Artificial Intelligence Benchmarking of Community Antimalarial Stock Reallocation Policies Under Unmeasured Cold-Chain Temperature Excursion Logging Gaps: A Healthcare Efficiency-Regret Evaluation

Community-level antimalarial stock reallocation is routinely evaluated as if cold-chain temperature histories were observed, yet in most district supply systems excursion logging is incomplete, so any policy that is benchmarked against a fully observed temperature record is being scored on information the deploying system will not have. the present analysis formalises that mismatch as an efficiency-regret problem and derives a benchmarking protocol for comparing reallocation policies under partial excursion observability. The protocol couples a latent excursion process, a logging-coverage parameter and a regret functional that separates the efficiency loss attributable to policy choice from the loss attributable to missing temperature evidence. Using NASA POWER monthly reanalysis air temperature for Accra, Ghana (5.6, -0.19) over 2000-2023, 312 monthly observations, mean 26.521 C, standard deviation 1.24 C and a warming trend of 0.1622 C per decade, the paper calibrates the ambient forcing that drives excursion hazard and shows how the resulting regret surface changes with logging coverage. The contribution is a coverage-indexed regret bound that lets benchmarkers report policy rankings together with the logging coverage at which those rankings become unreliable, rather than a single leaderboard computed under an unrealised observation regime.