The impact of improved data quality on the prevalence estimates of anthropometric measures using DHS datasets in India

Abstract

Abstract The importance of data quality to correctly determine prevalence estimates of child anthropometric failures has been a contentious issue among policymakers and researchers. Our research objective was to ascertain the impact of improved DHS data quality on the prevalence estimates of stunting, wasting, and underweight. The study also looks for the drivers of data quality. Using five data quality indicators based on age, sex, anthropometric measurements, and normality distribution, we arrive at two datasets of differential data quality and their estimates of anthropometric failures. For this purpose, we use the 2005–2006 and 2015–2016 NFHS data covering 311,182 observations from India. The prevalence estimates of stunting and underweight were virtually unchanged after the application of quality checks. The estimate of wasting had fallen 2 percentage points, indicating an overestimation of the true prevalence. However, this differential impact on the estimate of wasting was driven by the flagging procedure’s sensitivity and was in accordance with empirical evidence from existing literature. We found DHS data quality to be of sufficiently high quality for the prevalence estimates of stunting and underweight, to not change significantly after further improving the data quality. The differential estimate of wasting is attributable to the sensitivity of the flagging procedure.