Although severe hypophosphatemia has been associated with significant morbidity, little quantitative information is available on the magnitude of the risk. The objective of this study was to determine the frequency and timing of severe hypophosphatemia, the relative risk associated with various medical diagnoses, and the choice of therapy. Among 10,197 patients hospitalized over a 1-year period, the incidence of severe hypophosphatemia (serum phosphate less than or equal to 0.33 mmol/l) was 0.43%. The incidence was 0.91% in alcoholics, 2.42% in septic patients, 10.4% in patients with malnutrition, and 14.6% in patients with diabetic ketoacidosis. The odds ratio for developing the metabolic abnormality was 4.2 (95% confidence interval 2.2-8.1) in alcoholics, 6.4 (1.9-18.9) in septic patients, 46.9 (16.6-125.6) in patients with diabetic ketoacidosis, and 60.7 (31.4-117.5) in malnourished patients. There was a 4-fold increase in mortality in patients with severe hypophosphatemia (18.2 vs. 4.6%; p less than 0.001). The metabolic abnormality was not recognized or was inappropriately treated in 42% of patients. Severe hypophosphatemia is a relatively common finding in hospitalized patients, especially in those with certain well-defined medical conditions. It is associated with significant excess mortality. Its recognition and treatment remain suboptimal.