Prediction of Postoperative Hospital Stay with Artificial Intelligence (Deep Learning) in Neurosurgery

Rationalization of the use of internal resources of hospitals with predicting postoperative hospital stay in neurosurgery is an actual problem. This work includes two studies. The first study was conducted among 104 neurosurgeons in two neurosurgical centers (Moscow, Novosibirsk) which describes the results of the practices of rationing neurosurgical care, which affects the length of hospital stay, perioperative complications and increases the efficiency of the neurosurgeons. The second study was carried out with the aim of developing alternative methods of rationing neurosurgical care using machine-learning technologies based on a medical examination before surgery and an operative reports for 112 cases. The recurrent neuronal network (RNN-GRU) was tuned on new test dataset. Themean absolute error of prediction in the final test was 3.00 days. Our study demonstrated the usability of textual EHRs data for the prediction of the length of hospital stay in neurosurgery using deep learning.