Artificial Intelligence Counterfactual Sepsis Fluid Resuscitation Dosing Under Variable Intraoperative Blood Pressure Artifact: A Doubly Robust Off-Policy Evaluation of Healthcare Deterioration Outcomes

Intraoperative blood pressure artifact, arising from non-invasive cuff cycling, vasopressor boluses and motion, corrupts the state representation on which sepsis fluid resuscitation policies are trained and evaluated. The framework combines a measurement-error correction for the behaviour policy with an augmented inverse-propensity-weighted estimator of deterioration outcomes, and it establishes identification conditions under which the corrected estimator is consistent while the naive estimator is not. The empirical component uses real World Bank World Development Indicators for 24 African countries over 2000 to 2023 (N = 560 country-years) to estimate the macro-level relationship between access to electricity, urban population share and GDP per capita (constant 2015 US$); the ordinary least squares model yields R-squared = 0.591, adjusted R-squared = 0.5895, F = 184.764, and coefficients of 1291.55405 (SE 194.35159) for access to electricity, 113.13952 (SE 7.77493) for urban population and -104.41834 (SE 11.31314) for the linear time term. These estimates are used only to calibrate the scale of the covariate shift that measurement error induces in the policy inputs, not to claim a clinical effect.