Multilevel analysis of urban–rural variations of body weights and individual-level factors among women of childbearing age in nigeria and South Africa: A cross-sectional survey

Background: An unhealthy body weight is an adverse effect of malnutrition associated with morbidity among women of childbearing age. While there is increasing attention being paid to the body weights of children and adolescents in Nigeria and South Africa, a major surge of unhealthy body weight in women has received less attention in both countries despite its predom-inance. The purpose of this study was to explore the prevalence of body weights (underweight, normal, overweight, and obese) and individual-level factors among women of childbearing age by urban–rural variations in Nigeria and South Africa. Methods: This study used the 2018 Nigeria Demographic Health Survey data (n = 41,821) and 2016 South Africa Demographic Health Survey (n = 8514). Bivariate, multilevel, and intracluster correlation coefficient analyses were used to deter-mine individual-level factors associated with body weights across urban–rural variations. Results: The prevalence of being overweight or obese among women was 28.2% and 44.9%, respectively, in South Africa and 20.2% and 11.4% in Nigeria. A majority, 6.8%, of underweight women were rural residents in Nigeria compared to 0.8% in South Africa. The odds of being underweight were higher among women in Nigeria who were unemployed, with regional differences and according to breastfeeding status, while higher odds of being underweight were found among women from poorer households, with differences between provinces and according to cigarette smoking status in South Africa. On the other hand, significant odds of being overweight or obese among women in both Nigeria and South Africa were associated with increasing age, higher education, higher wealth index, weight above average, and traditional/modern contraceptive use. Unhealthy body weights were higher among women in clustering areas in Nigeria who were underweight (intracluster correlation coefficient (ICC = 0.0127), overweight (ICC = 0.0289), and obese (ICC = 0.1040). Similarly, women of childbearing age in clustering areas in South Africa had a lower risk of experiencing underweight (ICC = 0.0102), overweight (ICC = 0.0127), and obesity (ICC = 0.0819). Conclusions: These findings offer a deeper understanding of the close connection between body weights variations and individual factors. Addressing unhealthy body weights among women of childbearing age in Nigeria and South Africa is important in preventing disease burdens associated with body weights in promoting Sustainable Development Goal 3. Strategies for developing preventive sensitization interventions are imperative to extend the perspectives of the clustering effect of body weights on a country level when establishing social and behavioral modifications for body weight concerns in both countries.

Nigeria is the most populous country in Africa with an estimated population of 206 million. Although often pointed out as the “Giant Africa”, owing to its enormous population and economy, it is a multi-national state populated by more than 250 ethnic groups. Well-known with an extensive diversity of cultures, the three major ethnic groups—Hausa-Fulani in the North, Yoruba in the West, and Igbo in the East—comprise over 60% of the total population, and the country is home to Christian, Muslim, and indigenous religions [11]. Administratively, Nigeria is divided into states and 774 local government areas, defined by urban and rural areas. Presently, Nigeria is plagued with Boko Haram conflicts, poverty, malnutrition, diseases, and the burden of youth unemployment. In 2018, the NDHS report on maternal height and body mass index (BMI) showed a prevalence of women aged 15–49 years who were overweight/obese (28%), which is on the rise, compared with women who are underweight (12%) [11]. South Africa, officially called the Republic of South Africa (RSA), has over an estimated 59 million people in the southernmost part of Africa, covering an area of 1,221,037 square kilometers of land. South Africa has three capital cities: executive Pretoria, judicial Bloemfontein, and legislative Cape Town, with Johannesburg as the largest city [12]. South Africa is very racially diverse, where about 80% of South Africans are of Black African ancestry, with the remainder divided among other ancestry groups: European (White South Africans), Asian (Indian South Africans), and multi-racial (colored South Africans). South Africa is a multi-ethnic society with nine provinces, encompassing a wide variety of cultures, languages, and religions [12]. Even though South Africa is the most westernized country on the African continent, with a mixed economy and a relatively high gross domestic product (GDP) per capita, yet poverty and inequality remain widespread, with a major youth unemployment problem, as approximately one-quarter of the population is unemployed and living on less than 