Risk Factors of Coexisting Form of Undernutrition Among Under‐5 Children in Bangladesh: Evidence From BDHS 2022 Data in a Bayesian Setting

ABSTRACT

Background

Children in underdeveloped countries frequently suffer from malnutrition, which can have both short‐term and long‐term consequences, including cognitive impairment (i.e., stunting), underweight, wasting and ultimately mortality. It is necessary to understand the underlying causes and risk factors associated with this burden. Therefore, this study aims to investigate how risk factors affect coexisting forms of undernutrition (CFU) among Bangladeshi children under the age of 5, with the aim of suggesting possible strategies to enhance nutritional outcomes.

Methods

This study is based on secondary data extracted from the BDHS 2022 dataset, consisting of 4085 children aged under 5. Descriptive statistics, the chi‐square test, the Boruta algorithm and logistic regression are used in a Bayesian setting.

Results

Findings revealed that the prevalence of CFU is 19.44%, and the majority of them were from the Sylhet division. Children from lower wealth indices and those living in urban areas are at higher risk of CFU. Children of uneducated mothers have 3.4 times higher odds (95% CrI: 2.352, 5.001) of CFU than those of higher educated mothers; children of underweight mothers have 1.6 times higher odds (95% CrI: 1.338, 2.092); and children of overweight or obese mothers have 0.6 times higher odds (95% CrI: 0.532, 0.794) than children of mothers with a normal weight. The odds of having CFU are 1.2 and 1.4 times greater (95% CrI: 0.953, 1.553 and 1.103, 1.790) for children from the poorer and poorest households compared to children from middle wealth families and 0.8 and 1.7 times higher (95% CrI: 0.600, 1.145 and 1.405, 2.191) for children whose birth order is at least six and who are older than 23.9 months. A child's nutritional health is also influenced by a number of other important factors, including the child's age and birth order, the mother's age at first birth, breastfeeding status, TV access, toilet facilities, divisions and residence.

Conclusion

The authors suggest that, in addition to improving mothers' health and other socio‐demographic factors, targeted interventions should be developed to reduce the prevalence of CFU.

Summary

• Coexisting forms of undernutrition affect a substantial proportion (19.44%) of under‐5 children in Bangladesh, with notable geographic concentration in the Sylhet division, indicating persistent regional inequalities.

• Children of uneducated mothers have markedly higher odds of CFU, whereas maternal undernutrition significantly increases risk, highlighting the intergenerational nature of malnutrition.

• Children from poorer and poorest households are significantly more likely to experience CFU, reinforcing the role of poverty as a key structural driver of undernutrition.

• Higher birth order, older child age, breastfeeding practices, sanitation facilities, media access and maternal age at first birth are important predictors, suggesting multiple overlapping risk pathways.

1. Introduction

Malnutrition is an imbalance of essential nutrients, whether in deficiency or excess, that adversely affects body composition, physiological functions and overall health outcomes (Allen and Saunders 2023). It is a prevalent issue among children in developing countries, leading to both immediate and long‐term effects, including cognitive impairment (i.e., stunting), underweight, wasting and eventually death. Malnutrition can be classified into three main types: undernutrition (stunting, wasting and underweight), overnutrition (overweight or obesity) and micronutrient‐related malnutrition (resulting from deficiencies in essential vitamins and minerals) (World Health Organization, n.d.). Undernutrition is a type of malnutrition that results from inadequate intake of nutrients, which are essential for healthy growth (Tariku et al., 2026). It arises from inadequate intake of macronutrients and micronutrients. This may be due to insufficient food intake (undernourishment), poor nutrient absorption or ineffective biological utilization of consumed nutrients, often caused by recurring infections. It is assessed using anthropometric indicators that include being underweight for one's age, having stunted growth (being shorter than expected for one's age), experiencing severe thinness relative to height (wasting) and suffering from deficiencies in essential vitamins and minerals (micronutrient malnutrition), which impact both health status and development outcomes (Maleta 2007). The fact that these types of undernutrition frequently coexist makes it difficult to pinpoint the causes of each one. The coexistence of several types of undernutrition in children, which is sometimes called the ‘coexisting forms of undernutrition (CFU)’, necessitates a sophisticated comprehension of how socio‐economic, environmental and health‐related factors interact. These coexisting conditions typically include stunting (chronic undernutrition), wasting (acute undernutrition) and micronutrient deficiencies. This co‐occurrence is increasingly recognized as a major public health challenge, especially in developing and underdeveloped countries. In some literature, undernutrition is used synonymously with malnutrition. However, malnutrition encompasses both undernutrition and overnutrition, making it challenging to distinguish between these concepts when using the term alone.

