1 Background

Alcohol consumption poses a significant global public health challenge, contributing substantially to morbidity and mortality rates each year [1]. This issue is particularly concerning for women, as their tissues exhibit greater vulnerability to the detrimental effects of alcohol even at lower levels of alcohol exposure than men [2, 3]. Consequently, women who consume alcohol are at higher risk of developing various serious health conditions, including liver cirrhosis, early cardiomyopathies, hypertension, and stroke [4]. Moreover, alcohol consumption carries significant social repercussions, extending beyond its health implications leading to adverse effects on personal relationships and overall social integration [5].

Literature shows a correlation between alcohol consumption and various deviant behaviors [6]. For instance, a study conducted in Tanzania revealed that women with partners who consumed alcohol were more likely to experience controlling behaviors [7]. Male intimate partner controlling behavior is categorized as a moderate form of intimate partner violence (IPV) [8], and may serve as a precursor to engaging in highly risky substance-abuse behaviors [9]. While there exists literature highlighting women’s perpetration of controlling behaviors towards men [10, 11], the prevailing trend in research indicates that women are predominantly victims [12, 13].

In sub-Saharan Africa, research conducted by Aboagye et al. [14] demonstrated a positive association between a partner’s alcohol consumption and the likelihood of women experiencing various forms of violence, including physical, emotional, and sexual violence. Similarly, Shubina et al. [7] emphasized that hazardous drinking was linked to the perpetration of multiple forms of IPV, including sexual, physical, emotional, and financial abuse. Moreover, research has shown that IPV commonly co-occurs with substance use disorders like alcoholism, and abusive men who engage in severe alcohol misuse were found to have an increased risk of violent behavior [15,16,17]. Thus, a potential nexus emerges wherein male partner alcohol consumption and controlling behavior not only influence women’s experiences of violence but also interlace with their own alcohol consumption patterns [17]. This association is yet to be substantiated by empirical evidence.

Within the context of Ghana, different studies have reported varied alcohol consumption among women. One study reported a prevalence of 37.6 percent [18] among women of childbearing age, whereas another study limited to pregnant women reported 54.2 percent [19]. Another study conducted in Elmina, Ghana [20] revealed that 48.5 percent of women consumed alcohol. These statistics underscore the high alcohol consumption among women in Ghana. The existing literature on alcohol consumption in Ghana identifies urban residence, higher educational level, lower wealth index and being employed as factors with higher likelihood of consumption [20,21,22]. Kyei-Arthur and Kyei-Gyamfi [20] also argue that Ghanaian women consumed alcohol to forget marital problem, kill loneliness or just for fun. However, none of the existing studies conducted among Ghanaian women have explored the role of partner controlling behaviors. Evidence from sub-Saharan Africa suggests a link between alcohol intake and the perpetuation of IPV and controlling behaviors [16], however, a reverse association is yet to be established. As such, there is a lack of clarity and consensus on whether partner controlling behaviors predict alcohol consumption among women. Therefore, our study aims to extend the current understanding of the determinants of alcohol consumption among women in Ghana by investigating the association between partner controlling behavior and women’s past month alcohol consumption.

2 Methods

2.1 Data source and design

The present analysis utilized data sourced from the 2022 Ghana Demographic and Health Survey (GDHS), specifically drawing upon the individual recode file (IR). The GDHS is part of a broader initiative encompassing 85 low-and-middle-income countries (LMICs). The principal aim of the 2022 GDHS was to provide up-to-date estimates of key demographic and health indicators. To accomplish this objective, a meticulously crafted two-staged sampling strategy was implemented, resulting in a stratified representative sample consisting of 18,450 households distributed across 618 clusters [20]. This method ensured comprehensive representation at the national level, encompassing both urban and rural areas, as well as within each of Ghana’s 16 regions.

During the initial stage, 618 target clusters were identified utilizing a probability proportional to size method, taking into account urban and rural distinctions within each region [23]. Subsequently, an equal probability systematic random sampling approach was utilized to select the required number of clusters in both urban and rural settings. Transitioning to the second stage, subsequent to cluster selection, an exhaustive household listing and map updating process were carried out within all selected clusters, establishing a comprehensive roster of households for each cluster [23]. From each cluster, 30 households were randomly selected for interviews. For further insights into the design of the GDHS, please refer to: https://www.dhsprogram.com/pubs/pdf/FR387/FR387.pdf.

