In order to study the clinical consequences of postoperative hypomagnesemia, the serum magnesium (Mg) concentration was measured in samples of blood collected from 193 patients admitted to two postoperative ICUs. On admission to the ICU, 117 patients (61 percent) had hypomagnesemia (serum Mg less than 1.5 mEq/dl), 66 patients (34 percent) had normomagnesemia (1.5 to 2.0 mEq/dl), and ten patients (5 percent) had hypermagnesemia (greater than 2.0 mEq/dl). There were no correlations between the severity of illness score (r = 0.145) or the degree of hypoproteinemia (r = 0.01) and the postoperative serum Mg level. Patients with severe hypomagnesemia (serum Mg less than or equal to 1.0 mEq/dl) experienced hypokalemia more often (p less than 0.02) than the others in the study. Furthermore, those with severe hypomagnesemia had a higher mortality rate (7/17 or 41 percent) than the remainder of the population studied (22/176 or 13 percent) (p less than 0.02). Those with severe hypomagnesemia had received aminoglycosides more often (p less than 0.001) than those with normal serum Mg concentrations. The serum Mg level was not a sensitive (68 percent) or specific (37 percent) predictor of survival. Our conclusions were as follows: (1) hypomagnesemia is common in postoperative ICU patients; and (2) patients in the postoperative ICU who have severe hypomagnesemia have a higher mortality and more hypokalemia than similarly ill patients with normomagnesemia. Because of the association between aminoglycoside therapy and severe hypomagnesemia, we recommend measurement of this variable in those patients receiving aminoglycosides. Furthermore, Mg replacement therapy is recommended for those patients with serum Mg values of 1 mEq/dl or less.