Designing Robust Antimicrobial Supply Chains with Epidemiological Demand Uncertainty in Botswana: A Network Optimization Model

Abstract

Antimicrobial resistance (AMR) poses a growing threat to global health, with reliable access to essential medicines serving as a critical but often overlooked line of defense. This thesis develops a network optimization framework for antimicrobial supply chain management in Botswana, a country that declared a national public health emergency in August 2025 following widespread medicine shortages. The model incorporates epidemiological demand uncertainty through a Medical Demand Estimator (MDE) based on Negative Binomial regression, calibrated to facility-level catchment populations and national point-prevalence survey data. Three allocation policies are evaluated: deterministic, static robust, and adjustable robust optimization with affine decision rules (ARO-ADR), across both a Gaborone proof-of-concept trial and a national simulation spanning 631 facilities across 18 districts. The ARO-ADR policy reduces average unmet antimicrobial demand by 76% relative to the deterministic baseline and 48% relative to static robust, while incurring a 0.9% procurement cost premium. The framework further embeds a two-strain SEIR model to capture how distribution failures compound antimicrobial resistance emergence over time. The results suggest that investments in resilient, adaptive pharmaceutical logistics may yield substantial population-level health benefits, particularly in resource-constrained and procurement-based settings where access rather than innovation is the binding constraint.