Artificial Intelligence Governance of Community Health Insurance Fraud Detection Models Under Unmeasured Village Insurance Committee Exemption Authority: A Healthcare Political Robustness Evaluation

Community health insurance schemes in low- and middle-income countries routinely delegate exemption decisions to village insurance committees whose discretionary authority is never recorded in the administrative data that fraud detection models consume. the present analysis treats that unmeasured authority as a structural feature of the governance environment rather than as missing data, and asks what it does to the operating characteristics of the before-after-control-impact design that evaluators would use to judge whether an artificial intelligence fraud detection model improves scheme performance. The analysis is a design-based power evaluation, not an empirical study: it characterises the before-after-control-impact estimator under a declared variance structure with between-cluster standard deviation of 6.0, residual standard deviation of 4.5 and intraclass correlation of 0.64, across cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points, with 2000 replications per design cell. Power to detect a 5 percentage point interaction reaches 80 per cent at 54 clusters, a sample size that exceeds the administrative capacity of most district schemes. The political authority vested in village committees is not a nuisance parameter to be modelled away but a governance variable whose absence from the design constrains what any fraud detection evaluation can credibly claim, and it derives the conditions under which a robustness claim survives that absence.