The Role of Maternal Intimate Partner Violence Victimization on Neonatal Mortality in Ethiopia

Abstract

Background: Previous studies have suggested that when intimate partner violence (IPV) tends to be persistent across a woman’s life span, her newborn offspring have a higher risk of ill health and mortality. There is a high prevalence of both IPV and neonatal mortality in Ethiopia, but the issue of IPV has remained largely outside the focus of child survival programs in this country. One of the noticeable reasons is a lack of evidence regarding the effect of IPV on neonatal mortality. Therefore, this study investigated the effect of maternal IPV on neonatal mortality in Ethiopia. Method: This study used nationally representative data from the 2016 Ethiopian Demographic and Health Survey. A total of 2,863 currently married women of reproductive age who gave birth in the preceding 5 years were included in analysis. Regression models using propensity scores were used. Results: The prevalence of physical, emotional, and sexual IPV were 24.5%, 22.9%, and 12.0%, respectively. About 56% of women had also experienced at least one act of partner controlling behavior. Maternal IPV experience (a composite measure of physical, sexual, and emotional abuse) was associated with increased neonatal mortality (adjusted odds ratio [AOR] = 2.58, 95% confidence interval [CI] = [1.03, 6.45]). In addition, the odds of neonatal mortality were 2.75 times (AOR = 2.75; 95% CI = [1.05, 7.2]) higher among women who had experienced three or more partner controlling behaviors than women who had experienced less than three or none. Conclusion and implication: Maternal IPV is significantly associated with risk of neonatal mortality in Ethiopia. There is a clear need for IPV interventions in child survival programs. Therefore, existing neonatal survival strategies should focus beyond the direct causes of neonatal mortality, and they need to target IPV as an underlying factor to neonatal morbidities and mortality.

Keywords

Intimate partner violence partner controlling behaviour neonatal mortality inverse probability of treatment weighting propensity score Ethiopia

Introduction

Intimate partner violence (IPV) is the most common form of violence against women, and a global public health problem (Garcia-Moreno et al., 2006). Although it occurs across all socioeconomic and cultural contexts around the globe, the magnitude of the problem is higher in developing countries (Garcia-Moreno et al., 2006; Sunita & Kiersten, 2004). Due to a lack of effort in IPV prevention and rehabilitation for victims, the consequences for women in developing countries are serious (Emenike et al., 2008).

The consequences of IPV range from individual health problems to macro-level socioeconomic impacts (World Health Organization [WHO], 2013). IPV is considered a “health hazard” and plays a role in women’s inability to attain and enjoy a healthy life for themselves and their children (Sunita & Kiersten, 2004). IPV-induced maternal health problems could range from minor physical trauma to more serious disabilities and death (Dillon et al., 2013; Emenike et al., 2008; Sunita & Kiersten, 2004). When IPV tends to be persistent across a woman’s life span, the newborns of these mothers are more likely to suffer from ill health and mortality (Koenig et al., 2010; Taft et al., 2015).

There are different explanatory theories that link IPV and neonatal mortality. Generally, these theories can be explained in the way that IPV can lead to neonatal mortality either directly such as physical trauma (Dutton et al., 2006; Sarkar, 2013), vertical transmission of STDs due to sexual IPV (Taft et al., 2015), and poor emotional well-being (Asling-Monemi et al., 2003; Goldstein & Martin, 2004; Dutton et al., 2006). IPV can also affect neonatal mortality when husbands attempt to closely control and monitor their wives’ behavior, which may affect maternal decision-making autonomy (Furuta et al., 2016; Goldstein & Martin, 2004), health service access and utilization (Mohammed et al., 2017), fertility behaviors (Pallitto & O’Campo, 2004), and maternal general health and nutrition (Ahmed et al., 2006; Koenig et al., 2010).

There is a high prevalence of both IPV (Garcia-Moreno et al., 2006; Sunita & Kiersten, 2004; WHO, 2013) and neonatal mortality in developing nations (Lawn et al., 2005), yet the issue of IPV has remained largely outside the focus of child survival programs in these countries (Ahmed et al., 2006). A noticeable reason is the absence of strong evidence on the association between IPV and neonatal mortality (Koenig et al., 2010). Moreover, most available literature on IPV and its negative effects precluded defining partner controlling behavior as a form of IPV. However, there is evidence that partner controlling behavior is a reflection of power dynamics in an intimate relationship, increased vulnerability to abuse, and a warning sign of abuse (Gage & Hutchinson, 2006; Krantz & Nguyen, 2009). The WHO definition of IPV also includes “controlling behaviour” as a form of IPV (Krug et al., 2002, p. 89).

Neonatal mortality, a measure of the socioeconomic status and quality of life of a population, is high in Ethiopia. Neonatal mortality is defined as the death of a child before the first month of life (within 28 days after birth) (Central Statistical Agency [CSA], 2017). Neonatal mortality is the major contributor to under-five mortality in Ethiopia (Mekonnen et al., 2013), with the majority of under-five child death in Ethiopia occurring during the first month of life, and 74% of neonatal deaths occurring within the first week of life (CSA, 2017). Importantly, rates of neonatal mortality have not decreased as much as those recorded for infant and under-five mortality. The rate of neonatal mortality has decreased by 41% (from 49 to 29 deaths per 1,000 births) from 2000 to 2016, while under-five mortality has decreased by 60% (from 166 to 67 deaths per 1,000 live births) in the same time period (CSA, 2017). If this trend continues, it will be difficult to achieve the goal the country set to reduce the neonatal mortality rate to 10 per 1,000 live births by 2020 (Federal Ministry of Health [FMoH], 2015).

