The burden of anthropometric failure and child mortality in India

The public health burden of nutritional deficiency and child mortality is the major challenge India is facing upfront. In this context, using National Family Health Survey, 2015–16 data, this study estimated rate of composite index of anthropometric failure (CIAF) among Indian children by their population characteristics, across states and examined the multilevel contextual determinants. We further investigated district level burden of infant and child mortality in terms of multiple anthropometric failure prevalence across India. The multilevel analysis confirms a significant state, district and PSU level variation in the prevalence of anthropometric failures. Factors like- place of residence, household’s economic wellbeing, mother’s educational attainment, age, immunization status and drinking water significantly determine the different forms of multiple anthropometric failures. Wealth status of the household and mother’s educational status show a clear gradient in terms of the estimated odds ratios. The district level estimation of infant and child mortality demonstrates that districts with higher burden of multiple anthropometric failures show elevated risk of infant and child mortality. Unlike previous studies, this study does not use the conventional indices, instead considered the CIAF to identify the exact and severe form of undernutrition among Indian children and the associated nexus with infant and child mortality at the district level.


Scientific Reports
| (2020) 10:20991 | https://doi.org/10.1038/s41598-020-76884-8 www.nature.com/scientificreports/ present alarming situation of undernutrition among the Indian children 12 . To combat with the situation, policies at national and at sub-national level are formulated to reduce the burden of child undernutrition and the associated morbidities among the children. Since the International Conference on Population and Development in 1994 at Cairo, the paradigm shift in India's Population Policy embodied the National Population Policy in the year 2000 where child health and nutrition received much emphasis to be targeted. Government of India (GOI) took several initiatives to improve the living standard and socio-economic status of the population, to improve the maternal and child health with an emphasis to family planning and reproductive health services with a long term objective for sustainable economic growth, social development and environmental protection. Child mortality is another demographic phenomenon which is of immense public health concern globally. Child mortality is measured by two important indicators-under five mortality (probability of dying before reaching age five) and infant mortality (probability of dying before reaching age one). Under five mortality of any given population is one of the important indicators which reflects the health and socio-economic status of that population group in any given period of time 13 . Globally, under five mortality rate declined from 150 per 1000 live birth in 1970 to 67 per 1000 live births in 2010 13 . Parallel to global reduction in under-five mortality, India has also shown a consistent decline in infant mortality and under five mortality but India failed to meet the target of reducing under-five mortality by two-third to achieve the Millennium Development Goal four (MDG-4) by 2015. Though India has shown a decline from 190 per live births in 1990 to 64 per 1000 live births in 2011 14 . Since the Alma Ata declaration in 1978, Government of India took several initiatives to reduce child mortality. The National Diarrhoeal Disease Control Programme began in 1978. The Universal Immunization Programme and oral rehydration therapy (ORT) started in 1985. The Acute Respiratory Infection (ARI) Control programme started in 1990.
Poor nutritional status among children is associated with death and solely contributes to half of all the deaths among the children in developing countries [15][16][17] . Previous studies also established the association between malnutrition and childhood morbidity suggesting that children with anthropometric failure are at higher risk of childhood morbidity 7,[18][19][20] . Thus this study examined the undernutrition prevalence in terms of anthropometric failures which gives the exact estimate of undernutrition in any given period of time in the sub-population and the toll of infant and child mortality across districts of India. Whilst, it is important to estimate the exact pattern of undernutrition among the children for a geographically diverse country like India, few studies and no population level studies have investigated the potential linkages between composite index of anthropometric failure (CIAF) unlike the regular measures of undernutrition like stunting, wasting and underweight with infant and under five mortality across the districts of India. While conducting the experiment, this study hypothesized that districts with higher prevalence of multiple anthropometric failures carry higher burden of infant and child mortality. Beforehand, we identified the potential risk factors of anthropometric failure and further examined the variations in child and infant mortality across the districts in terms of the district level burden of multiple anthropometric failures.

