Artificial Intelligence Conformal Prediction of Community Clinic HIV Viral Load Suppression Under Unmeasured Assay Interruption: A Healthcare Coverage Guarantee Evaluation

Conformal prediction offers distribution-free coverage guarantees for individual predictions, yet its validity rests on the exchangeability of calibration and test data. In community HIV care, assay interruption—the unmeasured failure to obtain a viral load result—violates this assumption by censoring the outcome in a manner that is informative and non-random. The study drew 600 data sets of N = 500 from a declared data-generating process with intercept 12.0, exposure 0.42, group -3.1, severity 1.35, and error SD 4.2, and characterised the behaviour of the ordinary least squares estimator under the stated process. The analysis found that estimated bias was -0.03201 for the intercept and 0.00092 for the exposure coefficient, with RMSE of 0.90794 and 0.01615 respectively. Empirical coverage of the nominal 95% confidence intervals was 1.0 for both parameters. The results indicate that under the declared process, the estimator is approximately unbiased and the interval coverage is conservative. Conformal prediction under assay interruption requires explicit modelling of the censoring mechanism, and proposes a sensitivity framework that bounds the coverage guarantee as a function of the unmeasured interruption rate. The contribution is a formal characterisation of when conformal guarantees survive informative censoring and a practical protocol for reporting coverage under assay interruption in community HIV monitoring.