Prevalence and factors associated with sexual violence among rural women aged 15-49 years in Uganda: evidence from the Uganda demographic and health survey 2016

Background and Aim: Sexual violence is a serious global health problem with short and long-term physical, mental, and reproductive adverse-effects. We aimed to determine the prevalence and factors associated with sexual violence among rural women of reproductive age in Uganda. Methods: We used the Uganda Demographic and Health Survey (UDHS) 2016 data for 5,259 rural women aged 15 to 49 years. Multi-stage stratied sampling was used to select respondents, and data were collected using validated questionnaires. We used multivariable logistic regression to determine factors associated with sexual violence. All analyses we carried out in SPSS (version 25). Results: The overall prevalence of sexual violence was 24.3% (95% CI: 23.4-25.7), while that among pregnant women was 22.2% (95% CI: 19.1-25.6). Women of primary education had higher odds of sexual violence compared to those of tertiary education (AOR=3.01, 95% CI: 1.17-7.72), and those from the Western region compared to those in the Northern region (AORs=2.01, 95% CI: 1.41-2.87). Women of poorer wealth quintile also had higher odds of sexual violence compared to those of the richest quintile (AOR=2.06, 95% CI: 1.02-4.17). Higher odds of sexual violence were also found among women who justied beating (AOR=1.42, 95% CI: 1.12-1.79), those who were not involved in healthcare decision-making (AOR=1.38, 95% CI: 1.01-1.91), and those whose husbands often got drunk (AOR=2.07, 95% CI: 1.30-3.31), compared to their counterparts. Conclusions: The prevalence of sexual violence among rural women in Uganda was high and associated with factors such as education level, region, wealth index, justied beating, healthcare decision making, and husband drunk frequency. Thus, there is a need for improving livelihoods and income of rural women, promotion of girl-child education as well as strengthening protective laws and policies to curb this public health issue.


Introduction
Sexual violence (SV), a form of gender-based violence (GBV), is a serious global health issue that violates human rights [1]. Sexual violence has been de ned as any sort of harmful or unwanted sexual behaviour that is imposed on someone [2]. The act is not only devastating for survivors and their families but also involves signi cant social and economic costs [1]. Intimate partners are often the perpetrators of this violence [2,3]. According to the World Health Organisation (WHO), GVB affects 1 in 3 women, and over 35% of women worldwide have experienced sexual and physical violence by either intimate partners or non-partners in their lifetime [1,4].
A recent meta-analysis by Muluneh et al. reported the prevalence of sexual violence in sub-Saharan Africa to be 18.6%, with the East African region having a higher prevalence compared to other regions [5]. A previous study indicated that 25% of the ever-married women in Uganda experienced some form of sexual violence [6]. Despite the presence of laws and policies to protect victims and survivors, GBV, including sexual violence, increased by 4% between 2015 and 2016 in Uganda [7].
Sexual violence is known to have adverse physical, mental, and reproductive health outcomes to mothers and their children [8,9]. It also limits a woman's decision-making power regarding her reproductive health, putting her at risk for sexually transmitted infections (STIs) and unwanted pregnancies [10]. Sexual violence during pregnancy is associated with poor maternal and fetal outcomes, including miscarriages, low birth weight, preterm birth, and intrauterine growth restriction, among others [9,11,12]. In addition, children who witness violence have been reported to become future perpetrators of GBV [4]. Besides, several studies have reported several factors such as age, marital status, educational level, occupation, decision-making power, partners' alcohol intake and smoking, justi ed violence, among others to be associated with sexual violence [10,13,14]. Notably, an association has also been suggested between the area of residence (urban or rural) and sexual violence, and the same applies to social-economic status [5,15].
To date, no comprehensive study has focused on exploring sexual violence in rural communities of Uganda. Residents of rural areas are usually faced with limited or no access to quality social and health care services. Given the devastating physical, mental and reproductive effects of sexual violence, it is essential to understand the various factors associated with this form of violence in the Ugandan context using nationally representative data. Identifying the risk factors of sexual violence among rural women is also vital in nding targeted, practical solutions as well as informing policy-makers. This study thus aimed to determine the prevalence and factors associated with sexual violence among rural women in Uganda, using data from the UDHS 2016.

Study Data
The study used secondary data of the 2016 UDHS collected from June 2016 to December 2016 in a nationally representative survey [16]. The survey (UDHS) was implemented by the Uganda Bureau of Statistics (UBOS) with the technical assistance of the Inner City Fund (ICF) International through the USAID-supported MEASURE DHS project [16]. The survey inquired about household members' and individual characteristics using a household questionnaire, while sexual violence data was collected by the woman's questionnaire, which included the domestic violence module [16].

Study setting
As of July 2018, Uganda had a population of approximately 41 million people with 76.2% of the population residing in rural areas [17], and the country has 21.4% of the population living below the poverty line [18]. The average female expected years of schooling are 10.4, with the actual mean of 4.8 female years of schooling. The average age at rst marriage and females' birth in the country are 18.2 and 18.9 years, respectively [18]. The country's family setting rotates on the husband, who is regarded as the family head [18]. Uganda has a six-level health system that ranges from the national referral hospitals, as the highest level, to community-level health centres [19]. Agriculture is the primary source of income for Ugandans, and it contributes about 50% of export earnings [18].