Methods
Data source. Unit level data from National Family Health Survey of 2015-16 (NFHS-4) has been used in this study. The data is publicly available at https ://dhspr ogram .com/data/ and thus requires no ethical approval further. International Institute for Population Sciences (IIPS), India being the nodal agency approved all the survey protocols. National Family Health Survey round four, 2015-16 is one of the largest demographic and health survey being carried out in 640 districts of India. . The sample size of the survey constitutes of 6,99,686 women and 1,03,525 men from 6,01,509 households across India. The survey is intended to provide important indicators on population, health and nutrition. Necessary information on socio-demographic characteristics, marriage, fertility, children's immunizations and childcare, nutrition, contraception, fertility preference, sexual behaviour, attitudes towards gender roles, HIV/AIDS, anthropometric measurements are collected. NFHS-4 adopts two stage stratified probability proportional to size sampling design where census villages and urban blocks are the first stage unit for rural and urban areas respectively, and the households are the second stage unit 21

Unit level study variables
Outcome variable. The key outcome variable of the unit level analysis is the composite index of anthropometric Failure (CIAF), which is calculated using stunting, wasting, underweight status of the study children. The CIAF variable is generated using the definition of anthropometric failure 4  Note. As per definition, the grouping of CIAF does not contain a group of children with both stunted and wasted.
Exposure variables. District level percentages of all the concerned variables are estimated and validated through NFHS-4 report. The set of independent variables included in the district level study are the following: proportion of rural children, proportion of poor-the lowest two wealth quintiles-of the already calculated wealth index using household assets information in the NFHS-4 dataset 24,25 . Proportion of Scheduled Caste (SC) and Scheduled Tribe (ST) children, proportion of mother with no education, proportion of female child, proportion of home delivery, proportion of no/partial immunization, proportion of children with no access to safe drinking water and proportion of children with no access to improved sanitation.

Statistical analyses Multilevel analysis.
A four level hierarchical model is considered in this analysis because each of the level has a specific topographical, social and environmental importance that could potentially influence child's nutritional status 26,27 . Given a hierarchical structured data, multilevel modeling is always advantageous and helps to estimate the variance at different levels that are conceptualized under the study framework 27 The sampling design of NFHS data has its own hierarchy where children are apparently nested within PSUs, PSUs are nested within districts and districts are nested into states. According to NFHS-4, PSUs are the villages in rural areas and census enumeration blocks in urban areas and hence it is likely that children within PSUs possess similar characteristics and differs between PSUs. As per the Indian federal structure, district is the second level administrative area and state is the first-level administrative area where health specific governmental policies and programs are implemented and interventions take place. So taking care of the survey design and the hierarchical nature of the data, we employed a four-level multilevel structured modeling with children at level 1, nested within PSUs at level-2, nested within districts at level-3 and finally districts, nested within states at level 4 in order to account for the cluster sampling design and decompose the variation in child nutrition at child, PSU, districts and at state level 28 . Here we employed four level random intercept binary logistic model to estimate the corresponding probability (π ijkl ) for the i-th child from the j-th PSU, k-th district and l-th state who suffer from any specific forms of multiple anthropometric failure. And the corresponding probability could be denoted as π ijkl = Pr(y ijkl = 1) as: logit π ijkl = β 0 + β m X ijkl + f 0l + v 0kl + u 0jkl + e 0ijkl where m is the number of independent variables included in the model 29 . The parameter, β 0 , is the intercept which is the only fixed term in the model. And f 0l , v 0kl , u 0jkl and e 0ijkl are the random effects, the residual differentials at state, district, PSU and at child level respectively. Random effects are assumed to be independent and normally distributed with a mean of 0 and variance of σ 2 f0 σ 2 v0 , σ 2 u0 and σ 2 e0 respectively 29 . These variances quantify the between-states (σ 2 f0 ) , between-districts (σ 2 v0 ), between-PSUs (σ 2 u0 ) and between-children ( σ 2 e0 ) variations respectively in the log odds of the events under study, conditioned on the characteristics at different levels.
