Background: Sepsis remains a leading cause of mortality. Efforts to improve sepsis outcomes focus on early recognition and treatment. However, outcomes may also depend on patient and hospital characteristics. Our goal was to describe hospital-level variability in sepsis mortality and explore if this variability can be explained by common hospital characteristics.
Methods: Administrative claims data from Centers for Medicare and Medicaid Services (CMS) fee for service beneficiaries discharged from US hospitals in 2013 were used to retrospectively identify the incidence of death among discharges with ICD-9 codes for severe sepsis (995.92) or septic shock (785.52) by hospital. Hospital characteristics were obtained from the CMS Healthcare Cost Report Information System. We created a multivariable model to explain the variability in the incidence of death using hospital-level characteristics: teaching status, bed size, urban vs. rural location, volume of Medicare patients, case-mix index, annual number of sepsis discharges, average patient age, average patient co-morbidity score, and percent of sepsis hospitalizations discharged to hospice.
Results: Of 3054 acute care hospitals, 2340 (77%) had > 10 discharges with a code for severe sepsis/septic shock in 2013. Among these hospitals, the median number of discharges with codes for severe sepsis/septic shock was 77 (IQR: 35-147); the median incidence of death was 28 per 100 sepsis discharges (IQR: 21.6-35.1, Figure 1). Hospital-level characteristics explained 14.7% of the variability in the incidence of death; the majority explained by two variables: the percent of sepsis hospitalizations discharged to hospice (7.1%) and the annual number of sepsis discharges (3.4%).
Conclusion: There is significant hospital-level variability in the incidence of death among hospitalized Medicare patients coded with severe sepsis/septic shock. This variability is not explained by hospital-level characteristics such as bed size, teaching status, volume, geographic location, and case-mix index. Better understanding of hospital characteristics associated with variability in sepsis mortality may inform surveillance for sepsis and associated mortality, and potentially identify targets for intervention.
R. Dantes, None
A. Fiore, None
J. Jernigan, None
S. S. Magill, None
L. Epstein, None
See more of: Poster Abstract Session