AUTHOR=Tan Yvette , Young Michael , Girish Akanksha , Hu Beini , Kurian Zina , Greenstein Joseph L. , Kim Han , Winslow Raimond L , Fackler James , Bergmann Jules 

TITLE=Predicting respiratory decompensation in mechanically ventilated adult ICU patients

JOURNAL=Frontiers in Physiology

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2023.1125991

DOI=10.3389/fphys.2023.1125991

ISSN=1664-042X

ABSTRACT=Mechanical ventilation is a life-saving treatment in the Intensive Care Unit (ICU), but often causes patients to be at risk of further respiratory complication. We created a statistical model utilizing electronic health record and physiologic vitals data to predict the Center for Disease Control and Prevention (CDC) defined Ventilator Associated Complications (VACs). We constructed a random forest model to predict occurrence of VACs using health records and chart events from adult patients in the Medical Information Mart for Intensive Care III (MIMIC-III) database. We trained the machine learning models on two patient populations of 1921 and 464 based on low and high frequency data availability. Model features were generated using both basic statistical summaries and tsfresh, an automated python feature generation library. Classification to determine whether a patient will experience VAC one hour after 36 hours of ventilation was performed using a random forest classifier. Two different sample spaces conditioned on five varying feature extraction techniques were evaluated to identify the most optimal selection of features resulting in the best VAC discrimination. Each dataset was assessed using K-folds cross-validation (k = 10), giving average area under the receiver operating curves (AUROC) s and accuracies. After feature selection, hyperparameter tuning, and feature extraction, the best performing model used automatically generated features on high frequency data and achieved an average Area Under Receiver Operating Characteristic Curve (AUC) of 0·83 ± 0·11 and an average accuracy of 0·69 ± 0·10.