ARDSMLpred: A machine learning model for predicting SCAP-associated ARDS based on MIMIC-IV database - PubMed
3 hours ago
- #Machine Learning
- #ICU Care
- #ARDS Prediction
- Developed a machine learning model named ARDSMLpred to predict ARDS in severe community-acquired pneumonia (SCAP) patients using the MIMIC-IV database.
- Analyzed 3,807 SCAP patients, splitting them into training (2,664) and testing (1,143) sets, with ARDS incidence as the outcome.
- Selected 8 key variables (Charlson, Lactate, Stroke, Race, AG, ALB, Sepsis, ROX) via LASSO regression from initial 80, building 10 ML models.
- XGBoost showed the best predictive performance in the test set with an AUROC of 0.9466, among models like LGBM, CatBoost, and SVM.
- Created a web-based calculator (https://zxhzxh2000.shinyapps.io/ards/) for clinical use to predict SCAP-associated ARDS risk in ICU patients.