N/A
Background: Intimate partner violence (IPV) is major public health problem that affects many dimensions of women’s health. However, the role of IPV on women’s reproductive health in general and pregnancy loss in particular, is largely unknown in Ethiopia. Therefore, this study investigated the association between IPV and pregnancy loss in Ethiopia. Methods: A retrospective analysis of nationally representative data from the 2016 Ethiopian Demographic and Health Survey (EDHS) was conducted. Married women of reproductive age (15-49 years) who participated in the domestic violence sub-study of the survey were included in the analysis. Adjusted odds ratios were estimated using multilevel logistic regression models to represent the association of IPV with outcome variable. Results: Among 4167 women included in the analysis, pregnancy loss had been experienced by 467 (11.2%). In total, 1504 (36.1%) participants reported having ever experienced any form of IPV, with 25.1, 11.9, and 24.1% reporting physical, sexual and emotional IPV respectively. A total of 2371 (56.9%) women had also experienced at least one act of partner controlling behaviour. After adjusting for potential confounders, a significant association was observed between IPV (a composite measure of physical, sexual and emotional abuse) and pregnancy loss (Adjusted Odds Ratio (AOR) 1.54, 95% Confidence Interval (CI): 1.12, 2.14). The odds of pregnancy loss were also higher (AOR 1.72, 95% CI: 1.06, 2.79) among women who had experienced multiple acts of partner controlling behaviours, compared with women who had not experienced partner controlling behaviours. The intra-class correlation coefficient (ICC) indicated that pregnancy loss exhibits significant between-cluster variation (p < 0.001); about 25% of the variation in pregnancy loss was attributable to differences between clusters. Conclusion: IPV against women, including partner controlling behaviour, is significantly associated with pregnancy loss in Ethiopia. Therefore, there is a clear need to develop IPV prevention strategies and to incorporate IPV interventions into maternal health programs.
This study used data from the 2016 Ethiopian Demographic and Health Survey (EDHS), which was the year the domestic violence module was added. The EDHS was a national survey conducted from 18 January to 27 June 2016. The 2016 EDHS data was collected with five questionnaires (household, women, men, biomarker and health facility). The EDHS used 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 design 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 selected enumeration area were identified using systematic random sampling. All women aged 15–49 years in the household were eligible for the EDHS interview. Accordingly, 15,683 women, with a response rate of 95%, participated in the general survey [7]. For the domestic violence sub-study, only one married woman per household was interviewed. Of those women who were eligible, 97% (n = 5860) 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 did not reduce representativeness of the EDHS sample [7]. For this analysis, ever-married women who had complete data related to their pregnancy and birth history and responded to the IPV questionnaire were included. Women who had never been pregnant, who were missing either the outcome variable or IPV data were excluded from the analysis. Accordingly, 4167 (unweighted sample of 4372) women were included in the analysis. The outcome variable for this study was pregnancy loss. In the 2016 EDHS, women were asked a single question “Did you have any miscarriages, abortions or stillbirths that ended before 2011?” In addition, women were asked about their pregnancy and birth history during the 5 years (2011 to 2016) before the survey that provided information about whether the pregnancy was terminated or ended with a live birth [7]. Aggregating the responses from these two questions, women who had ever experienced pregnancy loss were identified. Accordingly, pregnancy loss was coded as ‘Yes’ if respondents reported ever having experienced a miscarriage, induced abortion, or stillbirth and ‘No’ if women had never experienced any of the three events. This method of defining pregnancy loss has been used in previous research [18, 20, 22]. The exposure variable was having ever experienced IPV (physical, emotional, and sexual violence, and partner controlling behaviour). IPV was measured based on women’s self-reported responses to questions asked whether or not they had experienced a number of violent acts within their relationship, perpetrated by their husband/partner for currently married women and recent husband/partner for previously married women (including widows). Physical IPV was assessed by asking participants seven questions regarding having: ever been pushed, shaken, or thrown something at her; slapped; her arm twisted or hair pulled; punched with fist or with something that could hurt; kicked, dragged, or beaten up; been choked or burnt on purpose; or been threatened or attacked with a knife, gun, or any other weapon. Three questions were asked to measure sexual IPV: having ever been physically forced to have sexual intercourse with her