Previous studies have documented multiple distal and proximal factors involved in the high prevalence of child mortality and neonatal mortality in Ethiopia (Mekonnen et al., 2013; Yaya et al., 2014). However, none of these studies investigated the potential effect of IPV as a risk factor for neonatal mortality, in spite of the high prevalence of IPV in Ethiopia. Evidence from around Ethiopia reveals that the lifetime experience of IPV by women ranges from 19% to 71%, compared with the global average of 35% (Abeya et al., 2011; Deribe et al., 2012; Garcia-Moreno et al., 2006; Semahegn & Mengistie, 2015). These high figures occur in a country where reporting IPV is considered shameful, where women hide the problem due to fear of further abuse, and where controlling behavior is considered an acceptable behavior for husbands in interactions with their wives (Abeya et al., 2012; Garcia-Moreno et al., 2006).

Evidence also suggests that women are not free of IPV during pregnancy (WHO, 2013). This is evident in Ethiopia as a proportionate number of women, compared with the general adult female population, experienced IPV during pregnancy (Abebe et al., 2016; Yimer et al., 2014). In some areas of the country, about half of pregnant women experienced IPV during their pregnancy (Abebe et al., 2016). For example, studies from Northern Ethiopia showed the prevalences of IPV during pregnancy were 44.5% (Abebe et al., 2016).

In summary, IPV has been associated with neonatal mortality in previous studies from developing countries, but has not been examined in Ethiopia, where the prevalence of both IPV and neonatal mortality are relatively high. Therefore, the aim of this study was to investigate the effect of maternal IPV experiences on neonatal mortality in Ethiopia. The association between partner controlling behavior, as one form of IPV, and neonatal mortality was also examined vis-a-vis the hypothesis that the concentration of partner controlling behaviors has a more significant effect than any single behavior.

Method

Data Source

This analysis used data from the 2016 Ethiopian Demographic and Health Survey, when the domestic violence module was added. This was a national survey conducted from January 18 to June 27, 2016, under the collaboration of the central statistical agency of Ethiopia, the Federal Ministry of Health, the United States Agency for International Development (USAID), and with financial support from other organizations. The 2016 Ethiopian Demographic and Health Survey used a de facto data collection method (all individuals who spent the night preceding the survey in the selected household were eligible for the survey), with five questionnaires (household, women’s, men’s, biomarker, and health facility) (CSA, 2017).

The woman’s questionnaire was used to collect a range of information on women’s and children’s health. Moreover, women were asked in detail about each reported pregnancy and birth 5 years before the survey (retrospectively). First, each selected woman was asked her experience with childbearing (number of children ever born). For each live birth, questions were asked concerning sex, date of birth (month and year), whether the birth was single or multiple, survival/death of each birth, and, if the child was deceased, the age at death was recorded (CSA, 2017).

Sample Size and Sampling Procedures

The Ethiopian Demographic and Health Survey uses 84,915 enumeration areas; each enumeration area has an average of 181 households from nine regions and two city administrations. A two-stage stratified cluster sampling was then implemented. First, 645 enumeration areas were selected from urban (202 enumeration areas) and rural (443 enumeration areas) areas based on proportional to size allocation. In the second stage, on average 28 households per each selected enumeration area were selected using systematic random sampling (CSA, 2017). All women aged 15 to 49 years in the household were eligible for the interview. Accordingly, 15,683 women, with a response rate of 95%, participated in the general survey (CSA, 2017).

For the domestic violence sub-study, only one married woman per household was interviewed. Of the 5,860 women who were eligible, 97% were interviewed, with 3% not involved mainly due to a lack of privacy. Background characteristics between selected women for the IPV sub-study and the general female population in the selected households was shown to be similar and have been determined to not cause problems in terms of representativeness (CSA, 2017).

For this analysis, only women who were married during the time of the interview, who gave birth in the last 5 years before the survey, and responded to the IPV questionnaire were included. Mothers without any birth recorded and those missing IPV data were excluded. If women had more than one under-five children, the most recent birth (index child) was included in the analysis. A total of 2,863 women were included in the analysis.

Measurement and Variables

For this analysis, there were three groups of variables: a dependent (outcome) variable, treatment (exposure) variable, and potentially control variables.

Dependent variable

The outcome variable was mortality status of the index child during the neonatal period, which is a dichotomous variable denoting whether a neonate was deceased or alive during the first month of life (within 28 days after birth) (CSA, 2017).

Treatment variables

The treatment variables of interest were IPV (defined in three ways) and partner controlling behavior. IPV was measured based on women’s self-reporting to questions designed and based on the modified Conflict Tactic Scales (CTS) of Straus, 1997 (as cited in CSA, 2017). Women were asked whether or not they had experienced a number of violent acts (listed in Supplemental Appendix I) within their relationship, perpetrated by their husband/partner for currently married women and recent husband/partner for previously married women (CSA, 2017). The tool has 13 questions with Questions 1 to 4 measuring less severe physical IPV, 5 to 7 severe physical IPV, 8 to 10 sexual IPV, and 11 to 13 emotional IPV. Respondent was categorized as having experienced lifetime IPV if she has experience of IPV since the age of 15 years (CSA, 2017). Likewise, partner controlling behavior was categorized as “yes” if one of the following behaviors were reportedly carried out on a woman by her husband: “being jealous if she talks to men,” “accusing her of being unfaithful,” “does not allow her to meet her friends,” “limits her contact with family,” and “tries to know where she is at all times” (CSA, 2017).

Control variables

Control variables were baseline covariates which needed to be controlled to examine the unbiased effect of the exposure on the outcome. These variables were identified based on an examination of the literature, both globally (Ackerson & Subramanian, 2009; Ahmed et al., 2006; Koenig et al., 2010; Rico et al., 2011; Silverman et al., 2011) and in Ethiopia (Debelew et al., 2014; Mekonnen et al., 2013; Yaya et al., 2014). Accordingly, 24 potential control variables were considered initially. The potential control variables included are age at birth (years), respondent’s educational status, respondent’s employment status, place of residence, region, number of living children, number of previous adverse events, respondent’s substance abuse, decision-making autonomy, partner’s education, partner’s occupation, access to media, and household wealth index.

