stewchang/nutrition-tb-modeling

Cross-model comparison of nutritional interventions on TB incidence (Odone 2014, Oxlade 2015, Mandal 2024, IDM EMOD India, Bhargava 2023 RATIONS trial) # Nutrition and TB Incidence: A Cross-Model Comparison **Stewart Chang** — Institute for Disease Modeling, Gates Foundation *Current affiliation: Model Stewardship LLC* This repository compares how reducing undernutrition affects tuberculosis (TB) incidence across three published modeling studies, an unpublished agent-based model, and an empirical trial result. The analysis digitizes and re-plots model outputs from each study on a common axis — relative incidence reduction (%) vs. proportion of the population that is undernourished (%) — enabling direct visual comparison of TB elasticity with respect to nutritional status. ## Main Figure The x-axis runs right-to-left (high to low undernutrition) to represent the trajectory of an improving nutritional intervention. Year labels mark the time dimension that is otherwise hidden when the two variables are plotted against each other. ## Data Sources | Study | Region | Baseline undernutrition | Intervention | Time period | |-------|--------|------------------------|--------------|-------------| | Odone et al. 201470164-0/abstract) | WHO SEARO region | 15.5% | Linear elimination of undernutrition by 2035 | 2013–2035 | | Oxlade & Murray 2015 | Central Eastern States of India | 25–50% (BMI-based estimate) | 11.7% per year relative reduction | 2011–2030 | | Mandal et al. 2024 | India overall (WHO SEARO) | 23.6% | 30% per year relative reduction | 2023–2030 | | IDM/EMOD TB-India model | India | 20% | 30% per year relative reduction; 2× disease progression risk for undernourished | 2025–2035 | | Bhargava et al. 2023 (RATIONS trial)01231-X/abstract) | Jharkhand, India | ~30% | Empirical RCT — nutritional supplementation to household contacts of TB patients | Single trial period | ## Methods ### Figure digitization Published TB incidence curves (baseline and intervention scenarios) were digitized from paper figures using **WebPlotDigitizer** (Rohatgi, 2022). For each study: 1. The relevant figure was captured as a PNG …