An uncertainty estimate of the prevalence of stunting in national surveys: the need for better precision

Abstract

Background Stunting is determined by using the World Health Organization (WHO) child growth standard which was developed using precise measurements. However, it is unlikely that large scale surveys maintain the same level of rigour and precision when measuring the height of children. The population measure of stunting in children is sensitive to over-dispersion, and the high prevalence of stunting observed in surveys in low and middle-income countries (LMIC) could partly be due to lower measurement precison. Objectives To quantify the incongruence in the dispersion of height-for-age in national surveys of < 5 y children, in relation to the standard WHO Multicenter Growth Reference Study (MGRS), and propose a measure of uncertainty in population measures of stunting. Methods An uncertainty factor was proposed and measured from the observed incongruence in dispersion of the height-for-age of < 5 y children in the MGRS against carefully matched populations from the Demographic Health Survey of 17 countries (‘test datasets’, based on the availability of data). This also allowed for the determination of uncertainty-corrected prevalence of stunting (height-for-age Z score < − 2) in < 5 y children. Results The uncertainty factor was estimated for 17 LMICs. This ranged from 0.9 to 2.1 for Peru and Egypt respectively (reference value 1). As an explicit country example, the dispersion of height-for-age in the Indian National Family Health Survey-4 test dataset was 39% higher than the MGRS study, with an uncertainty factor of 1.39. From this, the uncertainty-adjusted Indian national stunting prevalence estimate reduced to 18.7% from the unadjusted estimate of 36.2%. Conclusions This study proposes a robust statistical method to estimate uncertainty in stunting prevalence estimates due to incongruent dispersions of height measured in national surveys for children < 5 years in relation to the WHO height-for-age standard. The uncertainty is partly due to population heterogeneity, but also due to measurement precision, and calls for better quality in these measurements.

Co-occurrence of Overweight, Stunting, and Anemia among Adolescents (10–19 Years) in Tanzania Mainland: A School-Based Cross-Sectional Study

“…Moreover, the study reveals a significant difference in mean hemoglobin levels between female and male adolescents. The high prevalence of anemia and stunting among adolescents might be contributed to the consumption of low/poor dietary quality, especially during school break time, among other factors [ 19 ].…”

The Socio-Environmental Determinants of Childhood Malnutrition: A Spatial and Hierarchical Analysis

“…The accuracy of the observed stunted children is better, 78% (Nigeria) and 66% (Kenya). Other estimates for the overdispersion of heightfor-age z-scores suggest variance inflation factors as high as 110% [95].…”

The Potential Contribution of the Health System to Reducing Stunting in SUN Countries

“…We additionally recognize that LiST itself relies on many estimates, and therefore, some uncertainty has been introduced into our results. Additionally, stunting data from household surveys have large uncertainty, 35 and it is difficult to obtain accurate measurements of other nutrition indicators such as low birth weight. 36 Nevertheless, these are data that are available in household surveys, and the best estimates available at this time.…”