Artificial Intelligence Triage for Tuberculosis Treatment Failure Under Variable Sputum Conversion Definitions: A Survival-Model Calibration Evaluation of Regimen-Switch Timing Errors

Sputum conversion definitions for pulmonary tuberculosis vary across programmes, and the timing of regimen-switch decisions that depend on them is a calibration problem rather than a classification problem. the present analysis formalises artificial-intelligence triage for treatment failure as a survival-model decision rule whose operating characteristics depend jointly on the conversion threshold, the underlying hazard of failure, and the calibration of the predicted risk. A before-after-control-impact design is characterised analytically 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 effect sizes of 3, 5 and 8 percentage points. The intraclass correlation of 0.64 dominates the design's detectable-effect profile: 56 clusters are required for 80% power at a 5-percentage-point effect, and the minimum detectable effect at 80% power falls steeply as cluster count rises from 8 to 32. The contribution is a calibration-theoretic account of why regimen-switch timing errors are asymmetric under variable conversion definitions, and a design-level statement of the cluster counts that a credible evaluation of such triage would require.