks and n=1089 at 9 months postpartum. To facilitate interpretation, all continuous variables entered as predictors were centered prior to analysis. Generalized linear models were then used to assess interactions between IPV exposure and each of the putative effect modifiers (six interaction terms) in predicting postpartum emotional distress at 14 weeks and 9 months postpartum follow-up. For each follow-up, we first estimated a main effect model (Model 1) that included IPV exposure, the moderator variables, and all study covariates and then estimated a full model that additionally included all six interaction terms (Model 2). To control for type 1 error due to multiple comparisons for each of the two models an omnibus F test was conducted to evaluate the statistical significance of the set of interaction terms (Fletcher et al., 1989). If the omnibus test was significant, we examined the p-values for each of the individual interaction terms in Model 2 and probed significant interactions by estimating the model-predicted simple slope for the effect of IPV on postpartum distress at different levels (described further in results) of the moderator variable.

Results

At baseline, the average age of participants was 25.5 years (interquartile range [IQR]=21.3–28.6); average gestational age of pregnancy was 25 weeks (IQR=20.7–29.5); and average relationship length was 4.5 yrs (IQR=1.5–6.2). Approximately half of the participants (49%) reported their highest education as grade 7 or lower; over a third (39%) tested positive for HIV; 5% reported having experienced childhood abuse; and 18% reported having experienced any physical or sexual IPV in their current relationship. Only about a quarter (26%) of participants reported living with their current partner and 35% reported this was their first pregnancy. Attrition analyses were conducted to assess associations between baseline measures and loss to follow-up. Loss to follow-up was unrelated to baseline emotional distress, IPV exposure, child abuse exposure, family social support, religiosity, education, treatment group status, having had a previous pregnancy, relationship length or cohabitation status. Loss to follow-up was more likely among women who tested positive for HIV, were younger, were of lower SES, and were of greater gestational age at baseline. These variables were included in all analytic models to increase the plausibility of missing at random.

Bivariate correlations among key study variables including IPV exposure, putative moderators, and postpartum emotional distress at T1 and T3 are presented in Table 1. Notably, physical/sexual IPV exposure, baseline emotional distress, childhood abuse exposure, and family support, were each significantly correlated with postpartum distress in the expected direction at T2 and T3. Religiosity was negatively correlated with distress at T3 but not at T2. There was no bivariate association between baseline HIV diagnosis or SES and emotional distress at either follow-up.

Table 1.

Correlations among focal study variables

Variable		2.	3.	4.	5.	6.	7.	8.	9.
1.	Physical/sexual IPV exposure	−.09*a	.01a	−.10**a	.04b	.19*b	.25***c	.17***c	.24***c
2.	Family social support	--	.15***a	.03a	−.02a	−.11^a	−.09***d	−.07*d	−.07*d
3.	Socioeconomic status		--	−.02a	−.15**a	−.01a	.05*d	.004d	−.02d
4.	Religiosity			--	−.05a	.07a	−.03d	−.03d	−.07*d
5.	HIV positive				--	.04b	−.03c	.04c	.04c
6.	Childhood abuse exposure					--	.11***c	.09**c	.06*c
7.	T1 Emotional Distress						--	.43***d	.42***d
8.	T2 Emotional Distress							--	.44***d
9.	T3 Emotional Distress								--

^aPolychoric correlation

^bTetrachoric correlation

^cPoint biserial correlation

^dSpearman correlation

Note.

^{^}p<.10

^*p<.05

^**p<.01

^***p<.001

Tables 2 and 3 report parameter estimates for models estimating emotional distress at 14 weeks and 9 months postpartum respectively. Omnibus F tests assessing the joint significance of the set of interaction terms were statistically significant in models predicting distress at both 14 weeks, F(6, 1107)=2.79, p=.01, and 9 months, F(6,973)=2.77, p=.01, postpartum. As shown in Table 3, interactions between IPV and baseline distress (p=.04) and SES (p=.002) were both statistically significant in the model predicting distress at 14 weeks postpartum. As shown in Table 3, interactions between IPV and baseline distress (p=.007), SES (p=.02), and HIV status (p=.03) were each statistically significant in the model predicting distress at 9 months follow-up. No interactions emerged between IPV and childhood abuse, family support, or religiosity at either follow-up.

