Abstract

Introduction

Intimate Partner Violence (IPV) is the most prevalent form of violence against women globally and is more prevalent than rape or other violent attacks by strangers. Different observational studies have established a strong positive association between alcohol use and intimate partner violence. Even though there are a lot of studies that show the association between partner alcohol use and intimate partner violence limited studies were conducted that show the direct causative relations of partner alcohol use and IPV among reproductive-age women in East Africa. Therefore, this study aimed to determine the effect of partner alcohol use on intimate partner violence in East Africa’s recent Demographic and Health Survey (DHS) data with Propensity Score Matching (PSM).

Method

Community-based cross-sectional study design with a propensity score matching was used from the East African countries’ DHS data. A total of the weighted sample size of 72,554 reproductive-age women was used for this study. Propensity score matching analysis was conducted to determine the causal relation between partner alcohol use and intimate partner violence. Intimate partner violence was the outcome variable and partner alcohol use was the treatment variable. Propensity score matching was carried out through Stata software by using psmatch2 of the logit-based model. The assumption of common support was verified and achieved. Mantel-Haenszel boundaries have been used to investigate the possibility of hidden bias in the outcome.

Result

The prevalence of partner alcohol use and intimate partner violence from East African countries was 37.94 with a CI of (37.58%, 38.29%) and 41.45% with a CI (41.09%, 41.80%) respectively. Partner alcohol use contributed to a 2.78% increase in intimate partner violence according to the estimated average treatment on treated values in the treated and control groups were 59.41% and 31.51%, respectively. Ultimately, it was found that among all research participants, the average effect on the population as a whole was 25.33%.

Conclusion

We conclude that partner alcohol use has a direct cause for intimate partner violence. Therefore, controlling partner alcohol consumption can reduce the burden of intimate partner violence.

Introduction

The World Health Organization (WHO) defines intimate partner violence as the deliberate act of an intimate partner or former spouse that results in sexual misconduct, severe physical harm, emotional abuse, or dominating activities [1]. Intimate partner violence is the most prevalent type of violence against women, with major health consequences, and is more likely to occur in homes rather than on street level. Intimate partner violence increases the risk of gynecological, neurological, and stressful problems for women [2]. Alcohol is the most popular beverage in the world and a fluid that includes ethanol. Worldwide commonly men drinking alcohol is associated with numerous misconducts, including violence against their intimate partners [3]. Alcohol’s psychophysiological effects are considered to directly increase the risk of criminalizing IPV in those who consume alcohol [4]. The potential habit of alcohol addiction has made it difficult to determine whether there is a causal relationship between alcohol abuse and IPV [5].

Alcohol consumption is one of the common and well-established risk factors for intimate partner violence. Although different measures have been made to lessen intimate partner violence it remains a significant public health issue that requires additional work to address [6]. It’s unclear how the etiological theories put up to explain the connection between alcohol use and IPV have been tested in earlier studies conducted in low and middle-income countries [5]. Regardless of the consequences of alcohol consumption, some drinkers may purposefully act violently or aggressively toward their spouse in the hopes that their actions will be accepted as they were under the influence of alcohol drink [7].

The prevalence of IPV among reproductive-age women in the world including Africa varies greatly [2]. IPV is a serious public health issue and an attack on women’s human rights; globally, nearly one-third (27%) of women between the ages of 15 and 49 who were in a relationship have encountered sexual or severe physical assault at the expense of their intimate partner [8, 9]. One of the most prevalent forms of violence against women is intimate partner violence [10]. The prevalence of partner alcohol use and intimate partner violence in Sub-Saharan Africa ranges from 3 to 62% and 11–60% respectively. In other studies in Africa, the burden of partner alcohol use was 36.3% with a prevalence of IPV (9.7–25.0%) [11].

There are several factors in the previous observational study which show the association between partner alcohol use and intimate partner violence [12, 13]. The association of factors between alcohol use and intimate partner violence in developing countries was confounded by a wide range of factors that exist at the individual and community level variables [4]. The variables associated with IPV were incorporated: sex of male household head, age, occupation, educational status, marital status, mass media exposure, wealth status, and number of children [12, 14, 15].