Understanding and identifying CFU is crucial for effective nutritional interventions, as children experience multiple forms of undernutrition. The Composite Index of Anthropometric Failure (CIAF) is a tool developed to assess the overall burden of undernutrition by categorizing children based on combinations of stunting, wasting and underweight (Arora et al. 2025). This CIAF is commonly used to estimate the overall prevalence of undernutrition by aggregating all forms of anthropometric deficits. CIAF groups together children with single forms of undernutrition (stunting, wasting or underweight) and those with multiple concurrent failures, thereby limiting the ability to distinguish and analyse children experiencing multiple concurrent failures. The primary limitation of CIAF is that it does not differentiate between single and multiple failures; it combines children with relatively less severe conditions and those with compounded nutritional deficits into a single category. Such aggregation may obscure the identification of high‐risk groups, as children with multiple concurrent failures are known to experience substantially higher risk of morbidity and mortality compared to those with isolated conditions (Nandy and Svedberg 2012). A binary variable, CFU, is constructed for this study, where only children experiencing two or more simultaneous conditions (stunting, wasting and/or underweight) are classified as ‘yes’ and all others, including those with no or only one form of undernutrition, are classified as ‘no’. By focusing exclusively on overlapping anthropometric failures, this approach enables a more targeted assessment of risk factors associated with severe and compounded undernutrition. This approach aligns with the objective of this study of finding factors associated with multiple nutritional risks rather than overall prevalence, allowing a clearer identification of high‐risk subgroups for targeted interventions. The coexistence of undernutrition and overnutrition in the same population throughout the life course is known as double burden of malnutrition (DBM) (Seferidi et al. 2025), whereas the prevalence of multiple anthropometric failures (stunting, wasting and underweight) in the same child is known as CFU. This distinction is essential as the primary objective of this study is linked to CFU's emphasis on compounded forms of undernutrition.

Undernutrition remains a critical issue for children under 5 years old. In 2023, approximately 4.8 million children under 5 died, with nearly half of the deaths linked to undernutrition, according to UNICEF and WHO (World Health Organization, n.d.; Alkema and New 2014). Globally in 2022, 149 million children under 5 were estimated to be stunted (too short for age), and 45 million were estimated to be wasted (too thin for height). A nationally representative study in Bangladesh found that approximately 38.2% of children under 5 suffer from undernutrition, with 19.3% experiencing multiple concurrent forms and 18.9% affected by a single form (Chowdhury et al. 2021). Undernutrition predominantly affects low‐ and middle‐income countries, causing long‐term developmental, economic, social and health challenges for individuals, families, communities and nations (World Health Organization, n.d.). The coexistence of undernutrition results from a complex interplay of socio‐economic, demographic and environmental factors (Black et al. 2013). Previous research utilizing demographic and health survey data has identified various socio‐demographic and socio‐economic factors as key determinants influencing child malnutrition (Abdulla et al. 2023; M. M. Hossain, Abdulla, and Rahman 2023; Rahman et al. 2022; Panda et al. 2020). Several studies have highlighted that multiple factors are linked to CFU, such as a child's age, sex, birth order, birth size, diet, parental education, maternal age, occupation, family size, socio‐economic status and access to clean water and sanitation facilities (Worku et al. 2025; Manjong et al. 2021; Islam and Biswas 2020; Farah et al. 2021; Roba et al. 2021).

The majority of earlier research employed various statistical models to identify significant factors associated with each of the three malnutrition indicators. The prevalence, trends and contributing variables of coexisting forms of malnutrition in newborns, babies and children were also examined in earlier studies (Khura et al. 2023; Khaliq et al. 2021; Khaliq et al. 2022). In the context of Bangladesh, there is a dearth of literature on the factors influencing the prevalence of CFU, based on the most recent nationwide Bangladesh Demographic and Health Survey 2022 (BDHS 2022) data. Moreover, most previous studies identified risk factors using the frequentist approach. To fill this research gap, any one of the three indicators of CFU—wasted, underweight and stunted—was taken into consideration in a study. Therefore, investigating important socio‐demographic, economic, maternal and child health determinants affecting the presence of CFU under age 5 in Bangladesh is essential in the Bayesian setting. Thus, the authors aim to explore the association between CFU and child, maternal and household characteristics. This study also highlights regional inequalities by including division‐wise undernutrition rates of children under 5 in Bangladesh. This approach allows us to evaluate the efficacy of current nutritional interventions and to provide evidence‐based policy recommendations to enhance child nutrition. The authors believe that the findings of this research will help clarify the multifaceted impact of undernutrition on child growth and development, providing insights into risk factors and potential interventions to improve nutritional outcomes.

2. Methods and Materials

2.1. Data Source

This study considered a secondary dataset extracted from a country‐representative cross‐sectional survey, the BDHS 2022. It covers all enumeration areas (EAs) nationwide, detailing their locations, residence types (city corporation, non‐city corporation or rural) and estimated household counts, with mapped geographic boundaries. The survey followed a two‐stage stratified sampling method. First, 675 EAs (237 urban and 438 rural) were selected based on size. Mitra and Associates then conducted household listings to create a sampling frame. In the second stage, 45 households per EA were systematically chosen to ensure reliable demographic and health estimates. Eligible women were surveyed on their background and reproductive history. Thirty households per EA received a long questionnaire, and 15 were selected for biomarker measurements, including height and weight for children under 5 and for women aged 15–49. A total of 30 149 households were chosen from the selected EAs. Of these, 30 018 households completed the survey, yielding a high response rate. From these surveyed households, data were collected on 8784 children under the age of 5. After eliminating missing data, a final dataset of 4085 observations remained with an unweighted distribution of 2094 males and 1991 females. After applying statistical adjustments for representativeness, the weighted dataset included 2082 males and 2003 females. The weighted sample accounts for potential biases introduced during data collection, ensuring the final dataset accurately reflects our target population. The listing was conducted by 74 field staff, supervised by 12 quality control officers, in two phases from May to July 2022, after training in April (Mitra and Associates 2022).

