Data and code for: Nutritional quality scores overlook additive-driven ultra-processing globally

"Nutritional quality scores overlook additive-driven ultra-processing globally"Arslan, S., Yilmaz, S., & Gunal, A. M.Nature Food (2026). Data files: nutriscore_merged_full.csv — Product-level data for 944,316 food products with valid NutriScore grades, sourced from the Open Food Facts database (May 2025 dump). Each product includes nutritional composition per 100 g, NutriScore grade and numeric score, NOVA processing group, additive and ingredient counts, food category, country of sale, and merged country-level health and economic indicators. The analytic sample comprises 519,749 unique products with valid NOVA classification from 63 countries (≥50 products each). country_merged_dataset.csv — Country-level indicators for 86 countries with at least 100 products in the Open Food Facts database. Includes NCD mortality, diabetes prevalence, life expectancy, overweight rates, GDP per capita, health expenditure, smoking prevalence, alcohol consumption (World Bank World Development Indicators) and daily per capita calorie, protein, and fat supply (FAOSTAT Food Balance Sheets). Analysis code: R scripts for statistical analysis (multilevel modelling, health associations, visualisation) and Python scripts for data processing, machine learning (XGBoost with Optuna optimisation) and SHAP interpretability analysis. Full reproduction instructions are available at https://github.com/sy142/nutriscore-discordance-cross-national Data sources Source Variables Access URL Open Food Facts Product-level nutritional data, NutriScore, NOVA, additives, ingredients https://world.openfoodfacts.org/data World Bank WDI NCD mortality, diabetes, life expectancy, overweight, GDP, health expenditure, smoking, alcohol https://data.worldbank.org/ FAOSTAT Per capita calorie, protein, fat supply https://www.fao.org/faostat/en/#data/FBS USDA FoodData Central Validation sample (250 branded products) https://fdc.nal.usda.gov/