Study sampling and participants
UDHS employed a two-stage cluster sampling technique, where the census enumeration areas were the primary sampling units, and households were the secondary sampling units [16]. The enumeration areas were selected from the 2014 population and housing census sample frame [16]. A total of 18,506 women aged 15-49 years participated in the UDHS 2016, of which 9,232 women were sampled to respond to the domestic violence questions, and of these, 5,259 women resided in rural areas. One per cent (1%) of eligible women could not be successfully interviewed with the module because of the lack of privacy or other reasons. Weights were used to adjust for the selection of only one person per household and to ensure that the domestic violence subsample was nationally representative [16].

Dependent variable
The dependent variable in this study was the history of exposure to sexual violence among rural women in Uganda. During the survey, rural women were asked about their experience of any form of sexual violence at any time in their life, as a child or as an adult [16].

Independent variables
This study included determinants of sexual violence based on evidence from the available literature [13,14]. These included; individual level (age, pregnancy status, working status, justi ed beating, parity and educational level), household level (household size, husband's educational level, wealth index, healthcare decision-making, husbands age, and drinking frequency as well as husbands working status) and community level (region) factors. Wealth index is a measure of relative household economic status and was calculated by UDHS from information on household asset ownership using principal component analysis [16]. These quintiles were; the poorest, the poorer, the middle, the richer, and the richest quintiles [16]. Healthcare decision-making was classi ed as "woman alone," "with someone," or "not involved." Besides, "region" was categorised into four; Northern, Central, Eastern, and Western [20]. Level of Education was classi ed as no education, primary education, secondary and tertiary education. Age was categorised into 20-29, 30-39, and 40-49. Parity was classi ed as "less than 4" and "above 4", household size as "5 and below" and "above 5", while drunk frequency was categorised into three; never, sometimes, and often.

Statistical analysis
Statistical analysis was performed using SPSS complex samples software (version 25.0), which accounted for the multi-stage cluster study design. We used weighted counts in the analysis to account for the unequal probability sampling in different strata [21] and to ensure the representativeness of the survey results at the national and regional levels. We summarised selected categorical variables as frequencies and proportions.
Factors known to be associated with sexual violence, from the literature, were evaluated and those found signi cant at a p-value less than 0.2 at bivariable analysis [22] were included in the multivariable models. We presented two models; the primary model, which adjusted for only factors without missing data, and the secondary model, which included all the factors (with and without missing data) and adjusted for those with missing data. The crude odds ratio (COR), adjusted odds ratios (AOR), 95% Con dence Intervals (CI), and p-values were reported. The statistical signi cance level was set at p-value < 0.05 for multivariable logistic regression analysis. We assessed all covariates for collinearity, using a cut off value of above 10 variance in ation factor [23].

Results
A total of 5259 rural women were included in this study, and the prevalence of sexual violence was 24.3% (95% CI: 23.4-25.7) ( Table 1). Of the participants, 675 were pregnant, with a prevalence of sexual abuse of 22.2% (95% CI: 19.1-25.6). Majority of the women were below 35 years (64.2%) and not pregnant (87.2%).
60.1 % were from Western and Eastern Uganda, 83.3% were working, 49% had husbands with primary education as the highest level of education, and 60.7% took health care decisions alone or with someone. The majority (65.8%) had primary education as their highest level of education, and 71.5% belonged to the middle or a less wealth index, while 57% had a parity of less than 4, 78% justi ed beating, and 50.6% had household size of above 5. In addition, 40.9% had husbands who often or sometimes get drunk, 62.6% had husbands less than 45 years of age, and 80.4% had employed/ working husbands. The details are in Table 1.

Factors associated with sexual violence
Bivariable analysis results are shown in Table 2. In the primary logistic regression model, factors associated with sexual violence were: region, working status, justi ed beating, and household size. After adjusting for factors with missing data, the secondary model showed educational level, region, wealth index, justi ed beating, healthcare decision making, and husband drinking frequency as the signi cant factors associated with sexual violence (Table 3).

