Machine Learning-Based Early Prediction of Sepsis Using Electronic Health Records: A Systematic Review.

Area of Science:

• Medical Informatics
• Computational Biology
• Clinical Data Science

Background:

• Sepsis is a critical, life-threatening condition requiring prompt diagnosis.
• Early sepsis detection is vital for improving patient outcomes and preventing severe progression.
• Machine learning (ML) and deep learning (DL) offer potential for sepsis prediction using electronic health records (EHRs).

Purpose of the Study:

• To systematically review the application of ML/DL in predicting sepsis onset using EHRs.
• To evaluate the methodologies and findings of existing studies in this domain.
• To assess the quality and consistency of research on ML/DL for sepsis prediction.

Main Methods:

• A systematic literature search was conducted across PubMed, IEEE Xplore, Google Scholar, and Scopus.
• Studies utilizing ML/DL for adult sepsis detection or early prediction from EHRs were included.
• Data extraction, analysis, and quality assessment were performed on selected studies.

Main Results:

• 42 studies were selected from 1942 articles, primarily retrospective and US-based.
• Diverse datasets, sepsis definitions, and model parameters were noted, requiring data augmentation.
• Longitudinal EHR data facilitated early sepsis prediction, though study quality and funding correlation varied.

Conclusions:

• ML/DL methods are significant for sepsis detection and early prediction.
• EHR data combined with ML/DL holds considerable potential for clinical sepsis management.
• Further standardization and quality assessment are needed to optimize ML/DL applications in sepsis.