Artificial Intelligence Governance of Clinical Decision-Support Procurement Under Unmeasured Vendor Retraining Notifications: A Healthcare Auditability-Equivalence Evaluation

Clinical decision-support systems procured by hospitals are routinely retrained by their vendors after deployment, and the notifications that accompany those retraining events are unmeasured in the sense that no procurement instrument specifies what must be disclosed, in what form, or with what evidentiary force. the present analysis treats that gap as a governance problem of auditability equivalence: the condition under which a regulator, auditor or procuring authority can establish that a post-retraining system is evidentially equivalent to the system that was originally certified. A replication-based evaluation of the before-after-control-impact estimator is then used to characterise the operating characteristics of the design that a notification regime would need to support, under a declared variance structure with between-cluster SD of 6.0, residual SD of 4.5 and intraclass correlation of 0.64. 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 design attains 80 per cent power at 54 clusters for a 5 percentage point effect. Auditability equivalence is a procurement obligation that can be specified ex ante, and that unmeasured retraining notifications convert a tractable design problem into an intractable evidentiary one.