INTRODUCTION: The American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference defined severe sepsis as a systemic inflammatory syndrome in response to infection associated with acute organ dysfunction. Often, septicemia codes have been used in administrative datasets as a proxy for severe sepsis. However, these entities are not necessarily the same and one study found large differences in mortality between septicemia patients with and without severe sepsis (54% vs. 15%). We explored the accuracy of septicemia codes as identifiers of severe sepsis. METHODS: We selected all patients with septicemia or severe sepsis (bacterial or fungal infection plus acute organ failure) in the 1996 Florida hospital discharge database (n=1,936,479) and compared differences between the groups. Septicemia was defined using the ICD-9-CM code 038.xxx. We defined severe sepsis using a more sophisticated strategy previously validated against prospective clinical and physiologic criteria. RESULTS: We found 58,598 patients with severe sepsis, 53.4% of whom were in the intensive care unit (ICU). We found 57,875 patients with septicemia, 30.9% of whom were in the ICU. Patients with severe sepsis had a higher mortality (24.1% vs. 18.0%, p<0.001) and higher hospital costs ($20.4k vs. $14.4k, p<0.001). For patients with an ICU stay, hospital mortality was 14.4% among the 7,927 septicemia cases without severe sepsis, 23.5% among the 21,655 cases of severe sepsis without septicemia and 42.1% among the 9,970 cases of severe sepsis with septicemia. The mean hospital cost in these three groups was $18,381, $24,396, and $33,470 respectively. Of all selected patients (n=99,126), 17.5% met criteria for both septicemia and severe sepsis. The sensitivity and positive predictive values of septicemia codes as predictors of severe sepsis are 29.6% and 30.0% respectively. CONCLUSIONS: Septicemia codes are not accurate for identifying patients with severe sepsis.