The other control variables are antenatal care (ANC), delivery care, postnatal care (PNC), tetanus toxoid immunization, sex of neonate, desiredness of the pregnancy, multiple pregnancy, birth order, birth interval, birth weight, and mode of delivery. The definitions of these variables and their categorization are presented in Supplemental Appendix II.

Data Processing and Analysis

In this study, the prevalence of the outcome (neonatal mortality) was low 53 (1.9%), presenting challenges for multivariable regression modeling to adjust for control variables (CSA, 2017). When the outcome is binary and rare, the exposure is binary, and the number of covariates is many, propensity scores are proposed as an alternative to the more common multivariable regression analysis methods (Deb et al., 2016). A propensity score, also called a balancing score, is an estimate of the conditional probability (ranging from 0 to 1) of a participant being exposed to the exposure, given the baseline covariates (potential control variables) (Deb et al., 2016). There are several approaches to using propensity scores to estimate the independent effect of an exposure on an outcome. Inverse probability treatment weighting was used in this analysis (Austin, 2011). Compared with propensity score matching methods, Inverse probability treatment weighting helps with retaining data in the analysis, whereby unmatched samples are not removed from the analysis; rather, a pseudo sample is created by weighting the data based on the propensity score. Hence, all the data can be included in the analysis (Austin, 2011).

To account for the complex sampling procedures (multi stage stratified cluster sampling) used by this survey, which may result in unequal probability of selection for certain areas or subgroups, sampling (survey) weights were adjusted throughout the analysis procedures to allow generalization of results to the national level (CSA, 2017).

The analysis was conducted via the following five steps:

Step 1. Selecting control variables: Potentially, control variables were selected based on prior knowledge (literature) and context. A minimal adjustment set was then identified using a directed acyclic graph (DAG), constructed using DAGitty version 2.3 (www.dagitty.net). A DAG is a method of identifying confounding by mapping a presumed network of causal relationships (Greenland et al., 1999). Of the 24 variables initially considered (Supplemental Appendix II), a minimal adjustment set of nine was identified. These were respondent age at birth, respondent’s educational status, access to media, household wealth index, region of residence, urban/rural residence, respondent’s substance abuse, number of living children a woman had given birth to, and desiredness of the pregnancy.

Step 2. Generating the propensity scores: After identifying the minimal adjustment set of control variables, propensity scores (the probability of being selected as a treated or control group) were generated from a logistic regression of the binary exposure on the nine control variables.

Step 3. Generating the inverse probability of treatment weights: Propensity scores were used to generate weights, representing the inverse of the estimated probability of receiving the treatment a participant actually received. The weight for the treated was 1/propensity score and the weight for the untreated was 1/(1 − propensity score) (Austin & Stuart, 2015). Weights were subsequently stabilized to avoid undue influence by extreme (high or low) weights (Xu et al., 2010).

Step 4. Checking the balance of covariates: After weighting, the balance of control variables between the treatment group (women with IPV experience) and comparison group (women without IPV experience) was checked, using standardized mean differences (Cohen’s d) and variance ratios (Garrido et al., 2014). Balance was considered acceptable if Cohen’s d was below 0.25 and variance ratio was between 0.5 and 2.0, as previously recommended (Rubin, 2001).

Step 5. Estimating the treatment effect: Finally, the treatment effect was estimated using a logistic regression of the outcome (neonatal mortality) on the exposure in the inverse probability treatment weighting sample. The regression incorporated two weights: the survey weight and the stabilized weight, which was generated in Step 3. To consider both weights in the final treatment effect estimation, the grand weight was generated as the product of the two weights (Dugoff et al., 2014). Results are expressed as adjusted odds ratios (AORs) with 95% confidence interval (CI) and p-values. The threshold for statistical significance was set at .05. All analyses were performed using Stata version 15 (StataCorp, 2017).

Results

General Characteristics of Respondents

The mean age at first birth was 27.6 (SD ±6.3) years. About 61.0% of women were married before 18 years. Most of the study participants were illiterate (62.5%), married to an illiterate partner (49.7%), and living in a rural area (81.6%). About 30% had experienced at least one pregnancy loss or child mortality in their lifetime. About 35.5% of women did not receive any ANC for their most recent birth, only 36.9% births were assisted by a skilled provider, only 13.7% had PNC from a skilled provider, and about 42% received at least two doses of tetanus toxoid immunization during pregnancy. Regarding neonatal characteristics of the index child: 17.0% were undesired pregnancies, 3.0% were delivered by cesarean, 26.2% had their weight rated as “below average,” 50.2% had low birth order (≤3), 20.7% had a short birth interval (<2 years), and 1.6% were multiple pregnancies. Characteristics of relevant variables by neonatal mortality status are also displayed in Table 1.

Table 1.

Characteristics of Relevant Variables—Ethiopian Demographic and Health Survey, 2016.