World Health Organization suggested that primary prevention strategies aimed at minimizing alcohol-related harm could also potentially minimize IPV even though drinking alcohol can occur without IPV and IPV can occur without alcohol consumption [16]. Different literature evidence that it is very difficult to determine the degree of the associations between substance consumption and intimate partner violence [17]. The exact causal relation is difficult to determine by observational study due to the presence of other associated factors. Besides the variable observed, there are also unobserved variables and biases that prohibit the exact causal relations between alcohol use and intimate partner violence. Propensity Score Matching (PSM) analysis is the best technique to avoid bias through matching partner alcohol use (treatment group) and partner, not alcohol use (control group) among reproductive-age women with similar exposure to intimate partner violence. According to our best knowledge, there are no studies in East Africa that show the effect of partner alcohol use on intimate partner violence among reproductive-age women. Therefore, this study aimed to determine the impact of partner alcohol use on intimate partner violence in East Africa’s recent DHS data with propensity score matching.

Method and material

Study design and area

A community-based cross-sectional study was employed on the recent Demography and Health Survey (DHS) data of 12 East African countries (Burundi, Comoros, Ethiopia, Kenya, Madagascar, Malawi, Mozambique, Rwanda, Tanzania, Uganda, Zambia, and Zimbabwe) from 2016 to 2024. East African countries’ DHS data included fertility, reproductive health, maternal and child health, mortality, nutrition, and self-reported health reports. The DHS incorporated datasets are on men, women, children, births, and households for this survey and we have used the women’s data for this secondary data analysis. Reproductive-age women between the ages of 15 to 49 were included in this secondary data analysis. The total weighted sample size for this study was 72,554 with 1692 clusters. The detailed data can be accessed comprehensively by clicking on the official link [18].

Operational definitions

The revised version of the domestic violence questionnaire module was used. Study participants who have experienced intimate partner violence in their lifetime were considered in this study. The module of questions on GBV was administered following the World Health Organization’s guidelines on the ethical collection of information on gender-based violence (WHO 2001). From a total of nine revised versions of the domestic violence questionnaire scoring one and above were considered as having IPV.

Alcohol use: It was assessed by the simple question of alcohol drinking every day. Study participants whose husband drinks alcohol on an everyday basis and with harmful consequences even though the specific amount and frequency, it is the limitation of the study.

Outcome and treatment variables

Partner alcohol use is measured by a single question; did your husband/partner drink alcohol every day? The dependent variable for this study is intimate partner violence which includes severe physical violence, sexual violence, and emotional violence. It was measured by the following questions and participants who have said yes to at least one question are considered as they have IPV.

Severe physical violence

Ever been kicked or dragged by your husband?

Ever been strangled or burned by a husband?

Ever been threatened with a knife, gun, or another weapon?

Sexual violence

Ever been physically forced to have unwanted sex by your husband?

Ever been forced to do other sexual acts by your husband?

Ever been forced to perform sexual acts respondent didn’t want to?

Emotional violence

Ever been humiliated by your husband?

Ever been threatened with harm by your husband?

Ever been insulted or made to feel bad by your husband?

Data management and statistical analysis

When randomization was not an option, propensity score matching was frequently used to ascertain the effects of treatment in experimental designs. Because of bias caused by an imbalance in observable factors that modifies the causal influence of experience, study participants were randomly assigned to one of the groups. Adjust and rectify group inequality using a balancing score to fix the imbalance between the groups using PSM while confounding factors can be identified. The balance score indicates that the treatment group should not affect the observed variable which is partner alcohol use. After propensity score adjustments for the observed covariates, the difference in outcomes between those who experienced intimate partner violence and those who did not offer an objective measure of the impact of partner alcohol use on intimate partner violence becomes equal. The propensity score which always ranges from 0 to 1 is a conditional likelihood of receiving treatment (partner alcohol use). A higher propensity score indicates that women whose partners drink alcohol. The treatment variables of interest in propensity score matching need to be dichotomous. The imbalance of covariates between the treatment and control groups is assessed using a t-test for continuous factors and a chi-square for categorical components.

Based on the association between the outcome and treatment variables, three separate results were obtained from the observed covariant. However, the only one that can be added is PSM. The likelihood that a woman might have partner alcohol use is reduced to a propensity score for each woman based on the variables selected. A propensity score for each participant is generated with the selected confounders [19,20,21]. PSM approach was used for those women with partner alcohol use that wasn’t distributed randomly between the two groups and might be considerably impacted by both observable and non-observable factors. PSM covariates were incorporated as they have a strong association with partner alcohol use and intimate partner violence, including socio-demographic and behavioral factors. Variables included before PSM (sex of household head, age, education, currently working, wealth status, partner alcohol use, and husband education) had significant differences with a p-value of less than 0.05 with IPV among those who have alcohol user partners and do not have. The variables mentioned above showed no significant difference for IPV while participants with and without partner alcohol use were matched with a p-value of greater than 0.05. This suggests that PSM dramatically reduced the group’s observed variable difference.

