An upper estimate of the attributable mortality and cost of severe sepsis in surgical patients

Introduction

Although patients developing severe sepsis (SS) incur high mortality and costs, the extent to which these events are attributable to SS is unclear, especially in surgical patients. We conducted a retrospective analysis of the US Medicare hospital discharge database to estimate the resource use and mortality attributable to SS in patients > 65 years old.

Methods

We selected all major surgical prospective payment system (PPS) discharges from the FY 2000 Medicare hospital discharge database. Costs were calculated by multiplying charges by institution-specific cost to charge ratios. We defined SS as documented infection plus acute organ dysfunction using ICD-9-CM-based criteria. We calculated hospital mortality, length of stay (LOS), ICU LOS, and total cost for those with and without SS within each surgical DRG. Attributable outcomes were defined as differences in absolute outcomes between those with and without SS within DRG.

Results

There were 2.88 million major surgery PPS discharges, of whom 99.903 (3.4%) developed SS. There were 834,722 discharges that incurred ICU care, of whom 66,222 (7.9%) developed SS. In the entire SS cohort, the actual mortality was 26.4%, the expected mortality was 4.8%, and the potential attributable mortality was 21.6% (n = 21,628). In the ICU SS cohort, the actual mortality was 31.1%, the expected mortality was 8.4%, and the potential attributable mortality was 22.6% (n = 15,057). The additional SS deaths represented 23.7% of all surgical deaths and 25.3% of all surgical ICU deaths. SS was also responsible for 825,870 hospital days (4.2% of all), 443,113 ICU days (12.6% of all), and US$1,812,415,000 hospital costs (4.7% of all costs). This represents an additional 4.4 ICU days, 3.9 floor days, and US$18,142 per patient.

Discussion

The attributable mortality of severe sepsis in elderly surgical patients appears to be as much as 22%. The only factors reducing this rate would be concomitant conditions that covaried with SS and were not captured in our case-mix adjustment. Importantly, the attributable rate was so high that SS, although occurring in only a small number of patients, resulted in one-quarter of all surgical deaths. The attributable costs and resource use were also very high.