We estimate the health-expenditure cost of wintertime sulfur dioxide (SO2) released when power plants near Tehran burn mazut (heavy fuel oil) in place of natural gas during winter shortages. We assemble a daily panel of 1,330 five-digit postal zones over 2018–2023, some 500,000 postal–day observations, linking ground-station pollution monitors and meteorological records (IRIMO, ERA5) to administrative pharmacy transactions for respiratory medicines. Identification uses a source-aware wind instrument combining downwind alignment, wind speed and inverse-distance weights to two plants, restricted to the heating season. Ordinary least squares with postal and month–year fixed effects gives a 1.3 per cent rise in same-day respiratory-drug expenditure per microgram per cubic metre of winter SO2 (SE 0.3). Two-stage least squares yields a larger semi-elasticity of 0.078 (SE 0.010), consistent with attenuation from measurement error and with the instrument isolating high-emission episodes; the first stage is strong (Kleibergen–Paap rk F ≈ 38). Placebo tests randomising wind direction or rebuilding the instrument from fictitious plant locations, and a falsification test on the previous day's expenditure, are all insignificant. Effects accumulate beyond the exposure day, the cumulative fourteen-day response being 0.017 (SE 0.006). Placed in a social-cost framework in which two of four bridging multipliers are measured within the same data, the morbidity cost of the observed SO2 increment is 0.72 per cent of the affected population's health expenditure under the cumulative coefficient and 3.52 per cent under the instrumental-variables coefficient. Both are lower bounds: mortality and non-health damages are excluded.