Interpretable Bayesian Machine Learning Models for Predicting Undernutrition among under-Five Children

Interpretable Bayesian Machine Learning Models for Predicting Undernutrition among under-Five Children | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Interpretable Bayesian Machine Learning Models for Predicting Undernutrition among under-Five Children Ehsan Ahmed, Azizur Rahman This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8871095/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 22 You are reading this latest preprint version Abstract Background Child undernutrition remains a persistent public health challenge in Bangladesh, contributing substantially to growth retardation, infection risk, and early childhood mortality. The multiple biological, maternal, and socioeconomic factors that influence nutritional outcomes often interact in nonlinear ways, making prediction difficult through conventional statistical models. This study aimed to introduce an interpretable Bayesian machine learning (ML) model capable of predicting childhood undernutrition and identifying its most influential determinants using nationally representative data. Methods Data were extracted from 4,260 children under five years of age from the 2022 Bangladesh Demographic and Health Survey (BDHS). Three supervised ML algorithms—Logistic Regression, Bayesian Spike-and-Slab Regression (SSR), and Bayesian Additive Regression Trees (BART)—were trained and validated through ten-fold cross-validation. Model discrimination, calibration, and clinical utility were assessed using the area under the ROC curve (AUC), calibration curve and decision-curve analysis, respectively. SHapley Additive exPlanations (SHAP) values were applied to evaluate and visualize variable importance and to make the result interpretable. Results The Bayesian Spike-and-Slab model achieved the best overall performance in discremination (AUC = 0.63) and exhibited stable calibration. Decision-curve analysis further confirmed that this model provided the highest net clinical benefit across a wide range of threshold probabilities. SHAP interpretation identified small birth size, low maternal body mass index, limited maternal education, poor household wealth, inadequate sanitation, higher birth order, and older child age as the most influential factors associated with undernutrition. Conclusion The Bayesian Spike-and-Slab model provided a transparent and reliable framework for predicting childhood undernutrition in Bangladesh. SHAP analysis enhanced interpretability, clarified the contribution of major determinants, and offered evidence to support early detection and targeted nutrition strategies in resource-limited settings. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Health care Physical sciences/Mathematics and computing Health sciences/Medical research Health sciences/Risk factors Child undernutrition Machine learning Bayesian model SHAP Bangladesh Demographic and Health Survey Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Child undernutrition continues to pose a serious threat to public health in many low- and middle-income countries. It reflects a persistent shortage of calories, protein, and essential micronutrients that children need for normal growth and development [ 1 , 2 ]. When the nutritional requirements of a child are not met, growth slows, immune defense weakens, and the risk of infection rises sharply, leading to high levels of illness and death among children under five years of age [ 3 , 4 ]. Recent global reports indicate that millions of young children are affected by stunting, wasting, or being underweight, with the greatest concentration of these problems occurring in South Asia [ 5 ]. In Bangladesh, despite decades of targeted programs in food security and maternal–child health, undernutrition still represents a major obstacle to achieving several Sustainable Development Goals [ 6 ]. The determinants of undernutrition are complex and closely linked. Poverty, low maternal education, poor sanitation, and limited access to diverse foods all contribute to inadequate growth in early childhood [ 7 – 9 ]. Although many studies have explored these risk factors, their combined and interacting effects are not yet fully explained [ 10 ]. Traditional analytical approaches, such as logistic regression, assume linearity on the logit scale unless nonlinear transformations or interaction terms are explicitly specified. In practice, complex relationships may not always be fully explored, particularly in large multidimensional datasets [ 11 ]. Consequently, accurately identifying children who are most vulnerable to undernutrition remains difficult in many community and clinical settings [ 12 ]. In recent years, progress in computing and data science has created new opportunities for public-health research. Machine learning (ML) methods can analyze large, multidimensional datasets and detect complex patterns that traditional statistical tools might miss [ 13 ]. These approaches have already shown encouraging results in fields such as disease surveillance, nutrition monitoring, and mortality prediction [ 14 ]. Nevertheless, only a small number of studies have applied ML to predict child undernutrition, and many of those lack clarity about how specific factors influence the model’s output [ 15 – 17 ]. The limited interpretability of existing models reduces their usefulness for programme planners and policymakers [ 18 ]. Several recent studies have demonstrated the potential of machine learning methods for predicting childhood undernutrition. Talukder and Ahammed (2020) applied Random Forest, k-Nearest Neighbour, Support Vector Machine, and logistic regression models and found that ensemble classifiers provided better discrimination than conventional statistical models[ 19 ]. Islam et al. (2024) employed explainable ML techniques including XGBoost and SHAP analysis, achieving an AUC of about 0.80 for stunting and highlighting the importance of maternal education and household wealth as dominant features [ 20 ]. A comparative study using data from Bangladesh, India, and Nepal reported that Random Forest achieved the best overall accuracy for predicting child stunting across the region [ 21 ]. Similar advances have also been reported in Pakistan, where XGBoost and SHAP were used to uncover spatial inequalities in childhood stunting [ 22 ]. Earlier Bangladeshi work employing decision-tree and Random Forest models reported moderate accuracy (AUC ≈ 0.70) but limited interpretability [ 23 ]. Despite the growing number of ML studies, none have yet adopted a Bayesian machine learning framework to investigate child undernutrition. This gap is important because Bayesian models can combine prior knowledge with new data, handle high-dimensional and correlated variables, reduce overfitting, and provide predictions with uncertainty. In this study, two Bayesian models, Spike-and-Slab Regression (SSR) and Bayesian Additive Regression Trees (BART) were used with Logistic Regression to build an interpretable and reliable framework for identifying children at high risk of undernutrition. This combination allows flexible modeling of nonlinear relationships and account for prediction uncertainty while maintaining probabilistic inference and feature interpretability through SHAP analysis. Such an integrated approach can improve both predictive accuracy and policy relevance, enabling early identification of high-risk children in Bangladesh. Therefore, this study seeks to introduce and validate a Bayesian machine learning based predictive model to identify children at greatest risk of undernutrition. It also aims to interpret the contribution of individual variables using SHAP analysis so that the results can guide early detection, targeted nutritional interventions, and informed policy decisions in child-health programs. 2. Methods and Materials 2.1 Data Source This study used data from the most recent Bangladesh Demographic and Health Survey (BDHS) 2022, a nationally representative household survey conducted between May 25 and July 27, 2022 under the authority of the National Institute of Population Research and Training (NIPORT). The BDHS employed a two-stage stratified cluster sampling design to ensure representativeness across all administrative divisions and urban–rural strata. For this analysis, we focused on children under five years of age with complete anthropometric and household information. And the final analytic sample comprised 4,260 children. 2.2 Ethical Approval The study used publicly available, de-identified data from the 2022 Bangladesh Demographic and Health Survey (BDHS). The BDHS was approved by the National Institute of Population Research and Training (NIPORT) and the ICF Institutional Review Board. Because the dataset is anonymized and open access, no additional ethical approval was required. 2.3 Predictor variables The choice of predictors selected based on prior evidence on the determinants of child undernutrition in low- and middle-income countries [24–28]. To capture the different dimensions of risk, variables were grouped into five broad domains. Geographical and demographic factors included the child’s administrative division, urban–rural residence, and household size, reflecting regional and contextual disparities in nutrition[27,28]. Maternal characteristics included mother’s educational attainment, body mass index (BMI) category, occupation, and number of antenatal care (ANC) visits during pregnancy, factors repeatedly shown to influence child growth through maternal knowledge, health, and care-seeking behavior [24,25]. Household socioeconomic indicators comprised wealth index, type of sanitation and water facilities, and ownership of a radio, television, or mobile phone, which together serve as markers of socioeconomic position and access to health information[25,26]. Child-level characteristics included sex, age in months, birth order, birth interval, size at birth, breastfeeding status, type of birth (singleton or multiple), place of delivery, and delivery by caesarean section. These factors have been consistently associated with nutritional outcomes through both biological and caregiving pathways[29,30]. Recent morbidity was also considered, using maternal reports of diarrhea, fever, or cough in the two weeks preceding the survey, as acute illnesses are known to exacerbate nutritional deficits[31]. Together, these predictors provide a multidimensional representation of the biological, behavioral, and environmental influences on child nutrition. 2.4 Outcome Variable The primary outcome was child undernutrition, defined as a composite indicator encompassing stunting, wasting, or underweight. Stunting was defined as height-for-age Z-score (HAZ) < −2 standard deviations (SD) from the WHO reference population mean, wasting as weight-for-height Z-score (WHZ) < −2 SD, and underweight as weight-for-age Z-score (WAZ) < −2 SD. A child was classified as undernourished if they met at least one of these criteria[32,33] 2.5 Data Preprocessing The analytic sample was randomly split into training (70%) and testing (30%) subsets, with reproducibility ensured by fixing the random seed. Ten-fold cross-validation was applied within the training data to improve generalizability and reduce overfitting. 2.6 Statistical analysis and models Frequency and percentage distribution used to describe the characteristics of the data. Pearson’s chi-square tests were used to determine association between the outcome and explanatory variables. We applied three supervised learning models Spike and slab regression (SSR), logistic regression (LR), and Bayesian additive regression trees (BART) to predict the risk of undernutrition. These models were selected to balance interpretability, feature selection, and predictive flexibility. All statistical analyses were performed in R software (version: 4.5.1) and a p-value < 0.05 was considered significant. 2.6.1 Logistic regression (LR) Logistic regression serves as the conventional benchmark for binary classification problems and remains the most widely applied tool in public health and epidemiology[34]. It provides straightforward interpretability via odds ratios, making results easy to communicate to clinicians and policymakers. Although LR assumes linearity on the logit scale and does not readily capture nonlinear or higher-order interactions, its transparency and wide acceptance make it an essential comparator for evaluating the performance of more complex models. 2.6.2 Spike and slab regression (SSR) Spike and slab regression is a Bayesian shrinkage and variable selection approach that combines a “spike” prior, which shrinks irrelevant predictors toward zero, with a “slab” prior, which retains important predictors[35,36]. This dual mechanism makes SSR especially suited for both low and high-dimensional survey data where predictors such as socioeconomic, maternal, and child factors are often correlated. It simultaneously performs prediction and identifies the most influential risk factors, reducing the risk of overfitting compared with standard regression approaches [37]. 2.6.3 Bayesian additive regression trees (BART) BART is a nonparametric Bayesian ensemble method that represents the outcome as a sum over many regression trees [38,39]. For classification problems, BART uses a probit link function to estimate posterior probabilities of undernutrition. The method is highly flexible, capturing nonlinear effects and higher-order interactions that are difficult to pre-specify. In the context of undernutrition, where factors such as maternal BMI, child age, and wealth index may interact in complex ways, BART provides a robust predictive framework. Together, these three models allowed us to leverage complementary strengths: logistic regression for interpretability, SSR for parsimonious feature selection, and BART for flexibility in modeling nonlinearities and interactions. By comparing these complementary modeling approaches, we evaluated differences in predictive performance, calibration, and interpretability. 2.7 Evaluation Metrics Model performance was evaluated on the independent test dataset. Discrimination was quantified using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Additional performance measures included accuracy, sensitivity (recall), specificity, precision, F1 score, and balanced accuracy. Calibration was assessed by comparing predicted probabilities with observed outcomes through calibration plots. Decision curve analysis (DCA) was performed to estimate the net benefit of each model across a range of threshold probabilities. 