As the variance estimate of the lowest level cannot be obtained from the model, rest of the variances for the next higher levels are assumed to be a function of the binomial distribution. From the estimated variance of the random effects, proportion of variation known as the variance partitioning coefficients (VPCs) are calculated 29, 30 . As child is the lowest level of this study framework, we assumed the between-children variation to be the variance of the standard logistic distribution as π 2 /3 = 3.29 30,31 . Thus for any level x, the VPC can be calculated using the following formula: VPC x = σ 2 x0 /(σ 2 f0 + σ 2 v0 + σ 2 u0 + 3.29). First-order predictive (or penalized) quasi-likelihood (PQL) estimation is used to approximate the linearization with the help of a Taylor series expansion which transforms the discrete binary response model to continuous model 32  Prevalence of anthropometric failure. Table 2 shows the prevalence of different forms of multiple anthropometric failures among the children by their background characteristics. It is found that a significant portion of the children (44%) in India do not show any kind of anthropometric failure but still a large proportion (56%) of the children carry some form of anthropometric failure and in many cases (34%) multiple anthropometric failures. It is found that the prevalence of multiple anthropometric failures among urban children (36%) is comparatively higher than the rural children (27%). The wealth pattern of CIAF shows that, children from the poorest wealth quintile show the highest prevalence of multiple anthropometric failures. It is observed that 9% of the poorest children are suffering from wasting & underweight, 10% of them suffer from wasting, stunting and underweight simultaneously while 27% of them suffer from stunting and underweight. The education (mother's) pattern of anthropometric failure shows a decreasing prevalence among children with increasing educational attainment among their mothers. The prevalence of simultaneous occurrence of wasting, stunting and underweight is observed highest (9%) among those children whose mother did not achieve any formal educational qualification. Similarly, children of no educated mothers show highest prevalence (27%) in terms of concurrent occurrence of stunting & underweight and in terms of concurrent occurrence of wasting & underweight (9%). The social class pattern of CIAF shows that children from the scheduled caste and scheduled tribe category carry the higher burden of anthropometric failure than rest of the social groups. Birth order of the child shows differential in the prevalence of anthropometric failure. Higher ordered births show higher burden of multiple anthropometric failures. Around 66% of the children of the fourth or higher ordered births carry some form of the anthropometric failures. And it is observed that children of fourth or higher ordered births carry the highest (26%) prevalence of stunting and underweight. Male-female pattern of anthropometric failure does not show any substantial differential. Children aged between 2-5 years carry higher burden of CIAF than rest of the children. Similarly, the state pattern of anthropometric failures is shown in Table 3.

Risk factors of anthropometric failure in India, 2015-16.
Four level random intercept model proposed in the methodology is applied to assess the socio-economic and demographic correlates of multiple anthropometric failures among Indian children within a multilevel framework (Table 4). In all three cases the statistically significant random intercept indicates considerable variation in concurrent occurrence of stunting & underweight, wasting & underweight and wasting, stunting & underweight among Indian children between states, between districts, between PSUs and among children.
The concurrent occurrence of stunting & underweight is 1.09 (AOR = 1.086; p-value < 0.01) times more likely among the children residing in urban areas than rural areas ( Table 4). The odd is almost same for the other two categories of multiple anthropometric failures (AF) among the urban children. The economic status of a household is inversely associated with each of the indicators of multiple AF. Children from the richest wealth quintile are 61% less likely (AOR = 0.39; p-value < 0.01) to suffer from stunting & underweight, 35% (AOR = 0.65; p-value < 0.01) less likely to suffer from wasting & underweight and 54% (AOR = 0.46; p-value < 0.01) less likely to suffer from the simultaneous occurrence of wasting, stunting and underweight than those children from the poorest wealth quintile. Similarly, children from the middle and richer wealth quintiles are also substantially less likely to suffer from multiple AF. Among the children from different social groups; ST, OBC and Others show comparatively lower likelihood to multiple AF than the reference category which means children from SC category bear the highest risk of stunting & underweight, wasting & underweight and wasting, stunting & underweight. Mother's educational attainment shows a very strong statistical association with each of the indicators of www.nature.com/scientificreports/ multiple AF. Thus it is observed that children of higher educated mother are 48% (AOR = 0.52; p-value < 0.01) less likely to suffer from stunting & underweight, 26% (AOR = 0.74; p-value < 0.01) less likely to suffer from wasting & underweight and 39% (AOR = 0.61; p-value < 0.01) less likely to suffer from wasting, stunting and underweight than those children of no educated mothers. Children of secondary and primary educated mother are also showing lower odds to suffer from each of the types of multiple AFs. Children of higher birth orders show comparatively higher odds of multiple AF. Children of four or higher ordered birth show 17% (AOR = 1.17; p-value < 0.01) higher chance of stunting & underweight and 16% (AOR = 1.16; p-value < 0.01) higher chance of wasting, stunting and underweight than the first ordered births. Apparently, female children show lower likelihood to multiple AF than the male children. It is also found that fully immunized children carry higher likelihood to multiple AF than the rest of the children. Like, fully immunized children are 14% (AOR = 1.14; p-value < 0.01) more likely to be stunted & underweight and 22% more likely to be wasted, stunted & underweight. . District level patterns of infant and child mortality. www.nature.com/scientificreports/ teristics of the districts, proportion of rural, proportion of poor, proportion of no educated mothers, proportion of higher ordered births, percentage of full immunization and proportion of children with no access to improved sanitation show substantial differential in IMR and CMR across the districts.