2.2. Target Variable

CFU is the main outcome variable in this study and is generated from three anthropometric variables. The following z‐scores were calculated by entering these anthropometric measurements, the child's age and sex into the WHO 2006 child growth standards: Stunting, or low height‐for‐age $Z$ (HAZ) score, is a measure of chronic malnutrition in children. Wasting (low weight‐for‐height $Z$ [WHZ] score) is a metric for excessive nutrient insufficiency based on weight for height. Underweight (low weight‐for‐age $Z$ [WAZ] score) is a composite measure of both overweight and chronic illnesses based on weight for age (Sumon et al. 2023). Here, the $Z$ score for the ith child ($Z_i$) is defined as $Z_i=\frac{X_i-\mu }{\sigma }$, where $X_i$ represents underweight, stunting and wasting of the ith child and $\mu$ and $\sigma$ indicate the median and standard deviation (SD), respectively (Asmare and Agmas 2022). Stunted, wasted and underweight statuses were recoded into the following dichotomies once each child's $Z_i$ was determined:

This CFU indicates the presence of more than one of these three anthropometric failures in a child at the same population, based on WHO growth standards. This study recodes a binary variable indicating whether a child experienced multiple concurrent forms of undernutrition. Thus, the CFU is as follows:

2.3. Control Variable

This study investigates numerous economic and socio‐demographic factors, including wealth index (poorest, poorer, middle, richer and richest), TV access (yes/no), residence type (urban/rural), water source and toilet facility (improved/unimproved). Some of the factors include maternal information, such as maternal educational status (no education, primary, secondary and higher), working status (employed or not), age at first birth (15–19, 20–24, 25–34 and 35–49), pregnancy status and breastfeeding status (yes and no). Some of the attributes include the child's gender (male or female), birth order (1, 2–3, 4 or more), twin birth status (single birth, first of multiple births and second of multiple births), age in months (below 6, 6–23.9 and above 23.9) and maternal BMI classified using WHO standards: underweight (< 18.5 kg/m²), normal weight (18.5–24.99 kg/m²) and overweight/obese (≥ 25 kg/m²). Additionally, divisions (Barisal, Chattogram, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur and Sylhet) are considered essential for detecting regional differences. Several factors, including self‐consideration, appearance in the BDHS 2022 dataset and pertinent research, influence the selection of covariates used in this investigation.

2.4. Statistical Analysis

Descriptive statistics, including percentage and frequency, the chi‐square test and a Bayesian logistic regression model, are used in this study. Descriptive statistics (frequency and percentage) were used to characterize the distributions of the explanatory variables, whereas a chi‐square test and the Boruta algorithm were used as bivariate analysis methods to identify influential factors for CFU.

Boruta is a random forest classifier‐based feature selection technique. The Boruta feature selection algorithm's objective is to choose the set of characteristics that are most pertinent to the dependent variable rather than a specific model, in contrast to the objective of a general feature selection algorithm. In order to account for potential interactions among variables (Zhou et al. 2023), Boruta tests data in a multivariate context, whereas chi‐square tests data in a bivariate setting. Combining results from these two methods allows for more robust identification of relevant predictors for further analysis. Based on the Boruta feature selection algorithm and prior literature, variables such as mother age at first birth, TV access, home type and water source were excluded from the final model, despite having significant outcomes in chi‐square tests. Final variable selection was based on both statistical significance and prior literature relevance, which balances domain expertise and data‐driven selection.

A frequentist multivariate logistic regression analysis was conducted to identify the most significant predictors of CFU. Additionally, a Sankey plot was used to visualize the interrelationship among selected variables (Rahman et al. 2024; M. M. Hossain 2025). The data were statistically analysed using R (Version 4.3.2).