3. Results 3.1 Baseline Characteristics A total of 4,260 children under five years of age from the BDHS 2022 dataset were included in the analysis. Among them, 35.1% were classified as undernourished, having at least one anthropometric deficit (stunting, wasting, or underweight). Table 1 presents the baseline distribution of child, maternal, and household characteristics stratified by nutritional status. Marked regional disparities were observed across administrative divisions (p < 0.001). The prevalence of undernutrition was highest in Sylhet (44.0%) and Mymensingh (39.0%), whereas Khulna (29.3%) and Dhaka (30.6%) recorded the lowest proportions. Rural children exhibited slightly higher rates of undernutrition (36.1%) compared with those from urban areas (32.8%, p = 0.04). Maternal educational attainment and nutritional status showed strong associations with child nutrition (p < 0.001). Undernutrition declined steadily from 55.2% among children of mothers with no education to 24.5% among those whose mothers had higher education. Similarly, children of underweight mothers had a markedly higher prevalence of undernutrition (46.8%) compared with children of obese mothers (19.9%). Increasing numbers of antenatal care visits were linked to reduced undernutrition, from 53.0% among mothers with no visits to 28.0% among those with four or more. Household wealth and sanitation status also demonstrated significant gradients (p < 0.001). Undernutrition prevalence ranged from 47.4% in the poorest quintile to 24.6% in the richest. Children from households with improved sanitation and television ownership showed lower undernutrition prevalence than those without these facilities. Table 1: Baseline characteristics of children under five years of age according to undernutrition status Covariates Total Undernutrition p-value No (%) Yes (%) Division Barisal 496 315 (63.5) 181 (36.5) <0.001 Chittagong 732 479 (65.4) 353 (34.6) Dhaka 607 421 (69.4) 186 (30.6) Khulna 450 318 (70.7) 132 (29.3) Mymensingh 525 320 (61.0) 205 (39.0) Rajshahi 413 285 (69.0) 128 (31.0) Rangpur 491 323 (65.8) 168 (34.2) Sylhet 546 306 (56.0) 240 (44.0) Residence Urban 1349 906 (67.2) 443 (32.8) 0.04 Rural 2911 1861 (63.4) 1050 (36.1) Family Size =4 2836 1838 (64.8) 998 (35.2) Maternal Education No education 252 113 (44.8) 139 (55.2) <0.001 Primary 931 543 (58.3) 388 (41.7) Secondary 2129 1421 (66.7) 708 (33.3) Higher 770 581 (75.5) 189 (24.5) Maternal Age 15-19 370 257 (69.7) 113 (30.5) 0.081 20-24 1279 840 (65.7) 439 (34.5) 25-29 1181 778 (65.9) 403 (34.1) 30-34 797 510 (64.0) 287 (36.0) 35-39 366 223 (60.9) 143 (39.1) 40-44 79 46 (58.2) 33 (41.8) 45-49 10 4 (40.0) 6 (60.0) Body Mass Index Underweight 500 266 (53.2) 234 (46.8) <0.001 Normal weight 2272 1439 (63.3) 833 (36.7) Overweight 969 686 (70.8) 283 (29.2) Obesity 241 193 (80.1) 48 (19.9) Maternal Working Status No 3095 2016 (65.1) 1079 (34.9) 0.958 Yes 987 642 (65.0) 345 (35.0) No. of ANC Visits 0 181 85 (47.0) 96 (53.0) <0.001 1 355 214 (60.3) 141 (39.7) 2 444 303 (68.2) 141 (31.8) 3 472 332 (70.3) 140 (29.7) 4 & more 959 691 (72.0) 269 (28.0) Wealth Index Poorest 878 462 (52.6) 416 (47.4) <0.001 Poorer 802 492 (61.3) 310 (38.7) Middle 821 544 (66.3) 277 (33.7) Richer 784 559 (71.3) 225 (28.7) Richest 797 601 (75.4) 196 (24.6) Sanitation Facility Improved 2964 1999 (67.4) 965 (32.6) <0.001 Unimproved 695 395 (56.8) 300 (43.2) Source of drinking water Improved 3638 2379 (65.6) 1259 (34.6) 0.145 Unimproved 622 388 (62.4) 243 (37.6) Has Radio No 4228 2741 (64.8) 1487 (35.2) 0.052 Yes 32 26 (81.3) 6 (18.8) Has Television No 2198 1344 (61.1) 854 (38.9) <0.001 Yes 2062 1423 (69.0) 639 (31.0) Has Mobile No 43 23 (53.5) 20 (46.5) 0.113 Yes 4217 2744 (65.0) 1473 (34.9) Age of Child (in Months) <6 323 323(72.1) 125 (27.9) <0.001 6-8 177 177 (76.3) 55 (23.7) 9-11 1713 181 (73.3) 66 (26.7) 12-23 6483 503 (62.6) 300 (37.4) 24-35 6420 500 (60.7) 324 (39.3) 36-47 6647 534 (63.3) 309 (36.7) 48-59 6680 549 (63.6) 314 (36.4) Sex of Child Male 2097 1371 (65.4) 726 (34.6) 0.716 Female 1985 1287 (64.8) 698 (35.2) Birth Order First 1540 1056 (68.6) 484 (31.4) <0.001 Second 1425 939 (65.9) 486 (34.1) Third 712 456 (64.0) 256 (36.0) Fourth & more 263 140 (53.2) 123 (46.8) Size of child at birth Larger than average 355 255 (71.9) 100 (28.1) <0.001 Average 3321 2260 (68.0) 1061 (32.0) Smaller than average 584 315 (53.9) 269 (46.1) Type of Birth Single 4026 2626 (65.2) 1400 (34.8) 0.121 Multiple 234 141 (60.3) 93 (39.7) Had Fever in last two weeks No 2783 1799 (64.6) 984 (35.4) 0.281 Yes 1292 856 (66.3) 436 (33.7) Had Diarrhea in last two weeks No 3869 2510 (64.9) 1359 (53.9) 0.343 Yes 208 144 (69.2) 64 (30.8) Had Cough in last two weeks No 2923 1874 (64.1) 1049 (35.9) 0.093 Yes 1151 778 (67.6) 373 (32.4) 3.2 Model Performance and Comparison We evaluated the performance of three machine learning models—Spike and slab, Logistic Regression, and BART for predicting the risk of undernutrition in children. Figure 1 presents the discriminative performance of the three models using ROC curves. All models showed moderate ability to predict undernutrition. The Spike and slab model achieved the highest performance with an AUC of 0.63, followed by BART (AUC = 0.619) and Logistic Regression (AUC = 0.594). These findings suggest that Spike and slab provided better discrimination between undernourished and non-undernourished children compared with the alternative approaches. Table 2 summarizes detailed cross-validated performance metrics for the three models. The optimal cutoff values for each model were selected based on maximizing the performance metrics, with a cutoff of 0.303 for Spike and slab, 0.297 for Logistic Regression, and 0.301 for BART. The Spike and slab model exhibited superior overall performance, achieving the highest accuracy (60.43%), sensitivity (69.17%), and balanced accuracy (62.74%). Notably, Spike and slab also obtained the best F1 score (0.53), indicating an optimal balance between precision and recall. In comparison, BART demonstrated moderate performance (accuracy 58.91%, sensitivity 64.03%, F1 score 0.50), while Logistic Regression performed lowest across most indices (accuracy 58.66%, sensitivity 60.08%, F1 score 0.48). Table 2: Performances of the machine learning models for predicting Child Undernutrition Evaluation Metrics Models Spike and slab Logistic Regression BART Accuracy 60.43 % 58.66 % 58.91 % Sensitivity 69.17 % 60.08 % 64.03 % Specificity 56.32% 57.99 % 56.51 % Precision 42.68 % 40.21 % 40.91 % F1 Score 52.79 % 48.18 % 49.92 % Balanced Accuracy 62.74 % 59.04 % 60.27 % Calibration curves for the three models are displayed in Figure 2. The Spike and slab model showed the closest agreement between predicted probabilities and observed outcomes, particularly in the mid-range of risk thresholds, reflecting better reliability of its probability estimates. By contrast, Logistic Regression slightly underestimated observed risk at higher thresholds, while BART demonstrated mild overestimation at extreme probability ranges. The decision curve analysis (Figure 3) further demonstrated the net benefit of the Spike and slab model. Spike and slab provided the highest net benefit across a wide range of threshold probabilities, particularly around the 30% probability thresholds. These results highlight that Spike and slab not only deliver accurate predictions but also maximizes the clinical utility by identifying high-risk individuals with the most significant benefit. 3.3 Model Interpretability Analysis Figure 4A presents a comprehensive swarm plot of SHAP values from the Spike and slab model. The horizontal axis represents SHAP values, indicating the direction and magnitude of each feature’s contribution to undernutrition risk, while the vertical axis lists features ordered by their cumulative impact. Each point corresponds to a single child, with yellow shades denoting high contributions to undernutrition prediction and purple shades reflecting low influences. The size of the child at birth (M18) emerged as the strongest predictor. Children born smaller than average displayed positive SHAP values, indicating a higher contribution to undernutrition risk, whereas those reported as average or larger than average suggesting a protective effect. Child age category (Cage_cat) was the second most important predictor. In particular, children aged 24–35 months had the largest positive SHAP values, reflecting the increased vulnerability to Undernutrition during the transition from exclusive breastfeeding to complementary feeding. Maternal BMI status (BMI_cat) was also a critical determinant. Children of underweight mothers consistently showed positive SHAP values, indicating a greater risk of undernutrition. In contrast, children of overweight or obese mothers demonstrated negative SHAP contributions, suggesting a protective effect relative to the underweight group. Beyond these biological factors, household wealth index (V190) played a strong protective role, with children from wealthier households exhibiting negative SHAP values and thus a reduced risk of undernutrition. Additional but less influential predictors included maternal age group (V013), birth order category (Border_cat), administrative division (HV024), toilet facility type (Toilet_Tp), and maternal education level (V106). These features showed smaller SHAP contributions but were directionally consistent, with higher birth order, poorer sanitation, and lower education tending to increase risk. Figure 4B provides a detailed case study, demonstrating the model’s prediction process for an individual child. In this visualization, yellow indicators, such as being 24–35 months of age and having an underweight mother, increased the predicted risk, while protective purple contributions included larger birth size and higher household wealth. For this child, the balance of these factors produced a higher-than-baseline probability of undernutrition. 4. Discussion In this study, we developed and compared multiple machine learning models to predict the risk of childhood undernutrition by utilizing Geographical and demographic, Maternal characteristics, Child-level characteristics, and Household socioeconomic factors. Among the three models; Spike and slab, Bayesian Additive Regression Trees (BART), and Logistic Regression, the Spike and slab algorithm demonstrated superior predictive performance and calibration. The model achieved an AUC of 0.63 and provided the highest accuracy, sensitivity, and F1 score. The reliability of the model was further supported by calibration curves and decision curve analysis, which confirmed its net clinical benefit across a broad range of threshold probabilities. Recent advances in ML have transformed predictive modeling in public health, allowing researchers to uncover non-linear relationships that traditional regression often fails to detect [40]. The Spike and slab model, a Bayesian variable selection framework, effectively addresses model uncertainty by shrinking irrelevant predictors while retaining informative ones [41]. This enables the model to manage correlated features common in nutrition data, such as maternal and socioeconomic factors, and to avoid overfitting even with limited samples [42]. In contrast, the Logistic Regression model though interpretable relies on linear assumptions that may oversimplify real-world dynamics, and the BART model, while flexible, can overfit small datasets without proper regularization [43]. The calibration curve further illustrated that the Spike and slab model’s predicted probabilities closely matched observed outcomes, particularly at mid-range risk levels. This finding suggests that the model can produce stable and realistic probability estimates rather than overestimating or underestimating risk. Previous evidence supports that Bayesian-based models tend to maintain more stable calibration in health data with class imbalance [44]. In contrast, BART tended to overpredict undernutrition at extreme values, while Logistic Regression underestimated the risk in higher thresholds. These trends are consistent with previous comparative studies, where Bayesian and ensemble methods often demonstrated superior calibration when applied to unbalanced health datasets [45]. The decision curve analysis supported the clinical applicability of the Spike and slab model, showing the highest net benefit over a wide range of threshold probabilities, Across threshold probabilities between 20% and 40%. From a public health standpoint, this means that using the Spike and slab model could help community health workers identify high-risk children more effectively, thereby optimizing limited resources for targeted interventions. Such probabilistic frameworks are particularly relevant in low- and middle-income countries, where limited resources require prioritization of high-risk children [46]. To gain a deeper understanding of the model’s predictive process, we used SHAP visualization to interpret the contribution of each feature. The SHAP analysis revealed that eight major characteristics; birth size, child age, maternal BMI, household wealth index, maternal education, sanitation type, birth order, and maternal age had the most substantial influence on prediction. Among these, the size of the child at birth emerged as the strongest predictor, consistent with earlier epidemiological findings that low birth weight or small size at birth is a major risk factor for postnatal malnutrition and stunting [47]. Children born smaller than average were more likely to experience chronic growth deficits, limited nutrient absorption, and delayed recovery from infections, predisposing them to persistent undernutrition in early life. Child age category also played an essential role, particularly among children aged 24–35 months, who were identified as the most vulnerable. This period marks the transition from breastfeeding to complementary feeding, a phase when dietary inadequacy and infection exposure sharply increase [48]. Similar findings have been observed in studies conducted in sub-Saharan Africa and South Asia, where weaning-related nutritional gaps were strongly associated with growth faltering and wasting [49,50]. Maternal nutritional status, as indicated by BMI category, was another key determinant. Children born to underweight mothers showed higher predicted risks, reflecting the intergenerational transmission of malnutrition. Undernourished mothers are more likely to deliver low-birth-weight infants, have reduced breast milk quality, and face greater postpartum health challenges [51]. Socioeconomic and environmental variables also emerged as critical protective factors. Children from households in higher wealth quintiles or with improved toilet facilities exhibited significantly lower SHAP contributions toward undernutrition risk. These patterns reinforce the established association between economic stability, hygiene, and nutritional health. Improved sanitation reduces infection exposure, while wealth and education enhance access to quality food, healthcare, and awareness of child-feeding practices [52,53]. Collectively, these findings highlight that undernutrition is not driven by a single factor but results from the interaction of maternal, child, and household-level influences. Interestingly, administrative region and maternal age demonstrated smaller but directionally consistent effects. Younger mothers, particularly those below 20 years of age, tended to have higher predicted risks for their children. Early motherhood often coincides with limited education and economic dependence, which constrain childcare capacity and nutritional awareness. These contextual insights align with prior findings that maternal empowerment and household decision-making autonomy are protective against childhood stunting and wasting [54,55]. From a methodological standpoint, This study suggests that Bayesian shrinkage approaches such as Spike and Slab regression may provide improved sensitivity and calibration compared with standard logistic regression in this context. By incorporating prior information and probabilistic shrinkage, these models maintain parsimony while effectively capturing non-linear interactions [56,57]. Moreover, SHAP interpretability bridges the gap between technical model outputs and actionable insights, making predictive tools more transparent and clinically relevant [58]. 