Table 2.

Parameter estimates from models examining psychosocial moderators of the longitudinal association between physical or sexual intimate partner violence (IPV), measured during pregnancy, and emotional distress at 14 weeks postpartum.

	Model 1		Model 2
	Main effects		Interactions
	b (se)	p	b (se)	p
Physical/sexual IPV	0.08 (.04)	.03	0.19 (.07)	.003
T1 emotional distress	0.43 (.03)	<.001	0.40 (.03)	<.001
Childhood abuse exposure	0.09 (.06)	.15	0.14 (.07)	.05
HIV positive	0.06 (.03)	.03	0.06 (.03)	.048
Family social support	−0.01 (.01)	.38	−0.01 (.01)	.50
Socioeconomic status	−0.02 (.02)	.37	0.01 (.02)	.73
Religiosity	−0.02 (.02)	.23	−0.01 (.02)	.40
Interactions
IPV x T1 emotional distress	--	--	0.13 (.06)	.04
IPV x Childhood abuse exposure	--	--	−0.24 (.15)	.12
IPV x HIV positive	--	--	0.02 (.08)	.79
IPV x Family social support	--	--	−0.01 (.03)	.68
IPV x Socioeconomic status	--	--	−0.15 (0.5)	.002
IPV x Religiosity	--	--	−0.04 (.04)	.34

Note. Models control for age, education, gestational age, cohabitation status, relationship length, previous pregnancy, baseline to follow-up interval, and treatment condition. Bolded interactions are statistically significant at p<.05.

Table 3.

Parameter estimates from models examining psychosocial moderators of the longitudinal association between physical or sexual intimate partner violence (IPV), measured during pregnancy, and emotional distress at 9 months postpartum.

	Model 1		Model 2
	Main effects		Interactions
	b (se)	p	b (se)	p
Physical/sexual IPV	0.20 (.04)	<.001	0.18 (.07)	.007
T1 emotional distress	0.35 (.03)	<.001	0.31 (.03)	<.001
Childhood abuse exposure	0.01 (.07)	.83	−0.01 (.08)	.87
HIV positive	0.05 (.03)	.08	0.03 (.03)	.39
Family social support	−0.01 (.01)	.43	−0.01 (.01)	.42
Socioeconomic status	−0.03 (.02)	.08	−0.02 (.02)	.38
Religiosity	−0.02 (.02)	.20	−0.02 (.02)	.34
Interactions
IPV x T1 emotional distress	--	--	0.18 (.07)	.007
IPV x Childhood abuse exposure	--	--	0.10 (.16)	.53
IPV x HIV positive	--	--	0.17 (.08)	.03
IPV x Family social support	--	--	0.002 (.04)	.96
IPV x Socioeconomic status	--	--	−0.12 (.05)	.02
IPV x Religiosity	--	--	−0.05 (.04)	.28

Note. Models control for age, education, gestational age, cohabitation status, relationship length, previous pregnancy, baseline to follow-up interval, and treatment condition. Bolded interactions are statistically significant at p<.05.

Post-hoc probing of the interactions reported in Table 3 suggest that, as expected, the impact of IPV exposure on emotional distress at 14 weeks postpartum was stronger for women who reported greater baseline distress. The simple slope denoting the impact of IPV exposure on distress at 14 weeks postpartum was b=.13 (p=.09) for women who reported low levels of T1 distress (16^th percentile); b=.17 (p=.01) for women reporting average (50^th percentile) levels of distress; and b=.26 (p<.001) for women reporting high levels of T1 distress (84^th percentile). Also as expected, the impact of IPV on emotional distress at 14 weeks postpartum was weaker for women of higher SES. The simple slope denoting the impact of IPV exposure on distress at 14 weeks postpartum was b=.19 (p=.003) for women with low socioeconomic assets (poorest 40% on household assets index), but was not statistically significant for women with average or high socioeconomic assets (p>.05).

