Artificial Intelligence Sparse-Data Estimation of HIV Pre-Exposure Prophylaxis Discontinuation Under Variable Clinic Attendance Censoring: A Causal Survival Evaluation of Healthcare Retention Targeting

HIV pre-exposure prophylaxis (PrEP) programmes in sub-Saharan Africa lose a substantial fraction of initiates within the first year, yet the retention literature is built on clinic attendance records that are themselves generated by the same retention process they are used to measure. When attendance is irregular, the observation of discontinuation is censored by the very behaviour under study, and naive survival estimators conflate true discontinuation with unobserved follow-up. The framework combines a longitudinal targeted maximum likelihood estimator with a neural-network nuisance model that borrows information across sparse attendance histories, and it is evaluated through a replication-based design study. The design study characterises the operating characteristics of the estimator under a declared variance structure with between-cluster SD of 6.0, residual SD of 4.5 and intraclass correlation of 0.64, across 2000 replications per design cell. Power is estimated for cluster counts of 8, 16, 24 and 32 and effects of 3, 5 and 8 percentage points; 52 clusters are required for 80% power at a 5-percentage-point effect. The contribution is a formal account of when sparse-data causal survival estimation can recover retention effects that conventional Kaplan-Meier and Cox analyses systematically misattribute to loss to follow-up.