Artificial Intelligence replication of District Hospital Teleconsultation Bandwidth Allocation Under Unmeasured Store-and-Forward Latency Censoring: A Healthcare Queueing Robustness Evaluation

Store-and-forward teleconsultation in district hospitals depends on bandwidth allocation rules that are usually tuned against measured latency, yet the latency that matters clinically is frequently unmeasured because store-and-forward transfers are logged only at completion. the present analysis treats that missingness as right-censoring of the latency distribution and asks whether bandwidth allocation rules derived from complete-latency assumptions remain robust when the censoring mechanism is ignored. A queueing model of a district hospital teleconsultation link is formalised, the censoring-aware allocation problem is stated, and a replication-based evaluation of an ordinary least squares estimator of the allocation model is reported. Six hundred data sets of 500 observations were drawn from a declared data-generating process with intercept 12.0, exposure 0.42, group -3.1, severity 1.35 and error standard deviation 4.2, using seed 55634239. The estimator recovered the exposure coefficient with estimated bias 0.0001 and root mean squared error 0.01599, and the intercept with estimated bias 0.00261 and root mean squared error 0.89299; 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 essentially unbiased for the exposure coefficient and that interval coverage is conservative, which bounds the conditions under which censoring-naive allocation rules can be defended. The contribution is a formal statement of the censoring-aware allocation problem together with an explicit robustness bound on the estimator that any such rule must satisfy.