COVID-19 mortality prediction from deep learning in a large multistate EHR and LIS dataset: algorithm development and validation.

Abstract

BACKGROUND\nCOVID-19 is caused by the SARS-CoV-2 virus and has strikingly heterogeneous clinical manifestations with most individuals contracting mild disease but a substantial minority experiencing fulminant cardiopulmonary symptoms or death. The clinical covariates and the lab tests performed on a patient provide robust statistics to guide clinical treatment. Deep learning approaches on a dataset of this nature enable patient stratification and provide methods to guide clinical treatment.\n\n\nOBJECTIVE\nHere we report on the development and prospective validation of a state-of-the-art machine learning model to provide mortality prediction shortly after confirmation of SARS-CoV-2 infection in the Mayo Clinic patient population.\n\n\nMETHODS\nWe retrospectively constructed one of the largest reported and most geographically diverse laboratory information system (LIS) and electronic health record (EHR) COVID-19 datasets in the published literature, which included 11,807 patients with residence in 41 states, treated at medical sites across five states in three time zones. Traditional machine learning models were evaluated independently as well as in a stacked learner approach using AutoGluon, and various recurrent neural network (RNN) architectures were considered. The traditional machine learning models were implemented using the AutoGluon-Tabular framework, whereas the RNNs utilized the tensorflow keras framework. We trained these models to operate solely using routine laboratory measurements and clinical covariates available within 72 hours of a patient s first positive COVID-19 nucleic acid test.\n\n\nRESULTS\nThe GRU-D recurrent neural network achieved peak cross-validation performance with 0.938±0.004 AUROC. The model retained strong performance when reducing the follow-up time to 12 hours (0.916±0.005 AUROC), and leave-one-out feature importance analysis indicated the most independently valuable features were: age, Charlson score, minimum oxygen saturation, fibrinogen and serum iron level. In the prospective testing cohort this model provides an AUROC of 0.901 and statistically significant difference in survival (P<.001, hazard ratio for those predicted to survive: 95% CI [0.043,0.106]).\n\n\nCONCLUSIONS\nOur deep learning approach using GRU-D provides an alert system to flag mortality on COVID-19 positive patients, using clinical covariates and lab values within a 72-hour window after the first positive nucleic acid test.\n\n\nCLINICALTRIAL