BACKGROUND Urine specific gravity (USG) is often used by clinicians to estimate urine osmolality. USG is measured either by refractometry or by reagent strip. OBJECTIVE We studied the correlation of USG obtained by either method with a concurrently obtained osmolality. METHODS Using our laboratory's records, we retrospectively gathered data on 504 urine specimens on patients on whom a simultaneously drawn USG and an osmolality were available. Out of these, 253 USG's were measured by automated refractometry and 251 USG's were measured by reagent strip. Urinalysis data on these subjects were used to determine the correlation between USG and osmolality, adjusting for other variables that may impact the relationship. The other variables considered were pH, protein, glucose, ketones, nitrates, bilirubin, urobilinogen, hemoglobin, and leukocyte esterase. The relationships were analyzed by linear regression. RESULTS This study demonstrated that USG obtained by both reagent strip and refractometry had a correlation of approximately 0.75 with urine osmolality. The variables affecting the correlation included pH, ketones, bilirubin, urobilinogen, glucose, and protein for the reagent strip and ketones, bilirubin, and hemoglobin for the refractometry method. At a pH of 7 and with an USG of 1.010 predicted osmolality is approximately 300  mosm/kg/H(2)O for either method. For an increase in SG of 0.010, predicted osmolality increases by 182  mosm/kg/H(2) O for the reagent strip and 203  mosm/kg/H(2)O for refractometry. Pathological urines had significantly poorer correlation between USG and osmolality than "clean" urines. CONCLUSIONS In pathological urines, direct measurement of urine osmolality should be used.