Stunting and underweight are the most frequently used and internationally recommended indicators of undernutrition among children. Studies indicate that these two indicators may be biologically related i.e. increase in weight may be associated with an increase in length or height because of better nutrition (14, 20, 29, 30) . Since prior research pointed out several issues related to measuring children’s length/ height (20), through our analysis we assessed the correlation between the stunting and underweight using two national large scale surveys and developed a simple predictive model for the prevalence of stunting using underweight. Through our analysis we found a positive and significant correlations between underweight and stunting at the national level in both CNNS and NFHS-4. Further, using the district level prevalence results from NFHS-4, positive and significant correlation was found at the regional level, although at the regional level the correlation ranged from as low as 0.32 for the central region to 0.86 for the southern region
We found that nearly two-third of those stunted were also underweight. This indicates that in locations where the data quality of height measurement is questionable or where it is difficult to capture height, underweight may be able to assess the nutrition status of children to a large extent. This corroborates with prior research in other countries where underweight was used as an indicator to assess the nutrition status of children and also predict stunting (31, 32). However, this method may have limitations in severe forms of undernutrition, a lower proportion of those severely stunted also were severely underweight (<50%).
Regressing stunting on underweight resulted in similar slope at the national level, when using different survey data. However, the regression coefficients across regions were different. This indicates that the pattern of malnutrition and the relationship between the key nutritional indicators, underweight and stunting are unique to each region. In the southern region of the country, underweight explained 75% of the variability in stunting, whereas in the central region underweight only explained 10% of the variability in stunting. Prior research conducted in other countries as well as in India, highlights differences in the determinants of stunting and underweight (23, 33-37). The differences in the association between underweight and stunting, between different regions could be due to the differential influences of the determinants across regions i.e. different regional and local contextual factors such as, genetic differences, dietary habits, environmental factors, among others. To account for regional variations when deriving estimates the regional regression equation may be more appropriate. If state level contextual factors significantly and differentially influence the relationship between underweight and stunting, then it is important to consider state level regression equation and not the national or regional level equation. . We could not run state level regression models due to lack of sample size. Further research is needed to understand state level differences in the relationship between stunting and underweight.
Where the correlation is high, the equation can be used to assess if there is over estimation or under estimation of stunting prevalence in other surveys in the country, by examining the differences between observed and estimated % stunted. An example has been presented in table 5, using a survey that was conducted in four states of India during the period 2016-17 (28). The analysis indicates that in two out of four states, the survey reported lower prevalence of stunting compared to the regression model and in two states there was overestimation by the survey (Table 5). However, for the states belonging to the region (eastern India) where we found high correlation between underweight and stunting, the observed stunting prevalence was within the confidence interval of estimated stunting prevalence derived from the regression equation and in one of states from the central region, where we found low correlation between stunting and underweight, the observed stunting prevalence from the survey was outside the confidence interval of the estimated stunting prevalence (Table 5).
As indicated earlier in the paper, currently in India, growth monitoring method is changing from use of underweight to stunting and wasting. POSHAN Abhiyaan, which brought a lot of focus and momentum towards improving the nutritional status of children, was launched in 2018. As part of the program, growth monitoring devices including weighing scales and height measuring instruments - infantometers, and stadiometers were distributed. However, a recent report from NITI Aayog, a policy think tank of the Government of India, highlighted that many states in India in 2019 still do not have appropriate height measuring instruments (25). Anecdotal information from visits to Anganwadi Centres (AWC) in various states have highlighted data quality issues due to improper techniques used by the Anganwadi workers at the AWC, use of inappropriate equipment, and errors while entering the data. Till the time appropriate data quality for height is achieved, there is a need to use alternate growth monitoring indicator to estimate stunting. Prior research highlights the use of underweight, as in the absence of high wasting levels, underweight and stunting provide similar information about long-term health and nutritional experience (23). However, in India the prevalence of wasting is critically high in terms of severity index (8). Further, using only underweight may underestimate the magnitude of malnutrition in the country. Therefore, monitoring just weight-for-age may not be enough to address malnutrition, there is a need to also monitor length/ height-for-age. Therefore, our analysis presents a simple predictive model for estimating the prevalence of stunting using underweight.
While this study underlines the strong relationship between underweight and stunting, the results should be interpreted in light of certain limitations. First, in a country like India, there might be state level contextual factors that may influence the relationship between underweight and stunting. However, in the current analysis we could not undertake state level analysis due to small sample size. Second, predictive equations that are constructed depends on the quality of data used. In the current analysis we used CNNS and NFHS-4 data, both of which had rigorous quality control system put in place while collecting anthropometric data. However, issues related to the measurement of accurate and precise length/ height data may have led to some measurement errors that may have biased the correlation in some regions. Prior research in other countries indicated highest correlation coefficient for the 12-23 months age group, compared to other age groups among children. We did not undertake such analysis within each region due to limited sample size. Future research could provide critical insight regarding the relations between stunting and underweight for each age group. Further research may be required to derive state level regression equations, if state level contextual factor affect the relationship between stunting and underweight.