Mortality Prediction of Septic Patients in the Emergency Department Based on Machine Learning
Problem
Predicting mortality of suspected infected patients in hospital emergency departments, where sepsis is the most common cause of death.
Why it matters
Sepsis is the leading cause of death among suspected infected ED patients.
What the AI did
A deep learning structure predicted mortality for 42,220 suspected infected ED patients (2007-2013) and was compared against several machine learning methods and the SIRS and qSOFA screening tools.
Results
CNN+SoftMax accuracy 87.01% (72 h) and 81.59% (28 days), exceeding other machine learning methods, SIRS, and qSOFA.
Impact
Deep learning outperformed standard sepsis screening tools on retrospective ED data.
Record
| Venue | Journal of Clinical Medicine |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Jau‐Woei Perng, I-Hsi Kao, Chia‐Te Kung, Shih-Chiang Hung, Yi-Horng Lai, Chih-Min Su |
| Citations | 65 (OpenAlex, at import time) |
| References | 48 |
| Model or method | Convolutional Neural Network plus SoftMax |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/jcm8111906 |
| Open access | gold |
| Keywords | sepsis, mortality prediction, deep learning, CNN, emergency department |
Importance rating
significant — thousands to millions affected
Abstract states sepsis is the most common cause of death among suspected infected ED patients; mortality prediction outperforming SIRS/qSOFA addresses a deadly, high-stakes ED problem.