1.25 USD per day [23]. Furthermore, South African society continues to face steep challenges such as rising crime rates, ethnic tensions, great disparities in housing and educational opportunities, and the AIDS epidemic. In 2016, the SADHS documented a high prevalence of overweight/obesity among women aged 15–49 years (68%), with comparable findings in urban (68%) and non-urban (66%) areas [12]. The datasets used in this study, the 2016 SADHS and 2018 NDHS [11,12], were combined to maximize the sample size for each study area. In addition to increasing the number of observations, another advantage of combining two different surveys is that it is anticipated that increasing the overall sample size should lead to reduced sampling errors [24]. The 2016 SADHS was the third nationally representative cross-sectional survey in South Africa. The sampling frame of the survey was determined from the list of primary sampling units (PSUs) of the 2011 National Population and Housing Census (NPHC) [12]. It used a two-stage stratified sampling design, where the first stage consisted of 750 PSUs, with 468 in urban areas and 282 in non-urban areas, from a list of residential dwelling units (DUs) generated from the NPHC of South Africa [12]. The second stage of sampling involved a systematic selection of 20 DUs per residential dwelling unit. From the total of 15,292 surveyed households, one in every three households was randomly selected for anthropometric data, and samples were collected from all nine provinces of South Africa. The numbers of urban and non-urban women interviewed in the cross-sectional survey were 4805 and 3709, respectively, giving a total of 8514 women, yielding a response rate of 86% [12]. The survey collected information on various demographic, socio-economic, and health indicators, including individual characteristics and adult nutrition. The 2018 NDHS is the sixth nationally representative cross-sectional survey in Nigeria. The sampling frame of the survey was determined from the list of enumeration areas (EAs) of the 2011 National Population and Housing Census (NPHC) of the Federal Republic of Nigeria [11]. The study used a two-stage stratified sampling design, where the first stage consisted of 1400 EAs, with 580 in urban areas and 820 in rural areas being selected, with a probability proportional to the EA size generated from the NPHC of Nigeria [11]. The second stage of sampling involved an equal probability systematic sampling and selection of 30 households per EA in the household list [11]. From the total of 41,668 surveyed households, one in every three households was randomly selected for anthropometry measurements. Samples were drawn and collected from six zones of Nigeria, and the numbers of urban and rural women interviewed in the cross-sectional survey were 19,163 and 22,658 respectively, giving a total of 41,821 women, yielding a response rate of 99% [11]. The survey collected information on various demographic, socioeconomic, and health indicators, including individual characteristics, nutrition of children and women, and women’s nutritional status. The body weight derived from the body mass index (BMI) was the dependent variable. It was calculated by dividing the body weight in kilograms by height squared (m2). BMI was categorized into four categories: underweight (BMI < 18 kg/m2), normal weight (18 kg/m2 ≤ BMI < 25 kg/m2), overweight (25 kg/m2 ≤ BMI < 30 kg/m2), and obese (BMI ≥ 30 kg/m2), according to the World Health Organization’s (WHO) recommendation [25]. Anthropometric measurements on height and weight were recorded for all women aged 15+ years during home visits by trained field researchers using procedures standardized in survey settings. In South Africa, the women’s weights were measured using Seca 213 portable stadiometers, and height was measured in meters using an adjustable Seca 201 measuring tape [12]. In Nigeria, women’s weight measurements were also taken using Seca scales with a digital display, model number SECA 878U, and height was measured in meters using a Shorr Board® measuring board [11]. Thus, each measurement tool was calibrated to maintain accuracy with precision to the nearest one-tenth. The present study included individual-level factors such as the demographic, socio-economic, and geographical factors as independent variables to explore the body weights of women by urban–rural residence. Therefore, the major independent variables for the study were residence, education, employment status, wealth index, marital status, geographical zone, province, height, and weight. Other independent variables used in the study were identified from previous studies that established their relationship with body weight [26,27,28,29,30,31,32]. These variables included contraceptive method, breastfeeding, living with partner, currently working, and cigarette smoking. The individual-level variables were established on the basis of accepted genetically related importance, data structure, and published studies. The below