The Age Adjusted 12-Year Incidence and Mortality Rates of Sepsis Using the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) on MIMIC-III Data

Abstract

SESSION TITLE: Sepsis & Septic Shock SESSION TYPE: Original Investigation Slide PRESENTED ON: Sunday, October 29, 2017 at 01:30 PM - 03:00 PM PURPOSE: Prior epidemiologic studies defining sepsis incidence have been based on SIRS criteria and ICD-9 codes. While not considered a part of the definition of sepsis, the recent Sepsis-3 criteria are considered a less sensitive but more specific predictor of life threatening infection than the SIRS criteria. The objective of this study is to examine the differences of age adjusted incidence of sepsis in intensive care unit (ICU) patients when applying Sepsis-3 criteria on administrative claims data based on ICD-9 codes for sepsis. METHODS: Retrospective cohort study evaluating 36503 adult (>=18) ICU patients from 2001 to 2012 in the MIMIC-III dataset. Patients were divided into subgroups based on age at admission. The 12-year age adjusted incidence of sepsis was defined after applying Sepsis-3 criteria. Incidence was compared to two prior methods used in prior sepsis incidence studies (Angus method and ICD-9 explicit method). The independency and differences of incidences were tested using Fisher’s test and McNemar’s test, respectively. RESULTS: Within the MIMIC-III data, 43.9%, 27.0%, and 7.0% of the adult ICU patients were identified as defined sepsis when Sepsis-3 criteria, Angus method, and ICD-9 explicit method were applied, respectively. We observed the largest 12-year incidence rate in each of the 7 age groups when the Sepsis-3 definition was applied. When comparing Sepsis-3 with Angus definition, the smallest incidence difference occurred in the age 30-40 group (33.5% vs. 21.4% p<0.0001), while the largest occurred in the age 60-70 group (45.5% vs. 26.3% p<0.0001). When comparing Sepsis-3 with ICD-9 explicit, the smallest incidence difference also occurred in the age 30-40 group (33.5% vs. 5.6% p<0.0001), but the largest occurred in the age 70-80 group (48.2% vs. 7.3% p<0.0001). Meanwhile, the respective overall sepsis-related mortality rates for Sepsis-3, Angus, and ICD-9 explicit were 15.4%, 22.5%, and 38.6%, and within each of the 7 age groups, the mortality rate under Sepsis-3 was significantly smaller than those under two other definitions of sepsis. CONCLUSIONS: Sepsis-3 criteria yield a larger 12-year incidence of sepsis and a smaller 12-year sepsis-related mortality rate as compared to the two other ICD-9 claims definitions within all 7 age groups. CLINICAL IMPLICATIONS: An accurate understanding of discrepancies in sepsis burden among different definitions and ages has critical public health and policy implications if new criteria are to be considered in future sepsis definitions. DISCLOSURE: The following authors have nothing to disclose: Chen Feng, Shravan Kethireddy, Griffin Paul, Yajun Mei No Product/Research Disclosure Information