As shown in Table 3, interactions between IPV and baseline distress and between IPV and SES were sustained at 9 months postpartum such that impacts of IPV exposure were again stronger for women who reported greater baseline distress and weaker for women with greater socioeconomic assets. The simple slope denoting the impact of IPV exposure on emotional distress at 9 months postpartum was not significant for women who reported low (16^th percentile; p=.22) levels of baseline distress, but was significant for women reporting average (50^th percentile; b=.15, p=.03) and high (84^th percentile; b=.28. p<.0001) levels of baseline distress. The simple slope denoting the impact of IPV on emotional distress at 9 months postpartum was b=.18 (p=.007) for women with low socioeconomic assets (poorest 40% on household assets index) but was not statistically significant for women with average or high socioeconomic assets (p>.05). In addition, a new interaction between IPV and HIV status emerged at 9 months postpartum. Probing of this interaction suggest that, as expected, the impact of IPV exposure on emotional distress at 9 months postpartum was greater for women diagnosed during pregnancy as HIV positive (b=.36; p<.0001) compared to those who were diagnosed as HIV-negative (b=.18; p=.007). Figure 1 provides a visual depiction of each of the interactions retained in Model 2 (interactions from Model 1 are not depicted as they follow the same pattern as those from Model 2).

Figure 1.

Moderators of the association between physical or sexual intimate partner violence (IPV), assessed during pregnancy (T1), and emotional distress at 9 months postpartum (T2). Panels depict model-predicted postpartum distress as a function of IPV exposure at: (1) low (16^th percentile), average (50^th percentile) and high (84^th percentile) levels of T1 emotional distress (Top); (2) low, middle, and high SES; with categories defined by being in the poorest 40%, middle 40%, or wealthiest 20% on the asset index scale (Middle), and (3) positive and negative T1 HIV diagnosis (Bottom).

Discussion

Theory and research suggest that the mental health impacts of IPV exposure vary depending on individual and contextual factors that may buffer or exacerbate (i.e., moderate) associations. Few studies, however, have examined potential effect moderators and almost no research along these lines has been conducted in LMIC or during the perinatal period. In the current study we address these gaps by examining resiliency (SES, family social support, religiosity) and vulnerability factors (baseline distress, childhood abuse history, HIV diagnosis) as potential moderators of the longitudinal impact of physical/sexual IPV exposure on postpartum distress in a sample of South African women.

Consistent with expectations, we found that the strength of the association between IPV exposure and postpartum distress was stronger for women who reported greater levels of distress during pregnancy and weaker for women with greater socioeconomic assets. Further these effects persisted across the 14 weeks and 9 months postpartum follow-up assessments. The finding that socioeconomic status buffered the negative impacts of IPV exposure is consistent with research that has found that SES buffers against the negative impacts of trauma (e.g., Mock & Arai, 2011). This may be because greater SES enables IPV-exposed women to separate from abusive partners, permits access to counseling and treatment, and/or allows for more effective coping and adaptation in general during the perinatal period. Further, women of greater SES may be more socially connected and have access to greater practical support from social networks outside of the family. In a study of South African IPV survivors, Machisa et al. (2018) found that women who reported that they were able (vs. not able) to find money in an emergency were more likely to be resilient (i.e., under the threshold for depression and PTSD symptoms). Taken together, these findings suggest that IPV-exposed women of low socioeconomic status are thus an important target population for mental health promotion efforts during the perinatal period. Programs that provide socioeconomic supports to women during the perinatal period, such as cash transfer interventions (Glassman et al., 2013; Powell-Jackson et al., 2016), may be a particularly promising approach as such programs may buffer the impacts of IPV exposure on mental health. Future research is needed to evaluate the impacts of such interventions among pregnant and postpartum IPV-exposed women in LMICs.