Factor Group	Variable	Class	Neonatal Mortality
			Yes	No	p-value
			No. (%)	No. (%)	p-value
Maternal characteristics	Age at first cohabitation	<15	18 (33.96)	628 (22.35)	.122
		15–18	16 (30.19)	1,084 (38.58)
		≥18	19 (35.85)	1,098 (39.07)
	Respondent’s educational status	No education	33 (62.26)	1,755 (62.46)	.851
		Primary	13 (24.53)	747 (26.58)
		Secondary+	7 (13.21)	308 (10.96)
	Employment status	employed	33 (62.26)	1,626 (57.86)	.520
	Employment status	not employed	20 (37.74)	1,184 (42.14)	.520
	Type of place of residence	Urban	7 (13.21)	520 (18.51)	.324
	Type of place of residence	Rural	46 (86.79)	2,290 (81.49)	.324
	Region of residence	Tigray	2 (3.77)	273 (9.72)	.795
		Afar	5 (9.43)	255 (9.07)
		Amhara	5 (9.43)	314 (11.17)
		Oromia	9 (16.98)	419 (14.91)
		Somali	9 (16.98)	322 (11.46)
		Benishangul	4 (7.55)	243 (8.65)
		SNNPR	5 (9.43)	386 (13.74)
		Gambela	6 (11.32)	194 (6.90)
		Harari	4 (7.55)	159 (5.66)
		Addis Ababa	2 (3.77)	115 (4.09)
		Dire Dawa	2 (3.77)	130 (4.63)
	Number of living children	One or less	17 (32.08)	563 (20.04)	.133
		2–3	16 (30.19)	983 (34.98)
		4–5	9 (16.98)	716 (25.48)
		≥6	11 (20.75)	548 (19.50)
	Number of previous adverse events	No	27 (50.94)	1,979 (70.43)	.001
		One	16 (30.19)	554 (19.72)
		Two	9 (16.98)	171 (6.09)
		Three or more	1 (1.89)	106 (3.77)
Maternal characteristics	Substance abuse	No	36 (67.92)	1,699 (60.46)	.271
	Substance abuse	Yes	17 (32.08)	1,111 (39.54)	.271
	Decision-making autonomy	No	19 (35.85)	912 (32.46)	.601
	Decision-making autonomy	Yes	34 (64.15)	1,898 (67.54)	.601
Paternal characteristics	Partner educational status	No education	27 (50.94)	1,396 (49.68)	.207
		Primary	21 (39.62)	902 (32.10)
		Secondary+	5 (9.43)	512 (18.22)
	Partner’s occupation	Not working	5 (9.43)	255 (9.07)	.417
		Professional/business	4 (7.55)	426 (15.16)
		Agriculture	34 (64.15)	1,570 (55.87)
		Manual/services	9 (16.98)	414 (14.73)
		Others	1 (1.89)	145 (5.16)
Household characteristics	Access to media	No access	34 (64.15)	1,881 (66.94)	.796
		<Once a week	8 (15.09)	339 (12.06)
		At least once a week	11 (20.75)	590 (21.00)
	Wealth index	Poorest	20 (37.74)	970 (34.52)	.926
		Poorer	11 (20.75)	495 (17.62)
		Middle	7 (13.21)	426 (15.16)
		Richer	6 (11.32)	361 (12.85)
		Richest	9 (16.98)	558 (19.86)
Maternal health care variables	Antenatal care	No visit	25 (47.17)	990 (35.23)	.190
		At least one visit	14 (26.42)	856 (30.46)
		4+ visits	14 (26.42)	964 (34.31)
	Delivery care	No	37 (69.81)	1,770 (62.99)	.308
	Delivery care	Yes	16 (30.19)	1,040 (37.01)	.308
	Tetanus toxoid immunization	No	35 (66.04)	1,623 (57.76)	.226
	Tetanus toxoid immunization	Yes	18 (33.96)	1,187 (42.24)	.226
	Postnatal care	No	46 (86.79)	2,426 (86.33)	.923
	Postnatal care	Yes	7 (13.21)	384 (13.67)	.923
Neonatal-related characteristics	Sex	Male	35 (66.04)	1,432 (50.96)	.030
	Sex	Female	18 (33.96)	1,378 (49.04)	.030
	Desiredness of the pregnancy	Pregnancy desired	45 (84.91)	2,333 (83.02)	.718
	Desiredness of the pregnancy	Pregnancy not desired	8 (15.09)	477 (16.98)	.718
	Multiple pregnancy	Single	42 (79.25)	2,774 (98.72)	.001
	Multiple pregnancy	Twin	11 (20.75)	36 (1.28)	.001
	Birth order	3 or less	22 (41.51)	1,414 (50.32)	.204
	Birth order	Greater than 3	31 (58.49)	1,396 (49.68)	.204
	Birth interval	First birth	7 (13.21)	526 (18.72)	.212
		<24 months	14 (26.42)	468 (16.65)
		24–59 months	24 (45.28)	1,462 (52.03)
		≥60 months	8 (15.09)	354 (12.60)
	Birth weight	Above average	20 (37.74)	849 (30.21)	.145
		Average	16 (30.19)	1,228 (43.70)
		Below average	17 (32.08)	733 (26.09)
	Cesarean delivery	No	50 (94.34)	2,728 (97.08)	.244
	Cesarean delivery	Yes	3 (5.66)	82 (2.92)	.244

Note. Bold cases indicate significance; SNNPR = Southern Nations, Nationals and Peoples Region; all the descriptive statistics is based on unweighted sample.

IPV Prevalence

Table 2 shows the estimated prevalence of different forms of IPV, with 95% CI. The least prevalent form of IPV was sexual IPV (12%) and the most prevalent form was partner controlling behavior (56%). About two in every three women had experienced at least one form of IPV in their lifetime. There was also a high co-occurrence of multiple forms of IPV (Table 2).

Table 2.

Prevalence of the Different Forms of IPV—Ethiopian Demographic and Health Survey, 2016.

Form of IPV	Weighted Prevalence %	[95% CI]
Physical IPV	24.5	[21.8, 27.2]
Sexual IPV	12.0	[9.8, 14.2]
Emotional IPV	22.9	[20.1, 25.6]
Physical, sexual, or emotional IPV	35.9	[32.9, 38.9]
partner controlling behavior (single form)	56.1	[52.6, 59.6]
partner controlling behavior (three or more)	15.4	[12.9, 17.8]
All forms of IPV	65.0	[62.1, 67.9]

Note. IPV = intimate partner violence; CI = confidence interval.

