Interpretability Auditing of Clinical Decision Support Computer Models Under Unmeasured Clinician Override Documentation Gaps: A Healthcare Safety Evaluation

Clinical decision support (CDS) models are routinely audited for predictive performance, yet the audit typically treats the clinician's override as an observed event. In practice, overrides are documented incompletely and non-randomly, so the audit sample is a selected subset of the decisions the model actually influenced. the present analysis formalises that selection as a missingness mechanism operating on the override indicator and derives its consequences for interpretability auditing. A before-after-control-impact design is characterised under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64), with power estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effects of 3, 5 and 8 percentage points. The design attains 80% power at 54 clusters for a 5-percentage-point effect. The central result is a bracketing proposition: under documentation gaps that are ignorable given observed covariates, the audit's interpretability estimand is recoverable; under gaps that depend on the override itself, the estimand is only partially identified, and the identified set widens with the documentation rate.