Recurrent neural network models (CovRNN) for predicting outcomes of patients with COVID-19 on admission to hospital: model development and validation using electronic health record data

A University of Texas study proposed a machine learning model training framework that can flexibly adapt to the changing pandemic and requires minimal preprocessing. For convenience and practicality, the framework is designed to consume electronic health record data mapped to standard terminologies in common use without the need for specific feature selection or missing value imputation. Because they were trained and evaluated on large, heterogenous datasets collected from different health systems, the COVID-19 outcome prediction models (CovRNN) showed high accuracy in predicting three outcomes (in-hospital mortality, need for mechanical ventilation, and prolonged hospital stay), outperforming the prediction accuracy of state-of-the-art models in the literature, good calibration, and had a low risk of bias.