2.5. Bayesian Logistic Regression

The Bayesian framework includes three fundamental components for parameter estimation: the prior distribution, the likelihood function and the posterior distribution (provided in the Supporting Information). Bayes' rule formally combines these three components, as follows:

$$\mathit{\text{Posterior distribution}}=\mathit{\text{Prior distribution}}\times \mathit{\text{Likelihood function}}.$$

Given the chance of success (which in logistic regression varies from one subject to another, depending on their variables), the subject's likelihood contribution is binomial:

$$\mathit{\text{Likelihood}}=\pi {\left(x_i\right)}^{y_i}{\left(1-\pi \left(x_i\right)\right)}^{\left(1-y_i\right)},$$

where $\pi \left(x_i\right)$ denotes the likelihood of the event for a subject with a vector of covariates, and indicates whether the event is present or absent for that subject. Based on the classical logistic regression, $\pi \left(x_i\right)$ is given by

$$\pi \left(x_i\right)=\frac{e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}{1+e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}.$$

Given that individual subjects are believed to be independent from one another, the probability function across a data collection of subjects is as follows (Acquah 2013):

$$\mathit{\text{Likelihood}}=\prod _{i=1}^n{\left(\frac{e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}{1+e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}\right)}^{y_i}{\left(1-\frac{e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}{1+e^{{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_p{X}_p}}\right)}^{\left(1-y_i\right)}.$$

Bayesian analysis combines data from the likelihood function with prior knowledge about unknown parameters, which may come from other datasets or the modeller's intuition. This study utilized two forms of prior information: (a) a flat/non‐informative prior and (b) an informative prior (derived from previous BDHS survey data). The most frequent prior choice for logistic regression parameters is a normal distribution. For informative prior specification, this study utilized a previous survey dataset of BDHS 2017/2018 (i.e., Sumon et al. 2023; M. I. Hossain, Rahman, et al. 2023) and was supplemented with BDHS 2014 for selected variables (Sheikh et al. 2020). These datasets are comparable with similar sample methodology, population structure and variable definitions (e.g., maternal education, maternal BMI, wealth index and breastfeeding status) with the 2022 survey. Categorical variables were included using appropriate reference groups (e.g., higher education, middle wealth index and normal weight in BMI). A logistic regression model was fitted using maximum likelihood estimation to the earlier datasets. The normal prior distributions for each parameter were efficiently computed using the parametric bootstrap (with repeated sampling). The popular resampling technique was applied to obtain empirical distributions of the estimated coefficients for estimating summary statistics (e.g., mean or standard errors) of a population from a dataset sampled with replacement. The resulting coefficients were used as prior means, and standard errors were used as SDs.

To assess robustness, a sensitivity check was conducted by using weakly informative prior (Normal (0, 10)) and flat (non‐informative) priors. The resulting estimates were substantively similar (Table S1) in magnitude and direction, which shows that the findings are strong enough to hold up to prior specification. This indicates that the posterior estimates were primarily driven by the 2022 data and are robust to prior assumptions. This study uses informative prior Bayesian estimation to improve estimation stability, while incorporating prior evidence from previous BDHS datasets. The Bayesian logistic regression was implemented using the brms package in R.

Markov chain Monte Carlo (MCMC) is the most commonly used method for modelling a general posterior distribution (Anjullo and Haile 2018). This study used MCMC with Gibbs sampling to draw posterior samples. The MCMC method achieved convergence in four chains of 4000 iterations, including 1000 warm‐up (burn‐in) iterations. Once converged, its components can be viewed as a sample from the target posterior distribution. Convergence was assessed using trace plots and the potential scale reduction factor (R‐hat). The accuracy of posterior estimates is assessed using the Markov chain (MC) error, with simulations continuing until the error is less than 5% of the sample SD (M. I. Hossain, Rahman, et al. 2023; Gebrie 2021).

Throughout the entire analysis, sampling weights provided by BDHS were incorporated to correct for differential selection probabilities and reflect the total Bangladeshi population.

3. Results

Figure 1 represents the proportion of CFU with respect to the maternal BMI category. We can observe that across different BMI categories, the proportion of children with CFU varies. Children of normal‐weight mothers suffer CFU at a rate of 20.08%, whereas children of overweight or obese mothers experience CFU at a significantly lower rate of 12.58%. An additional CFU affects 28.79% of children born to underweight mothers.

FIGURE 1.

Association between coexisting forms of undernutrition and maternal BMI.

This study analyses the association between socio‐demographic and health‐related factors with CFU in children, using the chi‐square test (p ≤ 0.05 or 0.01 for significance). Findings presented in Table 1 depict that maternal education, maternal age at first birth, child's age, birth order, wealth index, breastfeeding status, place of residence, TV access, toilet facilities, geographical region and maternal BMI are the significant factors of CFU. It is observed that the CFU prevalence is highest among children of mothers with no education (39.42%) and decreases with higher education levels (secondary: 17.37%, higher: 10.37%), indicating that maternal education plays a crucial role in reducing child undernutrition. Teenage mothers (15–19 years) have the highest CFU prevalence (20.63%), whereas mothers aged 35–49 years have the lowest (5.97%). Mothers aged 25–34 years (13.97%) have lower CFU rates than those aged 20–24 years (17.92%), suggesting that delaying first births improves child nutritional outcomes. Undernutrition prevalence increases with age. It is lowest in infants (0–5.9 months: 10.65%) but rises among children aged 6–23.9 months (16.25%) and peaks at 24+ months (21.86%), likely due to food insecurity and cumulative nutritional deficiencies. Firstborn children have the lowest undernutrition prevalence (16.33%), whereas children of birth order four or higher face the highest (28.4%) due to resource dilution and economic constraints. CFU is most prevalent among children from the poorest households (28.7%) and decreases as economic status improves, reaching the lowest in the richest households (11.39%).

