Artificial Intelligence Benchmarking of Malaria Blood Smear Parasite Density Estimation Under Unmeasured Giemsa Stain Batch Variation: A Healthcare Calibration-Transfer Evaluation

Parasite density estimation from Giemsa-stained blood smears is a cornerstone of malaria diagnosis and trial endpoint measurement, yet artificial intelligence (AI) models trained on one stain batch frequently degrade when deployed on another because the batch identity is unrecorded. the present analysis treats stain-batch variation as an unmeasured cluster-level nuisance and asks what a benchmarking study can conclude about a model's calibration-transfer performance when batch labels are absent. The design is a before-after-control-impact (BACI) evaluation of a calibration-transfer estimator, characterised by replication over a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64. Power was estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points.54 clusters are required to detect a 5-percentage-point calibration-transfer effect at 80% power under this variance structure, and that the minimum detectable effect at 80% power falls steeply as cluster count rises from 8 to 32. The contribution is a design-theoretic account of what stain-batch confounding does to AI benchmark inference, and a reporting standard that requires stain-batch metadata to be recorded alongside model outputs so that calibration transfer can be estimated rather than assumed.