Artificial Intelligence Causal Inference of Community Clinic Tuberculosis Contact Tracing Coverage Under Unmeasured Household Stigma Confounding: A Healthcare Instrumental Variable Equity Evaluation

Contact tracing coverage for tuberculosis in community clinics is a key determinant of equity in case finding, yet household stigma is rarely measured and may confound both coverage and outcomes. The framework is evaluated through a replication-based replication study of 600 data sets of N = 500 drawn from a declared data-generating process with intercept 12.0, exposure 0.42, group -3.1, severity 1.35, and error SD 4.2. The ordinary least squares estimator is characterised under this process. Across replications, the estimated bias for the intercept was -0.02797 and for the exposure coefficient 0.00108; the root mean squared error was 0.8576 for the intercept and 0.01555 for the exposure coefficient. Empirical coverage of the nominal 95% confidence intervals was 1.0 for both parameters. The results indicate that under the stated process the estimator is approximately unbiased and that interval coverage is conservative. Instrumental variable designs can support equity-oriented causal inference in tuberculosis contact tracing when stigma is unmeasured, but that the credibility of such designs depends on assumptions that cannot be tested from the data alone.