Intimate Partner Violence against Women Living in Inadequate Conditions in Sub-Saharan Africa: A Meta-Analysis of Demographic and Health Survey Data

Intimate partner violence (IPV) against women is a major public health problem and is widespread in sub-Saharan Africa (SSA). However, little is known about its environmental determinants. This study aimed to investigate whether inadequate living conditions are associated with IPV victimization in women in SSA. We analyzed cross-sectional data for 102,714 women in 25 SSA countries obtained from the Demographic and Health Surveys Program. Logistic regression was used to estimate the country-specific effects of inadequate living conditions (housing with at least one of four characteristics of unimproved water, unimproved sanitation, insufficient space, and unfinished materials) on multiple forms of IPV. Random effects meta-analysis was used to combined the country-specific estimates. We found an association between inadequate living conditions and a higher likelihood of experiencing any (OR = 1.12, 95% CI 1.03 to 1.23, p = 0.012), sexual (OR = 1.18, 95% CI 1.05 to 1.34, p = 0.008), emotional (OR = 1.12, 95% CI 1.02 to 1.23, p = 0.023), and physical (OR = 1.15, 95% CI 1.03 to 1.28, p = 0.010) IPV. The associations were stronger for rural and less-educated women. These findings suggest that future research to establish a causal link between living conditions and IPV and to elucidate the underlying pathways is crucial to design IPV interventions in SSA.


Introduction
Intimate partner violence (IPV) is increasingly recognized as a significant social and public health problem. Globally, almost one-third (30%) of ever-partnered women have experienced physical and/or sexual IPV [1]. Most IPV research of risk factors and interventions focused on individual factors, such as a woman's or partner's wealth, education, and acceptance of violence and women's decision-making autonomy [1,2]. However, only a little attention has been paid to environmental determinants of IPV, such as drought [3] and armed conflict [4].
Inadequate living conditions, commonly characterized by unimproved water and sanitation, insufficient living area, and non-durable construction, are universal markers of a disadvantaged home environment [5]. Studies have revealed that inadequate living conditions are directly associated with poor health outcomes, such as injuries, infectious diseases, and malnutrition [6]. Inadequate living conditions have also been associated with depression and stress [7], which in turn may contribute to the occurrence of violence [8]. Indeed, associations of inadequate living conditions and/or its characteristics with multiple forms of violence have been documented, such as non-partner violence [9][10][11], homicide [12], and child abuse [13]. Thus, inadequate living conditions might also be linked to IPV toward women. However, few studies have investigated the direct associations between inadequate living conditions and multiple forms of IPV toward women.
Living conditions are particularly poor in sub-Saharan Africa (SSA), because of the high rate of population growth and lack of housing [14]. In 2015, 47% (53 million) of the urban population and 82% (595 million) of the rural population lived in unimproved housing in SSA [5]. Houses are deemed unimproved when they have at least one of the four characteristics of unimproved water, unimproved sanitation, more than three people per bedroom, and being made of unfinished materials [5]. SSA also has a high incidence of IPV, reflecting the situation on the African continent, where 36.6% of ever-partnered women are affected by IPV [1]. The association between inadequate living conditions and different forms of IPV merits examination in a multi-country setting, such as SSA.
In the present study, we used data collected by the Demographic and Health Surveys (DHS) Program [15] from 25 countries in SSA to test the hypothesis that, for married or cohabiting women, inadequate living conditions are associated with a higher likelihood of experiencing emotional, physical, and/or sexual violence. We conducted stratified analyses by residence type (rural/urban) and education level (less than secondary/secondary or higher education) of the respondent, which could help design targeted screening and interventions for vulnerable groups.

