Household Wealth Inequalities in High Body Mass Index among Women of Childbearing Age: Evidence from the Ghana Demographic and Health Survey

Background Ghana is currently experiencing higher body mass index (BMI), that is overweight and obesity, among reproductive-aged women. However, understanding the role of socioeconomic status in the high BMI among this cohort has not been studied extensively in Ghana and the few existing studies in the country have generated mixed results. This study aims to examine household wealth inequalities in high BMI among Ghanaian women of childbearing age. Methods The 2014 Ghana Demography and Health Survey (GDHS) dataset was analyzed. A univariable and multivariable regression model with a logit link function was specified to ascertain the effect of household wealth inequalities in high BMI among Ghanaian women. Furthermore, concentration index and curve were used to measure the degree of household wealth inequalities in high BMI among reproductive aged women. This study found high BMI prevalence of 35.9 percent with significant household wealth-related inequalities (Concentration index = 0.24, 95%CI (confidence interval): 0.22–0.26). The analysis revealed that high BMI is concentrated among wealthier women. Compared to poorest women, poorer (AOR (Adjusted odds ratio) = 2.18, 95%CI: 1.66–2.85), middle-class (AOR = 4.44, 95%CI: 3.24–6.09), richer (AOR = 7.75, 95%CI: 5.53–10.86) and richest (AOR = 11.03, 95%CI: 8.07–15.06) women were more likely to have high BMI. On top of that, socioeconomic characteristics including age, marital status, and education of reproductive-aged women were significantly associated with high BMI. The research revealed that a woman from a wealthier household had higher likelihood of having high BMI relative to those from a less wealthy household. Also, women who were educated and cohabiting, formerly or currently married had an increased risk of having high BMI. This observation suggests targeted policy interventions and programs that promote healthy body weight to reduce the high BMI prevalence among women of childbearing age in Ghana.


Introduction
High body mass index (BMI) which manifests as overweight and obesity is a global health issue [1,2].
Regrettably, the high BMI has emerged in sub-Saharan Africa, a region which has been plagued with underweight over the years [3,4]. In fact, a study estimated high BMI prevalence between 6.7-44.5% across sub-Saharan Africa [5]. Additionally, the rising level of high BMI has been reported by a study that posited that 1 in 10 west African adults are obese [6]. Furthermore, variation in high BMI between men and women has been observed [7,8]. An earlier study reported that west African women have about three-folds propensity to have high BMI compared to men [6], However, the surging levels of high BMI have both health and economic consequences [9][10][11][12]. For instance, several studies have attributed high BMI as a risk factor for chronic diseases such as diabetes, musculoskeletal disorders, cardiovascular diseases, hypertension among others [13][14][15][16].
Indeed, a study has projected a 27% rise in chronic diseases related mortality in Africa [17]. Besides, the high economic cost associated with high BMI has been evidenced in earlier studies [10,11].
Ghana has observed an elevated prevalence of high BMI [7,8,18]. A previous study in Ghana revealed that high BMI prevalence among adults ranged from 20 to 62 percent [18]. Another study stated that approximately 43% of Ghanaian adults had high BMI [7]. In addition, gender disparities in high BMI have been discerned. A study found that adult Ghanaian women have elevated risk of having high BMI relative to men (21.9% vs 6.0%) [7]. Particularly between 1993 and 2015, overweight (17.9-30.4%) and obesity (7.7%-22.0%) representing high BMI among urban Ghanaian women increased significantly [19]. Moreover, higher BMI-related economic burden and increasing prevalence of non-communicable diseases such as diabetes, hypertension among others have been documented in Ghana [20,21].
Though some studies have been conducted [22,23], the research is not exhaustive. Undoubtedly, it is necessary to unravel the underlying factors fueling the rising high BMI in Ghanaian adults. Though socioeconomic factors have been reported to play an important role in the episode of high BMI elsewhere [19,24] but investigations into these factors in Ghana are not exhaustive and the results of the limited studies are inconsistent. The findings from a systematic review in Ghana show that the majority of the studies did not consider some sociodemographic factors including marital status, religion and many others [7] in spite of their relevance in BMI status [25]. Also, most of the studies that attempted to investigate the socioeconomic factors associated with high BMI in Ghana were confined in a small geographical area [7], which has the potential to affect extrapolation of the findings to the entire Ghanaian population. More importantly, despite the discriminatory higher prevalence of high BMI among women yet just few studies in Ghana focused on only women [26][27][28][29].
These highlighted research gaps and thorough literature review points to an urgent need for a comprehensive study examining the association between socioeconomic status and high BMI among reproductive aged Ghanaian women. Therefore, this study aims to assess household wealth inequalities in high BMI among Ghanaian women of childbearing age. Our thorough literature search revealed that, this study will be the first to employ concentration curve and concentration index to ascertain the degree of household wealth inequalities in high BMI among women of reproductive age in Ghana. The findings from this study are expected to support policy interventions and guide the allocation of resources to help reduce the upsurge of high BMI in Ghana and other countries with similar context.

