Artificial Intelligence Triage for Stroke Thrombolysis Under Variable CT Scanner Age and Slice Thickness: A Domain-Shift Evaluation of Hemorrhage Exclusion Sensitivity Across Referral Hospitals

Artificial intelligence triage tools for stroke thrombolysis are typically trained on CT images from modern scanners and then deployed in referral hospitals where scanner age and slice thickness vary widely. the present analysis characterises the operating characteristics of a domain-shift evaluation design for haemorrhage exclusion sensitivity under such variation, using a replication-based power analysis of a before-after-control-impact estimator. No observational data are used or claimed. Under a declared variance structure with between-cluster SD = 6.0, residual SD = 4.5 and intraclass correlation = 0.64, power was estimated across 2000 design cells spanning 8, 16, 24 and 32 clusters and effect sizes of 3, 5 and 8 percentage points.54 clusters are required to detect a 5 percentage point change in sensitivity at 80% power, and that the minimum detectable effect at 80% power falls steeply as cluster count rises. The results establish the sample-size floor that any multi-site evaluation of scanner-age domain shift must meet, and they show that the intraclass correlation of 0.64 dominates the design's sensitivity. Domain-shift evaluations of AI triage for thrombolysis are underpowered at the cluster counts typically available to referral-hospital networks, and it specifies the design conditions under which a defensible sensitivity claim can be made.