Artificial Intelligence Cervical Cancer Screening Under Variable Human Papillomavirus Self-Sampling Kit Return Rates: A Causal Transportability Evaluation of Healthcare Recall Yield

Artificial intelligence (AI) triage models for cervical cancer screening are typically validated on populations with high human papillomavirus (HPV) self-sampling kit return rates, yet deployment often occurs where return rates are low and heterogeneous. the present analysis formalises the transportability of AI-guided recall yield as a causal estimand and characterises the operating characteristics of a before-after-control-impact (BACII) design intended to evaluate it. No observational data are used or claimed; the analysis is a replication-based design evaluation under a declared variance structure with between-cluster SD = 6.0, residual SD = 4.5, and intraclass correlation = 0.640. Power was estimated across 2000 design cells spanning 8, 16, 24 and 32 clusters and effects of 3, 5 and 8 percentage points. The design requires 56 clusters to detect a 5-percentage-point effect at 80% power. The results quantify the sample sizes at which BACII can distinguish AI-attributable recall gains from return-rate drift, and they identify the cluster counts at which the design becomes uninformative. The contribution is a transportability-aware design framework that treats kit return rate as a causal modifier rather than a nuisance covariate, and that specifies the conditions under which AI screening evaluations can support valid inference across heterogeneous return-rate regimes.