Artificial Intelligence Distributionally Robust Optimisation of Malaria Rapid Diagnostic Test Allocation Under Variable Batch Sensitivity Drift: A computed Healthcare Misdiagnosis Equity Evaluation

Rapid diagnostic tests for malaria are deployed in batches whose sensitivity can drift between manufacturing lots, yet allocation models typically assume a fixed and known sensitivity. the present analysis formulates the allocation of a fixed stock of rapid diagnostic tests across clinics as a distributionally robust optimisation problem in which batch sensitivity is uncertain within an ambiguity set, and evaluates the resulting misdiagnosis equity consequences through a replication study. The decision variable is the number of tests allocated to each clinic; the objective minimises the worst-case expected number of missed malaria cases subject to an equity constraint on the maximum clinic-level misdiagnosis rate. The ambiguity set is a Wasserstein ball around an empirical sensitivity distribution, and the robust counterpart is reformulated as a tractable linear program. A replication study with 2,000 replications per design cell evaluated the operating characteristics of the allocation rule under a before-after-control-impact design with an intraclass correlation of 0.64, between-cluster standard deviation of 6.0 and residual standard deviation of 4.5. The analysis characterised power across cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points, and identified 52 clusters as the number required for 80% power at a 5 percentage point effect. The contribution is a tractable robust allocation model and a design-characterisation of the evaluation that would be needed to detect its equity effects.