Artificial Intelligence Malaria Rapid Diagnostic Interpretation Under Variable RDT Lot Degradation: A Conformal Calibration Evaluation of Healthcare Treatment Initiation

Malaria rapid diagnostic tests (RDTs) degrade under tropical storage conditions, and artificial intelligence (AI) systems that interpret RDT images may therefore face lot-dependent shifts in test-line intensity that are not represented in their training data. The framework treats lot degradation as a distribution shift and uses conformal prediction to produce set-valued treatment recommendations with finite-sample coverage guarantees. A design study with 2,000 replications per cell across 8, 16, 24, and 32 clusters and effect sizes of 3, 5, and 8 percentage points, with an intraclass correlation of 0.64, between-cluster SD of 6.0, and residual SD of 4.5, shows that 54 clusters are required for 80% power at a 5-percentage-point effect. The analysis characterises the operating characteristics of the design rather than reporting empirical AI performance. The contribution is a calibration-aware evaluation architecture that links conformal set-valued predictions to cluster-level treatment-initiation outcomes, and a quantified statement of the design burden that lot-stratified RDT evaluation imposes on health systems.