Data Source
DHS are cross-sectional surveys with a focus on households in low-and middleincome countries (LMICs), and provide nationally representative data on demography, reproductive health, and child health [15]. We selected DHS conducted in SSA countries that (1) included information on the housing environment; (2) assessed IPV using the standard questionnaire in the domestic violence module; and (3) were publicly available before July 2021. For countries that had completed multiple rounds of the DHS, we selected the most recent survey data as they represent the latest situation of that country.
The surveys use a randomized multi-stage cluster-sampling design. In the first stage, clusters are selected with probability being proportional to cluster size, which is defined by the national population census. In the second stage, a predetermined number (approximately 25-30) of households are randomly sampled within each cluster. The target population is all women aged 15 to 49 years. Information on IPV is collected from one woman who is randomly selected among all the eligible women within each household. We restricted our sample to women who provided information on IPV and were currently married or cohabiting with a male partner.

Outcomes
The DHS contains the domestic violence module, which was developed according to the guideline set out by the World Health Organization (WHO) on the ethical collection of such sensitive information [16]. The IPV questionnaire in the module is modified from the Conflict Tactics Scales, which has good validity and reliability [17]. The questionnaire includes questions that ask women whether their current or most recent (if divorced, separated, or widowed) husband/partner ever perpetrated any of a number of behaviorally specific acts of violence and the frequency of perpetration in the 12 months preceding the interview.
The IPV questionnaire assesses sexual, emotional, and physical violence. Supplementary Text S1 presents the complete IPV questionnaire. If a woman reported experiencing at least one of the acts belonging to a form of IPV in the past 12 months, she is defined as having experienced a specific form of IPV. "Any IPV" was defined by as the experience of any form of IPV in the past 12 months [18]. We selected all three forms of IPV and "any IPV" as outcomes.

Exposures
Based on the definition of a slum household proposed by the United Nations [19], we defined inadequate living conditions as housing lacking any one of the following characteristics: durable construction, improved water, improved sanitation, and sufficient living space. The definition is consistent with previous studies that assessed the association of housing with child physical and mental health in LMICs [20][21][22].
We used the quality of building materials as a proxy for a house's durability. If the material of more than one out of three components of its floor, roof, and wall could be classified as "unfinished" based on the DHS criteria (Supplementary Table S2), then the house's building material was considered as unfinished [5,20]. We classified the household access to water and sanitation as 'improved' and 'unimproved' using the WHO and the United Nations definition (Supplementary Table S3) [23]. A house provides insufficient living space if more than three usual residents shared the room for sleeping [5].

A Priori Confounding Variables
The following covariates were included to control for the confounding bias: household wealth quintiles, residence type (rural/urban), age and education of the respondent, age of the respondent's partner, respondent's marital (married/cohabiting) and employment (employed/not employed) status, whether the age at the start of the first marriage or union of the respondent was less than 18 years (yes/no), and whether the partner consumes alcohol (yes/no).
Age was categorized into 15-19, 20-29, 30-39, and 40-49 years for the respondent and categorized into 15-19, 20-29, 30-39, 40-49, 50-59, and 60+ years for her partner. Education was categorized as "no education", "primary education", "secondary education", and "higher education", based on the highest education level achieved by the respondent. We defined cohabiting women as those currently living with unmarried partners, and employed women as those currently working, at the time of the interview. DHS generates wealth quintiles for all respondents based on their household ownership of a number of assets using principal component analysis [24].

Statistical Analysis
We conducted multivariate logistic regression to assess the association between inadequate living conditions and IPV in each country, adjusted for a priori confounding variables. The outcome of interest was each of the three forms of IPV and any IPV, and the exposures were one of the four inadequate living condition characteristics and inadequate living conditions as a combined variable. Then, for each of the four outcomes in each country, we included all four inadequate living condition characteristics in a single model and examined every characteristic's coefficients, which allows interactions between these characteristics to be taken into account.
We used the I 2 index to measure the consistency of the estimates across countries [25], and the DerSimonian and Laird method for meta-analyses with random effects to pool the estimates for all countries [26].
To reveal the heterogeneities in the effect of inadequate living conditions on IPV, we stratified our analyses by residence type (rural/urban) and education level (less than secondary/secondary or higher education) of the respondent, adjusted for all a priori confounding variables except for the stratifying variable. We then pooled the results using random effects meta-analyses.
The proportion of missing data is below 3% for all variables in all countries (Supplementary Table S4). Given the small proportion of missing data and a missing at random assumption, the estimation bias would be allowable if we ignored the missing data [27]. Therefore, only data from respondents without missing information on exposures, outcomes, and confounding variables were included in the analyses. We considered the complex sample design of DHS (stratification, clustering, and survey weights for the domestic violence module) when we estimated the standard errors.
The results were considered as statistically significant at a p-value lower than 0.05. We used R version 3.6.1 (R Foundation, Vienna, Austria) for all analyses.