2.2 Study variables

2.2.1 Outcome variable

Our outcome variable was past month alcohol consumption preceding the survey which was assessed through the question, “During the last month, on how many days did you have at least one drink of alcohol?”. The respondents responded by selecting either “Did not have one drink in the last month” or indicating the number of days they consumed at least one drink [24]. The responses were recoded to have all those who responded “did not have one drink” as ‘no alcohol consumption’ and the reported number of days as ‘alcohol consumed’ which was dichotomized as “No = 0” and “Yes = 1” respectively.

2.2.2 Main explanatory variable

The main explanatory variable was partner controlling behavior. This was constituted from the five questions, “Is the husband jealous if his wife talks with other men? Does he accuse his wife of being unfaithful? Does he refuse to permit his wife to meet her female friends? Does he try to limit his wife’s contact with her family? And does he insist on knowing where his wife is?” [25]. The initial responses were yes, no and don’t know. However, we dichotomized the responses into ‘Yes’ and ‘No’. To create the controlling behavior variable, we used the ‘egen’ command in STATA. Specifically, we generated an index by summing the responses to the five questions. Each question was recoded such that ‘Yes’ responses were coded as 1 and ‘No’ responses (including ‘don’t know’) were coded as 0.

2.2.3 Covariates

Nine factors were selected as covariates based on evidence from previous studies [7, 14,15,16,17, 20,21,22]. These factors included the experience of sexual, physical or emotional violence, frequency of internet use (i.e., never, rarely, often), age (15–49 years), residential location (rural and urban), level of education (e.g., no formal education, primary, secondary, higher), exposure to media (yes or no), wealth index (e.g., poorest, poorer, middle, richer, richest).

2.3 Statistical analyses

The sample weight (v005) for women provided in the DHS individual Recode file (IR file) was utilized to generate estimates. This was statistically important for producing accurate and reliable estimates as it accounts for non-response and oversampling of certain population segments in the survey data. The GDHS dataset included observations from 15,014 women aged 15–49 years. However, after dropping missing observations from the dataset, the remaining sample for analysis was 5137. Descriptive statistics were used to present the demographics and distribution of the prevalence of alcohol use across the variables. We then computed Pearson’s chi-square test to assess whether there were significant differences in observed distributions of the explanatory variables. A bivariate logistic regression model was fitted to test the hypothesis that partner controlling behavior significantly predicts alcohol consumption among women. We then fitted a multivariate logistic regression model to account for the effects of the covariates. We employed a backward stepwise selection method. This involved iteratively removing non-significant covariates from the model until only significant ones remained at a significance level of α = 0.05. By adopting this method, we aimed to streamline the model and enhance its interpretability by focusing on the most influential covariates while ensuring statistical rigor. Results from the multivariate logistic regression analysis were presented using adjusted odds ratios (AOR) along with a 95% confidence interval. To accommodate the complex sampling design of the GDHS, the survey design was incorporated by utilizing the complex survey command “svyset” in STATA, which incorporates information regarding strata and primary sampling units (PSUs). This approach ensures that the estimates from the analysis accurately reflect the population characteristics and account for the survey's sampling methodology. All statistical analyses for this study were conducted using STATA version 18 (StataCorp, College Station, TX, USA). The Akaike Information Criterion (AIC) was used to select the best fit model.

3 Results

3.1 Prevalence of alcohol consumption among women

Table 1 shows that 15.05% of women consumed alcohol; 37.31% of the respondent had partners who exhibited controlling behaviors. Alcohol consumption was high among women whose partners exhibited controlling behaviors, with 16.17% consuming alcohol compared to 14.39% among those whose partners did not display such behaviors. High alcohol consumption was observed among women who reported having experienced sexual violence (23.65%), physical violence (22.80%), and emotional violence (18.85%). Additionally, women aged 40–44 years had high proportion of alcohol consumption at 20.27%.

Table 1 Sample distribution and proportion of alcohol consumption across all variables
Full size table

3.2 Association between partner controlling behavior and women’s alcohol consumption

Table 2 shows a significant positive association between partner controlling behaviors and women’s likelihood to consume alcohol (AOR = 1.23; 95% CI 1.05–1.44). This association persisted even after controlling for the covariates, with women who reported experiencing partner controlling behaviors being 1.19 times more likely to engage in alcohol consumption compared to those who did not report such behaviors (AOR = 1.19; 95% CI 1.01–1.41). Additionally, the study identified other significant predictors of alcohol consumption among women, including experiences of sexual violation (AOR = 1.62; 95% CI 1.21–2.15) and richer wealth (AOR = 0.76; 95% CI 0.59–0.97). However, the experience of physical and emotional violence did not show significant associations with alcohol consumption. Other factors such as frequency of internet use, age, education level, exposure to media, and wealth did not exhibit significant associations.

