Dietary staple type, protein quality and child wasting and stunting across 127 LMICs: an ecological cross-sectional study — data and code
Description
This repository contains the analytical datasets, master food balance sheet dataset, and R code required to reproduce the primary analysis, sensitivity analyses, secondary cross-reference analyses, post-hoc maternal anthropometry analysis, and choropleth maps reported in the manuscript "Dietary staple type, protein quality and child wasting and stunting across 127 LMICs - an ecological cross-sectional study" (medRxiv preprint: https://doi.org/10.64898/2026.07.13.26357913). Data are de-identified, country-level aggregates drawn from publicly available sources: the Joint Malnutrition Estimates (JME, March 2025) for child wasting and stunting prevalence, the Global Burden of Disease study (GBD 2023) for disease and care covariates, and FAOSTAT (2019-2023 five-year means) for staple protein, animal protein, and caloric supply. No individual-level or identifiable data are included. The package includes: primary analytical dataset (n=127 LMICs for wasting, 126 for stunting); population-weighted dataset (used for sensitivity analysis S5); master food balance sheet dataset used for choropleth map generation (138 countries); secondary cross-reference datasets (Grellety & Golden WHZ-MUAC discordance data, n=46, and Kerac et al wasting-and-stunting in infants under 6 months, n=56); post-hoc maternal anthropometry cross-reference dataset (DHS StatCompiler, n=57); and nine R scripts covering data correction, primary robust MM-estimation and quantile regression analyses, five pre-specified sensitivity analyses, secondary cross-reference analyses, and choropleth map generation. Full documentation of file contents, column definitions, data corrections applied prior to analysis, and step-by-step instructions for reproducing all reported results is provided in README.txt and DATA_DICTIONARY.md within this repository.
Files
Steps to reproduce
Steps to reproduce: Download and extract all files into a single folder. Install R (v4.5.3 used) and required packages: robustbase, quantreg, dplyr, readr, readxl, ggplot2, ggrepel, sf, rnaturalearth, rnaturalearthdata, viridis, scales, cowplot, RColorBrewer, writexl. Set the R working directory to that folder: setwd("path/to/folder"). Run scripts in this order: run_primary_analysis_v3.R — primary robust MM-estimation, wasting quantile regression, and sensitivity analyses S1–S3 (input: validation_dataset_final_127LMIC_corrected.csv) run_stunting_QR_v3.R — stunting quantile regression run_S4_S5_v3.R — sensitivity analyses S4 (HIV exclusion) and S5 (population-weighted) run_secondary_crossref.R — Grellety & Golden (n=46 analytical) and Kerac et al (n=56) cross-reference analyses (input: S_Dataset_Secondary_CrossRef_v2.xlsx) run_maternal_anthro_analysis.R — post-hoc maternal anthropometry analysis (input: maternal_anthro_regression_dataset.csv) map_code_LGH_updated.R — generates all 15 choropleth maps (input: master_dataset_Food_balances_corrected.xlsx) 00_fix_dataset_NA_to_zero.R, run_figures_scatter.R, and grellety_oedema_plot.R are included for provenance/supplementary figures but don't need to be re-run to reproduce the primary reported results — the corrected data files already reflect their output. Full column definitions, data-correction notes, and model specifications are in README.txt and DATA_DICTIONARY.md inside the package.
Institutions
- SRM Medical College Hospital and Research CentreTamil Nadu, Kattankulathur