Artificial Intelligence Distributionally Robust Optimisation of Snakebite Antivenom Pre-Positioning Under Variable Cold-Chain Temperature Excursion Reporting: A computed Healthcare Stockout Equity Evaluation

Snakebite envenoming kills tens of thousands of people each year, and the antivenom that treats it is fragile: most products require storage between 2 and 8 degrees Celsius, and a single temperature excursion can render a vial unusable. Pre-positioning decisions therefore depend on cold-chain reliability that is reported unevenly across facilities, yet standard inventory models treat excursion rates as known constants. The framework is evaluated on a computed stockout-equity benchmark built from a real temperature record for Kampala, Uganda: 312 monthly observations from the NASA POWER MERRA-2 reanalysis for 2000 to 2023, with a mean of 22.443 degrees Celsius and a warming trend of -0.1564 degrees Celsius per decade. The analysis is methodological and computational; no field data on antivenom demand or facility reporting were collected. The contribution is a formal characterisation of how reporting-quality heterogeneity enters the ambiguity set, a proof that the resulting robust problem is tractable as a second-order cone program, and a replication protocol that scores pre-positioning policies on stockout equity rather than expected cost alone. The framework predicts that ignoring reporting heterogeneity systematically under-protects the facilities with the weakest cold-chain records, and it specifies the conditions under which equity-weighted robust policies close that gap.