Predicting mortality in intensive care units: imputation of incomplete data and prediction by deep learning neural networks.
The risk of mortality in intensive care units is currently assessed using scores based on admission data (SAPS, SOFA, etc.). Their performance is satisfactory when complications arise early after admission; however, they may become less relevant in the case of long hospital stays. In this study, we developed predictive models of short-term mortality in intensive care units based on longitudinal data, comparing the use of deep learning neural networks and a classical machine learning model.