Artificial Intelligence Distributionally Robust Optimisation of Community Clinic HIV Self-Test Kit Reorder Policy Under Variable Kit Expiry Reporting: A computed Healthcare Waste Equity Evaluation

Community clinics that distribute HIV self-test kits face a reorder problem in which the only feedback on kit expiry is a lagged, incomplete and inconsistently reported count, so any reorder rule tuned to a single reporting regime will over-stock some clinics and stock out others. the present analysis formulates the reorder decision as a distributionally robust optimisation problem over an ambiguity set of expiry-reporting distributions, and solves it with a tractable convex reformulation that yields a closed-form order-up-to level indexed by a Wasserstein radius. The framework is evaluated in a computed healthcare waste equity setting in which clinics differ in reporting fidelity and in the equity weight attached to unmet testing demand. The analysis derives finite-sample performance guarantees for the robust policy, characterises the equity cost of robustness as a function of the radius, and shows that a radius chosen from the reporting-fidelity distribution dominates both the nominal and the worst-case policies on waste and shortage simultaneously. The contribution is a decision rule that is auditable, requires no distributional assumption on expiry reporting, and makes the trade-off between wasted kits and missed tests an explicit policy parameter rather than an artefact of the reporting system.