Early Prediction of Sepsis Onset: Evaluating Supervised Machine Learning Techniques

Authors

  • Ava Berenji Harvard-Westlake School, Los Angeles, CA Author

Keywords:

sepsis prediction, machine learning, ICU, electronic health records (EHR), interpretable models, feature engineering, missing data imputation, comparative analysis

Abstract

Early detection of sepsis in intensive care units (ICUs) is critical for timely intervention and improved patient outcomes. This study evaluates the performance of supervised machine learning models in predicting sepsis onset using ICU electronic medical record (EMR) data. Four distinct feature engineering strategies were compared: Baseline Data (first recorded values), 24-Hour Baseline (first values within 24 hours of admission), 24-Hour Summary (baseline plus summary statistics), and 6-Hour Summary (summary statistics from the first 6 hours). This study provides insight into the impact of modeling choices on sepsis prediction performance and highlights practical considerations for the early detection of sepsis.

Downloads

Published

2025-07-31