partner even when she did not want to, physically forced to perform any other sexual acts she did not want to, or forced with threats or in any other way to perform sexual acts she did not want to. Likewise, emotional IPV was assessed by asking three questions: if the participant had ever been humiliated, threatened, or insulted or made to feel bad about herself. Those women who were married more than once were also asked about spousal violence committed by any other husband/partner with two questions that asked about having ever been hit, slapped, kicked or done something else to hurt her and ever been physically forced to have intercourse or perform any other sexual acts against her will. Respondents were categorized as having experienced lifetime IPV if they had experience of any single act of physical, sexual or emotional IPV since the age of 15 years [7]. Likewise, any single act of partner controlling behaviour was categorized as ‘yes’ if one of the following behaviours 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’. Where women reported two or more acts of partner controlling behaviour, the responses were coded as ‘multiple controlling behaviours’ [7]. Variables that needed to be controlled in order to estimate the unbiased effect of the exposure upon the outcome were identified based on an examination of previous literature [2, 3, 8, 13, 16, 18–22, 27]. Accordingly, current age of the respondent (15–19/20–24/25–29/30–34/35–39/40–44/45–49 years), age at first cohabitation (< 15/15–18/≥18 years), respondent’s educational status (uneducated/primary/secondary+), religion (Christian/Muslim/other), number of children ever born (≤1/2–3/≥4) were considered. In addition, respondent’s employment status, rurality (urban/rural), region (11 administrative regions), decision-making, wealth index, media access, substance abuse, and pregnancy intention were included. Respondent’s employment status was grouped as employed/not employed based on their response to “have you been employed in the last 12 months”. Decision-making autonomy was coded as ‘yes’ if women reported being involved in all decisions regarding her own health care, major household purchases and visits to her family or relatives. Household wealth index was measured based on the number and kind of goods households have and housing characteristics (drinking water, toilet facility, flooring material and availability of electricity), and was generated using principal component analysis and classified into quintiles from 1 (very poor) to 5 (very rich). Media access was measured as whether the respondent read a newspaper, listened to the radio, or watched television and was categorized as no access, access less than once a week, and access at least once a week. Substance abuse was classified ‘yes’ if respondent drinks alcohol, chews khat (a green plant consumed as a stimulant) or smokes tobacco and ‘no’ otherwise. Pregnancy intention of respondents was categorized into two as ‘unintended’ and ‘intended’. A respondent was defined as having an unintended pregnancy if she had a pregnancy in the past 5 years that was either mistimed (wanted the pregnancy to happen later i.e. after 2 years) or unwanted (did not want the pregnancy at all). Multilevel logistic regression models were fitted considering hierarchical nature of EDHS data (4167 women nested in 640 clusters). Multilevel analysis allows for the estimation of valid standard errors by adjusting for within-cluster correlation of the response variable [38]. Two models were constructed; Model I (the empty or unconditional model) and Model II (two independent models for IPV and partner control behaviours). In Model I, no independent variables were included. This model was used to estimate the random intercept at cluster level and the variation in pregnancy loss between clusters. Then, a second model was constructed by adding covariates and main independent variable (IPV or partner controlling behaviours) to Model I. Interactions between variables were assessed. Model fit was tested using Likelihood ratio test and the Akaike Information Criterion (AIC). Model II was the final model used to estimate measures of association between IPV and pregnancy loss. Adjusted odds ratios together with the 95% CI were used to report associations. Statistical significance was declared using a p-value < 0.05. The measure of variance (random effects) was reported in terms of the intra-class correlation coefficient (ICC). The ICC measures the extent to which women within the same cluster are more similar to each other in the outcome variable (i.e. pregnancy loss) than they are to women in different clusters [38]. All the analyses took into account the EDHS sampling weight and were based on the weighted sample (n = 4167). The sampling weights used in the EDHS account for the complex sampling procedures (multi-stage stratified cluster sampling) that might cause an unequal probability of selection for certain areas or subgroups either due to design or coincidence. Hence, sampling weights were adjusted for differences in probability of selection and interview that allow extrapolation of results to the national level of representativeness [7].
N/A