Epidemiology of sepsis and septic shock in critical care units: comparison between sepsis-2 and sepsis-3 populations using a national critical care database

TLDR: In an ICU database, Sepsis-3 identifies a similar sepsis population with 92% overlap and much smaller septic shock population with improved predictive validity, which shows significant risk-adjusted improvements in mortality over time.

Abstract: Background New sepsis and septic shock definitions could change the epidemiology of sepsis because of differences in criteria. We therefore compared the sepsis populations identified by the old and new definitions. Methods We used a high-quality, national, intensive care unit (ICU) database of 654 918 consecutive admissions to 189 adult ICUs in England, from January 2011 to December 2015. Primary outcome was acute hospital mortality. We compared old (Sepsis-2) and new (Sepsis-3) incidence, outcomes, trends in outcomes, and predictive validity of sepsis and septic shock populations. Results From among 197 724 Sepsis-2 severe sepsis and 197 142 Sepsis-3 sepsis cases, we identified 153 257 Sepsis-2 septic shock and 39 262 Sepsis-3 septic shock cases. The extrapolated population incidence of Sepsis-3 sepsis and Sepsis-3 septic shock was 101.8 and 19.3 per 100 000 person-years, respectively, in 2015. Sepsis-2 severe sepsis and Sepsis-3 sepsis had similar incidence, similar mortality and showed significant risk-adjusted improvements in mortality over time. Sepsis-3 septic shock had a much higher Acute Physiology And Chronic Health Evaluation II (APACHE II) score, greater mortality and no risk-adjusted trends in mortality improvement compared with Sepsis-2 septic shock. ICU admissions identified either as Sepsis-3 sepsis or septic shock and as Sepsis-2 severe sepsis or septic shock had significantly greater risk-adjusted odds of death compared with non-sepsis admissions (P Conclusions In an ICU database, compared with Sepsis-2, Sepsis-3 identifies a similar sepsis population with 92% overlap and much smaller septic shock population with improved predictive validity.