4.1 Limitations Despite the promising results, several limitations should be acknowledged. First, the dataset was cross-sectional, limiting causal interpretation of identified relationships. While the models can detect associations, they cannot determine temporal order or causality. Future longitudinal studies are needed to examine whether the identified predictors consistently precede and contribute to the onset of undernutrition. Second, some potentially influential variables, such as dietary diversity, and breastfeeding frequency were unavailable in the dataset. Their exclusion may have constrained the overall accuracy and reduced the model’s ability to account for behavioral and dietary influences. Finally, socioeconomic and regional disparities in data reporting may introduce bias, and improved standardization of national nutrition datasets would enhance future modeling efforts. 5. Conclusion We introduced a Bayesian machine learning model to predict the risk of childhood undernutrition. Among the compared models, the Bayesian Spike and Slab regression achieved the highest predictive performance and calibration reliability. Integration of SHAP analysis improved model interpretability, enabling a clearer understanding of the key determinants such as birth size, child age, maternal BMI, household wealth, and sanitation that contribute to undernutrition risk. By providing transparent and clinically meaningful intuitions, this interpretable machine learning framework offers a promising tool to support early identification of high-risk children and guide targeted nutrition interventions to expand child health outcomes. Declarations 6. Ethics approval and consent to participate Not applicable. 7. Consent for publication Not applicable for this study. 8. Clinical trial number Not applicable. 9. Availability of data The datasets analyzed during the current study are available in the Demographic and Health Surveys (DHS) Program repository https://dhsprogram.com/data/. DHS data are publicly available upon request. 10. Competing interests The authors declared no competing interests in this study. 11. Funding The authors received no funding for this study. 12. Acknowledgments The authors want to acknowledge the Measures DHS data archive for providing us with the datasets for further analysis. Author Contribution E.A.: Conceptualization, Methodology, Software, Formal analysis, Data curation, Validation, Visualization, Writing—original draft.A.Z.: Methodology refinement, Supervision, Interpretation of results, Writing—review & editing.Both authors approved the final manuscript. References World Health Organization. Malnutrition. (2024). https://www.who.int/news-room/fact-sheets/detail/malnutrition Amoadu, M. et al. Risk Factors of Malnutrition among In-School Children and Adolescents in Developing Countries: A Scoping Review. Children 11 , 476 (2024). Morales, F., Montserrat-de la Paz, S., Leon, M. J. & Rivero-Pino, F. Effects of Malnutrition on the Immune System and Infection and the Role of Nutritional Strategies Regarding Improvements in Children’s Health Status: A Literature Review. Nutrients 16 , 1 (2023). UNICEF. Malnutrition in children. (2024). https://data.unicef.org/topic/nutrition/malnutrition UNICEF, WHO & WORLD BANK. Level and trend in child malnutrition. World Health Organization 4 (2023). Global Nutrition Report. The burden of malnutrition around the world. (2023). https://globalnutritionreport.org/reports/global-nutrition-report-2018/burden-malnutrition Redwan, T. & Shabur, M. A. Analyzing the socio-economic factors of adolescent malnutrition in Bangladesh using Best Worst Method. Discover Sustain. 6 , 1–16 (2025). Sahiledengle, B. et al. Factors associated with childhood undernutrition in poor Ethiopian households: Implications for public health interventions. PLoS One 20 , (2025). Huda, T. M., Hayes, A., El Arifeen, S. & Dibley, M. J. Social determinants of inequalities in child undernutrition in Bangladesh: A decomposition analysis. Matern Child. Nutr. 14 , e12440 (2017). Momin Islam, M. & Noor, F. M. Determinants of Malnutrition among Under-Five Children in Bangladesh: Evidence from Multiple Indicator Cluster Survey, 2019 Data. Data Fortune J. Health Sci. 5 , 284–295 (2022). Das, S., Baffour, B. & Richardson, A. Prevalence of child undernutrition measures and their spatio-demographic inequalities in Bangladesh: an application of multilevel Bayesian modelling. BMC Public. Health . 22 , 1–21 (2022). Islam, M. M., Jahan Kibria, S., Kumar, N. M., Roy, S., Karim, M. R. & D. C. & Prediction of undernutrition and identification of its influencing predictors among under-five children in Bangladesh using explainable machine learning algorithms. PLoS One . 19 , e0315393 (2024). Kirk, D. et al. Machine Learning in Nutrition Research. Adv. Nutr. 13 , 2573–2589 (2022). Mccrory, M. A. et al. Interaction Analysis Based on Shapley Values and Extreme Gradient Boosting: A Realistic Simulation and Application to a Large Epidemiological Prospective Study. Front. Nutr. 9 , 871768 (2022). Hu, X. et al. An interpretable machine learning model for precise prediction of biomarkers for intermittent fasting pattern. Nutr. Metabolism . 21 , 1–8 (2024). Begashaw, G. B., Zewotir, T. & Fenta, H. M. A deep learning approach for classifying and predicting children’s nutritional status in Ethiopia using LSTM-FC neural networks. BioData Min. 18 , 1–24 (2025). Bitew, F. H., Sparks, C. S. & Nyarko, S. H. Machine learning algorithms for predicting undernutrition among under-five children in Ethiopia. Public. Health Nutr. 25 , 269 (2021). Kirk, D. Good practices and common pitfalls of machine learning in nutrition research. Proceedings of the Nutrition Society 1–14 (2024). 10.1017/S0029665124007638 Talukder, A. & Ahammed, B. Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh. Nutrition 78 , (2020). Islam, M. M., Jahan Kibria, S., Kumar, N. M., Roy, S., Karim, M. R. & D. C. & Prediction of undernutrition and identification of its influencing predictors among under-five children in Bangladesh using explainable machine learning algorithms. PLoS One . 19 , e0315393 (2024). Sara, S. S., Khan, M. S. & Talukder, A. Prediction of Child Stunting with Machine Learning Algorithms: A Cross-Country Study of Bangladesh, India, and Nepal. (2024). https://doi.org/10.21203/RS.3.RS-4696630/V1 doi:10.21203/RS.3.RS-4696630/V1. Zhang, X., Usman, M., Irshad, A. R., Rashid, M. & Khattak, A. Investigating Spatial Effects through Machine Learning and Leveraging Explainable AI for Child Malnutrition in Pakistan. ISPRS Int. J. Geo-Information 2024 . 13 , 330 (2024). Hemo, S. A. & Rayhan Md. I. Classification tree and random forest model to predict under-five malnutrition in Bangladesh. Biom Biostat Int. J. 10 , 116–123 (2021). Akombi, B. J. et al. Stunting and severe stunting among children under-5 years in Nigeria: A multilevel analysis. BMC Pediatr 17 , (2017). Das, S. & Rahman, R. M. Application of ordinal logistic regression analysis in determining risk factors of child malnutrition in Bangladesh. Nutr J 10 , (2011). Chowdhury, T. R., Chakrabarty, S., Rakib, M., Saltmarsh, S. & Davis, K. A. Socio-economic risk factors for early childhood underweight in Bangladesh. Global Health 14 , (2018). Chowdhury, M. R. K. et al. The prevalence and socio-demographic risk factors of coexistence of stunting, wasting, and underweight among children under five years in Bangladesh: a cross-sectional study. BMC Nutr. 8 , 1–12 (2022). Haque, S. et al. Nutritional status of under-five aged children of ready-made garment workers in Bangladesh: A cross-sectional study. PLoS One . 18 , e0284325 (2023). Kumar, P., Rashmi, R., Muhammad, T. & Srivastava, S. Factors contributing to the reduction in childhood stunting in Bangladesh: a pooled data analysis from the Bangladesh demographic and health surveys of 2004 and 2017–18. BMC Public. Health . 21 , 1–14 (2021). Hossain, M. I. et al. Risk or associated factors of wasting among under-five children in Bangladesh: A systematic review. Asia Pac. J. Clin. Nutr. 33 , 457 (2024). Akter, S., Siriphon, A., Ayuttacorn, A. & Boonchieng, W. Prevalence of ARI, fever, and diarrhea among under-five children and the influencing factors in southwestern coastal region of Bangladesh. BMC Public. Health . 25 , 2951 (2025). Halima, O. et al. Identifying Individual and Household Level Predictors of Undernutrition Among 6–59 Months Children in Bangladesh: A Multivariate Approach. Public. Health Challenges . 3 , e70007 (2024). Birhanu, F., Yitbarek, K., Bobo, F. T., Atlantis, E. & Woldie, M. Undernutrition in children under five associated with wealth-related inequality in 24 low- and middle-income countries from 2017 to 2022. Sci. Rep. 14 , 1–9 (2024). Hosmer, D. W., Lemeshow, S. & Sturdivant, R. X. Applied Logistic Regression: Third Edition. Appl. Logistic Regression: Third Ed. 1–510 10.1002/9781118548387 (2013). Ishwaran, H., Kogalur, U. B., Sunil Rao, J. & Spikeslab Prediction and variable selection using spike and slab regression. R J. 2 , 68–73 (2010). Ishwaran, H. & Rao, J. S. Spike and slab variable selection: Frequentist and Bayesian strategies. (2005). https://doi.org/10.1214/009053604000001147 33, 730–773 Hahn, P. R. & Carvalho, C. M. Decoupling shrinkage and selection in Bayesian linear models: a posterior summary perspective. (2014). http://arxiv.org/abs/1408.0464 He, L., Cao, L., Wang, T., Cao, Z. & Shi, X. A Bayesian Additive Regression Trees Framework for Individualized Causal Effect Estimation. Math. 2025 . 13 , 2195 (2025). Chipman, H. A., George, E. I. & McCulloch, R. E. BART: Bayesian additive regression trees. (2010). https://doi.org/10.1214/09 -AOAS285 4, 266–298. Vanshika, Gupta, N. & Machine Learning Applications in Healthcare. 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions), ICRITO 2022 (2022). https://doi.org/10.1109/ICRITO56286.2022.9964865 (2022) doi:10.1109/ICRITO56286.2022.9964865. O’Hara, R. B. & Sillanpää, M. J. A review of Bayesian variable selection methods: what, how and which. (2009). https://doi.org/10.1214/09 -BA403 4, 85–117. Sinha, S., Mallick, B. K., Kipnis, V. & Carroll, R. J. Semiparametric bayesian analysis of nutritional epidemiology data in the presence of measurement error. Biometrics 66 , 444–454 (2010). Zhu, Z. et al. Integrating Machine Learning and the SHapley Additive exPlanations (SHAP) Framework to Predict Lymph Node Metastasis in Gastric Cancer Patients Based on Inflammation Indices and Peripheral Lymphocyte Subpopulations. J. Inflamm. Res. 17 , 9551 (2024). Hong, Y. et al. Machine learning prediction of metabolic dysfunction-associated fatty liver disease risk in American adults using body composition: explainable analysis based on SHapley Additive exPlanations. Front. Nutr. 12 , 1616229 (2025). Steyerberg, E. W. Clinical Prediction Models. (2019). https://doi.org/10.1007/978-3-030-16399-0 doi:10.1007/978-3-030-16399-0. Talukder, A. & Ahammed, B. Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh. Nutrition 78 , (2020). Keats, E. C. et al. Effective interventions to address maternal and child malnutrition: an update of the evidence. Lancet Child. Adolesc. Health . 5 , 367–384 (2021). Victora, C. G. et al. Revisiting maternal and child undernutrition in low-income and middle-income countries: variable progress towards an unfinished agenda. Lancet 397 , 1388–1399 (2021). Anteneh, R. M. et al. Wealth-related inequalities in undernutrition among under-five children in sub-Saharan Africa. Sci. Rep. 15 , 1–14 (2025). Rahman, M. S., Howlader, T., Masud, M. S. & Rahman, M. L. Association of Low-Birth Weight with Malnutrition in Children under Five Years in Bangladesh: Do Mother’s Education, Socio-Economic Status, and Birth Interval Matter? PLoS One . 11 , e0157814 (2016). Aguayo, V. M., Nair, R., Badgaiyan, N. & Krishna, V. Determinants of stunting and poor linear growth in children under 2 years of age in India: an in-depth analysis of Maharashtra’s comprehensive nutrition survey. Matern Child. Nutr. 12 (Suppl 1), 121–140 (2016). Bitew, F. H., Sparks, C. S. & Nyarko, S. H. Machine learning algorithms for predicting undernutrition among under-five children in Ethiopia. Public. Health Nutr. 25 , 269 (2021). Lin, J. & Feng, X. L. Exploring the impact of water, sanitation and hygiene (WASH), early adequate feeding and access to health care on urban–rural disparities of child malnutrition in China. Matern Child. Nutr. 19 , e13542 (2023). Rahman, M. M., Saima, U. & Goni, M. A. Impact of maternal household decision-making autonomy on child nutritional status in Bangladesh. Asia Pac. J. Public. Health . 27 , 509–520 (2015). Paul, P. & Saha, R. Is maternal autonomy associated with child nutritional status? Evidence from a cross-sectional study in India. PLoS One . 17 , e0268126 (2022). Lu, Z. & Lou, W. Bayesian approaches to variable selection: a comparative study from practical perspectives. Int. J. Biostatistics . 18 , 83–108 (2022). Malsiner-Walli, G. & Wagner, H. Comparing Spike and Slab Priors for Bayesian Variable Selection. Austrian J. Statistics 40 , (2016). Ponce-Bobadilla, A. V., Schmitt, V., Maier, C. S., Mensing, S. & Stodtmann, S. Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. Clin. Transl Sci. 17 , e70056 (2024). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 04 May, 2026 Reviews received at journal 27 Apr, 2026 Reviews received at journal 26 Apr, 2026 Reviews received at journal 21 Apr, 2026 Reviews received at journal 19 Apr, 2026 Reviews received at journal 18 Apr, 2026 Reviewers agreed at journal 16 Apr, 2026 Reviewers agreed at journal 13 Apr, 2026 Reviewers agreed at journal 11 Apr, 2026 Reviews received at journal 11 Apr, 2026 Reviews received at journal 11 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 10 Apr, 2026 Reviewers agreed at journal 08 Apr, 2026 Reviewers invited by journal 08 Apr, 2026 Editor invited by journal 18 Feb, 2026 Editor assigned by journal 14 Feb, 2026 Submission checks completed at journal 14 Feb, 2026 First submitted to journal 13 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8871095","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":623111246,"identity":"887c804e-0d93-4c62-8d78-67e6e6907485","order_by":0,"name":"Ehsan Ahmed","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIiWNgGAWjYDACCTB5AIgTwCw5MPcB0VoOMBgYg7kJpGhJbGCA24gd8Eu3P/xcUHMnmp89gfnzh5o/6fPDDj8E2mInp9uAXYvknDPG0jOOPcud2fOATeLAMYPcjbfTDIBako3NDmDXYnAjh0Gah+1w7oYbCWwMB9iAWmYngLQcSNyGQ4v9jfTHv3n+Hc7dfyOB+cOBfwbphrPTP+DVYiCRYCbN2wa0RSKBQeJgm0GCvHQOflskbuSYWfP2Hc6dceZhm8TZPmPDDdI5BQcSDHD7hX9G+uPbPN8O5/a3Jx/+UPFNTl5+dvrmDx8q7ORwaUECjA0Qp4JVGhBUjgTkG0hRPQpGwSgYBSMBAABvi2ook1nWUAAAAABJRU5ErkJggg==","orcid":"","institution":"Tejgaon College","correspondingAuthor":true,"prefix":"","firstName":"Ehsan","middleName":"","lastName":"Ahmed","suffix":""},{"id":623111248,"identity":"be463716-c23d-4257-91f6-0dda7e90d78b","order_by":1,"name":"Azizur Rahman","email":"","orcid":"","institution":"University of Manitoba","correspondingAuthor":false,"prefix":"","firstName":"Azizur","middleName":"","lastName":"Rahman","suffix":""}],"badges":[],"createdAt":"2026-02-13 11:09:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8871095/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8871095/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107246406,"identity":"09d9f5ca-055d-40ec-bc2a-7f362906e85f","added_by":"auto","created_at":"2026-04-19 08:10:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":134043,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eROC Curves for selected three models\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8871095/v1/6f120bea22a6f335e2f2a66d.png"},{"id":107246405,"identity":"2129459e-51a6-4586-ba15-64102fe5a51b","added_by":"auto","created_at":"2026-04-19 08:10:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":143654,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCalibration Curve for selected three models\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8871095/v1/53ce1394e9821cab764ba11a.png"},{"id":107246498,"identity":"788561aa-75f9-41bf-addb-4794b2cc4a3d","added_by":"auto","created_at":"2026-04-19 08:10:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":48255,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDecision Curve Analysis for selected three models\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8871095/v1/23d5a68baa2445a18f21980e.png"},{"id":107246394,"identity":"9bd38fc2-70e8-454f-bb18-5cd932f44b81","added_by":"auto","created_at":"2026-04-19 08:10:16","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":123412,"visible":true,"origin":"","legend":"\u003cp\u003eVisually interpret spike and slab model using SHAP. \u003cstrong\u003eA\u003c/strong\u003e SHAP summary point. \u003cstrong\u003eB\u003c/strong\u003eSHAP force plot. M18: size of the child at birth, Cage_cat: Child age in month, BMI_cat: Maternal BMI status, V190: household wealth index, Border_cat: birth order category, HV024: administrative division, V013: maternal age group\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8871095/v1/bfc08434d7a41ccc8c96841f.png"},{"id":107246904,"identity":"99061c85-16e3-437f-aacc-f521bb7ef35b","added_by":"auto","created_at":"2026-04-19 08:10:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1347913,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8871095/v1/4a400314-4b9b-4930-a49f-5ec4e01c0364.