Study Population
Data were extracted from surveys in 25 countries dating from 2008 to 2020. Complete outcomes, covariates, and exposure data were available for 102,714 women ( Table 1). The median of the country-specific mean age of the women was 31.7 years, and that of their partners was 39.0 years. More than half of the women had less than secondary education in 22 out of the 25 countries. The percentage of unemployed women ranged from 11.1% in Ghana to 67.2% in Ethiopia. The highest proportion of women who were married or in a union before the age of 18 years was found in Chad (76.4%), followed by Ethiopia (62.3%) and Mali (61.0%).

Living Conditions
More than half of the women lived in inadequate living conditions in all countries, except for Gabon and Gambia. The median percentage of the women living in inadequate living conditions was 74.2%. In Ethiopia (94.5%), Chad (93.6%), and Democratic Republic of the Congo (DRC) (91.8%), the percentages were >90% ( Table 2). The median percentage of the women living in housing with unfinished material, unimproved water, unimproved sanitation, and insufficient room was 36.6%, 28.4%, 49.5%, and 27.8%, respectively. Ethiopia, Chad, and Mozambique had the highest percentage of the women who lived in housing with unfinished material (84.8%, 77.4%, and 76.2%, respectively), as well as the highest percentage of women who did not have access to improved sanitation (86.6%, 84.9%, and 72.9%, respectively). More than half of the women lived in housing with unimproved water only in DRC (51.6%). Crowdedness was particularly severe in Ethiopia (61.6%), Chad (41.6%), and Angola (38.4%).

Intimate Partner Violence
The median prevalence of any, emotional, physical, and sexual IPV toward the women was as follows: any 32.5% (ranging from 7.6% in Comoros to 50.9% in Sierra Leone), emotional 24.7% (ranging from 5.7% in Comoros to 39.2% in Sierra Leone), physical 18.1% (ranging from 3.9% in Comoros to 38.9% in Sierra Leone), and sexual 6.4% (ranging from 1.1% in Comoros to 19.8% in Burundi) ( Table 3).
Positive and significant associations between inadequate living conditions and IPV were demonstrated in six countries for any IPV, three countries for sexual IPV, four countries for emotional IPV, and four countries for physical IPV (Supplementary Tables S5-S8). In Uganda, inadequate living conditions were associated with all three forms of IPV and any IPV. In Nigeria, inadequate living conditions were protective for all outcomes, except for sexual IPV. In Mali, inadequate living conditions were protective for emotional and any IPV. Similar associations were found between insufficient space and any (OR = 1.10, 95% CI 1.03 to 1.18, p = 0.004), sexual (OR = 1.12, 95% CI 1.01 to 1.23, p = 0.030), emotional (OR = 1.12, 95% CI 1.06 to 1.20, p < 0.001), and physical (OR = 1.11, 95% CI 1.03 to 1.19, p = 0.007) IPV. Unimproved sanitation was associated with a 10% increase in the odds of experiencing physical IPV (OR = 1.10, 95% CI 1.01 to 1.20, p = 0.024). Unimproved water was associated with a 9% increase in the odds of experiencing any (OR = 1.09, 95% CI 1.02 to 1.17, p = 0.012) and emotional (OR = 1.09, 95% CI 1.03 to 1.16, p = 0.005) IPV.
The association between a specific characteristic of inadequate living conditions and IPV persisted after the additional adjustment for the other three characteristics, and the estimates were almost identical to those in the analyses that only adjusted for a priori confounding variables (Supplementary Table S9).
The association between inadequate living conditions and IPV was stronger among rural women and women with lower education levels ( Table 5). Inadequate living conditions were associated with an increase in the odds of experiencing emotional (OR = 1.13, 95% CI 1.01 to 1.27, p = 0.034), physical (OR = 1.12, 95% CI 1.01 to 1.24, p = 0.025), and any (OR = 1.13, 95% CI 1.02 to 1.25, p = 0.020) IPV in rural women, but not in urban women. Women with less than secondary education who lived in inadequate living conditions had a higher risk of experiencing sexual (OR = 1.20, 95% CI 1.02 to 1.41, p = 0.031), emotional (OR = 1.13, 95% CI 1.00 to 1.27, p = 0.043), and any (OR = 1.12, 95% CI 1.01 to 1.25, p = 0.026) IPV. Among women with secondary or higher education, inadequate living conditions were associated only with experiencing physical IPV (OR = 1.16, 95% CI 1.00 to 1.34, p = 0.047).

