Artificial Intelligence In Intensive Care

Early detection of septic shock is crucial for improving patient outcomes. This study aims to develop a machine learning model using XGBoost to predict septic shock six hours in advance. The model was trained on a public dataset comprising 40,336patients. It was tested on a portion of this set, achieving an accuracy of 0.97 and an AUC of 0.874. Predictions were also made for 8, 10 and 12 hours ahead, giving accuracies of 0.899, 0.891 and 0.8954, and AUCs of 0.867, 0.8639 and 0.8530, respectively.In addition, the model was tested on a local dataset from Fattouma Bourguiba University Hospital, comprising 30 patients. For prediction at 6 hours on the local dataset, the model achieved an accuracy of 0.89 and an AUC of 0.74. Predictions for 8, 10 and 12 hours ahead showed accuracies of 0.8861, 0.8772 and 0.8718, and AUCs of 0.73, 0.72 and 0.72, respectively. The XGBoost model shows potential for early detection of septic shock, but requires further testing and optimization for clinical application.