A Comparative Analysis of Sepsis Identification Methods in an Electronic Database

Area of Science:

• Critical Care Medicine
• Infectious Diseases
• Epidemiology

Background:

• Sepsis diagnosis relies on various criteria, impacting patient cohort identification and outcomes.
• Previous methods for sepsis identification may be influenced by coding practices and lack temporal context.

Purpose of the Study:

• To compare the relative validity of five sepsis identification criteria in an intensive care unit (ICU) database.
• To evaluate the Sepsis-3 criteria against other established sepsis definitions.

Main Methods:

• Retrospective cohort study of adult ICU admissions from 2008-2012 at a tertiary teaching hospital.
• Comparison of five sepsis identification algorithms, including Sepsis-3 criteria, using ICU data.
• Analysis of Sequential Organ Failure Assessment (SOFA) scores and infection suspicion for Sepsis-3 criteria fulfillment.

Main Results:

• Sepsis-3 criteria identified a larger cohort (49.1% of admissions) than administrative definitions.
• In-hospital mortality varied significantly across criteria, from 14.5% (Sepsis-3) to 30.1% (explicit codes).
• Agreement among the different sepsis identification criteria was acceptable (Cronbach's alpha, 0.40-0.62).

Conclusions:

• Sepsis-3 criteria identify a broader, less critically ill population than prior administrative definitions.
• Sepsis-3 criteria offer advantages in construct validity and reduced susceptibility to coding changes.
• Future sepsis research must acknowledge and contextualize findings based on the specific identification methods and resulting patient cohorts.