AMR-Associated and AMR-Attributable Burden of Escherichia coli in Ready-to-Eat Foods in West Africa: An Exposure-Based and Calibrated Probabilistic QMRA-DALY Study

Published: 12 November 2026| Version 1 | DOI: 10.17632/htvh89tff8.1
Contributors:
,

Description

This dataset supports the study "AMR-Associated and AMR-Attributable Burden of Escherichia coli in Ready-to-Eat Foods in West Africa: An Exposure-Based and Calibrated Probabilistic QMRA-DALY Study". It covers adults aged 15 years and older in Benin, Burkina Faso, Côte d'Ivoire, Ghana and Nigeria. The deposit contains five datasets: Supplementary Data 1 (E. coli concentrations): records from a quantitative systematic review of AMR and non-AMR E. coli concentrations in ready-to-eat foods (10 studies, 141 records, 42 food-country combinations), with the search strategy, PRISMA flow diagram, harmonization log and ready-to-model distributions. Supplementary Data 2 (food consumption): records from a quantitative systematic review of amounts eaten and eating frequencies (24 studies and datasets), with the search strategy, PRISMA flow diagram, harmonization log and ready-to-model inputs. Supplementary Data 3 (ProRisk model and inputs): the ProRisk model files of the base case and scenarios, with every input distribution and output formula. Supplementary Data 4 (ProRisk outputs): the outputs of the ProRisk runs. Supplementary Data 5 (Python model, new in version 2): an open Python implementation that reproduces the ProRisk base case and extends it to a two-dimensional Monte Carlo simulation. The outer loop samples parameter uncertainty and the inner loop simulates variability between adults. The model adds a prevalence term, a beta-binomial resistant fraction, infection from the total dose, AMR-associated and AMR-attributable burden, and calibration to WHO 2021 estimates of foodborne diarrheagenic E. coli. The folder holds the code, the output workbook and the manuscript figures. Supplementary Data 1-4 are unchanged from version 1. Every model input can be traced from Supplementary Data 3 and 5 back to the records and papers in Supplementary Data 1 and 2.

Files

Steps to reproduce

1. Download all five folders and keep them side by side in one parent folder, with their names unchanged. 2. Review data (Supplementary Data 1 and 2): each README describes the files. The search strategies and PRISMA counts document study selection. The harmonization logs trace every value from the source paper to the ready-to-model tables (Supplementary Data 1, Tables S3, S13 and S14; Supplementary Data 2, Tables S7a and S10). 3. ProRisk runs (Supplementary Data 3 and 4): open each .prorisk file (JSON) in ProRisk 2026 and run 100,000 iterations with seed 42. The base case and scenario outputs should match Supplementary Data 4. The workbook in Supplementary Data 3 lists every input and formula, so the model can also be rebuilt in other software. 4. Python model (Supplementary Data 5): install Python 3.13 with NumPy, SciPy, pandas, openpyxl and matplotlib. a. Open a terminal in Supplementary_Data_5_Python_model/code. b. Run "python3 validate_against_ProRisk.py". It confirms that the code reproduces the ProRisk base-case infection probabilities (differences below 1e-14). c. Run "sh run_all.sh". This regenerates all outputs and figures (about one hour). For a quick test, run "WAE_SCALE=0.05 sh run_all.sh". d. Results are written to the outputs and figures folders. With the fixed seeds, a full run reproduces the published workbook and manuscript figures exactly. 5. Any question about a value can be traced through the model input name (Supplementary Data 3, sheet Base_inputs) to the food row, the record and the source paper.

Categories

Public Health, Food Toxicology, Dose-Response in Food Toxicology, Antibiotic Resistant Bacteria

Licence