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Interpretable Bayesian Machine Learning Models for Predicting Undernutrition among under-Five Children","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eChild undernutrition continues to pose a serious threat to public health in many low- and middle-income countries. It reflects a persistent shortage of calories, protein, and essential micronutrients that children need for normal growth and development [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. When the nutritional requirements of a child are not met, growth slows, immune defense weakens, and the risk of infection rises sharply, leading to high levels of illness and death among children under five years of age [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Recent global reports indicate that millions of young children are affected by stunting, wasting, or being underweight, with the greatest concentration of these problems occurring in South Asia [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. In Bangladesh, despite decades of targeted programs in food security and maternal\u0026ndash;child health, undernutrition still represents a major obstacle to achieving several Sustainable Development Goals [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The determinants of undernutrition are complex and closely linked. Poverty, low maternal education, poor sanitation, and limited access to diverse foods all contribute to inadequate growth in early childhood [\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Although many studies have explored these risk factors, their combined and interacting effects are not yet fully explained [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Traditional analytical approaches, such as logistic regression, assume linearity on the logit scale unless nonlinear transformations or interaction terms are explicitly specified. In practice, complex relationships may not always be fully explored, particularly in large multidimensional datasets [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Consequently, accurately identifying children who are most vulnerable to undernutrition remains difficult in many community and clinical settings [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn recent years, progress in computing and data science has created new opportunities for public-health research. Machine learning (ML) methods can analyze large, multidimensional datasets and detect complex patterns that traditional statistical tools might miss [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. These approaches have already shown encouraging results in fields such as disease surveillance, nutrition monitoring, and mortality prediction [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Nevertheless, only a small number of studies have applied ML to predict child undernutrition, and many of those lack clarity about how specific factors influence the model\u0026rsquo;s output [\u003cspan additionalcitationids=\"CR16\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The limited interpretability of existing models reduces their usefulness for programme planners and policymakers [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSeveral recent studies have demonstrated the potential of machine learning methods for predicting childhood undernutrition. Talukder and Ahammed (2020) applied Random Forest, k-Nearest Neighbour, Support Vector Machine, and logistic regression models and found that ensemble classifiers provided better discrimination than conventional statistical models[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Islam et al. (2024) employed explainable ML techniques including XGBoost and SHAP analysis, achieving an AUC of about 0.80 for stunting and highlighting the importance of maternal education and household wealth as dominant features [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. A comparative study using data from Bangladesh, India, and Nepal reported that Random Forest achieved the best overall accuracy for predicting child stunting across the region [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Similar advances have also been reported in Pakistan, where XGBoost and SHAP were used to uncover spatial inequalities in childhood stunting [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Earlier Bangladeshi work employing decision-tree and Random Forest models reported moderate accuracy (AUC\u0026thinsp;\u0026asymp;\u0026thinsp;0.70) but limited interpretability [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Despite the growing number of ML studies, none have yet adopted a Bayesian machine learning framework to investigate child undernutrition. This gap is important because Bayesian models can combine prior knowledge with new data, handle high-dimensional and correlated variables, reduce overfitting, and provide predictions with uncertainty. In this study, two Bayesian models, Spike-and-Slab Regression (SSR) and Bayesian Additive Regression Trees (BART) were used with Logistic Regression to build an interpretable and reliable framework for identifying children at high risk of undernutrition. This combination allows flexible modeling of nonlinear relationships and account for prediction uncertainty while maintaining probabilistic inference and feature interpretability through SHAP analysis. Such an integrated approach can improve both predictive accuracy and policy relevance, enabling early identification of high-risk children in Bangladesh.\u003c/p\u003e \u003cp\u003eTherefore, this study seeks to introduce and validate a Bayesian machine learning based predictive model to identify children at greatest risk of undernutrition. It also aims to interpret the contribution of individual variables using SHAP analysis so that the results can guide early detection, targeted nutritional interventions, and informed policy decisions in child-health programs.\u003c/p\u003e"},{"header":"2. Methods and Materials","content":"\u003cp\u003e\u003cstrong\u003e2.1 Data Source\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study used data from the most recent Bangladesh Demographic and Health Survey (BDHS) 2022, a nationally representative household survey conducted between May 25 and July 27, 2022 under the authority of the National Institute of Population Research and Training (NIPORT). The BDHS employed a two-stage stratified cluster sampling design to ensure representativeness across all administrative divisions and urban\u0026ndash;rural strata. For this analysis, we focused on children under five years of age with complete anthropometric and household information. And the final analytic sample comprised 4,260 children.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.2 Ethical Approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study used publicly available, de-identified data from the 2022 Bangladesh Demographic and Health Survey (BDHS). The BDHS was approved by the National Institute of Population Research and Training (NIPORT) and the ICF Institutional Review Board. Because the dataset is anonymized and open access, no additional ethical approval was required.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.3 Predictor variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe choice of predictors selected based on prior evidence on the determinants of child undernutrition in low- and middle-income countries\u0026nbsp;[24\u0026ndash;28]. To capture the different dimensions of risk, variables were grouped into five broad domains. Geographical and demographic factors included the child\u0026rsquo;s administrative division, urban\u0026ndash;rural residence, and household size, reflecting regional and contextual disparities in nutrition[27,28]. Maternal characteristics included mother\u0026rsquo;s educational attainment, body mass index (BMI) category, occupation, and number of antenatal care (ANC) visits during pregnancy, factors repeatedly shown to influence child growth through maternal knowledge, health, and care-seeking behavior\u0026nbsp;[24,25]. Household socioeconomic\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eindicators comprised wealth index, type of sanitation and water facilities, and ownership of a radio, television, or mobile phone, which together serve as markers of socioeconomic position and access to health information[25,26]. Child-level characteristics included sex, age in months, birth order, birth interval, size at birth, breastfeeding status, type of birth (singleton or multiple), place of delivery, and delivery by caesarean section. These factors have been consistently associated with nutritional outcomes through both biological and caregiving pathways[29,30]. Recent morbidity was also considered, using maternal reports of diarrhea, fever, or cough in the two weeks preceding the survey, as acute illnesses are known to exacerbate nutritional deficits[31]. Together, these predictors provide a multidimensional representation of the biological, behavioral, and environmental influences on child nutrition.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.4 Outcome Variable\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe primary outcome was child undernutrition, defined as a composite indicator encompassing stunting, wasting, or underweight. Stunting was defined as height-for-age Z-score (HAZ) \u0026lt; \u0026minus;2 standard deviations (SD) from the WHO reference population mean, wasting as weight-for-height Z-score (WHZ) \u0026lt; \u0026minus;2 SD, and underweight as weight-for-age Z-score (WAZ) \u0026lt; \u0026minus;2 SD. A child was classified as undernourished if they met at least one of these criteria[32,33]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.5 Data Preprocessing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe analytic sample was randomly split into training (70%) and testing (30%) subsets, with reproducibility ensured by fixing the random seed. Ten-fold cross-validation was applied within the training data to improve generalizability and reduce overfitting.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.6 Statistical analysis and models\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFrequency and percentage distribution used to describe the characteristics of the data. Pearson\u0026rsquo;s chi-square tests were used to determine association between the outcome and explanatory variables. We applied three supervised learning models Spike and slab regression (SSR), logistic regression (LR), and Bayesian additive regression trees (BART) to predict the risk of undernutrition. These models were selected to balance interpretability, feature selection, and predictive flexibility. All statistical analyses were performed in R software (version: 4.5.1) and a p-value \u0026lt; 0.05 was considered significant.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.6.1 Logistic regression (LR)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLogistic regression serves as the conventional benchmark for binary classification problems and remains the most widely applied tool in public health and epidemiology[34]. It provides straightforward interpretability via odds ratios, making results easy to communicate to clinicians and policymakers. Although LR assumes linearity on the logit scale and does not readily capture nonlinear or higher-order interactions, its transparency and wide acceptance make it an essential comparator for evaluating the performance of more complex models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.6.2 Spike and slab regression (SSR)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpike and slab regression is a Bayesian shrinkage and variable selection approach that combines a \u0026ldquo;spike\u0026rdquo; prior, which shrinks irrelevant predictors toward zero, with a \u0026ldquo;slab\u0026rdquo; prior, which retains important predictors[35,36]. This dual mechanism makes SSR especially suited for both low and high-dimensional survey data where predictors such as socioeconomic, maternal, and child factors are often correlated. It simultaneously performs prediction and identifies the most influential risk factors, reducing the risk of overfitting compared with standard regression approaches\u0026nbsp;[37].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.6.3 Bayesian additive regression trees (BART)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBART is a nonparametric Bayesian ensemble method that represents the outcome as a sum over many regression trees [38,39]. For classification problems, BART uses a probit link function to estimate posterior probabilities of undernutrition. The method is highly flexible, capturing nonlinear effects and higher-order interactions that are difficult to pre-specify. In the context of undernutrition, where factors such as maternal BMI, child age, and wealth index may interact in complex ways, BART provides a robust predictive framework.\u003c/p\u003e\n\u003cp\u003eTogether, these three models allowed us to leverage complementary strengths: logistic regression for interpretability, SSR for parsimonious feature selection, and BART for flexibility in modeling nonlinearities and interactions. By comparing these complementary modeling approaches, we evaluated differences in predictive performance, calibration, and interpretability.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.7 Evaluation Metrics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eModel performance was evaluated on the independent test dataset. Discrimination was quantified using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Additional performance measures included accuracy, sensitivity (recall), specificity, precision, F1 score, and balanced accuracy. Calibration was assessed by comparing predicted probabilities with observed outcomes through calibration plots. Decision curve analysis (DCA) was performed to estimate the net benefit of each model across a range of threshold probabilities.