Geographical variation in multiple anthropometric failures.
To check the scatteredness of district level rates of infant and child mortality we plotted the two way scatter plots between the measures of mortality and the measures of multiple anthropometric failures. Figures 1, 2 District level risk factors of infant mortality. Extending the traditional regression models, the application of generalised linear model allows us to model the district level mean response of IMR based upon the explanatory variables under the study framework through the log link function which assumes the response variable to be a member of exponential family. As both the Shapiro-Wilk (W = 0.989; p-value = 0.0001) and the Shapiro-Francia (W = 0.992; p-value = 0.0017) normality test confirmed the district level distribution of IMR to be non-normal and we assumed the outcome of interest to be a member of the exponential family because the response variable is discrete and positively skewed. And at the same time, the exponential family gives a bet- www.nature.com/scientificreports/ ter model fit than rest of the assumptions. The generalised linear model with exponential family and log link assumes the multiplicative effects on the original outcome by the predictors. Model 1a gives the estimated coefficients for stunting & underweight (Table 6). Here the coefficients show the magnitude of increase or decrease in the log arithmetic mean of IMR. It is found that district level prevalence of concurrent occurrence of stunting & underweight significantly predicts the district level burden of infant mortality across India. The estimated coefficient (β) of "stunting & underweight" is 0.017 (p-value < 0.001)) which indicates a strong association between concurrent occurrence of stunting & underweight among the children and infant mortality across the districts of India. This suggests that the probability of infant mortality among the children under age one increases with the increase in the prevalence of concurrent occurrence of stunting & underweight among the children across the districts. It is evident that one unit increase in the prevalence of concurrent occurrence of stunting & underweight is statistically associated with 0.017 unit increase in the log arithmetic mean of IMR when adjusted for other variables. More specifically, the log arithmetic mean of IMR will be 1.017 times higher with every one unit increase in the rate of concurrent occurrence of stunting & underweight across the districts. Model 1b also shows a statistically significant association (β = 0.005; p-value = 0.008) between the concurrent occurrence of wasting & underweight and IMR. This indicates the districts with higher concurrent occurrence of wasting & underweight bear higher infant mortality. Similarly, the district level prevalence of concurrent occurrence of wasting, stunting & underweight does appear to be highly statistically significant predictor of district level IMR (Model 1c). From the model estimation, it is observed that district level variation in terms of each of the multiple anthropometric failures significantly predicts the district level variation in infant mortality in India.
Notably, variables like district level proportion of poor, proportion of four or higher ordered births, district level proportion of households with unimproved drinking water (except Model 1a) and unimproved sanitation show statistically significant and positive association with IMR throughout the models. This indicates that district level increase in poverty (proportion of poor) is associated with increase in (β = 0.003; p-value < 0.001)    www.nature.com/scientificreports/ IMR among children (Model 1a). Similarly, districts with higher percentages of four or higher ordered births carry higher burden of IMR. This suggests that infant mortality is much more prevalent among the children of higher ordered births. Drinking water and sanitation situation across the districts also predict the variation in infant mortality. The likelihood ratio test statistic found statistically significant which suggests that comparability between the full and reduced models providing the evidence of non-null parameter estimates against the null hypothesis, H 0 : β 1 = β 2 = …β n = 0.
District level risk factors of child mortality. Like IMR, we modeled child mortality within the same study framework and we found that prevalence of each of the forms of multiple anthropometric failures is significantly associated with CMR burden across India (Table 7). Model 2a shows an estimated coefficient for stunting & underweight to be 0.029 (p-value < 0.001). Similar to Model 1a, Model 2a also carries statistical evidence on anthropometric failure (stunting & underweight) and child mortality burden across the districts of India. Like, concurrent occurrence of stunting & underweight among children, concurrent occurrence of wasting & underweight and simultaneous occurrence of wasting, stunting & underweight also closely predicts the district level burden of CMR in India. This shows that districts with higher burden of multiple anthropometric failures are at more risk of higher child mortality burden. Unlike IMR, the district level prediction of CMR shows that mother's educational status and immunization status of the children substantially predict the district level variation in CMR. The estimated beta coefficients also suggest poverty association of district level CMR burden. The CMR   www.nature.com/scientificreports/ framework shows lower AIC values for each of the forms of anthropometric failures than the IMR. Similar to IMR model estimation, the likelihood ratio test for CMR provides the evidence of non-null parameter estimates against the null hypothesis, H 0 : β 1 = β 2 = …β n = 0 from the GLM estimation.