Table 2 Results from the binary logistic regression model
Full size table

4 Discussion

Undoubtedly, alcohol consumption (particularly heavy drinking) is a high risk factor for many adverse health outcomes including obesity [26], cardiovascular diseases [27], cancer [28], and liver diseases [29]. Yet, many women continue to consume more alcohol. Consequently, to promote healthy drinking habits and limit potential adverse health effects, there is a need to understand the determinants of alcohol consumption. In the present study, we were interested in investigating whether there is a significant association between partner controlling behavior and alcohol consumption in the past month. Our findings indicate that 15.05% of the participants consumed alcohol in the past month. The observed prevalence of past month’s alcohol consumption is somewhat similar to Dey et al.’s study [24] which reported a prevalence of 12.1 percent. It is also similar to Tampah-Naah and Amoah’s study [21] which reported a prevalence of 17.5 percent.

Our study supports the hypothesis that there is a significant association between partner controlling behavior and women’s alcohol consumption. Specifically, the study revealed that women whose partners exhibit controlling behaviors had a 19 percent higher likelihood of consuming alcohol in the past month. Plausible explanations for this association could include the psychological stress induced by controlling relationships, which may create an environment of heightened anxiety and emotional turmoil for women. In response to this stress, individuals may turn to maladaptive coping mechanisms such as increased alcohol consumption as a means of temporary relief or escape [30, 31]. Furthermore, the power dynamics inherent in controlling relationships may significantly impact women's autonomy and agency in decision-making processes [32], including those related to alcohol use. Women in such relationships may feel compelled to conform to their partner’s expectations or desires, including engaging in behaviors like alcohol consumption, as a means of maintaining the relationship or avoiding conflict. This imbalance of power and control can profoundly influence women's perceptions of their agency and contribute to patterns of alcohol use.

We found that women who had experienced sexual violence were at a 44 percent higher likelihood of consuming alcohol in the past month. This observation is similar to what has been reported in Haiti [33]. Our result is further corroborated by Firkey et al. [34] whose study shows that female victims of sexual violence are at a higher risk of consuming alcohol. Thus, aligning with Kyei-Arthur and Kyei-Gyamfi [20] also argue that Ghanaian women consumed alcohol to forget marital problems which may include sexual violence. We postulate that the trauma, psychological distress, societal stigma and shame surrounding sexual violence may inhibit women who experience sexual violence from seeking professional help or disclosing their experiences, leading them to rely on alcohol as a form of coping mechanism [35, 36]. We also found an association between inverse association between wealth status and alcohol consumption—an observation that is consistent with extant literature [21, 37, 38]. We posit that women with higher wealth status often have greater access to alternative leisure activities and resources for coping with stress (in this case partner controlling behavior), reducing their reliance on alcohol as a means of escape or relaxation.

4.1 Implications for policy and practice

Our findings highlight a need for initiatives that promote healthy relationship dynamics and prevent partner controlling behaviors, as these were significantly associated with increased alcohol intake. This may involve implementing education programs aimed at fostering equitable and respectful relationships, as well as providing support services for individuals experiencing controlling behaviors within their partnerships. Additionally, interventions targeting sexual violation experiences should be developed and implemented, focusing on trauma-informed approaches to address the underlying psychological distress that may contribute to alcohol use as a coping mechanism. Moreover, healthcare providers should be trained to screen for and address these various predictors of alcohol consumption during routine clinical encounters, offering tailored interventions and referrals to support services as needed.

4.2 Strengths and limitations

The GDHS follows a two-staged sampling methodology that ensures that the sample is representative at both the district and national levels. Hence, our findings are generalizable to women of reproductive age in Ghana. Also, our study addresses a significant knowledge gap not only in Ghana but transcending to the sub-Saharan African region. Nonetheless, there are some limitations. One limitation is the inability to establish a causal relationship between partner controlling behavior and women’s alcohol consumption. This is due to the cross-sectional nature of the GDHS. Additionally, we were limited to only selecting covariates available in the dataset. As such, potential confounders such as stress level, and witnessing alcohol consumption during one’s childhood, among others were not accounted for. Also, the self-reporting of alcohol consumption may mean that there is a potential underreporting or over-reporting of drinking behavior.

5 Conclusion

In conclusion, experiencing partner controlling behavior is associated with higher risk of alcohol consumption. Thus, underscoring the importance of addressing not only individual-level factors but also relational dynamics in understanding and addressing alcohol intake among women. Policy makers and program implementers must prioritize interventions aimed at promoting healthy relationship dynamics and addressing gender-based power imbalances. Furthermore, healthcare providers should be trained to recognize and address partner controlling behaviors and sexual violence as part of comprehensive interventions for reducing alcohol consumption among women.