\u003c/p\u003e"},{"header":"3. Results","content":"\u003cp\u003e\u003cstrong\u003e3.1 Baseline Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 4,260 children under five years of age from the BDHS 2022 dataset were included in the analysis. Among them, 35.1% were classified as undernourished, having at least one anthropometric deficit (stunting, wasting, or underweight). Table 1 presents the baseline distribution of child, maternal, and household characteristics stratified by nutritional status.\u003c/p\u003e\n\u003cp\u003eMarked regional disparities were observed across administrative divisions (p \u0026lt; 0.001). The prevalence of undernutrition was highest in Sylhet (44.0%) and Mymensingh (39.0%), whereas Khulna (29.3%) and Dhaka (30.6%) recorded the lowest proportions. Rural children exhibited slightly higher rates of undernutrition (36.1%) compared with those from urban areas (32.8%, p = 0.04).\u003c/p\u003e\n\u003cp\u003eMaternal educational attainment and nutritional status showed strong associations with child nutrition (p \u0026lt; 0.001). Undernutrition declined steadily from 55.2% among children of mothers with no education to 24.5% among those whose mothers had higher education. Similarly, children of underweight mothers had a markedly higher prevalence of undernutrition (46.8%) compared with children of obese mothers (19.9%). Increasing numbers of antenatal care visits were linked to reduced undernutrition, from 53.0% among mothers with no visits to 28.0% among those with four or more.\u003c/p\u003e\n\u003cp\u003eHousehold wealth and sanitation status also demonstrated significant gradients (p \u0026lt; 0.001). Undernutrition prevalence ranged from 47.4% in the poorest quintile to 24.6% in the richest. Children from households with improved sanitation and television ownership showed lower undernutrition prevalence than those without these facilities.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1: Baseline characteristics of children under five years of age according to undernutrition status\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCovariates\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 43px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUndernutrition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNo (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eYes (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDivision\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eBarisal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e496\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e315 (63.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e181 (36.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"8\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eChittagong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e732\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e479 (65.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e353 (34.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eDhaka\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e607\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e421 (69.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e186 (30.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eKhulna\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e450\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e318 (70.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e132 (29.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eMymensingh\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e525\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e320 (61.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e205 (39.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRajshahi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e413\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e285 (69.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e128 (31.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRangpur\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e491\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e323 (65.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e168 (34.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSylhet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e546\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e306 (56.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e240 (44.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eResidence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eUrban\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1349\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e906 (67.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e443 (32.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.04\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRural\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2911\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1861 (63.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1050 (36.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFamily Size\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u0026lt;4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1424\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e929 (65.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e495 (34.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.782\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u0026gt;=4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2836\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1838 (64.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e998 (35.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMaternal Education\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo education\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e252\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e113 (44.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e139 (55.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePrimary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e931\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e543 (58.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e388 (41.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSecondary\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2129\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1421 (66.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e708 (33.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eHigher\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e770\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e581 (75.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e189 (24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMaternal Age\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e15-19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e370\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e257 (69.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e113 (30.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"8\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.081\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e20-24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1279\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e840 (65.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e439 (34.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e25-29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1181\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e778 (65.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e403 (34.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e30-34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e797\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e510 (64.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e287 (36.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e35-39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e366\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e223 (60.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e143 (39.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e40-44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e46 (58.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e33 (41.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e45-49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e4 (40.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e6 (60.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBody Mass Index\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eUnderweight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e266 (53.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e234 (46.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNormal weight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2272\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1439 (63.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e833 (36.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eOverweight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e969\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e686 (70.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e283 (29.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eObesity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e241\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e193 (80.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e48 (19.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMaternal Working Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3095\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2016 (65.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1079 (34.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.958\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e987\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e642 (65.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e345 (35.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNo. of ANC Visits\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e181\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e85 (47.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e96 (53.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"5\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e355\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e214 (60.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e141 (39.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e444\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e303 (68.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e141 (31.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e472\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e332 (70.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e140 (29.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e4 \u0026amp; more\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e959\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e691 (72.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e269 (28.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eWealth Index\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePoorest\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e878\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e462 (52.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e416 (47.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"5\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003ePoorer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e802\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e492 (61.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e310 (38.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eMiddle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e821\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e544 (66.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e277 (33.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRicher\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e784\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e559 (71.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e225 (28.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eRichest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e797\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e601 (75.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e196 (24.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSanitation Facility\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eImproved\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2964\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1999 (67.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e965 (32.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eUnimproved\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e695\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e395 (56.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e300 (43.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource of drinking water\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eImproved\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3638\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2379 (65.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1259 (34.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.145\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eUnimproved\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e388 (62.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e243 (37.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHas Radio\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e4228\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2741 (64.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1487 (35.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.052\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e26 (81.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e6 (18.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHas Television\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2198\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1344 (61.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e854 (38.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2062\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1423 (69.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e639 (31.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHas Mobile\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e23 (53.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e20 (46.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.113\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e4217\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2744 (65.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1473 (34.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge of Child (in Months)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u0026lt;6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e323\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e323(72.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e125 (27.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e6-8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e177\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e177 (76.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e55 (23.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e9-11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1713\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e181 (73.