That IPV exposure impacts were greater for women reporting higher levels of emotional distress during pregnancy suggests that IPV-exposed women may be particularly likely to experience persistence (vs. discontinuity) of poor mental across the antenatal and postnatal periods. Consistent with diathesis-stress models, women who are emotionally distressed may be particularly vulnerable to the stressful impacts of IPV exposure, leading to prolonged distress across the perinatal period (Elwood et al., 2009). In terms of prevention implications, this finding suggests an important target population for programs that aim to improve postpartum mental health in South Africa are IPV-exposed women who also report high levels of antenatal emotional distress. Research evaluating such programs in LMICs is urgently needed. A recent systematic review of interventions for reducing domestic violence among pregnant women in LMICs found that only one of the five programs identified for inclusion assessed impacts on participants mental health and concluded there is “…narrow and inconclusive evidence with regards to the effectiveness of [domestic violence] interventions among pregnant women [in LMICs]” (Sapkota et al., 2019, p. 8).

At 14 weeks postpartum, having received an HIV-positive diagnosis during pregnancy was significantly positively associated with increased emotional distress (main effect), but no interaction was detected between HIV and IPV exposure. At nine months postpartum an interaction emerged between IPV exposure and HIV-status such that the impact of IPV exposure on postpartum distress was stronger for HIV-positive compared to HIV-negative women. This finding is consistent with syndemics models that posit that co-occurring health risks such as HIV and IPV may exert conjoint effects on health outcomes and suggests that “bundled” prevention efforts that jointly target both HIV and IPV may be most effective in preventing postpartum distress (Tsai & Venkataramani, 2016).

Notably, while childhood abuse exposure, religiosity, and family social support demonstrated significant bivariate correlations with postpartum distress, in adjusted analyses these variables were not significantly associated with distress at either follow-up and did not moderate the impact of IPV exposure. It may be that the impacts of these variables were explained by other constructs in the models. For example, the impacts of childhood abuse exposure may have been mediated by the baseline distress variable. Alternatively, measurement error may have attenuated our ability to assess the impacts of these constructs. Our measures of both childhood abuse exposure and religiosity were limited to two items and may not have captured the full dimensionality of these constructs. Further, while our assessment of family social support was comprehensive, research in this setting suggests that family social support may operate in complex and nuanced ways among IPV victims. For example, Machisa et al. (2018) found that the positive impact of seeking family support on resilience among IPV-exposed South African women depended on whether the woman reported positive or negative reactions to abuse disclosure.

The current study had several strengths including the a relatively large sample and the use of a longitudinal design that examined prospective impacts of IPV exposure on postpartum distress across two follow-up points while controlling for baseline distress levels. Further, the study drew on theory and empirical evidence to examine multiple potential effect modifiers while controlling for type one error due to multiple testing. Notwithstanding these strengths, the study also had several limitations including reliance on self-report and the use of a convenience sample of adult women (≥ 18 years old) recruited from one antenatal clinic, which limits the generalizability of findings to other populations (e.g., adolescent populations, women from rural areas). Future research should seek to replicate findings in other types of populations and build on the current study to examine other potential effect modifiers including health service-related and/or community factors, such as supportive attitudes towards abuse victims, that may confer resilience (Machisa et al., 2018). Such research should also examine other indicators of socio-economic support, such as the receipt of social grants, that may provide supports to IPV-exposed women that enable financial independence from abusive partners and buffer impacts on distress (Rosenburg et al. 2015). In addition, the current study assessed the mental health impact of having experienced serious physical or sexual IPV in the current relationship. Future research should build on our findings to examine moderators of the impacts of other forms of IPV and more broadly assess IPV exposures across the life-course.

Conclusion

To our knowledge this is the first longitudinal study to examine factors that may buffer or enhance the impacts of IPV exposure on postpartum distress among South African women. Findings suggest that SES, baseline distress, and HIV status work synergistically with IPV exposure to impact distress levels during the postpartum period. Mental health promotion efforts for women who screen positive for IPV during pregnancy should ensure that targeted services are offered to women of low SES and to those who report high levels of antenatal distress and/or are newly diagnosed as HIV positive.