Artificial Intelligence Inference of Community Clinic Neonatal Jaundice Treatment Thresholds Under Unmeasured Transcutaneous Bilirubin Device Calibration Drift: A Healthcare Distributionally Robust Safety Evaluation

Transcutaneous bilirubinometry is the primary screening modality for neonatal jaundice in community clinics, yet device calibration drift is unmeasured in routine practice and can silently shift measured bilirubin concentrations relative to the serum values on which treatment thresholds were established. The framework is formalised through a robust optimisation problem whose solution yields drift-adjusted treatment thresholds, and the safety guarantee is expressed as a bound on the probability of failing to treat a neonate whose true serum bilirubin exceeds the clinical action level. A design and power analysis of the before-after-control-impact estimator used to evaluate threshold revisions is reported: under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64, 54 clusters are required to detect a 5 percentage-point change at 80% power, and 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. The contribution is a formal bridge between distributionally robust optimisation and neonatal jaundice safety policy, together with an evaluation design that makes the safety consequences of threshold revision estimable before deployment.