Menu
Menu
Menu
Menu
Background Women exposed to Intimate Partner Violence (IPV) often do not utilize maternal health care optimally both because of stigma and other social problems. The current study aims to explore an association between maternal healthcare seeking and violence exposure among Ethiopian women and to assess if educational attainment and wealth status moderate this association. Methods The analyses included 2836 (weighted) currently married women with one live birth. We focus on the five years preceding the 2016 Ethiopian Demographic and Health Survey (EDHS) who participate, in the domestic violence sub-study. Exposure was determined by maternal reports of physical, emotional, sexual IPV or any form of IPV. The utilization of antenatal care (ANC) and place of delivery were used as proxy outcome variables for uptake of skilled maternal healthcare utilization. Women’s education attainment and wealth status were selected as potential moderators, as they can enable women with psychological and financial resources to counteract impact of IPV. Multilevel logistic regression analyses were used to explore the association between spousal IPV and maternal health outcomes. Moderation effects by education and wealth status were tested, and the data stratified. Using statistical software Stata MP 16.1, the restricted maximum likelihood method, we obtained the model estimates. Results About 27.5% of the women who reported exposure to any form of IPV had a health facility delivery. While 23.4% and 22.4% visited four or more antenatal care services among mothers exposed to emotional IPV and sexual IPV, respectively. After adjusting for potential confounding factors, only the association between maternal exposure to emotional IPV and adequate use of ANC was statistically significant (OR = 0.73, (95% CI:0.56-0.95)). But we found no significant association between IPV and utilization of health facility delivery. Some moderation effects of education and wealth in the association between IPV and maternal healthcare service utilization outcome were found. Conclusion Exposure to emotional IPV was associated with poor uptake of maternal health care service utilization for married Ethiopian women. While developing interventions to improve women’s maternal healthcare service use, it is crucial to consider the effects of socio-economic variables that moderate the association especially with the intersection of IPV.
The publically available data of the fourth nationally representative survey of the 2016 Ethiopia Demographic and Health Survey (EDHS) was collected between January–June 2016. The full details of the data collection methods and procedures as well as the standards for protecting the privacy of study participants have been published [13]. IPV information for ever-married women age 15–49 ever reporting exposure of spousal emotional, physical, or sexual violence was collected using a modified and abbreviated version of the Revised Conflict Tactics Scales (CTS2) [30]. After excluding missing values, a total of 3061 (unweighted) ever-married women in reproductive age were considered in this study [13]. Special domestic violence weights were used to make the survey data on violence nationally representative accounting for non-response [31]. The final study sample was further limited to those who were currently married and had at least one live birth in the five years preceding the survey (weighted, n = 2836). The analyses in the current study address two maternal healthcare binary outcomes: (1) adequate antenatal care (ANC) use; categorized into four or more visits (≥4) and less than four visits (<4, this included women with no visit), in accordance with the 2002 WHO ANC model [32], which was recommended by the Ethiopian Federal Ministry of Health (FMoH), at the time of initiating this study (This is not the current recommended ANC protocol by WHO which is based on WHO’s 2016 ANC Model prescribing a minimum of eight contacts.) and (2) place of delivery, categorized as home birth or birth at a health facility. The predictor variables were reported as exposure to emotional, physical, sexual IPV or any type of IPV. In the current study, emotional IPV is a composite binary variable based on responses to three questions: Had the husband ever: (1) said or did something to humiliate her in front of others; (2) threatened to hurt or harm her or someone she cared about; or (3) insulted or made her feel bad about herself, with yes (experiencing at least one of these); and (not experiencing any), [13, 28, 33]. Similarly, physical IPV is a composite binary variable based on women’s responses to the questions about whether the husband ever had done any of the seven following acts: (1) push, shake, or throw something; (2) slap; (3) twist arm or pull hair; (4) punch with fist or with something that could be harmful; (5) kick, drag, or beat her up; (6) tried to choke or burn her; (7) threaten or attacked with any material to deliberately hurt her at one point in lives [13, 28, 33]. Sexual IPV was, responding yes to any of these three questions: (1) physically forced to have sex; (2) forced to other sexual acts; (3) forced by threats when she did not want to [13, 28, 33]. Lastly, any IPV, was a composite dichotomous summary measure created from 13 questions (emotional IPV: 3, physical IPV: 7, and sexual IPV:3) to capture the women’s ever experience of any IPV (emotional, physical and/or sexual), grouped as: Yes (‘yes’ responses to any of these 13 questions), and No (‘no’ responses to all of the 13 questions). Based on the literature, two variables–women’s education level and household wealth status were considered as potential moderators [27, 28]. Education level of the woman was based on: the highest level of education attained by the respondent and grouped into two: as None, or Primary and