PURPOSE: The aim of the study was to compare estimates of hospitalizations, outcomes, and costs produced by 2 approaches for defining severe sepsis. METHODS: We used the Nationwide Inpatient Sample to study adults hospitalized in the United States in 2007. We defined severe sepsis using 2 previously published algorithms: (1) the presence of a principal or secondary diagnosis of septicemia combined with organ dysfunction or (2) the presence of a principal or secondary diagnosis of septicemia or another infection (eg, pneumonia) combined with organ dysfunction. For each approach, we calculated the weighted frequency of hospitalizations, population-based mortality rates, and geometric mean costs. RESULTS: A total of 719099 (SD, 16676) hospitalizations had a diagnosis of septicemia and a diagnosis of organ dysfunction. A total of 2.5 million hospitalizations were recorded, with a diagnosis code for either septicemia or infection combined with a diagnosis code for organ dysfunction. Hospitalizations without a diagnosis code for septicemia had lower rates of respiratory failure (35% vs 51%, P < .001) or shock (20% vs 46%, P < .001), lower in-hospital mortality (8% vs 29%, P < .001), and lower mean costs. CONCLUSIONS: An approach that requires a diagnosis code for septicemia and a diagnosis code for organ dysfunction yields estimates of disease burden and outcomes that are more consistent with chart-based studies. Copyright © 2012 Elsevier Inc. All rights reserved.