Propensity Score Balance

The kernel density plot presented in Figure 1 shows the propensity score balance in the exposed and unexposed groups before and after weighting. Plots “A” and “B” show the balance of propensity scores before and after weighting for the IPV model, respectively. However, plots “C” and “D” show the balance of propensity scores before and after weighting for the partner controlling behavior model, respectively. The plots clearly indicate that the balance of propensity scores are achieved after weighting, that is the red and blue lines are close together.

Figure 1.

Balance of propensity scores before and after weighting across treatment and comparison groups: (A) balance of propensity scores before weighting for the IPV model, (B) balance of propensity scores after weighting for the IPV model, (C) balance of propensity scores before weighting for the PCB model, and (D) balance of propensity scores after weighting for the PCB model.

Covariate Balance

Before weighting in the IPV model, eight of the nine control variables showed evidence of being unbalanced at least in one of their categories (Cohen’s d reaches up to 0.41). After weighting, however, all the covariates were balanced, with standardized mean differences ranging from 0.00 to 0.12, being <0.05 for most variable categories. Variance ratios after weighting ranged from 0.90 to 1.20, further indicating good balance.

Similarly, before weighting in the partner controlling behavior model, most categories of covariates were significantly different between women who experienced three or more partner controlling behaviors and those who experienced less than three or none at all. After weighting, all categories were balanced (Cohen’s d ranged from 0.00 to 0.16 and variance ratios ranged from 0.81 to 1.39).

Treatment Effect Estimation

After ensuring the balance of covariates, the association between neonatal mortality and each form of IPV was estimated using a logistic model in the weighted sample. Results are shown in Table 3. In the adjusted logistic model, women who had experienced lifetime IPV (a composite measure of physical, sexual, and emotional IPV) had over twice (AOR = 2.58, 95% CI = [1.03, 6.45]) the estimated odds of experiencing neonatal mortality, relative to women who did not experience IPV. In addition, the odds of neonatal mortality was higher among women who have experienced emotional abuse (AOR = 2.92, 95% CI = [1.15, 7.38]) than those who did not. There was also a significant relationship between maternal experience of three or more partner controlling behaviors and neonatal mortality (AOR = 2.75, 95% CI = [1.05, 7.20]) (Table 3).

Table 3.

Associations Between IPV and Neonatal Mortality—Ethiopian Demographic and Health Survey, 2016.

Treatment Variables	Neonatal Mortality		AOR	95% CI	p-value
Treatment Variables	No (%)	Yes (%)	AOR	95% CI	p-value
IPV
No	98.8	1.2	1.0
Yes	97.0	3.0	2.58	[1.03, 6.45]	.042
Partner controlling behavior
less than three	99.1	0.9	1.0
Three or more	92.7	7.3	2.75	[1.05, 7.20]	.039
Physical IPV
No	98.3	1.7	1.0
Yes	97.7	2.3	1.35	[0.49, 3.68]	.56
Sexual IPV
No	98.2	1.8	1.0
Yes	97.9	2.1	1.85	[0.61, 5.57]	.27
Emotional IPV
No	98.7	1.4	1.0
Yes	96.1	3.5	2.92	[1.15, 7.38]	.02
Partner controlling behavior
None	98.1	1.9	1.0
one or more	98.1	1.9	1.02	[0.44, 2.39]	.96

Note. Bold cases indicate significance; IPV = intimate partner violence; AOR = adjusted odds ratios; CI = confidence interval.

Discussion

This is the first national study to reveal evidence of a large, independent effect of IPV on neonatal mortality in Ethiopia. Advanced statistical methods were used to estimate the independent effect of different forms of IPV on neonatal mortality while controlling for a comprehensive range of potentially confounding variables.

The findings of this study are comparable with previous studies from other developed and developing countries that have reported significant associations between IPV and child mortality. For example, previous literature from Canada (Janssen et al., 2003), the United States (Coker et al., 2004), Brazil (Viellas et al., 2013), India (Ahmed et al., 2006; Koenig et al., 2010; Sarkar, 2013; Varghese et al., 2013), and Kenya (Emenike et al., 2008; Rico et al., 2011) have shown associations between IPV and child mortality. However, the available research differs in the form of IPV, the time when IPV was examined (lifetime or during pregnancy), the tool used to measure IPV, and the age at which child mortality was defined (perinatal, neonatal, infant, or under-five mortality).

This study has investigated a wide range of IPV experiences including lifetime experience of IPV (that reflects the long-term impact of IPV) and different types of abuse. This study is also focused on neonatal mortality instead of other types of child mortality because this is the period when most under-five mortality occurs in Ethiopia (CSA, 2017; Yared et al., 2013).

In this study, emotional IPV was found to be independently associated with neonatal mortality, while other single forms of IPV showed no evidence of an effect. This finding has suggested that only the emotional IPV, rather than any other type of IPV, might actually drive the overall significant effect of IPV (as a composite measure). This may be attributed to the fact that emotional IPV is often accompanied by other forms of IPV (WHO, 2013). The synergistic effect of these co-occurrences might lead to the strong association between emotional IPV and neonatal mortality. Contrary to the finding from this study, in a national survey from India, it was revealed that neither emotional nor sexual abuse was associated with infant mortality. On the contrary, physical abuse had a significant effect (relative risk = 1.24, 95% CI = [1.01, 1.53]) (Ackerson & Subramanian, 2009).

In this study, there was no significant relationship between a single partner controlling behavior and neonatal mortality. It is also noticed that as the number of partner controlling behaviors increased, there found to be significant association; the odds of neonatal mortality was 2.75 times higher among women who have experienced three or more partner controlling behavior than those who have experienced lesser or not at all. This evidence supports the theory that the concentration of controlling behaviors is more significant than any single behavior (Krantz & Nguyen, 2009). This can be explained in the way that the higher the number of partner controlling behaviors, the more severely she is being controlled, and the higher the level of stress she will experience. A woman whose husband controls one aspect of her life is quite different to one that controls three or more aspects. The former might be a chance occurrence, while the latter represents a pattern of behavior.

