Related Experiment Video
Updated: Jun 18, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.8K
Early predictive values of clinical assessments for ARDS mortality: a machine-learning approach
Ning Ding1,2, Tanmay Nath3, Mahendra Damarla4
1Division of Allergy and Clinical Immunology, Johns Hopkins University School of Medicine, 5501 Hopkins Bayview Circle, Baltimore, MD, 21224-6821, USA.
Scientific Reports
|August 1, 2024
Summary
Machine learning models predict mortality in acute respiratory distress syndrome (ARDS) using early clinical data. Day 3 data improved prediction accuracy, with Mean Airway Pressure being a key indicator for risk stratification.
Area of Science:
- Critical Care Medicine
- Data Science in Healthcare
- Respiratory Medicine
Background:
- Acute respiratory distress syndrome (ARDS) presents a significant challenge in critical care due to high morbidity and mortality.
- Early and accurate prognosis is crucial for effective patient management.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting ARDS mortality.
- To assess the predictive value of dynamic clinical indices at different time points.
Main Methods:
- Retrospective analysis of clinical data from the ARDSNet FACTT Trial (n=1000).
- Application of ML algorithms, specifically the random forest (RF) model, to predict mortality.
- Comparison of model performance using clinical data from baseline and Day 3.
Main Results:
- The RF model using Day 3 data achieved higher prognostication efficacy (AUC: 0.84) compared to baseline data (AUC: 0.72).
- Key predictors of mortality at Day 3 included Mean Airway Pressure (MAP), bicarbonate, age, platelet count, albumin, heart rate, and glucose.
- MAP was identified as the most significant feature for early risk stratification in ARDS patients.
Conclusions:
- Early clinical data, specifically from Day 3, significantly enhances the performance of ML models for ARDS mortality prediction.
- Dynamic clinical indices, particularly MAP, are vital for early risk stratification and prognosis in ARDS patients.
