Point-of-Care Artificial Intelligence Triage for Febrile Illness Under Malaria Rapid Diagnostic Test Lot Variability: A Calibration-Transfer Evaluation of Antibiotic Overprescription Across District Clinics

Malaria rapid diagnostic tests (RDTs) are the front-line confirmation tool for febrile illness in district clinics, yet their lot-to-lot variability in limit of detection is rarely propagated into the calibration of point-of-care artificial intelligence (AI) triage models that decide antibiotic prescribing. The framework is formalised as a bounded excess-risk decomposition in which the antibiotic overprescription rate is expressed as a function of the sensitivity and specificity of the RDT lot, the calibration slope of the triage model, and the prevalence of non-malarial bacterial illness. A design-based evaluation of the before-after-control-impact estimator used to audit overprescription is then characterised under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64), with power 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. The analysis indicates that 54 clusters are required to detect a 5-percentage-point reduction in overprescription at 80% power, and that lot-conditional recalibration of the triage threshold is the mechanism through which calibration transfer can be made safe. The contribution is a formal statement of the conditions under which point-of-care AI triage remains calibrated across RDT lots, and a design-level account of how much evidence is needed to detect the resulting change in prescribing.