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e66 (26.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e12-23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e6483\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e503 (62.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e300 (37.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e24-35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e6420\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e500 (60.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e324 (39.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e36-47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e6647\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e534 (63.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e309 (36.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e48-59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e6680\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e549 (63.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e314 (36.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex of Child\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2097\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1371 (65.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e726 (34.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.716\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1985\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1287 (64.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e698 (35.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBirth Order\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eFirst\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1540\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1056 (68.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e484 (31.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSecond\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1425\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e939 (65.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e486 (34.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eThird\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e712\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e456 (64.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e256 (36.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eFourth \u0026amp; more\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e263\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e140 (53.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e123 (46.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSize of child at birth\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eLarger than average\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e355\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e255 (71.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e100 (28.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eAverage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3321\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2260 (68.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1061 (32.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSmaller than average\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e584\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e315 (53.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e269 (46.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eType of Birth\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eSingle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e4026\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2626 (65.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1400 (34.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.121\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eMultiple\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e234\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e141 (60.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e93 (39.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHad Fever in last two weeks\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2783\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1799 (64.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e984 (35.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.281\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1292\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e856 (66.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e436 (33.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHad Diarrhea in last two weeks\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3869\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e2510 (64.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1359 (53.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.343\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e208\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e144 (69.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e64 (30.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHad Cough in last two weeks\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e2923\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e1874 (64.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e1049 (35.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e0.093\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 29px;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e1151\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20px;\"\u003e\n \u003cp\u003e778 (67.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\n \u003cp\u003e373 (32.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e3.2 Model Performance and Comparison\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe evaluated the performance of three machine learning models\u0026mdash;Spike and slab, Logistic Regression, and BART for predicting the risk of undernutrition in children. Figure 1 presents the discriminative performance of the three models using ROC curves. All models showed moderate ability to predict undernutrition. The Spike and slab model achieved the highest performance with an AUC of 0.63, followed by BART (AUC = 0.619) and Logistic Regression (AUC = 0.594). These findings suggest that Spike and slab provided better discrimination between undernourished and non-undernourished children compared with the alternative approaches.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2 summarizes detailed cross-validated performance metrics for the three models. The optimal cutoff values for each model were selected based on maximizing the performance metrics, with a cutoff of 0.303 for Spike and slab, 0.297 for Logistic Regression, and 0.301 for BART. The Spike and slab model exhibited superior overall performance, achieving the highest accuracy (60.43%), sensitivity (69.17%), and balanced accuracy (62.74%). Notably, Spike and slab also obtained the best F1 score (0.53), indicating an optimal balance between precision and recall. In comparison, BART demonstrated moderate performance (accuracy 58.91%, sensitivity 64.03%, F1 score 0.50), while Logistic Regression performed lowest across most indices (accuracy 58.66%, sensitivity 60.08%, F1 score 0.48).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2: Performances of the machine learning models for predicting Child Undernutrition\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEvaluation Metrics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\" style=\"width: 468px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModels\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpike and slab\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLogistic Regression\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBART\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e60.43 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e58.66 %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e58.91 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e69.17 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e60.08 %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e64.03 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e56.32%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e57.99 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e56.51 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e42.68 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e40.21 %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e40.91 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eF1 Score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e52.79 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e48.18 %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e49.92 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eBalanced Accuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e62.74 %\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e59.04 %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003e60.27 %\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eCalibration curves for the three models are displayed in Figure 2. The Spike and slab model showed the closest agreement between predicted probabilities and observed outcomes, particularly in the mid-range of risk thresholds, reflecting better reliability of its probability estimates. By contrast, Logistic Regression slightly underestimated observed risk at higher thresholds, while BART demonstrated mild overestimation at extreme probability ranges.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe decision curve analysis (Figure 3) further demonstrated the net benefit of the Spike and slab model. Spike and slab provided the highest net benefit across a wide range of threshold probabilities, particularly around the 30% probability thresholds. These results highlight that Spike and slab not only deliver accurate predictions but also maximizes the clinical utility by identifying high-risk individuals with the most significant benefit.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3 Model Interpretability Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFigure 4A presents a comprehensive swarm plot of SHAP values from the Spike and slab model. The horizontal axis represents SHAP values, indicating the direction and magnitude of each feature\u0026rsquo;s contribution to undernutrition risk, while the vertical axis lists features ordered by their cumulative impact. Each point corresponds to a single child, with yellow shades denoting high contributions to undernutrition prediction and purple shades reflecting low influences. The size of the child at birth (M18) emerged as the strongest predictor. Children born smaller than average displayed positive SHAP values, indicating a higher contribution to undernutrition risk, whereas those reported as average or larger than average suggesting a protective effect.\u003c/p\u003e\n\u003cp\u003eChild age category (Cage_cat) was the second most important predictor. In particular, children aged 24\u0026ndash;35 months had the largest positive SHAP values, reflecting the increased vulnerability to Undernutrition during the transition from exclusive breastfeeding to complementary feeding.\u003c/p\u003e\n\u003cp\u003eMaternal BMI status (BMI_cat) was also a critical determinant. Children of underweight mothers consistently showed positive SHAP values, indicating a greater risk of undernutrition. In contrast, children of overweight or obese mothers demonstrated negative SHAP contributions, suggesting a protective effect relative to the underweight group.\u003c/p\u003e\n\u003cp\u003eBeyond these biological factors, household wealth index (V190) played a strong protective role, with children from wealthier households exhibiting negative SHAP values and thus a reduced risk of undernutrition. Additional but less influential predictors included maternal age group (V013), birth order category (Border_cat), administrative division (HV024), toilet facility type (Toilet_Tp), and maternal education level (V106). These features showed smaller SHAP contributions but were directionally consistent, with higher birth order, poorer sanitation, and lower education tending to increase risk.\u003c/p\u003e\n\u003cp\u003eFigure 4B provides a detailed case study, demonstrating the model\u0026rsquo;s prediction process for an individual child. In this visualization, yellow indicators, such as being 24\u0026ndash;35 months of age and having an underweight mother, increased the predicted risk, while protective purple contributions included larger birth size and higher household wealth. For this child, the balance of these factors produced a higher-than-baseline probability of undernutrition.\u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIn this study, we developed and compared multiple machine learning models to predict the risk of childhood undernutrition by utilizing Geographical and demographic, Maternal characteristics, Child-level characteristics, and Household socioeconomic\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003efactors. Among the three models; Spike and slab, Bayesian Additive Regression Trees (BART), and Logistic Regression, the Spike and slab algorithm demonstrated superior predictive performance and calibration. The model achieved an AUC of 0.63 and provided the highest accuracy, sensitivity, and F1 score. The reliability of the model was further supported by calibration curves and decision curve analysis, which confirmed its net clinical benefit across a broad range of threshold probabilities.\u003c/p\u003e\n\u003cp\u003eRecent advances in ML have transformed predictive modeling in public health, allowing researchers to uncover non-linear relationships that traditional regression often fails to detect [40]. The Spike and slab model, a Bayesian variable selection framework, effectively addresses model uncertainty by shrinking irrelevant predictors while retaining informative ones [41]. This enables the model to manage correlated features common in nutrition data, such as maternal and socioeconomic factors, and to avoid overfitting even with limited samples [42]. In contrast, the Logistic Regression model though interpretable relies on linear assumptions that may oversimplify real-world dynamics, and the BART model, while flexible, can overfit small datasets without proper regularization [43].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe calibration curve further illustrated that the Spike and slab model\u0026rsquo;s predicted probabilities closely matched observed outcomes, particularly at mid-range risk levels. This finding suggests that the model can produce stable and realistic probability estimates rather than overestimating or underestimating risk. Previous evidence supports that Bayesian-based models tend to maintain more stable calibration in health data with class imbalance [44]. In contrast, BART tended to overpredict undernutrition at extreme values, while Logistic Regression underestimated the risk in higher thresholds. These trends are consistent with previous comparative studies, where Bayesian and ensemble methods often demonstrated superior calibration when applied to unbalanced health datasets [45].\u003c/p\u003e\n\u003cp\u003eThe decision curve analysis supported the clinical applicability of the Spike and slab model, showing the highest net benefit over a wide range of threshold probabilities, Across threshold probabilities between 20% and 40%. From a public health standpoint, this means that using the Spike and slab model could help community health workers identify high-risk children more effectively, thereby optimizing limited resources for targeted interventions. Such probabilistic frameworks are particularly relevant in low- and middle-income countries, where limited resources require prioritization of high-risk children\u0026nbsp;[46].