Artificial Intelligence Counterfactual Antiretroviral Adherence Support Allocation Under Variable Community Adherence Group Ledger Completeness: A Doubly Robust Evaluation of Healthcare Retention Targeting

Community adherence groups (CAGs) are a central mechanism for retaining people living with HIV in care across sub-Saharan Africa, yet the ledgers that record member attendance and support receipt are frequently incomplete, and the resulting missingness is not ignorable. The estimator combines an outcome regression with an inverse-probability-of-completeness weight, and its asymptotic bias is characterised under a general missingness mechanism. A replication-based evaluation of the before-after-control-impact design was conducted under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64; power was estimated across 2000 design cells.54 clusters are required to detect a 5 percentage-point effect at 80% power, and that the minimum detectable effect falls steeply as cluster count rises from 8 to 32. The contribution is a design and estimation framework that treats ledger completeness as a structured nuisance rather than a data-cleaning problem, and that specifies the cluster counts at which retention targeting can be evaluated credibly.