Intimate partner violence and depression among women in rural Ethiopia: a cross-sectional study

Study design and Study setting

This was a community-based cross-sectional survey conducted in Ethiopia in two districts called Meskan and Mareko. The study was part of the WHO multi-country study on women's health and life events and was conducted between January and December 2002 [24, 25]. The districts are about 130 km south of Addis Ababa, and have a total population of about 257,500, of which 51.1% are females, as projected from the 1994 census of Ethiopia. Women aged between 15 and 49 years make up 24.8% of the population [26]. The district is organized administratively into small units termed peasant associations in rural areas and urban dwellers' associations in semi-urban areas; both units are commonly referred to as kebeles. Most of the population (87%) live in rural villages in which households primarily depend on subsistence farming [26], the economic mainstay of the district, comprising more than 85% of the economy. The main staple foods are maize and ensete (false banana); khat, pepper and coffee are the main cash crops [27].

Within the area, there is a demographic surveillance site called the Butajira Rural Health Programme (BRHP). The Butajira Rural Health Programme is a Demographic Surveillance Site (DSS) that includes one semi-urban dwellers' association and nine rural peasant associations selected in 1986 from the Butajira district using the probability proportionate to size technique [27, 28]. The current study was conducted within the Butajira Rural Health Programme demographic surveillance site (BRHP), utilizing its household database as the sampling frame [27].

As part of the WHO multi-country study, women living in the Butajira Rural Health Programme (BRHP) were the source population. Women fulfilling the following inclusion criteria were selected: aged between 15 and 49 years, residents of the Demographic and Surveillance Site of the BRHP for at least the past three months, and included on the list of women in the database [25]. In this study we excluded women who were not currently married.

Sampling method

Based on the prevalence of physical violence in the study site of 45% [29], and the reported prevalence of depression among women in a previous study from the area [30], we assumed the prevalence of depression to be 7.0% among non-abused women and 10% among abused women. In order to obtain the maximum sample size at a 95% confidence level and 80–95% power certainty, to detect a lower associated odds of depression of about 1.4, a minimum of 3048 women were necessary. To compensate for possible non-response, an additional ca. 5%, or 152 women, were added. Thus, a total of 3200 women were required. We managed to recruit 3016 women between the ages of 15 and 49, a response rate of 94.25%. In this paper, data from 1,994 currently married women were used.

Since 85% of the population resides in rural areas, an equivalent percentage of the sample was drawn from rural peasant associations. Additionally, in order to ensure that the number of women was equally distributed in each clustered peasant association, we recruited participants in proportion to population size. The database of each kebele obtained from the BRHP provided a sampling frame that contains the women's name, a unique identity number and a household number. Women were selected by simple random sampling from the database of each Kebele using SPSS for Windows software. In order to find women at risk of violence by an intimate partner, all 1994 currently married women were included in the present study.

Data Collection and Management

Thirty women who had completed high school education and had experience of data collection were trained for a total of three months in the Composite International Diagnostic Interview (CIDI), version 2.1 [31], and a WHO standardized questionnaire of life events and health [25]. Data were collected after conducting a pilot test in villages outside the survey area.

Measurements

Cases of depressive episode were identified by interview using (Section E of) the Amharic version of the Composite International Diagnostic Interview (CIDI), version 2.1, which was validated in Addis Ababa [32]. In this study, to ascertain cases of depressive episode we used F32 of the International Classification of Diseases, 10^th edition (ICD-10) [33], which is equivalent to major depression in the Diagnostic and Statistical Manual (DSM-IV) [34] of mental disorders.

The Amharic translated version of questions with ambiguous terminologies used to measure intimate partner violence was discussed in two series of focus group discussions to ensure local understanding.

Sexual violence during childhood was measured anonymously using a picture showing a happy and a sad woman. The interviewees were instructed to mark the picture of the "sad woman" representing being sexually violated as a child before age 15 years, and to mark on the picture of the "happy woman" representing a woman who had not been sexually violated as a child. The enumerators were kept blind to the marks, and the marked pictures were placed in an envelope sealed by the interviewee.

To control for confounding effects and to assess any independent effect of intimate partner violence on depression, socioeconomic and sociocultural characteristics were included. Age was grouped into four categories; residency was grouped into urban and rural; and educational status was grouped as 'no formal education', 'elementary' for grades 1 to 6, and 'high school and above' for education longer than 6 years. Religion was categorized as Christian or Muslim, and occupational status was categorized into four groups: no job, selling/trading, seasonal work (working in any available unskilled job), and other. Poverty status of women was categorized into three groups, and khat chewing was categorized by frequency of use, as described elsewhere [35]. In addition, marriage type was categorized as polygamous or monogamous. As a predominantly Muslim community, most women did not respond to the question on alcohol use, so it was omitted from the analysis.

Data Analysis

The sociodemographic characteristics and experience of intimate partner violence of the participants were summarized using descriptive statistics. In summarizing the data, totals were indicated when figures did not exactly sum 1994. The 12-month status of association between intimate partner violence and depressive episode in the past 12 months was explored using crude and adjusted odds ratios and their 95% confidence intervals. To examine the presence of any association between various types of intimate partner violence and depressive episode, we used multiple regression. In the multiple regression, socioeconomic characteristics that were statistically associated with depressive episode or intimate partner violence were included to control for possible confounding effects. We set the criteria for entry and retention in the regression equation ∝ at 0.05.