TABLE 1.

Percentage distribution and chi‐square test among factors associated with CFU of children under 5 years old.

Variables	Labels	n (%)	CFU status, n (%)		p of χ 2
			Yes	No
Maternal educational status	No education	251.657 (6.16)	99.203 (39.42)	152.453 (60.58)	< 0.001**
	Primary	887.333 (21.72)	210.741 (23.75)	676.591 (76.25)
	Secondary	2223.200 (54.42)	386.231 (17.37)	1836.969 (82.63)
	Higher	723.191 (17.7)	74.988 (10.37)	648.202 (89.63)
Twin birth	Single birth	4031.353 (98.68)	757.862 (18.8%)	3273.491 (81.2)	0.483
	First of multiple births	27.181 (0.67)	5.937 (21.84)	21.243 (78.16)
	Second of multiple births	26.848 (0.66)	7.365 (27.44)	19.482 (72.56)
Working status	Yes	1036.598 (25.37)	194.437 (18.76)	842.160 (81.24)	0.946
	No	3048.784 (74.63)	576.727 (18.92)	2472.056 (81.08)
Maternal age at first birth	15–19	1933.129 (47.32)	398.791 (20.63)	1534.338 (79.37)	0.012*
	20–24	1836.168 (44.94)	329.082 (17.92)	1507.085 (82.08)
	25–34	305.171 (7.47)	42.640 (13.97)	262.530 (86.03)
	35–49	10.913 (0.27)	0.651 (5.97)	10.262 (94.03)
Child age (months)	Below 6	442.772 (10.84)	47.133 (10.65)	395.639 (89.35)	< 0.001**
	6–23.9	1287.864 (31.52)	209.233 (16.25)	1078.631 (83.75)
	Above 23.9	2354.745 (57.64)	514.798 (21.86)	1839.947 (78.14)
Birth order	1	1547.877 (37.89)	252.708 (16.33)	1295.168 (83.67)	< 0.001**
	2–3	2139.714 (52.37)	405.475 (18.95)	1734.239 (81.05)
	4 or more	397.7918 (9.74)	112.981 (28.4)	284.810 (71.6)
Recent fever	Yes	1272.932 (31.16)	249.509 (19.6)	1023.422 (80.4)	0.060
	No	2804.643 (68.65)	517.773 (18.46)	2286.870 (81.54)
	Do not know	7.807 (0.19)	3.882 (49.73)	3.924 (50.27)
Pregnancy status	Yes	195.983 (4.8)	40.393 (20.61)	155.589 (79.39)	0.587
	No	3889.399 (95.2)	730.771 (18.79)	3158.628 (81.21)
Wealth index	Poorest	845.640 (20.7)	242.738 (28.7)	602.902 (71.3)	< 0.001**
	Poorer	830.305 (20.32)	185.961 (22.4)	644.344 (77.6)
	Middle	851.705 (20.85)	156.213 (18.34)	695.492 (81.66)
	Richer	790.240 (19.34)	98.833 (12.51)	691.406 (87.49)
	Richest	767.490 (18.79)	87.418 (11.39)	680.071 (88.61)
Breastfeeding status	Yes	2323.710 (56.88)	407.055 (17.52)	1916.655 (82.48)	0.012*
	No	1761.672 (43.12)	364.110 (20.67)	1397.561 (79.33)
Child gender	Male	2082.200 (50.97)	386.616 (18.57)	1695.583 (81.43)	0.636
	Female	2003.182 (49.03)	384.549 (19.2)	1618.633 (80.8)
Residence type	Urban	1055.067 (25.83)	170.385 (16.15)	884.681 (83.85)	0.009**
	Rural	3030.315 (74.17)	600.779 (19.83)	2429.535 (80.17)
Water source	Improved	3662.423 (89.65)	694.127 (18.95)	2968.295 (81.05)	0.763
	Unimproved	422.959 (10.35)	77.037 (18.21)	345.922 (81.79)
TV access	Yes	1818.033 (44.5)	282.242 (15.52)	1535.791 (84.48)	< 0.001**
	No	2267.349 (55.5)	488.923 (21.56)	1778.426 (78.44)
Toilet facility	Improved	1240.285 (30.36)	198.841 (16.03)	1041.444 (83.97)	0.002**
	Unimproved	2845.097 (69.64)	572.324 (20.12)	2272.773 (79.88)
Divisions	Barisal	282.146 (6.91)	61.731 (21.88)	220.415 (78.12)	< 0.001**
	Chattogram	908.193 (22.23)	182.791 (20.13)	725.402 (79.87)
	Dhaka	986.055 (24.14)	143.550 (14.56)	842.504 (85.44)
	Khulna	405.997 (9.94)	63.918 (15.74)	342.079 (84.26)
	Mymensingh	349.197 (8.55)	77.165 (22.1)	272.031 (77.9)
	Rajshahi	410.980 (10.06)	71.780 (17.47)	339.200 (82.53)
	Rangpur	440.843 (10.79)	88.379 (20.05)	352.464 (79.95)
	Sylhet	301.967 (7.39)	81.848 (27.1)	220.119 (72.9)
Maternal BMI	Underweight (below 18.50)	518.681 (12.7)	149.352 (28.79)	369.328 (71.21)	< 0.001**
	Normal weight (18.50–24.99)	2306.507 (56.46)	463.230 (20.08)	1843.277 (79.92)
	Overweight/obesity (above 24)	1260.194 (30.85)	158.582 (12.58)	1101.611 (87.42)

