Artificial Intelligence Benchmarking of Speech-Based Tuberculosis Cough Screening Under Unmeasured Dialect-Confounded Audio Compression: A Healthcare Fairness-Transfer Evaluation

Speech-based cough screening for tuberculosis is frequently benchmarked on curated audio corpora that differ from deployment conditions in ways that are rarely measured. One such difference is audio compression, which varies across devices and transmission channels and may interact with dialect to alter the acoustic features on which classifiers depend. The framework treats the before-after-control-impact estimator as the unit of analysis and characterises its operating characteristics under a declared variance structure with an intraclass correlation of 0.64, a between-cluster standard deviation of 6.0 and a residual standard deviation of 4.5. Power was estimated across 2000 design cells spanning cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points.54 clusters are required to detect a 5 percentage point effect at 80 per cent power, and that the minimum detectable effect at 80 per cent power declines steeply as cluster count rises from 8 to 32. The framework distinguishes compression-dialect confounding from ordinary domain shift and specifies the conditions under which fairness-transfer claims are identifiable. The contribution is a methodological apparatus for benchmarking fairness transfer in resource-constrained screening pipelines, not a report of field performance.