Hospitalizations, costs, and outcomes of severe sepsis in the United States 2003 to 2007

Objectives: To assess trends in number of hospitalizations, outcomes, and costs of severe sepsis in the United States. Design: Temporal trends study using the Nationwide Inpatient Sample. Patients: Adult patients with severe sepsis (defined as a diagnosis of sepsis and organ dysfunction) diagnosed between 2003 and 2007. Measurements and Main Results: We determined the weighted frequency of patients hospitalized with severe sepsis. We calculated age- and sex-adjusted population-based mortality rates for severe sepsis per 100,000 population and also used logistic regression to adjust in-hospital mortality rates for patient characteristics. We calculated inflation-adjusted costs using hospital-specific cost-to-charge ratios. We identified a rapid steady increase in the number of cases of severe sepsis, from 415,280 in 2003 to 711,736 in 2007 (a 71% increase). The total hospital costs for all patients with severe sepsis increased from $15.4 billion in 2003 to $24.3 billion in 2007 (57% increase). The proportion of patients with severe sepsis and only a single organ dysfunction decreased from 51% in 2003 to 45% in 2007 (p < .001), whereas the proportion of patients with three or four or more organ dysfunctions increased 1.19-fold and 1.51-fold, respectively (p < .001). During the same time period, we observed 2% decrease per year in hospital mortality for patients with severe sepsis (p < .001), as well as a slight decrease in the length of stay (9.9 days to 9.2 days; p < .001) and a significant decrease in the geometric mean cost per case of severe sepsis ($20,210 per case in 2003 and $19,330 in 2007; p = .025). Conclusions:The increase in the number of hospitalizations for severe sepsis coupled with declining in-hospital mortality and declining geometric mean cost per case may reflect improvements in care or increases in discharges to skilled nursing facilities; however, these findings more likely represent changes in documentation and hospital coding practices that could bias efforts to conduct national surveillance. (Crit Care Med 2012; 40:754-761)