Session: Poster Session: Blood Stream Infections
Monday, October 27, 2008: 12:00 AM
Room: Hall C
Background: National Hospital Discharge Survey data reportedly show declining case-fatality rates for septicemia between 1979-1984 and 1995-2000, but hospital stays also shortened, and discharge data do not capture deaths out of hospital. Methods: The 1991-2005 Medicare Provider Analysis and Review (MEDPAR) 5% sample files were searched for patients 65 years or older discharged from short-stay hospitals after admission for septicemia or bacteremia (120,899 discharges). The Healthcare Cost and Utilization Project (HCUP) National Inpatient Sample provided supplemental data on patients of all ages (5,063,830 discharges). Results: Most cases fell under Diagnosis Related Groups (DRGs) 416 or 415. Declines in mean length of stay for MEDPAR cases under DRG 416 from 10.6 to 7.8 d in 1991-1996 were counterbalanced by a 65% rise in annual discharges. Case-fatality rates observed in hospital declined in parallel with length of stay, but survival to 28 days after admission did not improve. After 1996, with length of stay stable at 7.2-7.5 d, annual discharges first fell 22% through 2001 and then rebounded by 47% by 2005. Meanwhile, deaths by 28 d adjusted for inpatient days climbed 91% over the 15 y. Multivariable analysis suggested that longer stays protected against 28-d fatality in patients discharged alive. Results for septicemia cases in DRG 415 were similar. MEDPAR data also showed that etiologically specific principal diagnoses were progressively displaced by "unspecified septicemia" and "bacteremia." HCUP data did not include survival after discharge but otherwise showed trends like those observed in MEDPAR. Conclusions: Right-censoring confounded survival data from hospital records. Fatalities at 28 d per 1000 patient-days have risen year after year, at least in aged patients. Criteria for admission, timing of discharge, and effectiveness of care for septicemic patients should be reassessed. The possible role that nonspecific diagnoses play in poor prognosis should also be examined.