Artificial Intelligence replication of District Hospital Computed Tomography Triage Under Unmeasured Contrast Media Viscosity Drift: A Healthcare Safety Evaluation

Computed tomography (CT) triage in district hospitals increasingly relies on automated contrast-enhanced protocols whose safety margins depend on contrast media viscosity, a property that is rarely measured at the point of care and that drifts with storage temperature and formulation variability. The framework couples a pharmacokinetic contrast-transport model to a triage classifier and derives a viscosity-drift sensitivity bound on triage error. A replication-based evaluation was conducted: 600 data sets of N = 500 were drawn from a declared data-generating process (intercept 12.0, exposure 0.42, group -3.1, severity 1.35, error SD 4.2), and the behaviour of the ordinary least squares estimator was characterised under the stated process. The estimator recovered the exposure coefficient with negligible bias (-0.00108) and empirical coverage of the nominal 95% confidence interval of 1.0 for both intercept and exposure, while the intercept bias was 0.07473 with RMSE 0.92702. The framework proposes that viscosity drift acts as a structured perturbation whose safety consequence is bounded by the classifier's margin distribution rather than by its average accuracy, and it identifies the conditions under which triage safety claims remain valid without direct viscosity measurement.