Prevalence and Determinants of Intimate Partner Violence Against Women in Burundi: Evidence From 2016-17 Demographic and Health Survey

The This study consists of the secondary analysis of the 2016-17 Burundi Demographic and Health Survey. The study population consisted of 6014 married women, interviewed using the domestic violence questionnaire. The research applied Chi-square tests and binary logistic regression to identify the factors inuencing IPVAW in Burundi, at a 95% statistical signicance.


Background
Intimate Partner Violence against Women (IPVAW) refers to "the behaviour within an intimate relationship that causes physical, sexual or psychological harm, including acts of physical aggression, psychological abuse and controlling behaviours" (1) against a woman. Although domestic violence cuts across age and sex, violence against women is the most widespread. Globally, 30% of all women who have been in a relationship have experienced intimate physical and or sexual abuse. Husband/partners constitute the perpetrators for as many as 38% of all murders of women (2).
Spousal violence against women is a rights violation and a serious public health issue, thus a limitation to the woman's capacity to participate in the family, community, and society at large. Target 5.2 of the Sustainable Development Goal 5 calls for United Nations Member States to "eliminate all forms of violence against all women and girls in public and private spheres." From a public health perspective, victims of IPVAW suffer adverse consequences such as injuries, chronic physical health problems, excess deaths, particularly neonatal, infant and child deaths. A regional study, which included Burundi, recommended considering IPVAW as an urgent priority in all programs and policies aimed at maximizing the infant and child survival and the safety and wellbeing of the mothers, in East Africa (3).
A WHO study (2) found out that IPVAW was also associated with increased sexually transmitted infections, including HIV, depression, anxiety, post-traumatic stress disorders, self-harm, and suicide.
Women victims of IPVAW were almost twice as likely to experience depression, and in some regions, they were 1.5 times more likely to contract HIV.
Some of WHO's (4) ndings in its multi-country study on women's health and domestic violence were that secondary education, high social-economic status, and formal marriage were protective factors. In contrast, alcohol abuse, cohabitation, young age, attitudes supportive of wife-beating, having outside sexual partners, growing up with domestic violence, increased the risk of IPVAW.
Also, a study from Nigeria (5) identi ed parity as a key explanatory variable of IPVAW. It found out that the likelihood of IPVAW was lower among childless women than women with 2-3 children. On its part, a study from Angola (6) detected the religious background and the frequency of attending church as the most important predictors of IPVAW.
A study from 28 international surveys aimed at assessing the relationship between women's asset ownership and experience of IPVAW (7) found out that property had no signi cant association in 20 countries but was negatively associated with IPVAW in 3 and positively associated in 5 countries. The husband's characteristics were more important than the woman's emancipation. For example, many studies identi ed the husband's alcohol drinking as an essential factor in the occurrence of IPVAW. In Peru, women married to heavy drinkers' husbands were almost nine times as likely to be victims of abuse (8).
Studies on IPVAW in Burundi are scarce. However, Burundi is part of some regional and global studies, and information on the prevalence and risk factors of IPVAW is ndable in different reports of multicountry studies [3]. However, the literature research could not identify any research giving a broad overview of the prevalence and determinants of IPVAW at the national level for Burundi, hence the need for this study.

