Artificial Intelligence Counterfactual Audit of Community Clinic Triage Algorithm Updates Under Temporal Drift in Presenting Case Mix: A Healthcare Sequential Inference Stability Evaluation

This is the authors' abstract. We don't add key points for this paper.

Community clinic triage algorithms are periodically updated to track drifting case mix, yet the counterfactual question of whether an update improves or degrades sequential inference stability is rarely posed formally. No observational data are used or claimed: the analysis characterises the design under a declared variance structure with between-cluster SD = 6.0, residual SD = 4.5 and intraclass correlation = 0.640, 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 BACIP interaction estimator was evaluated for power, minimum detectable effect at 80% power and interval coverage. Power rose with both cluster count and effect size, and 54 clusters were required to detect a 5 percentage point effect at 80% power under the declared variance structure. The framework separates the counterfactual stability question from the estimation problem and shows that the audit's inferential reach is bounded by cluster count rather than by replication. The contribution is a formalisation of counterfactual triage audit as a sequential inference problem and a design-theoretic account of what BACIP can and cannot establish about algorithm updates under temporal drift.