Clinical Application of Artificial Intelligence in Emergency and Critical Care Medicine, Volume I

This research topic focuses on the use of machine learning models to predict outcomes in mechanically ventilated patients in intensive care units. Several studies have identified important predictor variables, such as admission from the emergency department, insertion of arterial lines, prior quinolone use, hemodialysis, and admission. The XGBoost model has been found to be reliable in predicting outcomes with an AUROC value of 0.89. Additionally, gaze patterns of intensive care nurses have been studied to understand the need for training and education with new tools in critical care.

A perioperative RBCs transfusion predictive model based on a machine learning algorithm has been developed and validated with an accuracy of 81.82%. A combination of disease severity scores and other features available on the first day of admission have been used to establish prediction scores for mechanically ventilated patients in the ICU. Three machine learning models have been used to predict mortality, severity, and length of ICU stay in 2,224 patients with sepsis. A parsimonious model has been developed to accurately identify two phenotypes within COVID-19 patients.

A machine-learning model called Categorical Boosting (CatBoost) has been developed to predict extubation failure in ICUs. AI-derived algorithms have been used to develop a variety of Clinical Decision Support Systems to predict, diagnose, subphenotype, assess prognosis, and manage sepsis. An explainable machine learning model has been developed to predict successful weaning in patients requiring prolonged mechanical ventilation. A practical nomogram has been developed to predict the risk of 28-day mortality in sepsis-induced coagulopathy (SIC) patients. Machine learning algorithms have been used to accurately predict in-hospital death of critically ill patients and identify factors contributing to the prediction power. An automated search strategy has been developed to retrospectively identify ARDS patients and a machine learning model to predict the need for red blood cell transfusion during or after liver transplantation surgery. Two machine-learning models have been developed to dynamically predict the risk of Sepsis-induced coagulopathy (SIC) in critically ill patients with sepsis. A risk stratification score for mortality prediction in sepsis-3 patients has been developed. A tool to classify sepsis into two distinct immune endotypes has been developed and the novel