Study population and data source
The research was carried out in Ghana, a lower-middle-income sub-Saharan African country. This current study used the 2014 Ghana Demographic and Health Survey (GDHS) dataset was conducted by the Ghana Statistical Service, Ghana Health Service and Inner-City Fund (ICF) International [8]. The survey participants were recruited based on two steps stratified sampling technique with details published elsewhere [8]. In total 9,396 women between the ages of 15-49 years were interviewed representing a response rate of about 97 %. General sociodemographic and health-related information about the women were collected using a pretested structured questionnaire after consenting to participate in the GDHS. Among the anthropometric measurements that were taken, Body Mass Index (BMI) was computed for 4,750 women in the GDHS [8].

Outcome variable
Overweight and obesity, an abnormal or unhealthy fat buildup [16,30], have been studied using BMI [3,13,19]. Conventionally, BMI is estimated as the ratio of a person's weight (Kg) against the square of the individual's height (metres) and have mainly been grouped into four: less than 18.5 kg/m2 (underweight), 18.5-24.9 kg/m2 (normal weight), 25.0-29.9kg/m2 (overweight) and at least 30 kg/m2 (obesity) [19,31,32]. As shown in Table 1, the BMI status of women which is the dependent variable was classified into two groups. Overweight and obese women were coded '1' representing high BMI whereas women who were not overweight or obese were coded '0' indicating not high BMI.

Primary independent variable
Household wealth index was classified into the poorest, poorer, middle, richer and richest in the 2014 Ghana Demographic and Health Survey data [8] and was coded in this current study as (1) poorest (2) poorer (3) middle (4) rich (5) richest (Table 1). According to the Ghana Statistical Service, Ghana Health Service and Inner City Fund (ICF) International; this variable was derived from the output of the principal component analysis (PCA) of household factors and assets [33].

Control variables
The independent variables examined in this research were age, religion, marital status, ethnicity, parity, employment status, place of residence, educational level. In this study, maternal age was categorized into three groups: 15-24 years, 25-34 years and 35-49 years. The religious affiliation of women was grouped by traditional /other believers, Muslims, and Christians. Women's marital status was classified as never married, cohabiting, formerly married and currently married. With respect to ethnic background, women were identified as Akan, Northern tribe, Ewe, Ga, and other groups. The employment status of women was categorized as either employed or unemployed. The area women reside were grouped as rural and urban. The educational attainment of women was categorized as uneducated, primary and at least secondary education (Table 1).

Study data analysis and modeling
This study employed a logistic regression model and concentration index after considering sampling weight and clustering in the 2014 GDHS dataset to estimate household wealth inequalities in high BMI among Ghanaian women A univariable association between high BMI and a study predictor that recorded a p-value of less than 0.25 were selected for inclusion in the multivariable model. This lenient p-value threshold was adopted to retain statistically relevant study variables whose effect could be influenced by confounders and effect modifiers [34,35]. Further, the multivariable logistic regression was specified using a manual backward selection approach [34]. Initially, all the selected study variables were included in the adjusted model. The study predictors were excluded from the adjusted model based on importance one at a time until all the variables in the final model have p-value ≤0.05.
Multicollinearity check was done, study predictors with variance inflation factor (VIF) of >2.5 as well as tolerance of < 0.4 were regarded as collinear. All possible two-way interactions among significant study variables in the multivariable model were tested. Also, more than 20% percentage change in the regression coefficient of study variables between unadjusted and adjusted models was deemed confounders. On model diagnostics, smaller Akaike's Information Criterion (AIC) was adjudged a parsimonious model [36]. Also, the area under the receiver-operating characteristic (ROC) curve and Hosmer-Lemeshow goodness of fit tests were checked [37]. Odds ratios and 95% confidence interval were generated from the crude and the adjusted models, and the statistical package used for the analyses was R version 3.6.3.
Further, this study used concentration index to quantify the changes in the wealth inequalities in prevalence of high BMI among reproductive aged women [38]. Concentration index ranges from -1 to +1, a positive concentration index show that high BMI is concentrated among wealthier women while negative concentration signifies aggregation of high BMI among poorer women. The concentration index of zero infers that there is no household wealth related inequality in high BMI among women of childbearing age [38]. The concentration curve was graphed by the cumulative proportion of high BMI among Ghanaian women against the rank of household wealth status. A curve below the diagonal line shows greater concentration of high BMI among affluent women whereas a curve above the diagonal line highlights concentration of high BMI among poorer women. In this study, concentration index and concentration curve of household wealth inequality in high BMI were done using STATA 14 (StataCorp LP, College Station, TX, USA) command 'conindex' [38]. Finally, concentration index of household wealth inequality in high BMI were computed across study variables that were significant in the final model.

