Data source
Data of non-institutional residing Bangladeshi adults from the Bangladesh Demographic Health Surveys (BDHS) 2014 and 2017/18 were used for this study. The BDHS collects health and nutritional indicators’ data using a standard questionnaire with 99% response rate on average. Details of the survey questionnaire, sample design, data collection procedure can be found in BDHS 2014 and 2017/18 [5,13].
The BDHSs surveys use two-stage stratified sampling techniques to select primary sampling units (PSUs) and households using probability proportional to their size and an equal probability systematic sampling technique, respectively [5,13]. The enumeration areas (clusters) were taken from the 2011 censuses compiled by the Bangladesh Bureau of Statistics and were considered as the PSUs [5,13]. Children born from January 2009 or later and aged under five years at the time of the survey were considered eligible for height and weight measurements. A total of 7,886 (BDHS 2014) and 8,759 (BDHS 2017/18) children met the eligibility criteria, and 6,610 (BDHS 2014) and 7,357 (BDHS 2017/18) children complete and credible anthropometric and socio-demographic data (Figure 1).
Outcome variables and operational definitions
The primary outcome was coexistence of stunting, wasting and underweight among under-5 children in Bangladesh. A child was considered to be stunted (short stature for age), wasted (dangerously thin) and underweight (underweight for age) if their height-for-age, weight-for-height, and weight-for-age indices were ≤ -2 standard deviations (SDs) of the World Health Organization (WHO) reference population median [14]. Stunting, a cumulative effect of chronic malnutrition indicates the failure to receive adequate nutrition over a long period. Wasting is a form of acute malnutrition resulting from poor dietary intake or frequent infections like diarrhea. Underweight is an indication of overall nutritional health and a composite index of stunting and wasting [15]. Implausible values while estimating child malnutrition was defined based on the WHO 2006 standards flag limits of unitless z-score: stunting: <-6 or >6; wasting: <-5 or >5; and underweight: <-6 or >5 [14]. A child is considered malnourished, if he/she is stunted, if he/she is wasted and if he/she is underweight. For each of the questions/indicators, responses were re-coded dichotomously: 1=malnourished (i.e. stunting, wasting and underweight) and 0 = normal. Thereafter, the responses of all three malnutrition indicators were added which resulted in a score ranging from 0 to 3. The scores were again recategorized as 0 for normal, 1 stands for children with single dimension of malnutrition (either stunting, wasting or underweight), 2 for children with co-occurrence of two forms malnutrition (i.e. either stunting and wasting, stunting and underweight or wasting and underweight) and 3 for children with coexistence of stunting, wasting and underweight.
Independent variables
A selection of socio-demographic variables or risk factors of interest were identified from relevant literature [3,4,8,]. The variables were mother’s education (no education, primary, secondary, higher); mother’s working status (currently not working, currently working); mother’s body mass index (underweight, normal, overweight); children’s age (0-11 month, 12-23 month, 24-35 month, 36-47 month, 48-59 month); sex of child (male, female); birth order (first, second, third, fourth and above); breastfeeding initiation (within 1 hour, after 1 hour); birth weight (normal/average, small, not weighted); watching television (not at all/do not know, less than once a week, at least once a week); wealth index (poorest, poorer, middle, richer, richest), place of residence (urban, rural).
In low income countries, babies are often born at home without proper measurement of birth weight. Actual weight at birth was reported for less than 50% cases [15]. Therefore, all DHS in developing countries retrospectively collect information on baby’s size at birth based on mother’s perception as proxy of birth weight by asking the question “was the newborn very large, larger than average, average, smaller than average or very small?” Approximately 75% mothers can correctly report their baby's size at birth, therefore mother’s recall is a valid proxy measure of birth weight [16,17,18]. The wealth index or socio-economic status was constructed using information about household assets that were collected in BDHSs. The data on household assets included ownership of durable goods (e.g. televisions and bicycles) and dwelling characteristics (e.g. source of drinking water, sanitation facilities, and construction materials). Principal component analysis was performed to assign individual household wealth scores. These weighted values were then summed and rescaled to range from 0-1, and each household was assigned into quintiles: the first quintile: poorest; the second quintile: poorer; the third quintile: middle class; the fourth quintile: richer and the fifth quintile: richest [5].
Statistical analysis
Descriptive statistics were used to describe socio-demographic characteristics. The prevalence of coexistence of stunting, wasting and underweight was estimated using Chi-square test. Prevalence estimates considered the complex survey design and sampling weights. In all analyses, the significance level was set at P<0.05 (2-tailed). Adjusted models were developed to analyze the appropriate binary value for the adverse nutritional outcome of children under-5 (i.e. coexistence of stunting, wasting and underweight). All independent variables except those were found insignificant in the bivariate analysis (Chi-square test) were simultaneously entered into the negative binomial regression models for adjustment. A negative binomial regression model was used due to unequal dispersion property, i.e. mean ≠ variance and for the occurrence of rare cases (<10%). The strength of associations was assessed using incidence rate ratios (IRR). 95% confidence intervals (CIs) were used for significance testing. All statistical analyses were performed using Stata version 14.2 and sample weighting based on the complex design of the BDHSs was considered.