Other studies have failed to find a significant association between maternal IPV experience and child mortality. A study of four low-income countries revealed that there was no association between maternal exposure to physical and sexual IPV and under-two child mortality (Egypt, Honduras, Malawi, and Rwanda) (Rico et al., 2011). In addition, in studies from India (Silverman et al., 2011) and Malawi (Rao et al., 2017), researchers failed to find a significant effect of IPV on neonatal death.

Overall, there is no consistent evidence across studies about the effect of IPV on child mortality and hence evidence has to be interpreted and reflected in the context of where the study is conducted. This research has used nationally representative comprehensive evidence, which addressed marked social, economic, and cultural differences and can be used in the context of Ethiopia for future decision-making.

Strength and Limitations of the Study

As this study addressed a marked social, economic, and cultural diversified community of Ethiopia, where over 80 ethnic groups with diversified contexts live, the findings might be generalizable to groups outside the sample. The use of nationally representative data, application of advanced statistical analysis, and covariate selection guided by DAG are also the strengths of this study. The study has provided important evidence on the effects of IPV on neonatal mortality, which has never been investigated in Ethiopia, where neonatal mortality is one of the highest in the world (CSA, 2017). This study has also revealed a new body of knowledge on partner controlling behavior as a form of IPV and its strong relationship with neonatal mortality. Furthermore, this study has demonstrated the inverse probability of treatment weighting approach as a rigorous analytical approach to estimate the effect of an exposure on the outcome by allowing one to balance and account for differences across treatment and control groups. This is especially a useful method when there are too few observations and too many covariates. In addition, this study exhibits the use of inverse probability of treatment weighting with complex survey designs.

Despite the above strengths, there are potential limitations. One of the obvious limitations is the cross-sectional nature of the study that might not reflect the temporal relationship between variables; there is a possibility that neonatal mortality might lead to maternal IPV victimization. This is especially the case in rural communities where neonatal mortality might be thought of as the fault of the woman. However, future research will be needed to determine whether postneonatal mortality IPV represents the onset of IPV, or is just a continuation of preexisting IPV.

Another limitation, which is often the case in IPV research, is underreporting due to social desirability bias (Abeya et al., 2012; Garcia-Moreno et al., 2006). Nevertheless, Demographic and Health Surveys follow standard procedures and precautions such as maintaining confidentiality, ensuring privacy, the multiple items asked to measure IPV (with participants given an option of re-thinking their response), and the careful training of data collectors (Rutstein & Rojas, 2006; Sunita & Kiersten, 2004) that might lessen underreporting.

Another potential limitation is recall bias. Women might not remember or wish to recall their experiences of IPV, nor the date at age at birth and age at death of their child. To minimize this, only the latest child was considered in the analysis, which enables an opportunity that the younger the child is, the more likely the mother to remember the stated dates. Finally, this study did not address the mechanisms through which IPV causes neonatal mortality and this points future research directions.

Conclusion and Implications

This study has indicated a strong effect of lifetime IPV experience (a composite measure of physical, sexual, and emotional IPV) on neonatal mortality. The findings suggest the need for the consideration of IPV interventions in maternal and child health programs in the future. The existed neonatal survival strategies should focus beyond the direct causes of neonatal mortality and they need to target IPV as an underlying factor to neonatal morbidities and mortality. There should also be intersectoral collaborations between justice, social, and the health system to mitigate domestic violence and promote women’s health for the betterment of neonatal outcomes. Incorporating gender issues in the health care system; empowering women through education, increasing the age at first marriage, and involving women in decision-making has important role in decreasing IPV and its potential consequences.

One of the problems for preventing the effects of IPV is that it is considered as normal part of relationship; therefore, creating awareness about the consequences of IPV through school-based programs, community conversations and media are needed. Involving men in IPV intervention efforts might also help reducing the problem. We also suggest further research to investigate the pathways that may entangle the relationship between IPV and neonatal mortality.

Supplemental Material

Appendix_3rd_1 – Supplemental material for The Role of Maternal Intimate Partner Violence Victimization on Neonatal Mortality in Ethiopia

Supplemental material, Appendix_3rd_1 for The Role of Maternal Intimate Partner Violence Victimization on Neonatal Mortality in Ethiopia by Tenaw Yimer Tiruye, Catherine Chojenta, Melissa L. Harris, Elizabeth Holliday and Deborah Loxton in Journal of Interpersonal Violence

Footnotes

Acknowledgments

We are grateful to the central statistical agency of Ethiopia and measure Demographic and Health Survey (DHS) program, which allowed us to access and use the data freely. We are also thankful for women who have participated in the survey and shared their intimate partner violence experiences. We thank the University of Newcastle, Hunter Medical Research Institute, and the Research Center for Generational Health and Aging for creating a quality research environment for us to accomplish this work.

Author’s Note

All authors agreed on the submission of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was partially funded by the University of Newcastle, which has provided a scholarship for the student researcher and supported him in obtaining statistical support and trainings. Dr. Melissa L. Harris was funded by an Australian Research Council Discovery Early Career Research Award.

ORCID iD

Tenaw Yimer Tiruye

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Tenaw Yimer Tiruye is a PhD student at the University of Newcastle, Australia. His research focuses on the determinants, consequences, and responses to intimate partner violence. He is interested in evaluating multidimensional effects of intimate partner violence on maternal and child health and generating evidence for evidence based decision-making.

Catherine Chojenta is a postdoctoral research fellow at the University of Newcastle. She is a public health researcher with a particular focus on women’s health and well-being across the life course.

