Prediction of mortality and length of stay with deep learning

Abstract

Predicting the mortality and the length of stay of patients during ICU stay is important for better acute care and planning of ICU resources. The recent advancements in deep learning and with the Electronic Health Record data becoming available for researchers, there has been an increasing interest in the healthcare domain. While patient's structural data are frequently used, most of the studies in the literature do not use the clinical notes due to the complex nature. In this study, we use the clinical notes besides time-series features to improve our predictions. The results have shown that the proposed deep learning based multimodal approach outperforms on all clinical tasks over baseline models. © 2021 IEEE.