above. Household wealth index; a composite index of household possessions, assets, and amenities, derived using principal component analysis (PCA), and ranked as poorer; poor; middle; rich; and richest. For our analysis, we re-categorized wealth into three categories (poor, middle, rich) [34, 35]. Based on the current literature, we included several potential confounding variables. The woman’s self-reported age at the time of the survey, was categorized as younger (15–24 years); middle (25–34 years) and older (35–49 years) as age affects health seeking behaviors, [36]; the order of the last birth closes to the time of the survey; education level of the partner reported as none, primary and above; exposure to mass media (composite variable based on the access to and frequency of use of radio and/or television at least once a week), [37]; decision-making autonomy in making three household decisions (access to health care; large household purchases; and freedom to visit families and relatives), grouped into, low autonomy (no participation in any decision making), medium autonomy (participation in 1 or 2) and high autonomy (participation in all decision making); attitude towards wife beating was created using scenarios: (1) she burns the food; (2) she argues with him; (3) she goes out without telling him; (4) she neglects the children and (5) she refuses to have sex with him. A woman was regarded as accepting violence if she said it was justified for any of these five reasons and as rejecting if she reported that beating was not justified for any reasons [38], and the place of residence at the time of the survey categorized as urban or rural. The regions were defined according to the FMoH as agrarian (Tigray, Amhara, Oromia and SNNPR), pastoralist (Somali, Afar, Gambella, and Benishangul Gumuz regions) and city dwellers (Addis Ababa, Dire Dawa, and Harar). We used bivariate analyses to describe the characteristics of the women in relation to the outcome of interest and each type of IPV along with the Pearson Chi-square (X2) test of independence to examine whether there were any significant differences in the sociodemographic characteristics, and the associated p-value calculated. Sampling weights were applied for the data when we computed both the bivariate and multivariate analysis to manage the unequal probability of selection between the strata defined by geographical location and for non-responses. We fitted separate random-effects multilevel logistic regression models, for each outcome of interest (ANC and delivery care) using only the variables that were significantly associated with each outcome and type of IPV in the bivariate models. We used a binary logistic multilevel regression model, as the data was clustered at the survey level (2836 women nested in 626 clusters). Univariate logistic regression was performed to estimate the crude odds ratios (COR), (See, S1 Table). And the 95% confidence intervals (95%CI) of facility delivery or not, and if she had at least four ANC visits or not. Potential predictors and confounders significantly associated with the outcome variables in the univariate analysis were entered in the multilevel logistic regression analysis. We conducted four separate fully adjusted models for each type of IPV (emotional, physical, sexual, and any type of violence) for each outcome variable while controlling for confounders to identify the association between spousal IPV on the use of maternal healthcare services. To assess any moderating effect of education and/or wealth in the association between exposure to spousal IPV and maternal healthcare services, interactions were checked (interaction between IPV and education/wealth). Finally, analyses of the association between exposure to spousal IPV and maternal healthcare services were stratified by level of education and household wealth status. Prior to the multivariate regression analysis, multi-collinearity was checked using variance inflation factors (VIF) which indicates that there was no multi-collinearity since all variables have VIF <2 and all were considered for the subsequent analysis. In addition, we computed an estimate of intra-cluster correlation coefficient (ICC), which described the amount of variability in the response variables attributable to differences between the clusters. We examined the model fit measured using the Akaike information criteria (AIC). A lower AIC value represents a better model fit [39]. All the statistical analyses were performed using complex sample analysis procedure to allow for adjustment for the sampling weight, stratification, the cluster sampling design, and the calculation of standard errors of the large survey data [31, 40]. We conducted tests for correlations among the types of IPV. IBM SPSS 26.0 was used for data preparation and the model parameter estimates were obtained in the statistical software Stata MP 16.1 using the restricted maximum likelihood method (REML). The level of significance was set at 0.05. The study adhered to national and international ethical guidelines for biomedical research involving human subjects [41], including the Helsinki declaration. The study protocol was reviewed and approved by the Regional Committee for Medical and Health Research Ethics (Code number: 2016/967/REK sør-øst A) and the Norwegian Centre for Research Data (Code number: 48407) at the University of Oslo. Our team also requested permission and access to the data from the CSA in Ethiopia and Inner City Fund (ICF) international by registering online on the website www.dhsprogram.com [42] and submitting the study protocol by highlighting the objectives of the study as part of the online registration process, (See, S1 File). The ICF Macro Inc, removed all information that could be used to identify the respondents; hence, anonymity of the data was maintained.
