The double burden of malnutrition is becoming more prevalent among Bangladeshi women. Underweight, overweight, and obesity were examined among women aged 15–49 years using the 2017–2018 Bangladesh Demographic and Health Survey (BDHS). A dataset of 20,127 women aged 15–49 years with complete Body Mass Index (BMI) measurements were extracted and categorized as underweight, normal weight, overweight, and obesity. A multiple logistic regression that adjusts for clustering and sampling weights was used to examine underweight, overweight, and obesity among reproductive age Bangladeshi women. Our analyses revealed that the odds of being overweight and obese were higher among women who completed primary and secondary or more levels of education, rich households, breastfeeding women, and women exposed to media (newspapers and television (TV). Women from the poorest households were significantly more likely to be underweight (AOR = 3.86, 95%CI: 2.94–5.07) than women from richer households. The likelihood of being underweight was higher among women with no schooling, adolescent women, and women not using contraceptives. Conclusions: Overweight and obesity was higher among educated and affluent women while underweight was higher among women from low socioeconomic status, indicating that tailored messages to combat overweight and obesity should target educated and affluent Bangladeshi women while improving nutrition among women from low socioeconomic status.
The 2017–2018 BDHS represents a national survey because it encompasses the whole population living in Bangladeshi non-institutional housing units. The BBS 2011 sampling framework used in the survey included enumeration areas (EAs) of the 2011 Population and Housing Census, which are provided by the Bangladesh Bureau of Statistics (BBS). The Primary Sampling Unit (PSU) (i.e., clusters) for the survey includes an EA of about 120 households on average. Each cluster was considered as a community based on previous studies [2,31]. Bangladesh is made up of eight divisions, including Barishal, Chattogram, Dhaka, Mymensingh, Khulna, Rajshahi, Rangpur, and Sylhet. There are zilas (districts) for each division, and each zila is subdivided further in Upazilas (sub-district). Each urban area of Upazila is split into wards, further divided into Mohallas, whereas each rural area in Upazila is separated into parishes of union (UPs). There are Mouzas in UPs, and all these divisions enable the division into rural and urban areas. Figure 1 presents the sampling procedure. The survey included a stratified two-stage sample of households: 675 EAs with probability proportions to the size of EA were selected in the first phase. In urban areas, there were 250 EAs, and in rural areas, 425 EAs. The sample was drawn by BBS in the first stage, in accordance with the specifications of the DHS team. The selection resulted in 20,250 residential households in accordance with this design. Approximately 20,100 married women aged 15–49 years are expected to complete interviews. The survey report contains details of the sample structure, including the sample framework and the sample implementations [5]. Sampling frame, BDHS2017–2018. Weight measurement using the lightweight scale SECA787 (with digital screens). Heights were measured using an adjustable wood measuring board, designed specifically to provide an accurate reading of 0.1 cm to take the developed countries into account. Through weight and height measurements, their BMI was calculated. The survey also included information on fertility, contraceptive use, maternal and child health, mother’s nutritional status, women’s empowerment, and sociodemographic characteristics. Data for 20,127 women 15 to 49 years of age who were not pregnant have been used following exclusion. The dependent variables were ordered as normal, underweight, overweight, and obesity which was based on the WHO classes for BMI: underweight (<18.50 kg/m2), normal (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), and obesity (≥30.0 kg/m2) (48). In order to ensure the quality, all weight/height were recorded, BMI has been continuously extended in conservative form. The independent variables were the individual-, household- and community-level factors identified in the conceptual framework. The individual-level variables included all relevant attributes of the respondents, including maternal work status, parents’ level of education, mother’s marital status, mother’s age, mother’s literacy status, access to health care services (autonomy to health care), access to the media (newspaper, radio and Television (TV)), and power over family income. Household-level variables consisted of the source of drinking water (improved or unimproved) and household wealth index into five categories [poorest, poorer, middle, richer, and richest]. In creating a wealth index [32], principal component analysis was used to estimate the index weights based on acquired information on various household assets, including ownership of different means of transport and other sustainable domestic goods. This index was divided into five categories, and one of five categories was allocated to each household. Variables at the community level included residence (urban/rural) and geographical area (Barishal, Chattogram, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur, and Sylhet). All analyses were conducted using Stata version 14.1 (Stata Corp 2015, College Station, TX, USA). The command ‘Svy’ has been used for the adaptation of the cluster sampling design, weights, and the Taylor series linearized procedure was used to calculate the standard errors. The dependent variable was always expressed as binary, with number ‘1’ assigned as underweight, overweight, and obesity, while 0 was normal weight. Frequencies or proportions were used to show the prevalence of overweight and obesity and their 95 percent confidence intervals, using descriptive statistics and surveying tabulation. Logistics regression was adjusted using the cluster and survey weights. Multivariable logistic regression analysis was performed to obtain the association of each independent variable with the dependent variable (i.e., underweight, overweight, and obesity), utilizing a normal BMI range as the reference value. Crude and adjusted regression models were built and variables with a pre-specified significance value of <0.2 in the unadjusted model were eligible for inclusion in the final adjusted multivariable models [27]. Association results of multivariable regression analysis were presented by odds ratio (OR) at 95% confidence intervals (CIs). Statistical significance was considered with a p-value < 0.05. Our final model was tested for any co-linearity. The adjusted regression models’ odds ratios and the 95% confidence intervals (CI) were determined.
N/A