Discussion
According to our literature search, the present research is the first multi-country study to examine the association between inadequate living conditions and IPV. Overall, we found that inadequate living conditions were associated with 13% higher odds of any IPV, 19% higher odds of sexual IPV, 12% higher odds of emotional IPV, and 15% higher odds of physical IPV. We also found that insufficient space was individually associated with experiencing one or more forms of IPV. The magnitudes of associations of unimproved water and unimproved sanitation with IPV were weak, although some were statistically significant. There were no statistically significant associations between unfinished material and any form of IPV.
Our findings add to the body of literature documenting the associations of inadequate living conditions and their characteristics with IPV in LMICs. Evidence from DHS data in Nepal suggests that unimproved household water sources elevated women's risk of experiencing physical and emotional IPV [28]. A study conducted in Brazil suggests that 20-59-year-old women living in households with more than two people per bedroom were more likely to report experiencing IPV compared with those living in less crowded housing [29].
However, contrary to our hypothesis, we found two countries, namely Mali and Nigeria, where inadequate living conditions were protective for at least one form of IPV. One potential explanation is that, in these two countries, partners' attitudes toward IPV are an important factor for the occurrence of IPV [30,31], which could reverse or mute the relation between inadequate living conditions and IPV [32]. In Nigeria, the protective effect of women's education levels could be reversed if the community's social norms justify IPV [31]. In Mali, the cash transfer program has been reported to reduce household poverty, but offer limited effects on reducing IPV, as cash might allow men to enact authority [30]. Further research is needed to elucidate the heterogeneity in these findings. Specifically, additional evidence may reveal the mechanisms underlying the negative association between inadequate living conditions and IPV.
The association between living conditions and IPV might be driven by several potential mechanisms: (1) Inadequate living conditions may act as direct stressors. For example, household crowding might lead to stress and depression by making it difficult for women to escape negative interactions [7,33], which are associated with both IPV perpetration and victimization [8,34]. (2) Inadequate living conditions might be directly associated with poor physical health outcomes [6], leading to health-related economic losses, and even household food insecurity [35], which is well established as causing negative emotional well-being among household members [36][37][38]. (3) Inadequate living conditions might affect occupants' social interactions, which have been recognized as protective factors associated with reduced IPV [39]. The home is commonly regarded as a place for the establishment and maintenance of social ties [40]. Individuals living in an unimproved home might also be reluctant to invite guests into their homes because of the stigma associated with poor housing environments [41], or socially withdraw from others to reduce excessive and unwanted social interactions, which is a prominent feature of a crowded home [42]. Individuals living in crowded homes are less likely to seek support from peers or offer support to peers in need [42].
We observed that rural women were more susceptible to the effects of inadequate living conditions on IPV compared with urban women. Research has reported higher rates of violence among rural couples than among urban couples in SSA [43], potentially owing to the fewer community resources, sparse distribution of law enforcement, and more accepting attitude toward IPV in rural areas [44]. Therefore, rural women are more vulnerable to IPV, which might be exacerbated by inadequate living conditions. We also noted that the association between inadequate living conditions and IPV remained statistically significant only in women with less than secondary education, who were more likely to experience IPV compared with women with secondary or higher education in SSA [43]. A lower education level might create an increased risk of IPV victimization through negative mental status, such as low self-esteem, stress, and depression [45], thereby affecting IPV through synergism with inadequate living conditions. In addition, educated