Machine Learning and Deep Learning Models for Early Sepsis Prediction: A Scoping Review.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support

Background:

• Sepsis detection is critical for patient survival, but traditional methods have limitations.
• Artificial intelligence (AI), including machine learning (ML) and deep learning (DL), offers advanced solutions for sepsis prediction.
• Electronic Health Records (EHRs) provide a rich data source for developing AI-driven predictive models.

Purpose of the Study:

• To conduct a scoping review of ML and DL models for sepsis prediction published between 2022 and 2025.
• To provide clinicians with a comprehensive update on recent advancements in AI for sepsis detection.
• To analyze features, data processing, performance, and clinical integration of these models.

Main Methods:

• A PubMed search was conducted on March 11, 2025, identifying 13 relevant studies.
• The review focused on ML and DL models developed for adult sepsis prediction using EHR data.
• Studies published from 2022 to 2025 were included.

Main Results:

• Supervised ML was the predominant approach, with some studies exploring DL and hybrid models.
• Models utilized standard clinical data (vitals, labs) and extended features (demographics, ECG).
• AI models demonstrated superior predictive performance (AUROC, sensitivity, specificity) compared to traditional methods and clinicians, with innovations like federated learning and EHR integration.

Conclusions:

• AI, ML, and DL models show significant potential for early sepsis detection.
• Clinical adoption requires further real-world validation and enhanced model interpretability.
• Standardization of AI tools is crucial for practical implementation in healthcare settings.