The most widely accepted PSM hypotheses are a selection of unobservable variables and common support that have been evaluated statistically and graphically. Throughout the study, the common support option was taken into account to limit the balance of propensity to mothers with treatment (partner alcohol use) whose propensity score for IPV was within the ranges of propensity scores for controls. We tested two types of matching methods: nearest neighbor matching with and without replacement and radius matching with calipers ranging from 0.01 to 0.05. Stata psmatch2 was used to calculate the Average Treatment Effect for treated (ATT), Average Treatment effect on Untreated (ATU), and Average Treatment effect for the whole population (ATE) for the matching technique that produced the most effective matches. Standard supported option was also used to generate higher-quality matches. The basis for assessing the quality of matching was the balance of the variables between the treated and control groups. To determine the degree of matching, the standardized bias before and after matching was calculated. The difference in percentages is used to compute this bias.

The percentage of the square root of the average sample variances in both groups was used to check the percentage difference between the sample means in the matched control and treatment groups. Although, there is no hard and fast rule on the degree of standardized difference to indicate an imbalance variation of less than 10% is considered a low variation. The pseudo-R2 and likelihood ratio tests were used to examine the joint importance of all the covariates from the logit estimation of the conditional treatment probability before and after matching. A sensitivity analysis was used to evaluate the PSM estimations’ reliability [22]. Because the outcome variable was binary, the Mantel-Haenzel (MH) test statistic was used to assess whether the PSM estimates were sensitive to the hidden bias [23]. The gamma coefficient quantifies the unobserved confounding or hidden bias that affects how the treatment is allocated to the treated and control groups. Using the mhbounds STATA command, the gamma value ranges from 1 to 2 with a 0.05 increment.

Results

Study participant descriptive characteristics

A total of 72,554 reproductive-age women aged from 15 to 49 were used from the East African countries DHS data. The prevalence of partner alcohol use and intimate partner violence from East African countries was 37.94 with a CI of (37.58%, 38.29%) and 41.45% with a CI (41.09%, 41.80%) respectively. The characteristics of study participants before matching were described according to partner alcohol use (TableÌý1). From this propensity score matching analysis women’s age, partner age, maternal occupation, media exposure, sex household, wealth index, paternal education, and residence were significantly associated with a p-value of less than 0.05 with partner alcohol use before matching (TableÌý2).

Table 1 Characteristics of study participants included the effect of partner alcohol use on intimate partner violence from East Africa DHS

Table 2 The association between covariates and intimate partner violence

Estimations of propensity score

The estimations of the association’s orientation, power, and significance aligned with the findings of other researchers (TableÌý2). The minimum variability with mean propensity score among the intervention and control groups was 1.44. The range of propensity scores varied from 0.08 to 1.24 which showed the common support assumption was satisfied. Reproductive-age women whose propensity scores fell below the range of common support were dropped from either the treatment or control groups.

Impact of partner alcohol use on intimate partner violence

The unmatched estimate indicates that women who have partner alcohol users are 2.89% more likely to have IPV than women who have not. The nearest neighbor matching had the best matching quality with a caliper width of 0.01. IPV increased by 2.78% as a result of partner alcohol consumption, according to an estimated average treatment on treated values of 59.41% in the treated and 31.51% in the control group. Similarly, the estimated average treatment effect on untreated values in the control group and treated group was 30.46% and 54.24%, respectively. This finding indicated that if the women who hadn’t partner alcohol use had been encountered with partner who uses alcohol the chance of developing IPV would have increased by 23.77%. Ultimately, it was found that among all research participants, the average effect on the population as a whole was 25.33% (TableÌý3).

Table 3 The impacts of partner alcohol use on IPV in East African Countries using PSM method

Quality of matching

Common support

Only one woman was eliminated because of off-support (TableÌý4). The propensity score distributions for both groups are almost identical when plotted on PSM after matching (Fig.Ìý1). The significant overlap between the treatment and control groups’ features validates the common support assumption.

Table 4 Common support

Fig. 1

Propensity score histogram by treatment status (partner alcohol use)

Balancing test

The test’s significance level was established and the t-test was utilized to evaluate the difference between the matched and unmatched pairs. Almost all factors displayed no significant mean difference following matching, despite a significant mean difference across all covariates (TableÌý5). This proved that for every variable in the model, the treated and control groups were appropriately balanced.

Table 5 Performance of the propensity score matching: quality measurements

Standardized bias

The pstest’s mean and median biases considerably lowered once the intervention and control groups were matched. The mean absolute bias in the unpaired sample decreased from 15.5 to 1.1% after the treated and control groups were matched. This is less than the 5% threshold and shows that the model’s quality matching has improved. The median bias decreased from 13.8% in the unmatched to 0.3% after matching (TableÌý5).

