Clinical Computer Vision for Dengue Warning-Sign Recognition Under Variable Capillary Refill Assessment Skill: A Human-in-the-Loop Deferral Evaluation of Artificial Intelligence Triage Safety

Dengue warning-sign recognition depends on capillary refill assessment, a bedside skill whose reliability varies widely among clinicians. Computer vision systems that estimate refill time from smartphone video could standardise triage, but their safety depends on how they interact with the variable skill of the humans who defer to them. The framework characterises triage safety through a deferral operating characteristic and derives the conditions under which machine deferral reduces, rather than redistributes, triage error. A design-based evaluation of the before-after-control-impact estimator was conducted under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64, 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 effect at 80 percent power. The framework proposes that deferral safety is governed not by model accuracy alone but by the covariance between model and clinician errors, and that this covariance is itself a function of clinician skill. The contribution is a formal account of when human-in-the-loop deferral improves triage safety for dengue warning-sign recognition under variable capillary refill assessment skill.