^*Significant at a 5% level of significance (p < 0.05).

^**Significant at a 1% level of significance (p < 0.01).

Breastfed children have lower CFU prevalence (17.52%) than non‐breastfed children (20.67%), emphasizing the importance of exclusive breastfeeding for optimal child nutrition. Undernutrition is more common in rural areas (19.83%) than in urban areas (16.15%), possibly due to limited access to healthcare, sanitation and maternal education. Children in households without TV access (21.56%) have higher undernutrition rates than those with access (18.21%). Similarly, children in households with unimproved toilets (21.56%) have a higher CFU prevalence than those in households with improved facilities (16.03%), highlighting the role of hygiene and awareness in child nutrition. CFU prevalence varies by region, with the highest rates in Barisal (21.88%) and Mymensingh (22.1%) and the lowest in Sylhet (7.39%), suggesting regional disparities in healthcare access and food security (Table 1).

Thin vertical rectangles in Figure 2 indicate each step in the socio‐demographic and undernutrition coexistence process, a road that contains proportionate information. This Sankey diagram shows how the wealth index contributes to the CFU, in contrast to mothers in poorer families (5.1%), who are more likely to have primary (5.4%) or no education (1.5%) (lower education), richer families' mothers (4.8%) typically have secondary (13.6%) or higher education (4.4%). With greater education levels connected to mothers who are overweight or obese (7.7%) and lower education levels linked to underweight mothers (3.2%), education seems to have an impact on BMI. Children from normal‐weight and overweight/obese mothers are more likely to fall into the ‘no undernutrition’ category, whereas children from underweight mothers are most likely to suffer from undernutrition (Figure 2).

FIGURE 2.

Sankey diagram of the relationship between wealth index, maternal educational status, maternal BMI and coexisting forms of child undernutrition.

In logistic regression, the significant risk factors of CFU are included. These risk factors were identified by the chi‐square test and the Boruta algorithm. The Boruta algorithm findings are presented in Figure S1 and suggest that mother's educational status, wealth index, birth order, maternal BMI, breastfeeding status and child age are risk factors for CFU, consistent with the chi‐square test results.

The results of the logistic regression analysis in the Bayesian framework are presented in Table 2. Here, the 95% credible interval indicates a 95% probability that the true parameter lies within this reported interval. Findings revealed that children's maternal education has a substantial impact on undernutrition, with those whose mothers have no education having 3.4 times higher odds than those with higher education. Again, across various wealth indices, the poorest and richest have a greater effect than the middle wealth index. Compared to the firstborn child, children with birth orders of at least four have 1.5 times higher odds (AOR: 1.562; 95% credible interval: 1.190, 2.046), indicating a 95% probability that the true odds ratio lies within this interval. In contrast, those with birth orders of two or three have 1.2 times greater odds. Mothers who are underweight or overweight/obese have significantly different results than mothers who have a normal weight. Underweight mothers have higher odds (AOR: 1.675; 95% credible interval: 1.338, 2.092), whereas overweight/obese mothers have lower odds (AOR: 0.651; 95% credible interval: 0.532, 0.794) of having CFU in their children. In both cases, the credible intervals do not include 1, suggesting statistically meaningful associations with CFU. On the other hand, mothers who are actively breastfeeding have no significant influence compared to their counterparts. The study found that children aged 24 and above have a 1.7 times higher risk (OR: 1.752; 95% credible interval: 1.405, 2.191) of developing undernutrition compared to those aged 6–23.9, whereas children under 6 months old have a decreased risk (Table 2). The results of the multivariate logistic regression analysis (Table S2) were consistent with the direction of the Bayesian estimates, with maternal education status, maternal BMI, child age in months (above 23.9) and birth order (at least four) being the most significant predictors of CFU.

TABLE 2.

Bayesian logistic regression among different determinants of CFU.