Model significance

The overall significance of the model was assessed using the LR and pseudo R2 tests. The pseudo-R2 was less than 0.001 and the LR-chi2 test had become negligible (p = 1.0), suggesting that there was no systematic variation in the covariate distribution between the treated and control groups (TableÌý6).

Table 6 Performance of the propensity score matching: quality measurements

Sensitivity analysis

The Mantel-Haenszel finding indicated that the overestimation of partner alcohol use effect on IPV was not significant at 5% of the significance level. However, the statistical significance of the underestimating of partner alcohol use impact was established at a 5% level of significance. As gamma increases, the probability of underestimating the effect of partner alcohol use on IPV increases, suggesting a reduction in the possibility of heterogeneity due to unobserved factors (TableÌý7).

Table 7 Mantel-Haenszel bounds for sensitivity analysis

Discussion

The main objective of this study was to determine the relationships between lifelong experiences of partner alcohol use and intimate partner violence among reproductive-age women. Secondary data analysis was conducted from the recent East African countries’ DHS data.

Based on PSM approach, partner alcohol use contributed to a 2.78% increase in intimate partner violence. The estimated average treatment on treated values in the treated and control groups were 59.41% and 31.51%, respectively. Comparably, the treated group’s estimated average treatment effect on untreated values was 54.24%, while the control group’s estimated average treatment effect was 30.46%. According to this research, the number of women experiencing IPV would have increased by 23.77% if they had met partner alcohol use instead of none. In the end, it was discovered that the average effect on the population as a whole for all research participants was 25.33%. This finding is comparable with other propensity score matching analyses that show the impact of partner alcohol use on IPV [4].

A significant positive association has been shown in a lot of studies between alcohol use and intimate partner violence. However, because people may misreport their alcohol abuse and because there may be reversed causality from IPV to alcohol abuse, it has been challenging to determine the causal relationship between alcohol abuse and IPV [5]. Additionally, there is a potential endogeneity issue, which suggests that those who are more likely to engage in excessive drinking are also more likely to engage in IPV due to an undetected third factor. Previous studies have demonstrated a correlation between colonization and alcohol consumption as a coping mechanism for being emotional of rage, avoidance, grief other factors [24].

These temporal correlations between frequent alcohol use by partners and IPV may be explained by several factors: Men who drink alcohol often may have poor judgment which makes it harder for them to recognize their fault and violation towards their intimate partners [25]. The effect of an intoxicated partner due to drinking alcohol was a great concern to cause intimate partner violence. Once men have been intoxicated after using alcohol their cognitive function entirely deteriorates which causes intimate partner violence among reproductive-age women in East Africa. The other justification for this association could be the effect of husbands who drink alcohol being easily tempered and aggressive toward their wives [26].

There is a frequent association between alcohol use and incidents of IPV among reproductive-age women. Although the idea that IPV causes alcohol usage cannot be completely ruled out, there is a lack of long-term data to support most previous studies [27]. This study determines the direct causative relation between partner alcohol use and intimate partner violence among reproductive-age women by 2.78%. This study estimates the correlation between alcohol consumption and IPV besides seeking to determine a causal relationship. According to DHS data, intimate partner violence is influenced by the partner’s alcohol consumption among women of reproductive age from East African countries.

Despite the presence of limitations, this study has several advantages. This is the first study to estimate bias through the determination of the causative relation of partner alcohol consumption on IPV into account using propensity score matching in East Africa. Nationally representative DHS data from 12 East African countries with a large sample size of 72,544 served as the foundation for this study and was used with a high response rate. The weakness of this study is the sensitivity of intimate partner violence results under-report their case. The variables that were observed provided the framework of the matching there might be a chance of the occurrence of residual confounding. Additionally, we have used DHS data with cross-sectional research that might have a social desirability and recall bias.

Conclusions and recommendations

Although it has been suggested that treating and preventing alcohol abuse is a good way to prevent IPV this guidance is not implemented widely in East African countries. These results highlight the necessity of using alcohol consumption reduction as a potential target for IPV prevention efforts and as a key correlate of IPV. These results imply that structural, macro-level actions may be able to reduce the causative association of alcohol use on IPV. When taken as a whole, these results emphasize the necessity of assessing multilayer intervention techniques to reduce or mitigate the causative association of alcohol use with intimate partner violence. Focusing on decreasing the partner’s alcohol consumption to mitigate the burden of intimate partner violence is our best recommendation.