Ethical Considerations

The study was approved by the ethics and publication committees of the Medical Faculty of Addis Ababa University, the Ethiopian Science and Technology Commission and Umeå University in Sweden. As part of the multi-country study on violence against women, the current study followed WHO ethical and safety guidelines [24]. Informed consent was obtained from each woman participating in the study and from the local formal and informal leaders. Privacy and confidentiality were also maintained throughout the data collection. During the data collection period and for one year after data collection, a counselling service was instituted in the area for those in need.

Sample

Prevalence

Intimate partner violence and depressive episode

Women who had experienced a depressive episode in the last 12 months were more likely to report experience of sexual violence as a child than women who were not depressed. However, there was no association between depressive episode and sexual violence by an intimate partner (Table 4).

Women with depression reported severe or moderate forms of emotional violence to a higher extent than those without depression. The prevalence of depressive episode was significantly higher among women who reported being totally or moderately controlled by their spouses compared to those who reported not being controlled (Table 4).

Prevalence of Depressive episode

The 12-month prevalence of depressive episode of 4.8% among currently married women in the present study was relatively low compared to studies from the high income countries, such as the study in six European countries with a six-month prevalence of 17% [36] and a lifetime prevalence of major depression of 13.3% in the Cagliari district of Sardinia [37]. However, the prevalence of depression in the current study is similar to findings from a survey conducted in Australia with a 12-month prevalence of depression of 5.8% [38], as well as a 30-month prevalence of 3.5% [39]. It is also comparable to a study conducted using CIDI in 1996 in Butajira, the same location as the current study, where the lifetime and one-month prevalences of depression among the female population were 4.5% and 3.8% respectively [8]. The relatively low prevalence of depression in this study site may have several explanations, one of which may be the existence of high social capital. It may also be influenced by the instrument used to ascertain depressive episode, with terminologies relating to symptoms of mental illness having different social and cultural contextual meanings [40].

Prevalence of intimate partner violence

The results of our study indicate that nearly one half of the women surveyed had experienced physical violence at least once in their marital life. Compared to other studies that were part of the multi-country study using a similar instrument, the women in Butajira were more subjected to violence. This could be due both to a lack of policies and legislation relating to intimate partner violence and to cultural norms. Our interview technique was carefully geared to include culturally sensitive terminologies based on focus group discussion with local women, which might have contributed to better recording of intimate partner violence. However, our prevalence result is comparable to an earlier study conducted in the same area in 1996, which found a lifetime prevalence of physical violence by intimate partner of 45% [29], and a Nicaraguan study where 52% reported similar experiences [23].

Intimate partner violence and depression

Our result of an association between physical violence and depressive episode is consistent with findings from several studies in the high income countries. For example, a study in Australia among 1257 female primary care patients reported that depressed women were significantly more likely to have experienced severe forms of abuse than women who were not depressed [41]. Similarly, a study in a Chinese community in the USA showed a strong association between intimate partner violence and postnatal blues/depression [42].

There are several possible explanations. Firstly, violence by an intimate partner is likely to decrease self-esteem and underpin a negative view of self, the world and the future. Secondly, in Ethiopia, women who experience physical violence are usually humiliated in front of their neighbours, friends and relatives [29], resulting in feelings of shame and withdrawal behaviour, which might further increase isolation and stigmatization [23]. In addition, emotional violence is a form of violence that directly erodes a person's self-esteem, especially intimidation in front of other people. Violence is known to result in recurrent fear and learned helplessness leading to impairment of problem-solving ability. Physical violence combined with emotional trauma can result in feelings of frustration, motivational impairment including passivity, and low self-esteem may lead to depression [21]. In this study, physical violence that begin a year prior to the date of interview had a higher probability of depressive episode than violence that begin within a year of the interview; this finding could be a clue to the probability of depression increasing in parallel with time or with frequency of violence.

Our finding of an association between experiencing sexual violence as a child and depression is consistent with previous studies [43]. Childhood sexual violence is well documented as a risk factor for depression and other mental illnesses in adult life. However, our study failed to show an association between experience of sexual violence by an intimate partner and depressive episode. This might be explained by its hidden nature: it does not take place in public, making the women less stigmatized in the eyes of neighbours and others. The lack of association may also be due to the contextual meaning of sexual violence by an intimate partner, which may not be deemed as violence in a community where men are considered to have the right to have sex with their wives regardless of the woman's consent. This was confirmed by a qualitative study conducted in northwest Ethiopia by Tegbar and co-workers in 2006, which showed that a woman is considered to experience sexual violence only if she is sexually violated by someone other than her spouse (personal communication).

Spousal control and depressive episode

In this study, reporting being controlled by their spouse was associated with depressive episode in women. This is in accordance with other studies, suggesting that a lack of decisional control among women might be a factor in leading to depression to be more prevalent among women compared to men [5]. Similar to physical or emotional violence, women controlled by their spouses may suffer from feelings of frustration, motivational impairment including passivity and low self-esteem, which could result to adverse mental health consequences including depression [21].

Limitations and strengths of the study

Our study is limited first by its cross-sectional design with difficulties proving a temporal relationship between explanatory and outcome variables. Another limitation might be recall bias when lifetime experience is included. To minimize this limitation, we used 12-month prevalences of the dependent variables. Even though the research group made great efforts in cultural adaption of the diagnostic instrument CIDI [31], we cannot exclude that CIDI performs less well in this Ethiopian setting. Other possible shortcomings include absence of HIV status, and lack of recording of other hassles of life events which may confound a relationship between intimate partner violence and depressive episode. However, the strengths of our study include minimizing bias by its community-based design utilizing a high-quality study base and a high response rate of 94.6%, one of the highest of the countries participating in the multi-country study [24]. Meticulous training, regular daily supervision of the data collectors, and checking completeness and accuracy are further strengths.