Variable categories	Coexisting forms of undernutrition (ref = no)
	Estimate	Odds ratio (OR)	95% credible interval for OR
			Lower, upper
Intercept	−2.45	0.086	0.058, 0.127
Maternal educational status (ref = higher)
No education	1.231	3.426	2.352, 5.001
Primary	0.602	1.825	1.342, 2.492
Secondary	0.402	1.494	1.139, 1.974
Wealth index (ref = middle)
Poorer	0.196	1.216	0.953, 1.553
Poorest	0.340	1.405	1.103, 1.790
Richer	−0.278	0.757	0.547, 1.000
Richest	−0.216	0.806	0.599, 1.801
Birth order (ref = 1)
2–3	0.184	1.202	1.005, 1.440
At least 4	0.446	1.562	1.190, 2.046
Maternal BMI (ref = normal weight)
Underweight	0.516	1.675	1.338, 2.092
Overweight/obese	−0.429	0.651	0.532, 0.794
Breastfeeding status (ref = no)
Yes	0.072	1.075	0.876, 1.316
Child age (months) (ref = 6–23.9)
Below 6	−0.183	0.833	0.600, 1.145
Above 23.9	0.561	1.752	1.405, 2.191

The authors check the performance of the posterior samples of the parameters of the Bayesian logistic regression. The trace, autocorrelation and density plots of the posterior samples of the selected parameters are presented in Figure 3. Plots of other parameters are presented in Figure S2. It is observed that the posterior realizations are stable and exhibit no significant autocorrelation.

FIGURE 3.

Trace, density and ACF plots of the posterior samples for the selected parameters.

The division‐wise disparities in the prevalence of CFU in Bangladesh are depicted in Figure 4. Findings show that the CFU in this case is 27.1% in the Sylhet division, the highest among all divisions, and 22% in Mymensingh and Barisal, the second highest. However, Dhaka (14.6%) and Khulna (15.8%) have the relatively lowest prevalence of CFU (Figure 4).

FIGURE 4.

Prevalence of coexisting forms of undernutrition among different divisions of Bangladesh.

4. Discussion

This study uses BDHS 2022 data to identify socio‐demographic and socio‐economic factors associated with coexisting undernutrition in Bangladesh's under‐5 population. Results show an association between mothers' BMI and CFU prevalence, indicating consistent health status differences between undernourished and healthy children. This study shows that only a tiny percentage of children under 5 (19.44%) had concomitant deficiencies in muscle mass, weight and height. Although it is higher than the 14.09% reported in the study on 26 country‐based DHS (Worku et al. 2025), the observed prevalence of coexistence among the three dimensions of undernutrition (stunting, wasting and underweight) in this study is comparable to that reported in a study conducted in Bangladesh (Chowdhury et al. 2021). Findings depict that a mother's education significantly impacts her child's undernutrition status. Educated mothers are more likely to use healthcare facilities, practise healthy childcare and follow appropriate feeding habits. Uneducated mothers have a 3.4 times increased chance of developing CFU, consistent with a Tanzanian study revealing three undernutrition indicators among children aged 0–23 months. The study emphasizes the need for focused nutrition programmes involving mothers without formal education to reduce the risk of undernutrition in children (Chowdhury et al. 2021; Pallangyo et al. 2025). However, a study conducted in Bangladesh suggests that education alone is insufficient to significantly lower child malnutrition (Rahman et al. 2016). The findings are also supported by a Nepali study that found early pregnancy is positively correlated with CFU, contradicting the Nepal study's findings that the prevalence of CFU increases with the number of birth orders (Wells et al. 2022). In contrast to national estimates and earlier research among Indigenous children in India, Malaysia and Guatemala and non‐Indigenous children in Cameroon, Ghana, Ethiopia and Zambia, a study of Mbororo children in Cameroon revealed that stunting was more common in female children than in male children. Our study found that gender was not a significant factor (Manjong et al. 2021).

Our study indicates a significant association between maternal age at first birth and CFU, with children born to adolescent mothers more likely to experience CFU. Adolescent mothers often have limited nutritional knowledge, lower education levels and limited healthcare access, which may contribute to adverse nutritional outcomes in their children (Rahman et al. 2019). They may be less biologically prepared for childbirth and childbearing (Black et al. 2013). Additionally, children with underweight mothers have 1.6 times higher odds and children with overweight/obese mothers have 0.6 times lower odds of having CFU compared to children of mothers of normal weight. Results of both bivariate and multivariate analyses show an association between family wealth and CFU in children, with children from poorer households having higher rates of undernutrition than those from wealthier households. Poor households are more likely to experience CFU among children under 5 than those in wealthier households. This aligns with previous studies suggesting that children from higher wealth quintiles are less likely to be stunted or underweight (M. M. Hossain, Abdulla, and Rahman 2023; Chowdhury et al. 2018; Hasan et al. 2020). Children from the richest wealth index had 44% lower odds of being underweight compared to those from the poorest wealth index. Previous studies have corroborated the idea that economic deprivation reduces access to nutrient‐rich foods, sanitation and maternal healthcare services, contributing to poor child growth outcomes (Chowdhury et al. 2020; Huda et al. 2018). Breastfeeding was found to be associated with a lower prevalence of CFU in this study. However, this finding differs from a previous study that analysed data from 33 studies and found breastfeeding to be inversely associated with overall undernutrition (Mardani et al. 2022).

