Artificial Intelligence Causal Inference of Community Clinic Snakebite Antivenom Allocation Under Unmeasured Species-Identification Delay: A Healthcare Instrumental Variable Mortality Equity Evaluation

Community clinics in sub-Saharan Africa allocate snakebite antivenom without knowing which species caused the bite, because species identification by clinical syndrome is unreliable and laboratory confirmation is rarely available at the point of care. The framework treats the delay between bite and species-confirmed diagnosis as an unmeasured confounder of the allocation–mortality relationship and proposes clinic catchment distance as an instrument that affects allocation timing but not mortality except through that timing. Identification is established under a conditional independence condition and a monotonicity condition on the instrument, and the local average treatment effect is derived for the subpopulation of patients whose allocation is shifted by catchment distance. A sensitivity bound is derived for the case in which the exclusion restriction is only approximately satisfied. The framework is then applied to a real World Bank panel of 560 country-year observations from 24 African countries over 2000–2023, in which gross domestic product per capita is regressed on access to electricity and urban population share; the fit is R-squared = 0.591 and the coefficient on access to electricity is 1291.55405 (SE = 194.35159). Instrumental-variable designs can make species-identification delay an explicit object of inference rather than an unacknowledged source of allocation bias, and it specifies the conditions under which antivenom equity claims are identifiable from routine clinic data.