\u003c/p\u003e\n\u003cp\u003eTo gain a deeper understanding of the model\u0026rsquo;s predictive process, we used SHAP visualization to interpret the contribution of each feature. The SHAP analysis revealed that eight major characteristics; birth size, child age, maternal BMI, household wealth index, maternal education, sanitation type, birth order, and maternal age had the most substantial influence on prediction. Among these, the size of the child at birth emerged as the strongest predictor, consistent with earlier epidemiological findings that low birth weight or small size at birth is a major risk factor for postnatal malnutrition and stunting [47]. Children born smaller than average were more likely to experience chronic growth deficits, limited nutrient absorption, and delayed recovery from infections, predisposing them to persistent undernutrition in early life.\u003c/p\u003e\n\u003cp\u003eChild age category also played an essential role, particularly among children aged 24\u0026ndash;35 months, who were identified as the most vulnerable. This period marks the transition from breastfeeding to complementary feeding, a phase when dietary inadequacy and infection exposure sharply increase [48]. Similar findings have been observed in studies conducted in sub-Saharan Africa and South Asia, where weaning-related nutritional gaps were strongly associated with growth faltering and wasting [49,50].\u003c/p\u003e\n\u003cp\u003eMaternal nutritional status, as indicated by BMI category, was another key determinant. Children born to underweight mothers showed higher predicted risks, reflecting the intergenerational transmission of malnutrition. Undernourished mothers are more likely to deliver low-birth-weight infants, have reduced breast milk quality, and face greater postpartum health challenges [51].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSocioeconomic and environmental variables also emerged as critical protective factors. Children from households in higher wealth quintiles or with improved toilet facilities exhibited significantly lower SHAP contributions toward undernutrition risk. These patterns reinforce the established association between economic stability, hygiene, and nutritional health. Improved sanitation reduces infection exposure, while wealth and education enhance access to quality food, healthcare, and awareness of child-feeding practices\u0026nbsp;[52,53]. Collectively, these findings highlight that undernutrition is not driven by a single factor but results from the interaction of maternal, child, and household-level influences.\u003c/p\u003e\n\u003cp\u003eInterestingly, administrative region and maternal age demonstrated smaller but directionally consistent effects. Younger mothers, particularly those below 20 years of age, tended to have higher predicted risks for their children. Early motherhood often coincides with limited education and economic dependence, which constrain childcare capacity and nutritional awareness. These contextual insights align with prior findings that maternal empowerment and household decision-making autonomy are protective against childhood stunting and wasting\u0026nbsp;[54,55].\u003c/p\u003e\n\u003cp\u003eFrom a methodological standpoint, This study suggests that Bayesian shrinkage approaches such as Spike and Slab regression may provide improved sensitivity and calibration compared with standard logistic regression in this context. By incorporating prior information and probabilistic shrinkage, these models maintain parsimony while effectively capturing non-linear interactions [56,57]. Moreover, SHAP interpretability bridges the gap between technical model outputs and actionable insights, making predictive tools more transparent and clinically relevant [58].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.1 Limitations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDespite the promising results, several limitations should be acknowledged. First, the dataset was cross-sectional, limiting causal interpretation of identified relationships. While the models can detect associations, they cannot determine temporal order or causality. Future longitudinal studies are needed to examine whether the identified predictors consistently precede and contribute to the onset of undernutrition.\u003c/p\u003e\n\u003cp\u003eSecond, some potentially influential variables, such as dietary diversity, and breastfeeding frequency were unavailable in the dataset. Their exclusion may have constrained the overall accuracy and reduced the model\u0026rsquo;s ability to account for behavioral and dietary influences. Finally, socioeconomic and regional disparities in data reporting may introduce bias, and improved standardization of national nutrition datasets would enhance future modeling efforts.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eWe introduced a Bayesian machine learning model to predict the risk of childhood undernutrition. Among the compared models, the Bayesian Spike and Slab regression achieved the highest predictive performance and calibration reliability. Integration of SHAP analysis improved model interpretability, enabling a clearer understanding of the key determinants such as birth size, child age, maternal BMI, household wealth, and sanitation that contribute to undernutrition risk. By providing transparent and clinically meaningful intuitions, this interpretable machine learning framework offers a promising tool to support early identification of high-risk children and guide targeted nutrition interventions to expand child health outcomes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e6. Ethics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e7. Consent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e8. Clinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e9. Availability of data\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analyzed during the current study are available in the Demographic and Health Surveys (DHS) Program repository https://dhsprogram.com/data/. DHS data are publicly available upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e10. Competing interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declared no competing interests in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e11. Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors received no funding for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e12. Acknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors want to acknowledge the Measures DHS data archive for providing us with the datasets for further analysis.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eE.A.: Conceptualization, Methodology, Software, Formal analysis, Data curation, Validation, Visualization, Writing\u0026mdash;original draft.A.Z.: Methodology refinement, Supervision, Interpretation of results, Writing\u0026mdash;review \u0026amp; editing.Both authors approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWorld Health Organization. Malnutrition. (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/news-room/fact-sheets/detail/malnutrition\u003c/span\u003e\u003cspan address=\"https://www.who.int/news-room/fact-sheets/detail/malnutrition\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmoadu, M. et al. Risk Factors of Malnutrition among In-School Children and Adolescents in Developing Countries: A Scoping Review. \u003cem\u003eChildren\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 476 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorales, F., Montserrat-de la Paz, S., Leon, M. J. \u0026amp; Rivero-Pino, F. Effects of Malnutrition on the Immune System and Infection and the Role of Nutritional Strategies Regarding Improvements in Children\u0026rsquo;s Health Status: A Literature Review. \u003cem\u003eNutrients\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 1 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUNICEF. Malnutrition in children. (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.unicef.org/topic/nutrition/malnutrition\u003c/span\u003e\u003cspan address=\"https://data.unicef.org/topic/nutrition/malnutrition\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUNICEF, WHO \u0026amp; WORLD BANK. Level and trend in child malnutrition. \u003cem\u003eWorld Health Organization\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGlobal Nutrition Report. The burden of malnutrition around the world. (2023). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://globalnutritionreport.org/reports/global-nutrition-report-2018/burden-malnutrition\u003c/span\u003e\u003cspan address=\"https://globalnutritionreport.org/reports/global-nutrition-report-2018/burden-malnutrition\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRedwan, T. \u0026amp; Shabur, M. A. Analyzing the socio-economic factors of adolescent malnutrition in Bangladesh using Best Worst Method. \u003cem\u003eDiscover Sustain.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, 1\u0026ndash;16 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSahiledengle, B. et al. Factors associated with childhood undernutrition in poor Ethiopian households: Implications for public health interventions. \u003cem\u003ePLoS One\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e, (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuda, T. M., Hayes, A., El Arifeen, S. \u0026amp; Dibley, M. J. Social determinants of inequalities in child undernutrition in Bangladesh: A decomposition analysis. \u003cem\u003eMatern Child. Nutr.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, e12440 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMomin Islam, M. \u0026amp; Noor, F. M. Determinants of Malnutrition among Under-Five Children in Bangladesh: Evidence from Multiple Indicator Cluster Survey, 2019 Data. \u003cem\u003eData Fortune J. Health Sci.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, 284\u0026ndash;295 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDas, S., Baffour, B. \u0026amp; Richardson, A. Prevalence of child undernutrition measures and their spatio-demographic inequalities in Bangladesh: an application of multilevel Bayesian modelling. \u003cem\u003eBMC Public. Health\u003c/em\u003e. \u003cb\u003e22\u003c/b\u003e, 1\u0026ndash;21 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam, M. M., Jahan Kibria, S., Kumar, N. M., Roy, S., Karim, M. R. \u0026amp; D. C. \u0026amp; Prediction of undernutrition and identification of its influencing predictors among under-five children in Bangladesh using explainable machine learning algorithms. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e19\u003c/b\u003e, e0315393 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirk, D. et al. Machine Learning in Nutrition Research. \u003cem\u003eAdv. Nutr.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 2573\u0026ndash;2589 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMccrory, M. A. et al. Interaction Analysis Based on Shapley Values and Extreme Gradient Boosting: A Realistic Simulation and Application to a Large Epidemiological Prospective Study. \u003cem\u003eFront. Nutr.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 871768 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, X. et al. An interpretable machine learning model for precise prediction of biomarkers for intermittent fasting pattern. \u003cem\u003eNutr. Metabolism\u003c/em\u003e. \u003cb\u003e21\u003c/b\u003e, 1\u0026ndash;8 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBegashaw, G. B., Zewotir, T. \u0026amp; Fenta, H. M. A deep learning approach for classifying and predicting children\u0026rsquo;s nutritional status in Ethiopia using LSTM-FC neural networks. \u003cem\u003eBioData Min.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 1\u0026ndash;24 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBitew, F. H., Sparks, C. S. \u0026amp; Nyarko, S. H. Machine learning algorithms for predicting undernutrition among under-five children in Ethiopia. \u003cem\u003ePublic. Health Nutr.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 269 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirk, D. Good practices and common pitfalls of machine learning in nutrition research. \u003cem\u003eProceedings of the Nutrition Society\u003c/em\u003e 1\u0026ndash;14 (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1017/S0029665124007638\u003c/span\u003e\u003cspan address=\"10.1017/S0029665124007638\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTalukder, A. \u0026amp; Ahammed, B. Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh. \u003cem\u003eNutrition\u003c/em\u003e \u003cb\u003e78\u003c/b\u003e, (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam, M. M., Jahan Kibria, S., Kumar, N. M., Roy, S., Karim, M. R. \u0026amp; D. C. \u0026amp; Prediction of undernutrition and identification of its influencing predictors among under-five children in Bangladesh using explainable machine learning algorithms. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e19\u003c/b\u003e, e0315393 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSara, S. S., Khan, M. S. \u0026amp; Talukder, A. Prediction of Child Stunting with Machine Learning Algorithms: A Cross-Country Study of Bangladesh, India, and Nepal. (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21203/RS.3.RS-4696630/V1\u003c/span\u003e\u003cspan address=\"10.21203/RS.3.RS-4696630/V1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e doi:10.21203/RS.3.RS-4696630/V1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, X., Usman, M., Irshad, A. R., Rashid, M. \u0026amp; Khattak, A. Investigating Spatial Effects through Machine Learning and Leveraging Explainable AI for Child Malnutrition in Pakistan. \u003cem\u003eISPRS Int. J. Geo-Information 2024\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e, 330 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHemo, S. A. \u0026amp; Rayhan Md. I. Classification tree and random forest model to predict under-five malnutrition in Bangladesh. \u003cem\u003eBiom Biostat Int. J.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 116\u0026ndash;123 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkombi, B. J. et al. Stunting and severe stunting among children under-5 years in Nigeria: A multilevel analysis. \u003cem\u003eBMC Pediatr\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDas, S. \u0026amp; Rahman, R. M. Application of ordinal logistic regression analysis in determining risk factors of child malnutrition in Bangladesh. \u003cem\u003eNutr J\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChowdhury, T. R., Chakrabarty, S., Rakib, M., Saltmarsh, S. \u0026amp; Davis, K. A. Socio-economic risk factors for early childhood underweight in Bangladesh. \u003cem\u003eGlobal Health\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChowdhury, M. R. K. et al. The prevalence and socio-demographic risk factors of coexistence of stunting, wasting, and underweight among children under five years in Bangladesh: a cross-sectional study. \u003cem\u003eBMC Nutr.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 1\u0026ndash;12 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaque, S. et al. Nutritional status of under-five aged children of ready-made garment workers in Bangladesh: A cross-sectional study. