Artificial Intelligence Distributionally Robust Optimisation of mRNA Thermotherapy Cold-Chain Routing Under Variable Grid Voltage Telemetry: A Healthcare Wastage Evaluation

MRNA thermotherapy products must be held within a narrow temperature band from manufacture to infusion, and the cold-chain routes that carry them are powered by electrical grids whose voltage telemetry is intermittent and biased. the present analysis formulates the resulting routing problem as a distributionally robust optimisation over an ambiguity set of grid-voltage trajectories, and couples it to a wastage model in which thermal excursions translate into dose loss. The formulation is developed as a methodological contribution: a Wasserstein ambiguity set is constructed from telemetry samples, a tractable convex reformulation is derived, and the wastage functional is shown to be Lipschitz in the voltage trajectory, which bounds the out-of-sample wastage gap. Because no field data were available, the operating characteristics of the design were characterised analytically and through a replication-based evaluation under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64). The robust routing policy dominates the deterministic nominal policy in worst-case wastage, that the price of robustness is bounded by the telemetry sample size, and that the intraclass correlation of 0.64 implies that cluster-level voltage heterogeneity, not within-route noise, governs the achievable wastage reduction. The contribution is a formal bridge between distributionally robust routing and healthcare wastage accounting for thermotherapy cold chains.