Alternative Metrics of Linear Growth for Tracking Global Progress in Child Undernutrition (P10-001-19)

Abstract

Listen

Translate article

ObjectivesCompare under-5y stunting prevalence to alternative height-for-age z-score (HAZ)-based measures of location (MoLs) in terms of the strength of their relationship with key population-level indicators known to be associated with HAZ; and, examine the modification of these associations by anthropometric data quality. MethodsData from 140 Demographic and Health Surveys (DHS) from 63 low- and middle-income countries (LMICs) (2000 to present) were used to assess pairwise associations between seven alternative HAZ-based MoLs and three population-level indicators, using absolute values of correlation coefficients (corr) and Akaike’s Information Criterion (AIC) from linear mixed effects models. Extended models with interaction terms were used to assess how survey quality (defined using a score based on principal component analysis) modified each HAZ MoL - population-level indicator relationship. All analyses were performed under three flagging approaches for excluding HAZ outliers: WHO (cut-off at -6/ + 6 SD of the reference mean); SMART (cut-off at -3/ + 3 SD of the observed mean); and no flagging. ResultsAll HAZ MoLs were highly correlated with stunting (range: 0.92 to 0.98). Correlations between HAZ MoLs and population-level indicators were moderate and similar across metrics (range: 0.44 to 0.65). Stunting and model-predicted HAZ at 3y frequently had among the highest corr and lowest AIC values. Extended models consistently showed that associations between HAZ MoLs and population-level indicators attenuate as survey quality decreases. This modifying effect was weakest for stunting and predicted HAZ at 3y across flagging scenarios, and under conditions of WHO flagging across most MoLs. ConclusionsPreliminary analyses suggest that stunting is at least as or more robust compared to most other HAZ-based MoLs based on the strength of its correlation with population-level indicators, and insensitivity of those associations to variations in survey quality. Of the MoLs that represent central tendency, predicted HAZ at 3y performed comparably to stunting, and may have conceptual advantages for comparing child undernutrition within and across LMICs. Funding SourcesSickKids Hospital Growth and Development Fellowship Program. United States Agency for International Development (USAID) through the DHS Program. Supporting Tables, Images and/or Graphs▪