A Method for Decompensation Prediction in Emergency and Harsh Situations.

To save more lives, critically ill patients need to make timely decisions or predictive diagnosis and treatment in emergency and harsh conditions, such as earthquakes, medical emergencies, and hurricanes. However, in such circumstances, medical resources such as medical staff and medical facilities are short supply abnormally. So, we propose a method for decompensation prediction in emergency and harsh conditions. The method includes components such as patient information collection, data selection, data processing, and decompensation prediction. Based on this, this paper demonstrates the method using MIMIC-III data. Firstly, we tried a series of machine learning models to predict physiological decompensation. Secondly, to detect patients whose condition deteriorates rapidly under severe and limited circumstances, we try to reduce the essential physiological variables as much as possible for prediction. The experimental results show that the Bi-LSTM-attention method, combined with eleven essential physiological variables, can be used to predict the decompensation of severe ICUs patients. The AUC-ROC can reach 0.8509. Furthermore, these eleven physiological variables can be easily monitored without the need for complicated manual and massive, costly instruments, which meets the real requirements under emergency and harsh conditions. In summary, our decompensation prediction method can provide intelligent decision support for saving more lives in emergency and harsh conditions.