Discussion
Almost two-fifth of all the children in India suffers from chronic undernutrition and still infant and under five mortality is substantially high in India 12 . Though there are different measures of undernutrition; still, due to some demerits of the given scientific measure, we may fail to identify the group of children with multiple anthropometric failures. In this context, this study examined the pattern and severity of multiple anthropometric failures among under-five children by population characteristics and across states of India. Additionally, the district level burden of infant and child mortality is predicted in terms of the different forms of anthropometric failure prevailing across India. Among the three anthropometric measures of undernutrition-stunting is defined as the chronic and long term undernutrition and wasting being the acute form of undernutrition, while the measure of underweight is assumed to be a combination of stunting and wasting both and indicates a possible occurrence of short and long term of undernutrition 37 . Thus in a group of children undernourished, the possible anthropometric failures could be multiple and severe which needs a careful identification of the exact type of growth faltering among the under-five children. In this context, this study provides a detailed understanding of the current scenario on multiple anthropometric failures prevailing among under-five children in India using the most recent household based survey data available for India. Thus we examined the risk factors of multiple anthropometric failures within a multi-level framework taking care of the hierarchy of the survey dataset. Additionally, we performed a district level analysis to predict the district level burden of infant and child mortality in terms of the prevalence of multiple anthropometric failures. A significant proportion of the children across India carry different forms of anthropometric failures and concurrent occurrence of "stunting and underweight" is much more prevalent followed by "wasting & underweight" and "wasting, stunting & underweight". Of the different types of anthropometric failures, the prevalence of concurrent occurrence of "stunting & underweight" shows high national average with sharp differential by place of residence (rural/urban), wealth quintile, mother's education, birth order, place of delivery and child age. Notably, no substantial gender difference in terms of any of the identified anthropometric failures is observed. Parallel to the concurrent occurrence of "stunting & underweight" it is also observed that the prevalence of "only stunting" is also quite high by different socio-economic characteristics of the children. It is observed that the burden of "only underweight" shows the lowest prevalence than any other forms of anthropometric failures. The state level pattern of CIAF among under-five children shows significant variation. And the largest variation is observed in terms of "stunting & underweight". More than one-fifth of the total children from the states like Bihar, Uttar Pradesh, Jharkhand and Madhya Pradesh are both stunted and underweight. And at the same time it is also alarming to note that 8-11% of the total children from the states like Jharkhand, Madhya Pradesh, Bihar and in Gujarat suffer from stunting, underweight and wasting simultaneously.
Among Indian children, the risks of different forms of multiple anthropometric failures are significantly determined by their socio-economic and demographic characteristics. Within the multilevel framework, we find the hierarchy in the variations of multiple anthropometric failures. It is found that children from the urban areas carry higher odds to suffer from multiple anthropometric failures-stunting & underweight, wasting & underweight and wasting and stunting and underweight. The wealth pattern of multiple anthropometric failures show a gradual reduction in odds across the richer wealth quintiles and children with higher economic wellbeing www.nature.com/scientificreports/ show lower odds in terms of all the three types of anthropometric failures. This wealth gradient of anthropometric failure clearly suggests a very sharp differential in the burden of anthropometric failures and children from the lower wealth quintiles are extremely vulnerable to multiple anthropometric failures. Similar to wealth gradient, the education (mother's) gradient is also evident. From the estimated AOR values it is clear that higher educational attainment among the mothers significantly reduces the likelihood to suffer from multiple anthropometric failures among their children-especially for the concurrent occurrence of stunting & underweight and concurrent occurrence of wasting, stunting and underweight. Apparently, higher ordered births show higher likelihood to anthropometric failure. The odds of multiple anthropometric failures show persistent increase among the second, third and four or higher ordered births compared the first ordered births. Previous studies also suggest poor nutritional health among children of higher ordered births subject to food insecurity of the households 38,39 . It is argued that, households with low or no food security may substantially face food shortage causing the children to starve and thus malnourished [40][41][42] . Though observed gender pattern of anthropometric failure does not show any differential but once adjusted for other covariates, it is found that female children are comparatively less likely to suffer from multiple anthropometric failures. Though in Indian setting, it is profoundly argued on the anti-female bias of food allocation 43 across the households of different socio-economic and demographic characteristics but the present study carries the evidence of lopsided low likelihood to anthropometric failure among female children than the males which is a consistent finding at par with the previous studies 44 . This study also finds that children with no access to improved source of drinking water show higher likelihood to different forms of anthropometric failures, consistent with the findings of previous studies showing the linkages between child malnutrition and drinking water 45,46 .