Melissa L. Harris is a postdoctoral research fellow at the University of Newcastle. Dr Harris has an interest in chronic diseases management, including the impacts of psychological factors on physical health and health care outcomes and contraceptive practices of women.

Elizabeth Holliday is an associate professor and senior biostatistician in the School of Medicine and Public Health at the University of Newcastle. Her particular focus is promoting statistical excellence in medical research.

Deborah Loxton is a professor of public health at the University of Newcastle. She is a co-director of research center for generational health and aging and co-director of Australian longitudinal study on women’s health. She has interest on examining the health and well-being of women who have lived with violent partners, and the impact of reproductive health options and choices on women.

References

Abebe

Admassu

Tilahun

(2016). Determinants of intimate partner violence during pregnancy among married women in Abay Chomen district, Western Ethiopia: A community based cross sectional study. BMC Women’s Health, 16, Article 16. https://doi.org/10.1186/s12905-016-0294-6

Abeya

S. G.

Afework

M. F.

Yalew

A. W.

(2011). Intimate partner violence against women in western Ethiopia: Prevalence, patterns, and associated factors. BMC Public Health, 11, Article 913. https://doi.org/10.1186/1471-2458-11-913

Abeya

S. G.

Afework

M. F.

Yalew

A. W.

(2012). Intimate partner violence against women in west Ethiopia: A qualitative study on attitudes, woman’s response, and suggested measures as perceived by community members. Reproductive Health, 9, Article 14. https://doi.org/10.1186/1742-4755-9-14

Ackerson

L. K.

Subramanian

S. V.

(2009). Intimate partner violence and death among infants and children in India. Pediatrics, 124(5), e878–e889. https://doi.org/10.1542/peds.2009-0524

Ahmed

Koenig

M. A.

Stephenson

(2006). Effects of domestic violence on perinatal and early-childhood mortality: Evidence from north India. American Journal of Public Health, 96(8), 1423–1428. https://doi.org/10.2105/AJPH.2005.066316

Asling-Monemi

Pena

Ellsberg

M. C.

Persson

L. A.

(2003). Violence against women increases the risk of infant and child mortality: A case-referent study in Nicaragua. Bulletin of the World Health Organization, 81(1), 10–16.

Austin

P. C.

(2011). An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research, 46(3), 399–424.

Austin

P. C.

Stuart

E. A.

(2015). Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Statistics in Medicine, 34(28), 3661–3679.

Central Statistical Agency. (2017). Ethiopia Demographic and Health Survey 2016 full report.

10.

Coker

A. L.

Sanderson

Dong

(2004). Partner violence during pregnancy and risk of adverse pregnancy outcomes. Paediatric and Perinatal Epidemiology, 18(4), 260–269. https://doi.org/10.1111/j.1365-3016.2004.00569.x

11.

Deb

Austin

P. C.

J. V.

D. T.

Mazer

C. D.

Kiss

Fremes

S. E.

(2016). A review of propensity-score methods and their use in cardiovascular research. Canadian Journal of Cardiology, 32(2), 259–265. https://doi.org/10.1016/j.cjca.2015.05.015

12.

Debelew

G. T.

Afework

M. F.

Yalew

A. W.

(2014). Determinants and causes of neonatal mortality in Jimma Zone, Southwest Ethiopia: A multilevel analysis of prospective follow up study. PLOS ONE, 9(9), Article e107184.

13.

Deribe

Beyene

B. K.

Tolla

Memiah

Biadgilign

Amberbir

(2012). Magnitude and correlates of intimate partner violence against women and its outcome in Southwest Ethiopia. PLOS ONE, 7(4), Article e36189. https://doi.org/10.1371/journal.pone.0036189

14.

Dillon

Hussain

Loxton

Rahman

(2013). Mental and physical health and intimate partner violence against women: A review of the literature. International Journal of Family Medicine, 2013, Article 313909.

15.

Dugoff

E. H.

Schuler

Stuart

E. A.

(2014). Generalizing observational study results: Applying propensity score methods to complex surveys. Health Services Research, 49(1), 284–303.

16.

Dutton

M. A.

Green

B. L.

Kaltman

S. I.

Roesch

D. M.

Zeffiro

T. A.

Krause

E. D.

(2006). Intimate partner violence, PTSD, and adverse health outcomes. Journal of Interpersonal Violence, 21(7), 955–968.

17.

Emenike

Lawoko

Dalal

(2008). Intimate partner violence and reproductive health of women in Kenya. International Council of Nurses, 55, 97–102.

18.

Federal Ministry of Health. (2015). Health Sector Transformation Plan (HSTP): 2015/16 - 2019/20.

19.

Furuta

Bick

Matsufuji

Coxon

(2016). Spousal violence and receipt of skilled maternity care during and after pregnancy in Nepal. Midwifery, 43, 7–13. https://doi.org/10.1016/j.midw.2016.10.005

20.

Gage

A. J.

Hutchinson

P. L.

(2006). Power, control, and intimate partner sexual violence in Haiti. Archives of Sexual Behavior, 35(1), 11–24.

21.

Garcia-Moreno

Jansen

H. A. F. M.

Ellsberg

Heise

Watts

C. H.

(2006). Prevalence of intimate partner violence: Findings from the WHO multi-country study on women’s health and domestic violence. The Lancet, 368(9543), 1260–1269.

22.

Garrido

M. M.

Kelley

A. S.

Paris

Roza

Meier

D. E.

Morrison

R. S.

Aldridge

M. D.

(2014). Methods for constructing and assessing propensity scores. Health Services Research, 49(5), 1701–1720.

23.

Goldstein

K. M.

Martin

S. L.

(2004). Intimate partner physical assault before and during pregnancy: How does it relate to women’s psychological vulnerability? Violence & Victims, 19(4), 387–398.

