OpenSep: a generalizable open source pipeline for SOFA score calculation and Sepsis-3 classification

Area of Science:

• Medical Informatics
• Clinical Research

Background:

• Electronic health record (EHR) research on sepsis faces challenges due to varied definitions, hindering study comparability and generalizability.
• Standardized sepsis phenotyping is crucial for reliable clinical research and improved patient outcomes.

Purpose of the Study:

• To develop and validate an open-source pipeline, OpenSep, for consistent sepsis phenotyping based on the Sepsis-3 definition.
• To establish a Minimal Sepsis Data Model for applying the pipeline to diverse EHR data sources.
• To accurately determine sepsis onset time and Sequential Organ Failure Assessment (SOFA) scores.

Main Methods:

• Developed OpenSep, an open-source computational pipeline for sepsis identification and scoring.
• Created the Minimal Sepsis Data Model to facilitate data harmonization across different EHR systems.
• Validated OpenSep's accuracy using MIMIC-IV data, comparing results with existing methods.

Main Results:

• OpenSep demonstrated high reliability in determining sepsis onset and SOFA scores when compared to established methods.
• The Minimal Sepsis Data Model enhanced the pipeline's applicability to a wider range of EHR data.
• Validated results indicate OpenSep's potential for consistent sepsis research.

Conclusions:

• OpenSep provides a standardized, open-source solution for sepsis phenotyping in EHR research.
• The Minimal Sepsis Data Model broadens the utility of OpenSep beyond specific datasets.
• This approach promotes more reliable and comparable sepsis research, advancing clinical understanding and treatment.