Methods
This study aims to determine the lifetime prevalence and underlying factors among women who were married/in-union at the time of the survey, to guide the formulation of strategies and programs targeting the reduction of IPVAW. The research hypothesis is that IPVAW results of the interplay of different factors: demographic, socio-economic and cultural.
This study consists of the secondary analysis of the 2016-17 Burundi Demographic and Health Survey (BDHS), which is the third DHS conducted in the country. For the rst time, it included a module on domestic violence. It was a population-based survey, nationally representative, carried out from October 2016 to March 2017. It sampled about 16,620 households and interviewed 17,269 women of reproductive age, using a standard structured questionnaire, which is available on MeasureDHS website [9] The sample was doubly strati ed by province and urban/rural residence, and clustered by Enumeration Areas (EAs) of the 2008 national Census. The nal report of the survey gives the details of the sampling methodology. Out of the 17, 269 women interviewed using the women questionnaire; the survey selected 6401 ever-married/ever-partnered women, who reported having 'ever' experienced any form of violence perpetrated by the husband/partner. The analysis used three approaches: Firstly, the univariate analysis used frequencies and percentages to describe the variables. Secondly, the study of the relationship between the dependent and the independent variables used cross-tabulations and chi-square tests. Thirdly, the multivariate analysis used logistic regressions to examine the association of the independent variables and the three dependent variables: physical, sexual, and emotional violence. All calculations used SPSS, version 26.0, with a 5% level of statistical signi cance.
The analysis used three dependent variables to capture the three forms of IPVAW: Physical violence was assessed from the responses to questions on whether the husband ever (i) pushed, shocked, or threw something at her; (ii) slapped; (iii) punched by a st or hit by something harmful; (iv) kicked or dragged; (v) strangled or burnt; (vi) threatened with a knife/gun or another weapon. A respondent had experienced intimate partner physical violence if she answered yes to at least one act of any of the items of physical violence mentioned above.
Emotional violence was assessed from responses to three questions whether the husband (i) ever humiliated her, (ii) ever threatened with harm and (iii) ever insulted or made to feel bad. A respondent had experienced emotional violence if she answered yes to at least one act of any of the psychological violence aspects.
The sexual violence was assessed through questions whether the husband (i) ever physically forced into unwanted sex, (ii) ever forced into other unwanted sexual acts and (iii) ever forced into other sexual acts, the respondents did not want. A respondent had experienced sexual violence if she answered yes to at least one item of any of the forms of sexual violence.
Items for physical violence had a Cronbach's alpha (α) of 0.69; those for sexual and emotional violence α = 0.53 and α = 0.62, respectively; any physical, sexual, or emotional violence with α = 0.80, indicating an overall excellent test performance of the interview questions.
A respondent had experienced IPVAW if she answered yes to at least one act of any form of violence (physical, sexual, or emotional).

Independent Variables
Independent variables included: women's characteristics (mothers' age, education, children ever-born, type of marriage, work status), husband characteristics (age, drink alcohol, controlling behaviour, desire for children) and societal factors (province, type of residence, religion, wealth, sex of household head). These independent variables were selected for inclusion in the study based on their assumed association with IPVAW or importance in previous studies of IPVAW, as depicted in the literature review.
Some variables were regrouped from their original categories in the dataset, to make analysis and interpretation simpler and more meaningful. Recorded variables included: Woman's age: self-reported age at the time of the interview, regrouped into < 30, 30-49 and 40-49 years.
Woman's education: The highest level of education attained by the respondent, regrouped into No education, primary, secondary + Religion: Self-reported religious a liation at the interview. Catholic, Protestant, Muslim, and others.
Type of marriage: Self-reported kind of union, monogamous or polygamous. Husband's desire for children: Reported number of children the woman and the husband want. The variable intends to re ect the linkage between fertility preferences and IPVAW. Three categories: Husband wants fewer children, the same number of children, more children.
Father's abuse: Self-reported childhood history of witnessing father beating the mother: Y/N Alcohol drinker: Respondent reporting whether the husband/partner drinks alcohol: Y/N Husband's controlling behaviour: A composite variable re ecting respondent self-reporting of ve controlling behaviour displayed by the husband/partner. The respondent had to answer (i) whether the husband was jealous when the woman talks with other men; (ii) the husband accuses her of unfaithfulness; (iii) does not permit her to meet her friends; (iv) tries to limit her contact with family; insists on knowing where she is always Women responding Y to one or more questions were considered as having a partner/ husband with controlling behaviour. Those who answered N to all the items had no control issues.
Husband's education: The highest level of education attained by the respondent's husband/partner, grouped into no education, primary education, secondary+ education.