The district level aggregated analysis of the data brings forth the pattern of infant and child mortality in terms of the prevailing district level burden of multiple anthropometric failures among the children under age five. The bivariate analysis clearly demonstrates higher infant and child mortality in those districts where the prevalence of anthropometric failure is comparatively higher. GLM estimation of the district level rates of infant and child mortality indicates that districts with higher prevalence of multiple anthropometric failures show elevated risk of infant and child mortality. Each of the forms of multiple anthropometric failures shows strong statistical association to explain the district level variation in IMR and CMR. Previous literatures suggest that poor nutrition is associated with immuno-deficiency against the infectious diseases among children and undernourished children are at higher risk of dying from diseases like malaria, measles, diarrhea and acute respiratory infections 7,47,48 . Low plasma complement, reduction in exocrine secretion of protective substances, poor gut function and reduction in antibodies produced after vaccination are common among malnourished children who are susceptible to higher risk of death. Although, cause-specific mortality is not known but this study confirmed that districts with higher prevalence of multiple anthropometric failures are at higher risk of experiencing increased infant and child mortality.
Apart from the children's nutritional status, other district level characteristics like-poverty situation across districts shows substantial association with IMR. It is also evident that IMR is much more prevalent among the children of higher ordered births across the districts of India. Availability of clean and safe drinking water also shows an interplay with IMR across the districts and districts with higher portion of children with no access to improved source of drinking water show increased risk of IMR. In case of CMR, other than poverty, mother's education, children's immunization status, drinking water and sanitation condition show substantial association with CMR. At the aggregated level, the gender pattern of infant and child mortality shows no statistical significance. Although, district level proportion of rural shows negative association with IMR; while, the association with CMR has been found positive and suggests that among the rural children the risk of CMR is higher. Contrary to the CMR situation, the risk of infant mortality is found lower among the rural children than the urban children when adjusted for other district level risk factors.
The study has few limitations. First, this study is based upon a cross-sectional study whereby we cannot draw any causal inference. Second, though within a cross-sectional framework we checked the pattern of CIAF prevalence by population characteristics and examined the determinants yet the data information was limited. Like, child's nutritional status directly depends upon their previous disease pattern however; there is no appropriate information available on this aspect of the children from the dataset. Third, this study used the last five years birth history and included all the children surveyed but for a significant portion of the sample, we have either the anthropometry data missing or the children's height/weight measure are out of the plausible limits and thus we had to drop a number of 54,147 many children from the data set. Additionally, checking the pattern of missing we found that the missing pattern is not random (MAR) and thus we could not impute. Though we followed the DHS guidelines, another limitation of vaccination data from National family Health Survey is that it is based upon mother's recall when vaccination card not found at the time survey, which introduces a non-sampling bias in the estimation process 50,51 . Additionally, the dataset lacks the information on cause specific mortality and in a country like India, diarrhoeal infection, acute respiratory infection and low utilization of post natal care largely contributes to child mortality 37, 52, 53 with state and regional variation to which we could not adjust while compiling the infant and child mortality rates to execute the district level meso scale analysis.

Conclusion
This study assessed the multilevel contextual determinants of anthropometric failures among Indian children and further explored the district level variation in IMR and CMR in terms of the existing burden of multiple anthropometric failures. This study shows that a significant portion of children under age five in India suffers from multiple anthropometric failures. And socio-economic and demographic characters of the children show larger variation in different types of anthropometric failures-especially in terms of the concurrent occurrence of stunting & underweight in India. Analysis suggests that household's economic wellbeing and mother's education www.nature.com/scientificreports/ plays a significant role in child's nutritional status. The multilevel analysis of CIAF confirms significant geographical variations in terms of each of the anthropometric failures across districts, across PSUs within districts and among the children within PSUs. The children level variation in multiple anthropometric failures is found to be the largest. This reiterates the fact that anthropometric failure among children in India is a micro-level phenomenon than a subject at the meso scale. At the child level, mother's education, birth order of the child and sex of the child substantially determine the different forms of anthropometric failures. At the same time, household's economic status in terms of wealth plays a crucial role on child's nutrition. Although it has shown a reduction over time, Indian districts are still burdened with high levels of IMR and CMR and this study propagates the key message that districts with higher burden of multiple anthropometric failures are vulnerable and exposed to higher burden of infant and child mortality in India.