Discussion
This study examined the prevalence and factors associated with sexual violence among Ugandan rural women of reproductive age. The prevalence of sexual violence among our participants was 24.3%. This prevalence is slightly higher than the national prevalence of 22% [16]. Rural settings have been linked with higher rates of physical and sexual violence than urban areas, which can be attributed to the difference in the social-dynamics that increase chances of domestic violence in rural areas [10,24,25]. Compared to urban settings, the strong cultural believes, ties and norms, as well as higher poverty levels in rural areas, could explain the higher prevalence of sexual violence [6,10,25].
The study identi ed several factors that were signi cantly associated with sexual violence, including women's educational level, region, wealth index, justi ed beating, healthcare decision-making, and husband drinking frequency. Women with a tertiary educational level and richest wealth quintile were less likely to experience sexual violence than the less educated and poor ones. Higher (post-secondary) level of education implies more awareness about individual rights, and educated women are more nancially stable and more empowered compared to the poor less educated [15]. The results agree with previous studies that also reported higher educational level and richest wealth quintile as protective factors against not only sexual violence but also gender inequality [6,15,26]. This implies that more efforts and targeted interventions are needed to improve the livelihoods of rural women to reduce nancial dependence on their husbands. In addition, promotion of girl-child education is a vital tool for women empowerment, improving the social status and reducing gender-based violence.
Women from Western and Eastern Uganda had the highest chances of having experienced sexual violence, compared to those in the Northern region. The ndings are in line with Wandera et al., who reported the highest rate of sexual violence in the Eastern region, followed by Western, Central, and Northern regions [6]. However, the results deviate from a recent meta-analysis that reported Northern Uganda having the highest rate of sexual violence (50%) within Sub-Saharan Africa [5]. Although the Northern region has a record of past civil wars that were characterised by the violation of human rights, including sexual abuse [27], our study found no signi cant chances of sexual violence among women in this region. The reasons underlying the higher risks of sexual violence in the Western and Eastern regions of the country are not clear, calling for further investigation. However, the higher sexual violence risks in these regions may be linked to social-cultural norms such as women subordinate, justi ed violence, among others [28,29].
In addition, women who justi ed beating showed higher odds of sexual violence, unlike their counterparts. The results agree with previous studies that highlight justi ed violence as a risk factor for sexual violence [28,29]. Strict adherence to cultural norms that emphasise women's inferiority to men or those which objectify women affects the mindset and awareness of women about their rights [29]. Justi ed violence against women is deeply rooted in cultural norms and practices like payment of bride price, grooming women to be submissive and subservient, among others. All these disempower women from taking part in sexual decision-making while mounting men with in nite sexual contact with their wives [28]. Therefore, there is a need for culture-oriented sensitisation programs to condemn negative cultural norms and practices that violet women's rights.
The study also noted that women who were involved in healthcare decision-making had lower chances of experiencing sexual violence compared to those who did not. The ability to make independent healthcare decisions imply knowledge of one's rights, as well as empowerment. Women who do not take part, at all, in their healthcare decision making are more likely to be less empowered and so are prone to not only sexual violence but also physical violence [25]. In addition, joint decision making (wife and husband) could imply better communication and fewer incidences of misunderstandings, thus less chances of sexual violence in a home. The results agree with Semahegn's study [30], which also indicated joint decision-making as a protective factor against not only sexual violence but also other forms of domestic violence.
The study revealed that women with husbands who often get drunk were more likely to be sexually abused compared to those whose husbands don't get drunk. The results agree with previous studies that also highlighted the husband's behaviour like alcohol consumption as a risk factor for not only sexual violence but also physical violence [6,28]. Excessive alcohol consumption has been linked with aggressive behaviour and increases the chances of misunderstandings and con ict among partners leading to sexual violence, especially when the woman does not drink [28].
Notably, although it became non-signi cant on controlling for missing data, working status also had some association with sexual violence. Employed women showed higher odds of sexual violence compared to the unemployed ones. This nding is similar to other studies that have also reported working women having more chances of being sexually abused [25,31,32]. Although employment directly translates to women empowerment [26], it has been shown, in several studies, not to be a protective factor against sexual violence among women [25,32,33]. This could be due to the work-related sexual abuse faced by women, depending on the type of employment, especially in the informal sector. For example, some studies have also reported higher rates of sexual violence among female sex workers as well as casual labourers [34,35,36]. This implies a need to strengthen the country's available protective policies to ensure the safety of working women.
Strengths and Limitations of the study Standardised procedures are a requirement of DHS surveys in data collection, and validated questionnaires are used, which ensures the internal and external validity of the results. Secondly, we used the most recent nationally representative sample and weighed the data for analysis, and therefore, our results are generalised to all Ugandan women aged 15 to 49 years. However, the study had some limitations, such as the possibility of information bias, as well as recall, and interviewer biases since most data on the predictors were based on self-reporting, which could not be veri ed through records, and respondents gave answers about events that occurred in the past. In addition, data on explanatory variables such as wealth and working status re ected the women's conditions at the time of the survey and not at the time experiencing sexual violence. Hence, women may have moved from one category of classi cation into another. Such non-differential misclassi cation may have reduced the strengths of observed associations.

Conclusions
The study has found a considerable prevalence of sexual violence among rural Ugandan women of reproductive age. Several factors have also been found to be associated with sexual violence among these people, including educational level, region, wealth index, justi ed beating, healthcare decision making, husband drunk frequency, and working status. Thus, the study results indicate a need for strategies aimed at improving rural women's livelihoods and income, as well as promoting education of the girl-child as a tool for women empowerment and protection against not only sexual violence but also gender-based inequalities. Culturally-oriented community sensitisation programs and strengthening protective laws and policies against sexual violence are also key in addressing this public health problem in Uganda.

Declarations
Author contributions QS and JK initiated the topic, analyzed the data and wrote the rst draft of the paper. DM, ASS and MM made critical input and comments on the draft manuscript. QS and DM contributed to the interpretation and writing of the paper. All authors read and approved the nal manuscript.