people more often have a stable social support network [46], which might mitigate the erosion of social relationships by inadequate living conditions. Thus, policy makers should prioritize rural and less-educated women when gathering resources for housing interventions and IPV screening for families living in inadequate living conditions. This study has several limitations. First, owing to the cross-sectional design of DHS, our analysis could not derive a causal relation between inadequate living conditions and IPV. Second, the psychological and social mediators that we hypothesized had not been assessed in the DHS; therefore, we could not conduct a mediation analysis. Third, other characteristics of inadequate living conditions such as toxins (e.g., lead, solvents), noise, and insufficient daylight might also result in behavioral disturbances (e.g., aggression and violence) [47]. However, sufficiently detailed measures of these characteristics are not included in the DHS data. Fourth, although the validity of the IPV questionnaire in DHS has been confirmed [18], women might underreport IPV for fear of victimization or stigmatization [48]. Non-differential misclassification generally biases the results to null [49]. Fifth, in some countries, the percentages of inadequate living conditions were high, even above 90%, which could reduce the precision of the estimates. We believe that a metaanalysis would compensate for this limitation by calculating the inverse variance-weighted average of the country-specific estimates. Sixth, our analysis only included married and cohabitating women aged 15-49 years in SSA countries; therefore, the generalization of our findings to other populations should be taken with caution, especially given the cultural differences in IPV disclosure and the subjective acceptability of crowdedness across regions and countries [48,50].
The strength of this study is that we investigated the associations of inadequate living conditions with IPV using multiple nationally representative datasets and analyzed the data in each country in a unified manner. Compared with a single-country study, our work included 25 countries across SSA, and could thus provide a more generalizable estimation. Further, instead of merging data from all countries and conducting a one-stage analysis, we used meta-analyses, which allowed the effect of covariates to vary across countries [51].
Both current WHO guidelines and the technical package for preventing and responding to IPV from the Centers for Disease Control and Prevention (CDC) have primarily focused on legal reform, advocacy of gender equity, and the promotion of the empowerment of women, giving little attention to the living environment [52,53]. Although the CDC's technical package suggests creating a protective environment as a prevention strategy, it does not specify physical living conditions [53]. However, given the relatively small magnitude of the associations found in this study, we do not recommend that scarce resources be shifted away from IPV interventions that have already been shown to be effective in SSA, such as educational meetings or workshops addressing expectations about gender roles and behavior in communication and conflict resolution [54]. Instead, we suggest conducting additional research with a higher level of evidence to provide a better understanding of the effect of living conditions on IPV.

Conclusions
To the best of our knowledge, this study is the first in SSA to demonstrate the direct associations of inadequate living conditions, especially insufficient space, with three forms of IPV. Specifically, our findings highlighted that the association is most apparent in rural women and women with less than secondary education. Therefore, research with rigorous study designs and measurements of the exposures, outcomes, and mediators is needed to establish a causal link between living conditions and IPV to elucidate the pathways behind this link and design cost-effective IPV screenings and interventions in SSA.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/ijerph181910138/s1, Table S1: The intimate partner violence (IPV) questionnaire in the Demographic and Health Surveys.   Institutional Review Board Statement: No human participant work or field research was conducted as part of this secondary data analysis. All data used are publicly available and fully de-identified. Ethical approval for the DHS was provided by individual review boards within each participating country at the time of survey implementation.