Abstract:
Wasting among Under-5 children is an important indicator of acute malnutrition, and is also a predictor of under-five mortality. A wasted child has increased risk of mortality as well as future episodes of stunting/wasting. In India, the prevalence of wasting is quite high as compared to international standards, with every one in five children found to be wasted. National level estimates obtained from large scale cross-sectional surveys do not account for seasonality, mainly due to their complex logistical operations. Data collected across different months and across different states introduces seasonal bias in these estimates. Thus, special methods are required to account for seasonality and analyze the effect of data collection by month on these estimates. This paper thus attempts to analyze the effect of seasonality on wasting among Under-5 children in India.
Methods: We used data from four nationally representative cross-sectional surveys conducted in India, namely the National Family Health Survey or the NFHS-3 (2005-06), NFHS-4 (2015-16), NFHS-5 (2019-21), and the Comprehensive National Nutrition Survey (2016-18). To account for the seasonal bias in prevalence estimates of wasting, we performed a series of unadjusted and adjusted Ordinary Least Squares (OLS) and logistic regression models on the four surveys. Mean Weight-for-height (WHZ) z-scores and prevalence of wasting were estimated from these regression models controlling for timing using month dummy variables and adjusted for individual, maternal, household and community level covariates. Further, we also fitted a seasonal trend to present linear trends of wasting from 2005 to 2021, and compared the survey results to the fitted trends.
Results: Results from the OLS regression shows that the mean WHZ starts at its highest in January, and falls at its lowest in the June-August period, and then returns to its peak at the end of the year across all the surveys. The prevalence of wasting is at its lowest in the beginning of the year, and reaches its peak by June/August across all the four surveys. The adjusted monthly estimates closely overlap with the fitted trend for every cross-sectional survey, indicating that the unadjusted national level estimates of mean WHZ and prevalence of wasting miss out on the seasonal pattern.
Conclusion: This study shows that the direct comparisons of acute malnutrition across surveys should not be made without considering the effect of seasonality. To avoid biased estimates, the seasonality bias should be addressed before the planning, implementation, or analysis of cross-sectional surveys. Eliminating the seasonal variation in wasting could reduce the prevalence by half and can pave the way for further reduction of acute malnutrition.