The study reveals that the Sylhet division has the highest CFU rate among children under 5, whereas the Dhaka division has the lowest. This disparity is influenced by socio‐economic conditions, healthcare access and maternal practices like breastfeeding and nutrition education. The findings from a previous study also indicated that, despite national improvements, divisions such as Barisal and Sylhet continued to exhibit higher levels of malnutrition compared to other regions (Hasan et al. 2018). Another study, which utilized data from the 2014 BDHS, also found that the Sylhet division was notably lagging in controlling malnutrition among children (Das and Gulshan 2017). The found regional discrepancies could be attributed to socio‐economic conditions, infrastructure and medical services.

Mothers and children who reside in remote areas with inadequate transportation and communication have less access to medical facilities. The nutritional status of mothers and children is impacted by the low coverage of the Expanded Program on vaccine (which offers vaccine, vitamin A and zinc supplements) and exclusive breastfeeding practices in these areas, according to research. In the context of climate change and natural disasters, these districts with the highest rates of undernutrition are also regarded as food‐insecure and vulnerable areas.

Additionally, the high degree of undernutrition in these communities is linked to low female literacy and primary school enrolment rates (Mia et al. 2018).

Findings depict that rural areas face higher undernutrition rates due to limited access to healthcare, education and nutritious foods; inadequate toilet facilities; and waterborne diseases, which can spread infections. Previous research involving 30 154 children under 5 found that children from households without basic sanitation facilities had higher rates of undernutrition. Specifically, the absence of basic sanitation was associated with a 13% increase in stunting and a 19% increase in underweight cases (Hasan et al. 2023). Our study finds that children in households without TV access have higher undernutrition rates, which is consistent with previous research suggesting that households with TV access have higher awareness of nutrition and health practices, potentially improving child nutritional outcomes (Dhawan et al. 2023). All of these findings reflect associations between CFU and related variables rather than causal relationships as this study is based on cross‐sectional data. Due to the nature of the data, this study cannot establish causal relationships but only identify associations. The absence of temporal ordering makes it difficult to determine whether the exposure variables preceded the outcome CFU variable. For this reason, the results are interpreted by emphasizing associations rather than causal effects.

4.1. Limitations of the Study

The study's main disadvantage is recall bias, as self‐reporting was used for most BDHS characteristics, limiting verification of data‐collection procedures and potentially affecting the consistency or completeness of certain variables. This study has further limitations that point to areas worthy of further research. One of them is that only specific CFU parameters were assessed; other factors, such as skin‐to‐skin contact with others and deficiencies or intoxications of vitamin A, vitamin D, folic acid, iodine and calcium, could be a more detailed area for future study. Additionally, as our study only looked at children under 5, we do not know enough about how CFU affects children beyond that age range and how they cope with it. Although the survey weights the entire analysis in this study, clustering and stratification within the Bayesian framework are not taken into account, which could have an impact on variance estimations.

5. Conclusion

It is observed that the CFU is a serious health issue for Bangladeshi children under 5, with Sylhet having the highest impact and Dhaka having the least. Particularly at high risk for CFU are children whose mothers are illiterate, underweight, currently breastfeeding and have had children at a young age. Once more, children who are older than 23.9 years, have a higher birth order and come from relatively poorer families and live in rural areas are also at a higher risk of developing CFU. The authors recommended that the government of Bangladesh should focus on specific regions and development projects, especially in rural areas. The study emphasizes the need to use traditional medical therapy for women and calls on medical professionals to understand the issue and provide comprehensive care, as well as for mothers to be aware. The authors believe the study's findings will help policymakers reduce the prevalence of comorbid undernutrition among children under 5. Moreover, it would be beneficial to conduct a future study examining the prevalence, patterns and causes of CFU among Bangladeshi children under 5.

Author Contributions

Afrida Tasnim: data curation, methodology, software, writing – original draft, visualization. Md. Moyazzem Hossain: conceptualization, methodology, validation, writing – review and editing, supervision. Onnesha Saha: software, writing – original draft, methodology. Moinur Rahman: conceptualization, methodology, software, writing – original draft, formal analysis, data curation.

Funding

This research received no specific grant from any funding agency, commercial or not‐for‐profit sectors.

Ethics Statement

The Demographic and Health Surveys (DHS) Program website (https://dhsprogram.com/data/) provided a secondary dataset used in this investigation. Formal ethics approval was not necessary because the dataset was utilized for research and was accessible to the public. However, Inner City Fund (ICF) International and an in‐country Institutional Review Board (IRB) have approved DHS surveys to guarantee that protocols are compliant with U.S. Department of Health and Human Services regulations for the protection of human subjects and that the research was carried out with the highest respect for ethical research practices.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Sensitivity analysis of informative and non‐informative prior.

Table S2: Multivariate logistic regression examining factors associated with coexisting forms of undernutrition (CFU).

Figure S1: Variable selection using Boruta algorithm.

Figure S2: Trace, density and ACF plots of the selected variables.

Data S1: Bayesian model specification.

Data Availability Statement

The data are publicly available and can be accessed through the DHS website (https://www.dhsprogram.com/Data/) after registration.