explanatory variables for Nigeria and South Africa were included in the multi-level analysis model of this study. The independent variables used in this study are described in Table 1. Summary of independent variables measured in this study. Source: Authors’ compilation. Data analyses were conducted and singled out for the study countries on the basis of the socio-demographic factors/variables featured in the 2018 NDHS and 2016 SADHS. The data were weighted for under-sampling and over-sampling errors as per the survey design using the ‘stata svyset’ command before data analyses. All the analyses were based on women’s body weights by urban–rural differences. Subsequently, the analysis of the data involved univariate analysis of the study population characteristics, as well as the prevalence of women’s body weights and prevalence of BMI categories by urban–rural differences. The descriptive statistics reported the frequencies and percentages to summarize the categorical data extracted from the Nigeria and South Africa DHS, while continuous data were measured in averages (±SD). The women’s body weights were measuring using the BMI classification by adopting the World Health Organization’s (WHO) internationally recognized criteria-based BMI: underweight (BMI < 18 kg/m2), normal weight (18 kg/m2 ≤ BMI < 25 kg/m2), overweight (25 kg/m2 ≤ BMI < 30 kg/m2), and obese (BMI ≥ 30 kg/m2). In addition, bivariate analyses of all the independent variables and women’s body weights were carried out using binary logistic regression that reported the adjusted odds ratio (AOR) in order to ascertain if significant associations existed between women’s body weights and individual-level factors. Lastly, multilevel logistic regression analyses (mixed effect) were used to estimate the effect of all the independent variables on the outcome variable. The regression coefficients of the independent variables were unadjusted (U) and expressed as odds ratio (OR) and 95% confidence interval (CI) for body weight categories (underweight, overweight, and obese) using normal weight as the reference category to evaluate predictors of underweight, overweight, and obese in women by urban–rural variations. A null (random intercept only) model was first fitted to evaluate the intracluster correlation coefficient (ICC) to assess the contributions of residence type to each body weight category. The multi-level analysis adjusts for dependency in data owing to variations in communities surveyed, regions/provinces, and other clustered areas. Therefore, adjusting estimates for this dependency is more accurate than measuring within- and between-cluster variations. Women’s households were nested within the urban–rural residence. Hence, the multilevel model was expressed as where In (Pij1−Pij) is the probability of belonging to one of the body weight categories (BMI), β1X1j+β2X2j+…+β15X15j are the model predictors, β0K is the addictive function, γ00 is the grand mean, and u0K is the level 2 random intercept term. Similarly, significant factors in the bivariate analysis were included in the multivariate model when the variables perfectly predicted the outcome (multicollinearity), while those without an observation set in the model were dropped. The final model of each body weight featured fewer significant predictors, which was established on urban–rural differences, and a variable with odds ratio greater than 1.00 implied that the variable increased the likelihood of the outcome (body weight) while the opposite was true when the OR was less than 1.00. Moreover, the intracluster correlation coefficient (ICC) was employed to account for the relatedness of clustered data by comparing the variance within clusters with the variance between clusters [33]. The ICC is expressed as where Sb2 is the between-cluster variance, and S2w is the within-cluster variance in the outcome variable. Only intercept multi-level regression models were used to produce estimates of the ICC [34,35], and the explanatory models were not included in the intercept-only models. Theoretically, as S2w (within-cluster variance) moves toward 0 (zero), ICC gets closer to 1 (one). All statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS) version 14.0 and Stata version 15.0 (StataCorp, College Station City, TX, USA) with the ‘svy’ command to adjust for sampling weights, clustering effects, and stratification; the 95% CI, with a 5% alpha level of significance, was determined. The 2018 Nigeria and the 2016 South Africa Demographic Health Surveys can be downloaded from the website and are free to use by researchers for further analysis. In order to access the data from the Demographic Health Survey (DHS) MEASURE, a written request was submitted to the DHS MACRO and electronic permission was granted to use the dataset for this study; this was received from the ICF in May 2021. The DHS ensured international ethical standards of confidentiality, anonymity and informed consent, and availability of de-identified DHS datasets.

N/A

• Dima Health