Although various formulas predicting plasma sodium level ([Na]) are proposed for correction of hyponatremia, it seems that an anticipated [Na] frequently exceeds or falls below the measured [Na], especially in syndrome of inappropriate antidiuretic hormone secretion (SIADH). The causative factors of the fluctuation have never been investigated clearly. The aim of this study was to identify the determining factors for accurate prediction of [Na] by comparing data from previously proposed formulas and a novel osmotic compartment model (O-C model). The O-C model, which simulates the amounts of osmoles in extracellular and intracellular fluids, can estimate resultant osmotic water shift (OWS) and [Na]. The accuracy of representative formulas was verified in a point-to-point study using blood and urine samples obtained every 4 hours from 9 patients. Among 161 measurement points, a large fluctuation of urine volume and urine sodium level was observed. The gap between anticipated and measured [Na] in the widely used Adrogue-Madias formula was -0.5 ± 0.1 mEq/L/4 h (mean ± standard error), showing a marked tendency to underestimate [Na]. The gap in the O-C model including OWS was 0.1 ± 0.1 mEq/L/4 h, and that in the O-C model eliminating OWS was 1.9 ± 0.2 mEq/L/4 h, indicating that measurement of urine output and estimation of resulting OWS are essential for a superior prediction of [Na] in SIADH. A simulation study with the O-C model including OWS unveiled a distinctive correction pattern of [Na] dependent on the urine volume and urine sodium level, providing a useful choice for the proper type and rate of infusion.