Study setting and population
We conducted a cross-sectional analysis using the Ethiopian Demographic and Health Survey (EDHS) 2016. All child-mother pairs that participated in EDHS 2016 from all regions of Ethiopia were included. The analysis included mother-child pairs where 6–23 months aged children with mothers who were ever in a committed partnership and interviewed for domestic violence were involved.
Sample size and sampling procedure
For EDHS 2016, the census frame was a complete list of 84,915 enumeration areas (EAs) created for the 2007 census. Each region was stratified into urban and rural areas and samples of EAs were selected in each stratum in two stages. In the first stage, a total of EAs were selected with probability proportional to EA size, and a household listing operation was performed in the selected EAs. The resulting lists of households served as a sampling frame for the selection of households in the second stage. Details of the recruitment of study participants found elsewhere . From the involved households, 15,683 women completed the interview, and 5860 of them were interviewed for domestic violence. Only one eligible woman per household was randomly selected for domestic violence interview, and it was omitted if privacy could not be obtained. There were 10,641 children under the age of 5 and 3105 of them were in the age group between 6 and 23 months.
Data collection procedures and quality control
The EDHS is a standardized report prepared from data collected using standardized tools. The data were collected through face-to-face interviews with the child’s mothers/caregivers. The 2016 EDHS questionnaires have parts divided into three; households, woman’s, and man’s questionnaires. Primary caregivers were interviewed to get data about the household and children. Generally, the questionnaires capture information on socio-demographic variables, reproductive health issues, domestic violence, health service characteristics, and others. The woman’s questionnaire had parts for children, which capture information on child immunization, health, and nutrition. The dietary diversity data were collected using a 24-h recall method; that is, mothers were asked to recall all foods given to their child in the past twenty-four hours before the survey.
Variables of the study
The outcome of interest was getting minimum acceptable diet and the primary exposure was maternal exposure to intimate partner violence. The covariates considered were child characteristics (age in months, birth order, child sex); healthcare characteristics (place of delivery, number of antenatal visits, PNC counseling); parental characteristics (mother’s age, place of residence, father’s education, mother’s marital status, mother’s work status); household characteristics (wealth index, number of children under the age of 5, exposure to media).
Minimum dietary diversity (MDD) was defined as the proportion of children aged 6–23 months who received foods minimum acceptable diet from four or more food groups out of the seven food groups during the previous day/ within 24 h [9, 19]. The seven food groups are the following: starchy staples (foods minimum acceptable diet from grain, roots, or tubers); 2) legumes and nuts; 3) dairy products (milk other than breast milk, cheese, or yogurt); 4) Flesh foods (meat, fish, poultry and liver/organ meats; 5) vitamin A-rich fruits and vegetables (pumpkin; red or yellow yams or squash; carrots or red sweet potatoes; green leafy vegetables; fruits such as mango, papaya, or other local vitamin A-rich fruits); 6) other fruits and vegetables (or fruit juices); 7) eggs. Minimum meal frequency (MMF) was defined as  at ages 6–8 months, the child was breastfed and received two or more daily feedings of solid, semi-solid or soft foods; or  at ages 9 to 23 months, the child was breastfed and received three or more daily feedings of solid, semi-solid or soft foods; or  at ages 6 to 23 months, the child was not breastfed and received four or more daily feedings of solid, semi-solid, or soft foods. Minimum acceptable diet: A child was considered to receive at least the minimum acceptable diet for health if the MDD and the MMF criteria were met.
Intimate partner violence was defined as ever in committed partnership women who have experienced one or more of the specified acts of spousal physical violence or sexual violence or emotional violence by their current or most recent husbands/partners in the 12 months preceding the survey . Intimate partner physical violence was defined as ever in committed partnership women who have experienced one or more of the specified acts of spousal physical violence by their current or most recent husbands/partners in the 12 months preceding the survey. Likewise, intimate partner sexual violence and emotional violence were defined as ever experiencing one or more of the specified acts of spousal sexual or emotional violence by their current or most recent husbands/partners in the 12 months preceding the survey.
Exposure to mass media: in the EDHS 2016 survey, respondents were asked how often they read a newspaper, listened to the radio, or watched television. Those who were exposed to any of the media at least once a week were considered to have adequate media exposure.
Data management and analysis
The data were weighted considering enumeration areas as a cluster and place of residence as a stratum. Initial descriptive analyses provided general information on the characteristics of the study populations. A bivariate logistic regression analysis was done. Variables with P-value < 0.2 were included in the multivariable logistic regression model to identify factors independently associated with a minimum acceptable diet.
A propensity score matching analysis was performed to identify the effect of intimate partner violence on a minimum acceptable diet. A propensity score was the probability of being exposed to the IPV, given a set of observed covariates (residence, mother educational status, wealth index, mother work status for the last 12 months, child desire, media exposure), and estimated using the logistic regression model. Nearest neighborhood matching was used in the analysis, which matches a given child of a woman who had intimate partner violence to a child of a woman who had no intimate partner violence whose propensity score is closest to that of the first subject or vice versa. The method is used to balance the two groups so that a direct comparison would be possible for evaluating the effects of intimate partner violence on the minimum acceptable diet. The average effect on children of women who had intimate partner violence was computed by averaging the difference between the outcomes of the two groups. The level of significance was defined at a P-value of less than 0.05.