Results
The women's median age in this study was 31.9 years; most of the respondents (93.9%) were Christians with 3.9% Muslims and 2.2% belonging to other religions (Table 1), distributed in the 18 provinces of the country.
Educational level was higher for men/partners than women: Forty-six per cent (46.4 %) of the women had no education, 41.1% had primary education and only 12.4% had secondary and higher education. More than half (51.7%) of the respondents had not heard about conjugal violence in the media.
The majority of the married women resided in the rural areas (83.8%), had an active status (87.4%) and were in monogamous unions (93.4%). About 40.8% of the women were in the lowest quintile of the wealth index, six women (59.7%) out of ten reported having witnessed their father beating the mother in their childhood, and 82.7% of the respondents lived in male-headed households. Only 4.3% of the women had no children, 29.3% had 1-2 children ever born, 31.4% had 3-4 children ever born, and 34.9% had ve or more children ever born.
Slightly more than two-thirds (67.7%) of women had husbands who drunk alcohol, and only a third (31.9%) had husbands with controlling behaviours. The study population showed that 61.3% of the husbands desired the same number of children as their wives, 20.0% desired fewer children, whereas 18.7% of the husbands desired more children. On average, the husbands below 30 years old represented 24.2%, those between 30 and 40 years old 41.5%, and those above 40 years old 34.2%. The husbands' mean age was 36.5 years, which shows an average of 4.6 years difference between men and women, favouring men. Men were also more educated than women, with 49.3% having primary school against 41.1%.  Figure 2, on its part, shows the distribution of the different acts of abuse for the three forms of IPVAW. Slapping, forced unwanted sex and insults are the most common forms found in either category.
Bivariate analysis of and background variables Table 2 presents the results of cross-tabulations of IPVAW with each one of the independent variables taken individually.
It shows that physical violence is signi cantly associated with women's age, the highest rate being among women aged 40-49 (43.1%). The distribution by province shows large disparities, women in Kirundo having the highest prevalence (60.2%) and those in Bujumbura-Mairie the lowest (22.1%).
The analysis found higher proportions of physical violence among rural women (38.6%), women whose husband is the household-head (38.6%); women in the poorest (44.9%) and the poorer (43%) categories of the wealth quintiles and women with an active status (38.6%). Higher proportions also characterized women with no education (43.0%), with primary education (35.4%), women a liated to 'other religions' (52.1%) and those who witnessed parental abuse in their young age (46.8%).
The bivariate analysis established that psychological violence was positively associated with the respondent's age: The more the age increases, the more the risk of experiencing emotional abuse. The highest prevalence rate (26.6%) was in the age group 40-49. Regional disparities are visible, Rumonge province holding the highest prevalence rate (41.7%) and Mwaro the lowest (12.5%).
The bivariate analysis has shown that sexual violence increased with the age of the women. The proportions of victims raised from 22.9% among women less than 30 years old to 26.6% among the aged 49+. Regional disparities were observable with the highest prevalence rates of victims in Muramvya province (44.7%) followed by Rumonge (42.1%) and the lowest in Bujumbura Mairie (12.2%).
Other factors related to sexual violence were the rural residence (25.9%), the male household head (25.1%), the low wealth status. In all wealth categories, except the richest (16.8%), a woman out of four or less (poorest: 27.9%; poorer: 26.0%; middle: 26.1% and richer: 25.6%) was subject to sexual abuse.
Non-active women (25.8%), those with ve children ever born and those with no education (25.2%) or with only primary education (23.3%) displayed high proportions of sexual abuse. Women a liated to 'other religions' (33.1%), those in polygamous union (36.2%); women whose husband desired fewer children (33.1%) or more children (27.2%) and those who had experienced father-mother abuse in their childhood (29.2%) were more likely to be the victim of sexual abuse.
Women whose husband's demonstrated controlling behaviour (41.8%), women whose husband drunk alcohol (28.4%), the married to older men in the age group 40-49 (25.9%), or having a husband with no education (27.1%) or with a primary school only (24.7%), had high prevalence rates of sexual violence.

Multivariate analysis
Multivariate logistic regression was applied to analyze further the relationship between IPVAW and the independent variables of interest, after controlling for confounders. were at a higher risk of experiencing physical abuse.

