Global prediction of anemia, stunting, and wasting using household surveys and geospatial covariates
This MSc project aims to generate high-resolution global maps of anemia, stunting, and wasting by combining biomarker and anthropometric data from DHS/MICS with globally consistent environmental, agricultural, and socioeconomic covariates. Using scalable machine-learning methods, the study will predict malnutrition outcomes across diverse regions where survey data are sparse and unevenly distributed. The project also investigates which covariates best explain global patterns of micronutrient deficiencies and child growth failure and identifies regions where multiple forms of malnutrition overlap.
Objectives:
- To compile global household survey data and link them with environmental, agricultural, and socioeconomic covariates.
- To model global anemia, stunting, and wasting prevalence using data-driven methods capable of handling sparse survey coverage.
- To identify key global predictors of malnutrition and map areas where multiple forms of undernutrition co-occur.
Methodology:
The study compiles anemia, stunting, and wasting indicators from global household surveys and links them to a harmonized set of environmental, agricultural, and socioeconomic covariates at consistent spatial resolution. Machine-learning models capable of capturing non-linear relationships are trained using these survey-linked covariates, with spatial cross-validation used to ensure global robustness. The final models are applied to global gridded datasets to produce high-resolution predictions and uncertainty estimates.
Reference:
Kinyoki, D. K., Osgood-Zimmerman, A. E., Pickering, B. V., Schaeffer, L. E., Marczak, L. B., Lazzar-Atwood, A., Collison, M. L., Henry, N. J., Abebe, Z., … Local Burden of Disease Child Growth Failure Collaborators. (2020). Mapping child growth failure across low- and middle-income countries. Nature, 577(7789), 231–234. https://doi.org/10.1038/s41586-019-1878-8
This topic fits the SIS domain by integrating global household surveys with diverse geospatial covariates and addressing issues of data provenance, spatial uncertainty, and large-scale predictive mapping. It contributes to TE through the analysis of environmental and biophysical processes that shape global patterns of malnutrition. Insights generated are relevant to SPG, as the resulting maps support evidence-based decision-making for governments and international agencies working to address nutritional inequality.