24.

Greenland

Pearl

Robins

J. M.

(1999). Causal diagrams for epidemiologic research. Epidemiology, 10(1), 37–48.

25.

Janssen

P. A.

Holt

V. L.

Sugg

N. K.

Emanuel

Critchlow

C. M.

Henderson

A. D.

(2003). Intimate partner violence and adverse pregnancy outcomes: A population-based study. American Journal of Obstetrics & Gynecology, 188(5), 1341–1347.

26.

Koenig

M. A.

Stephenson

Acharya

Barrick

Ahmed

Hindin

(2010). Domestic violence and early childhood mortality in rural India: Evidence from prospective data. International Journal of Epidemiology, 39(3), 825–833.

27.

Krantz

Nguyen

D. V.

(2009). The role of controlling behaviour in intimate partner violence and its health effects: A population based study from rural Vietnam. BMC Public Health, 9, Article 143. https://doi.org/10.1186/1471-2458-9-143

28.

Krug

E. G.

Dahlberg

L. L.

Mercy

J. A.

Zwi

A. B.

Lozano

(2002). World report on violence and health. World Health Organization.

29.

Lawn

J. E.

Cousens

Zupan

(2005). 4 million neonatal deaths: When? Where? Why? The Lancet, 365(9462), 891–900. https://doi.org/10.1016/S0140-6736(05)71048-5

30.

Mekonnen

Tensou

Telake

D. S.

Degefie

Bekele

(2013). Neonatal mortality in Ethiopia: Trends and determinants. BMC Public Health, 13, Article 483. http://www.biomedcentral.com/1471-2458/13/483

31.

Mohammed

B. H.

Johnston

J. M.

Harwell

J. I.

Tsang

K. W.

Haidar

J. A.

(2017). Intimate partner violence and utilization of maternal health care services in Addis Ababa, Ethiopia. BMC Health Services Research, 17(1), Article 178.

32.

Pallitto

C. C.

O’Campo

(2004). The relationship between intimate partner violence and unintended pregnancy: Analysis of a national sample from Colombia. International Family Planning Perspectives, 30(4), 165–173. https://doi.org/10.1363/ifpp.30.165.04

33.

Rao

Norris Turner

Harrington

Nampandeni

Banda

Norris

(2017). Correlations between intimate partner violence and spontaneous abortion, stillbirth, and neonatal death in rural Malawi. International Journal of Obstetrics & Gynaecology, 138(1), 74–78.

34.

Rico

Fenn

Abramsky

Watts

(2011). Associations between maternal experiences of intimate partner violence and child nutrition and mortality: Findings from Demographic and Health Surveys in Egypt, Honduras, Kenya, Malawi and Rwanda. Journal of Epidemiology & Community Health, 65(4), 360–367. https://doi.org/10.1136/jech.2008.081810

35.

Rubin

D. B.

(2001). Using propensity scores to help design observational studies: Application to the tobacco litigation. Health Services and Outcomes Research Methodology, 2(3), 169–188.

36.

Rutstein

Rojas

(2006). Guide to DHS statistics: Demographic and health surveys methodology. United States Agency for International Development.

37.

Sarkar

N. N.

(2013). The cause and consequence of domestic violence on pregnant women in India. Journal of Obstetrics & Gynaecology, 33(3), 250–253. https://doi.org/10.3109/01443615.2012.747493

38.

Semahegn

Mengistie

(2015). Domestic violence against women and associated factors in Ethiopia; Systematic review. Reproductive Health, 12, Article 78. https://doi.org/10.1186/s12978-015-0072-1

39.

Silverman

J. G.

Decker

M. R.

Cheng

D. M.

Wirth

Saggurti

McCauley

H. L.

… Raj

(2011). Gender-based disparities in infant and child mortality based on maternal exposure to spousal violence: The heavy burden borne by Indian girls. Archives of Pediatrics & Adolescent Medicine, 165(1), 22–27. https://doi.org/10.1001/archpediatrics.2010.261

40.

StataCorp. (2017). Stata statistical Software: Release 15.

41.

Sunita

Kiersten

(2004). Profiling domestic violence: A multi-country study. ORC Macro.

42.

Taft

A. J.

Powell

R. L.

Watson

L. F.

(2015). The impact of violence against women on reproductive health and child mortality in Timor-Leste. Australian & New Zealand Journal of Public Health, 39(2), 177–181. https://doi.org/10.1111/1753-6405.12339

43.

Varghese

Prasad

J. H.

Jacob

K. S.

(2013). Domestic violence as a risk factor for infant and child mortality: A community-based case-control study from southern India. National Medical Journal of India, 26(3), 142–146.

44.

Viellas

E. F.

Gama

S. G.

Carvalho

M. L.

Pinto

L. W.

(2013). Factors associated with physical aggression in pregnant women and adverse outcomes for the newborn. Jornal de Pediatria, 89(1), 83–90. https://doi.org/10.1016/j.jped.2013.02.013

45.

World Health Organization. (2013). Global and regional estimates of violence against women: Prevalence and health effects of intimate partner violence and nonpartner sexual violence.

46.

Ross

Raebel

M. A.

Shetterly

Blanchette

Smith

(2010). Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value in Health: The Journal of the International Society for Pharmacoeconomics and Outcomes Research, 13(2), 273–277.

47.

Yaya

Eide

K. T.

Norheim

O. F.

Lindtjørn

(2014). Maternal and neonatal mortality in South-West Ethiopia: Estimates and socio-economic inequality. PLOS ONE, 9(4), Article e96294.

48.

Yimer

Gobena

Egata

Mellie

(2014). Magnitude of domestic violence and associated factors among pregnant women in Hulet Ejju Enessie District, Northwest Ethiopia. Advances in Public Health, 2014, Article 484897.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.37 MB