Emotional violence
Except for Muramvya, women in all other provinces enjoyed a protective effect against emotional violence, compared to the province of Rumonge, which was chosen as the reference, as shown in Table 3.
Women who had never had a child or had 1-2 ever-born children were less exposed to emotional violence   (10). The differences could mostly be attributable to differences in social, economic and cultural conditions in the various countries.
IPVAW levels seem to increase with the mother's age, women in the age group 40+ having higher proportions of 'ever' experienced physical violence. The high prevalence observed among older women could be due to increased exposure to violence, as most of these could be old and been married for a longer time. This relation has also appeared in other studies, like in Nepal (15).
The prevalence of IPVAW depends in no small extent on the province of residence. Studies conducted in other countries such as Malawi (16), Zambia (17) have also demonstrated variations of IPVAW by region.
A Rwandan's research revealed that IPVAW was highest in the Northern Province, where the rate was 45.6%, and the lowest in Kigali City with a 28.8% prevalence rate (12). Regional differences in IPVAW could be attributable to the differences in the regions' social and economic conditions. These conditions contribute to the characteristics of the population residing in each province. Studies indicate regional disparities in the distribution of infrastructures such as roads, health, and educational facilities in most developing countries. It can be argued that in regions that are better equipped with such facilities and could be said to be more developed, the risk for IPVAW is lower than in the less developed areas. In certain regions, IPVAW may re ect different cultural practices that may reinforce speci c values and norms.
The study found that IPVAW was signi cantly associated with the sex of head of household. It was high among women residing in households headed by men. The importance of the sex of head of household is also underscored in a study in Bangladesh even though in that study woman living in a female-headed home had higher odds of experiencing IPVAW (18). Two plausible explanations could be responsible for this relationship. First, it can be argued that in female-headed households, there is no intimate partner to abuse the women. Second, it is also possible that in female-headed homes, women are empowered to voice their concern when the intimate partner abuses them.
The study found out that IPVAW was positively associated with a childhood experience of abuse. In particular, women who had witnessed their father abusing their mother are more likely to report domestic violence against women. This nding is similar to what research revealed in other countries: Ghana (14), Ethiopia (19), Nigeria (20). Another explanation is that growing up in an atmosphere of domestic violence may induce an attitude of acceptance toward spousal abuse.
The study also found that IPVAW was signi cantly associated with wealth index and work status. It was high among working women and women with low wealth status. This nding is consistent with results from the Democratic Republic of Congo (10), Malawi (16), Zambia (17) and Zimbabwe (21). One possible explanation is that when women work and have an income of their own, it challenges the patriarchy system, undermining the husband's superiority, which induces domestic violence against women.
Another factor that is signi cantly associated with IPVAW in Burundi is the number of children ever born.
The study found out that IPVAW increases with the number of children ever born. In general, women who have more children ever born were more likely to experience IPVAW. Studies conducted in other countries have also revealed that IPVAW was signi cantly associated with high parity: Ethiopia (19), Nigeria (20), Nepal (15), Zimbabwe (21). This nding is no surprise given that as the number of children increases, there is increased pressure on the family, especially fathers who in patriarchal societies are the head of household, to fend for their families. This pressure may result in con icts and violence.
The study has also found that the incidence of IPVAW was high among women in polygamous than monogamous marriages. This nding is consistent with results from other countries: Afghanistan (22), Democratic Republic of Congo (23), Mozambique(23), Ghana(24), Uganda (11), Nigeria(20), Rwanda (12). One possible explanation for this is that there is increased competition for resources and attention among wives living in the same household, resulting in pressure on the husbands to manage the situation. Both the competition among wives and pressure on the husband may result in tension and con ict.
Furthermore, the study indicates that IPVAW in Burundi was signi cantly associated with the husband's behavioural characteristics. Women whose husband drunk alcohol displayed high odds of sexual, emotional and physical violence. This nding is consistent with results from other countries: Democratic Republic of Congo (10), Ethiopia (19), Ghana (24), India (26), Malawi (16), Rwanda (12), Nigeria (20), Haiti (27), Zambia (17,28). The available literature suggests that alcohol consumption can weaken mental and physical function and cause in delity and aggressive behaviour (17). Also, alcohol consumption competes with other household expenses such as food and clothing, which may give rise to marital tension and con ict (14,17).
The study also identi ed the husband's desire for children as a critical factor in uencing both forms of violence. Discordance among the couple about the number of children they want may result in tension and con ict in the family. Violence enables some men to resolve disagreements over the use of contraception, for example, by imposing their fertility preferences on their partners (29).
The study also found that IPVAW in Burundi was high among women whose husband had controlling behaviour. Besides, about a third of women in Burundi reported having husbands with controlling behaviour. Women with such partners were at increased risk of experiencing intimate partner violence, which is similar to what other researchers have reported in Haiti (27), Nigeria(30), the Philippines (31). Other commentators on this subject have argued that husbands often use violence to control their wives (32).

Strengths and Limitations
The present study has some limitations. First, the study design is cross-sectional; the analyzed variables can only con rm a statistical association's existence but not a causal relationship between the variables and IPVAW. Second, there may be a possibility of recall bias as the study entailed self-reported data without any means of veri cation on the woman's entire lifetime. Also, secondary data restricted the potential to su ciently assess the in uence of some characteristics, such as cultural practises as drivers of IPVAW. Nonetheless, the use of a large nationally representative data set constitutes a strength for the study.

Conclusion
In conclusion, this study has found out that IPVAW was widespread in Burundi. The research has demonstrated that domestic violence against women in the country was associated with various social and economic characteristics of both men and women. There is a need to conduct further studies to explore additional cultural factors.
The study recommends addressing the social and economic imbalances between men and women that perpetuate patriarchy. Focusing on the economic empowerment of women alone may not only increase but also unlikely to stop IPVAW. Economic interventions need to include husbands, families and communities to change societal gender norms and values. There is also a need to strengthen Information, Education and Communication activities on IPVAW, paying particular attention to men.  (7):1332-8. Figure 1 Prevalence of IPVAW in Burundi