Artificial Intelligence Pediatric Sepsis Alerting Under Variable Capillary Refill Time Annotation Disagreement: A Label-Noise-Robust Evaluation of Healthcare Escalation Accuracy

Capillary refill time (CRT) is a bedside component of pediatric sepsis screening, but its annotation is subjective and disagreement among clinicians is common. the present analysis characterises the operating characteristics of a label-noise-robust evaluation design for AI-based pediatric sepsis alerting when CRT labels are discordant. The design is a before-after-control-impact (BACI) estimator with a declared variance structure: between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.640. Power was estimated across 2000 design cells for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The analysis is a replication-based evaluation; no observational data are used or claimed. The design achieves 80% power at 54 clusters for a 5-percentage-point effect. The paper derives the bias and variance of the BACI estimator under label noise, shows that attenuation is governed by the disagreement rate and the intraclass correlation, and proposes a noise-adjusted estimator. The contribution is a formal account of how CRT annotation disagreement propagates into escalation-accuracy estimates and a design template for label-noise-robust evaluation of pediatric sepsis alerting.