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e18\u003c/b\u003e, e0284325 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumar, P., Rashmi, R., Muhammad, T. \u0026amp; Srivastava, S. Factors contributing to the reduction in childhood stunting in Bangladesh: a pooled data analysis from the Bangladesh demographic and health surveys of 2004 and 2017\u0026ndash;18. \u003cem\u003eBMC Public. Health\u003c/em\u003e. \u003cb\u003e21\u003c/b\u003e, 1\u0026ndash;14 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHossain, M. I. et al. Risk or associated factors of wasting among under-five children in Bangladesh: A systematic review. \u003cem\u003eAsia Pac. J. Clin. Nutr.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e, 457 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkter, S., Siriphon, A., Ayuttacorn, A. \u0026amp; Boonchieng, W. Prevalence of ARI, fever, and diarrhea among under-five children and the influencing factors in southwestern coastal region of Bangladesh. \u003cem\u003eBMC Public. Health\u003c/em\u003e. \u003cb\u003e25\u003c/b\u003e, 2951 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHalima, O. et al. Identifying Individual and Household Level Predictors of Undernutrition Among 6\u0026ndash;59 Months Children in Bangladesh: A Multivariate Approach. \u003cem\u003ePublic. Health Challenges\u003c/em\u003e. \u003cb\u003e3\u003c/b\u003e, e70007 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBirhanu, F., Yitbarek, K., Bobo, F. T., Atlantis, E. \u0026amp; Woldie, M. Undernutrition in children under five associated with wealth-related inequality in 24 low- and middle-income countries from 2017 to 2022. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 1\u0026ndash;9 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHosmer, D. W., Lemeshow, S. \u0026amp; Sturdivant, R. X. Applied Logistic Regression: Third Edition. \u003cem\u003eAppl. Logistic Regression: Third Ed.\u003c/em\u003e \u003cb\u003e1\u0026ndash;510\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/9781118548387\u003c/span\u003e\u003cspan address=\"10.1002/9781118548387\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIshwaran, H., Kogalur, U. B., Sunil Rao, J. \u0026amp; Spikeslab Prediction and variable selection using spike and slab regression. \u003cem\u003eR J.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e, 68\u0026ndash;73 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIshwaran, H. \u0026amp; Rao, J. S. Spike and slab variable selection: Frequentist and Bayesian strategies. (2005). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1214/009053604000001147\u003c/span\u003e\u003cspan address=\"https://doi.org/10.1214/009053604000001147\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e 33, 730\u0026ndash;773\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHahn, P. R. \u0026amp; Carvalho, C. M. Decoupling shrinkage and selection in Bayesian linear models: a posterior summary perspective. (2014). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://arxiv.org/abs/1408.0464\u003c/span\u003e\u003cspan address=\"http://arxiv.org/abs/1408.0464\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe, L., Cao, L., Wang, T., Cao, Z. \u0026amp; Shi, X. A Bayesian Additive Regression Trees Framework for Individualized Causal Effect Estimation. \u003cem\u003eMath. 2025\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e, 2195 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChipman, H. A., George, E. I. \u0026amp; McCulloch, R. E. BART: Bayesian additive regression trees. (2010). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1214/09\u003c/span\u003e\u003cspan address=\"https://doi.org/10.1214/09\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cem\u003e-AOAS285\u003c/em\u003e 4, 266\u0026ndash;298.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanshika, Gupta, N. \u0026amp; Machine Learning Applications in Healthcare. \u003cem\u003e10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions), ICRITO 2022\u003c/em\u003e (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ICRITO56286.2022.9964865\u003c/span\u003e\u003cspan address=\"10.1109/ICRITO56286.2022.9964865\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022) doi:10.1109/ICRITO56286.2022.9964865.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO\u0026rsquo;Hara, R. B. \u0026amp; Sillanp\u0026auml;\u0026auml;, M. J. A review of Bayesian variable selection methods: what, how and which. (2009). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1214/09\u003c/span\u003e\u003cspan address=\"https://doi.org/10.1214/09\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cem\u003e-BA403\u003c/em\u003e 4, 85\u0026ndash;117.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSinha, S., Mallick, B. K., Kipnis, V. \u0026amp; Carroll, R. J. Semiparametric bayesian analysis of nutritional epidemiology data in the presence of measurement error. \u003cem\u003eBiometrics\u003c/em\u003e \u003cb\u003e66\u003c/b\u003e, 444\u0026ndash;454 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu, Z. et al. Integrating Machine Learning and the SHapley Additive exPlanations (SHAP) Framework to Predict Lymph Node Metastasis in Gastric Cancer Patients Based on Inflammation Indices and Peripheral Lymphocyte Subpopulations. \u003cem\u003eJ. Inflamm. Res.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 9551 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHong, Y. et al. Machine learning prediction of metabolic dysfunction-associated fatty liver disease risk in American adults using body composition: explainable analysis based on SHapley Additive exPlanations. \u003cem\u003eFront. Nutr.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 1616229 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteyerberg, E. W. Clinical Prediction Models. (2019). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-16399-0\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-16399-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e doi:10.1007/978-3-030-16399-0.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTalukder, A. \u0026amp; Ahammed, B. Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh. \u003cem\u003eNutrition\u003c/em\u003e \u003cb\u003e78\u003c/b\u003e, (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeats, E. C. et al. Effective interventions to address maternal and child malnutrition: an update of the evidence. \u003cem\u003eLancet Child. Adolesc. Health\u003c/em\u003e. \u003cb\u003e5\u003c/b\u003e, 367\u0026ndash;384 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVictora, C. G. et al. Revisiting maternal and child undernutrition in low-income and middle-income countries: variable progress towards an unfinished agenda. \u003cem\u003eLancet\u003c/em\u003e \u003cb\u003e397\u003c/b\u003e, 1388\u0026ndash;1399 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnteneh, R. M. et al. Wealth-related inequalities in undernutrition among under-five children in sub-Saharan Africa. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1\u0026ndash;14 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahman, M. S., Howlader, T., Masud, M. S. \u0026amp; Rahman, M. L. Association of Low-Birth Weight with Malnutrition in Children under Five Years in Bangladesh: Do Mother\u0026rsquo;s Education, Socio-Economic Status, and Birth Interval Matter? \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e11\u003c/b\u003e, e0157814 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAguayo, V. M., Nair, R., Badgaiyan, N. \u0026amp; Krishna, V. Determinants of stunting and poor linear growth in children under 2 years of age in India: an in-depth analysis of Maharashtra\u0026rsquo;s comprehensive nutrition survey. \u003cem\u003eMatern Child. Nutr.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e (Suppl 1), 121\u0026ndash;140 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBitew, F. H., Sparks, C. S. \u0026amp; Nyarko, S. H. Machine learning algorithms for predicting undernutrition among under-five children in Ethiopia. \u003cem\u003ePublic. Health Nutr.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 269 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin, J. \u0026amp; Feng, X. L. Exploring the impact of water, sanitation and hygiene (WASH), early adequate feeding and access to health care on urban\u0026ndash;rural disparities of child malnutrition in China. \u003cem\u003eMatern Child. Nutr.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, e13542 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahman, M. M., Saima, U. \u0026amp; Goni, M. A. Impact of maternal household decision-making autonomy on child nutritional status in Bangladesh. \u003cem\u003eAsia Pac. J. Public. Health\u003c/em\u003e. \u003cb\u003e27\u003c/b\u003e, 509\u0026ndash;520 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaul, P. \u0026amp; Saha, R. Is maternal autonomy associated with child nutritional status? Evidence from a cross-sectional study in India. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e17\u003c/b\u003e, e0268126 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu, Z. \u0026amp; Lou, W. Bayesian approaches to variable selection: a comparative study from practical perspectives. \u003cem\u003eInt. J. Biostatistics\u003c/em\u003e. \u003cb\u003e18\u003c/b\u003e, 83\u0026ndash;108 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalsiner-Walli, G. \u0026amp; Wagner, H. Comparing Spike and Slab Priors for Bayesian Variable Selection. \u003cem\u003eAustrian J. Statistics\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePonce-Bobadilla, A. V., Schmitt, V., Maier, C. S., Mensing, S. \u0026amp; Stodtmann, S. Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. \u003cem\u003eClin. Transl Sci.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, e70056 (2024).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Child undernutrition, Machine learning, Bayesian model, SHAP, Bangladesh Demographic and Health Survey","lastPublishedDoi":"10.21203/rs.3.rs-8871095/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8871095/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eChild undernutrition remains a persistent public health challenge in Bangladesh, contributing substantially to growth retardation, infection risk, and early childhood mortality. The multiple biological, maternal, and socioeconomic factors that influence nutritional outcomes often interact in nonlinear ways, making prediction difficult through conventional statistical models. This study aimed to introduce an interpretable Bayesian machine learning (ML) model capable of predicting childhood undernutrition and identifying its most influential determinants using nationally representative data.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eData were extracted from 4,260 children under five years of age from the 2022 Bangladesh Demographic and Health Survey (BDHS). Three supervised ML algorithms\u0026mdash;Logistic Regression, Bayesian Spike-and-Slab Regression (SSR), and Bayesian Additive Regression Trees (BART)\u0026mdash;were trained and validated through ten-fold cross-validation. Model discrimination, calibration, and clinical utility were assessed using the area under the ROC curve (AUC), calibration curve and decision-curve analysis, respectively. SHapley Additive exPlanations (SHAP) values were applied to evaluate and visualize variable importance and to make the result interpretable.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe Bayesian Spike-and-Slab model achieved the best overall performance in discremination (AUC\u0026thinsp;=\u0026thinsp;0.63) and exhibited stable calibration. Decision-curve analysis further confirmed that this model provided the highest net clinical benefit across a wide range of threshold probabilities. SHAP interpretation identified small birth size, low maternal body mass index, limited maternal education, poor household wealth, inadequate sanitation, higher birth order, and older child age as the most influential factors associated with undernutrition.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe Bayesian Spike-and-Slab model provided a transparent and reliable framework for predicting childhood undernutrition in Bangladesh. SHAP analysis enhanced interpretability, clarified the contribution of major determinants, and offered evidence to support early detection and targeted nutrition strategies in resource-limited settings.\u003c/p\u003e","manuscriptTitle":"Interpretable Bayesian Machine Learning Models for Predicting Undernutrition among under-Five Children","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-19 08:09:20","doi":"10.21203/rs.3.rs-8871095/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-04T18:47:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-28T02:29:09+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-26T04:50:32+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-21T17:03:27+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-19T17:34:57+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-18T17:24:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"118381919139130110891499039695192323353","date":"2026-04-17T02:22:21+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"244827139987633391260801564518864568748","date":"2026-04-13T13:12:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"102298232569817261210416626891577721043","date":"2026-04-11T09:30:17+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-11T06:37:14+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-11T04:35:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"270339238680879995165094350838578689918","date":"2026-04-11T01:01:57+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"151706520797468796252739301138856071246","date":"2026-04-10T23:08:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"245290122293794864122639160608352749121","date":"2026-04-10T20:07:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"209496768562652437834541709114717745929","date":"2026-04-10T19:59:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"66118408856353412633326184285892759028","date":"2026-04-10T19:31:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"117964703768553488828816842236637430193","date":"2026-04-09T01:18:17+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-08T21:22:17+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-18T12:16:01+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-14T11:38:23+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-14T11:36:33+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-02-13T11:01:17+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"94207cd7-b77a-48e9-ad0c-f3db4d10d362","owner":[],"postedDate":"April 19th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-04T18:47:32+00:00","index":95,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":66312729,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":66312730,"name":"Health sciences/Diseases"},{"id":66312731,"name":"Health sciences/Health care"},{"id":66312732,"name":"Physical sciences/Mathematics and computing"},{"id":66312733,"name":"Health sciences/Medical research"},{"id":66312734,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2026-04-19T08:09:23+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-19 08:09:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8871095","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8871095","identity":"rs-8871095","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}