Artificial Intelligence replication of District Hospital Medical Gas Pipeline Leak Localisation Under Unmeasured Acoustic Sensor Placement Bias: A Healthcare Detection-Latency Safety Evaluation

Acoustic leak localisation in hospital medical gas pipelines depends on sensor placement that is rarely measured and almost never recorded, so the detection latency reported by replication studies may be optimistic in ways that no field calibration can reveal. the present analysis treats placement bias as an unmeasured nuisance parameter and asks what happens to localisation error and alarm latency when the estimator is fitted under a declared data-generating process but evaluated under displaced sensor geometry. A replication-based evaluation was conducted: 600 data sets of N = 500 were drawn from a declared process with intercept 12.0, exposure 0.42, group -3.1, severity 1.35 and error SD 4.2, and the ordinary least squares estimator was characterised across replications. The estimator recovered the exposure coefficient with negligible bias (0.00025) and RMSE 0.01569, while the intercept carried bias -0.02335 and RMSE 0.87586; nominal 95% coverage was 1.0 for both parameters. Coefficient-level fidelity of this kind is a necessary but insufficient condition for safety, because placement bias enters through the design matrix rather than through the noise term, and it derives the conditions under which latency bounds remain valid. The contribution is a formal statement of when acoustic localisation replications can be trusted for district hospital safety cases, and when they cannot.