Artificial Intelligence Causal Inference of Community Clinic Referral Transport Voucher Allocation Under Unmeasured Road-Impassability Confounding by Seasonal Flooding: A Healthcare Instrumental Variable Equity Evaluation

Community clinic referral transport voucher programmes allocate scarce resources to patients who must reach higher-level care, but seasonal flooding makes road impassability an unmeasured confounder that biases naive equity comparisons. The design parameters were clusters of 8, 16, 24 and 32, effect sizes of 3, 5 and 8 percentage points, and 2000 replications per cell, with a between-cluster standard deviation of 6.0, a residual standard deviation of 4.5 and an intraclass correlation of 0.64. The analysis is a replication-based evaluation of the before-after-control-impact estimator; no observational data are used or claimed. Power curves, minimum detectable effect curves, kernel density estimates, residual diagnostics and coefficient plots were generated from the declared variance structure. The framework proposes that the interaction instrument identifies the local average treatment effect of voucher allocation among compliers whose road access is plausibly exogenous to unobserved clinic-level demand, and the analysis is designed to identify the conditions under which the design attains 80 per cent power. The contribution is a formal identification argument and a design-characterisation protocol for equity evaluation of transport vouchers under seasonal road-impassability confounding.