Descriptive and univariable model findings
As shown in

Household wealth-associated inequality in high BMI
As depicted in Figure 3, the concentration curve falls below the diagonal line indicating a concentration of high BMI among wealthier women.
From Table 4 Also, a significant difference in household wealth inequality was detected among the women's education groups (p-value=0.003). This study found wealth-related inequalities in high BMI as demonstrated in the concentration curve.

Discussion
The concentration index computed in this research was almost the same as what was estimated in a study conducted among 23 sub-Saharan countries (Concentration index = 0.2285) [5]. This finding supports the literature that high BMI is a pro-rich condition [5]. Furthermore, the findings from this study support the general body of knowledge that wealthier women have a greater risk of having high Also, the marital status of women and high BMI were found to have a significant association in this study. Women who were cohabiting, formerly and currently married had increased risk of having BMI than women who never married. This finding agrees with other studies conducted in Ghana [20,27,39,45]. This worrying trend could be due to the perception that having high BMI body figure is a sign of having a responsible partner and good living [42]. Further, the social roles linked to being in a union can alter dietary consumption patterns and at the same time, these engagements tend to reduce physical activity schedules as posited in a similar study [46].
Concerning education, most studies have reported a significant positive association between high BMI and education [20,28,39,40,47] consistent with the findings from this research. In contrast with this study, a similar study conducted on only urban women in the Greater Accra region found no significant effect of education on high BMI [27]. The probable explanation to the higher odds of high BMI among educated women identified in this research could be that educated women can easily adopt western and sedentary lifestyles due to the ability to access available information through social media among others [48].
Of the study predictors that were investigated, women's religion, ethnicity, place of residence and employment status did not have a significant association with high BMI. A similar study in Nigeria found no significant effect of religion on high BMI [49] as reported in this research. Although some studies have associated ethnicity with high BMI [7,28,45] contrary to the findings of this study, the magnitude and direction of the effect need clarity. Therefore, future studies should concentrate on investigating the cultural practices of the ethnic groups that facilitate or discourage high BMI in women. Besides, though previous studies have shown significant association between place of residence and high BMI [40,47], the insignificant effect identified in this study may be attributed to general penetration of western lifestyle in both urban and rural areas [43] and the perceived ideal body size of women in Ghanaian communities [26]. The influence of women's employment status on high BMI has generated mix results, some authors agree with this study [49] while other similar studies found a significant association between employment status and high BMI [45,47]. Further research that would investigate the employment type of women is needed to assess the effect of sedentary work-related lifestyle on high BMI.
Notwithstanding, this study was had both strengths and limitations. Firstly, this research used a population-based dataset that was large and nationally representative to explore the association between socioeconomic factors and high BMI. On top of that, several potential socio-economic and demographic confounding variables were considered in this present to ensure accurate parameter estimates. Finally, this study is the first to quantify household wealth related inequalities in high BMI among reproductive aged Ghanaian women using concentration index. However, the study findings may be limited because the self-reported variables used in this study could be affected by recall bias.
Also, causality cannot be assumed because a cross-sectional study design was used.

Conclusion
The study analyses found household wealth inequalities in high BMI among Ghanaian women.      Concentration curve of household wealth inequalities in high body mass index (BMI) among Ghanaian women