Prediction and feature importance analysis for severity of COVID-19 using artificial intelligence: A nationwide analysis in South Korea.

BACKGROUND The number of deaths from COVID-19 continues to surge worldwide. In particular, if the patient's condition is sufficiently severe to require invasive ventilation, it is more likely to lead to death than to recovery. OBJECTIVE To analyze the factors of severe COVID-19 patients and develop an artificial intelligence (AI) model to predict the severity of COVID-19 at an early stage. METHODS We developed an AI model that predicts severity based on data from 5,601 COVID-19 patients from all national and regional hospitals across South Korea as of April, 2020. The clinical severity has two categories: low and high severity. The conditions of patients in the low-severity group correspond to no limit of activity, oxygen support with nasal prong or facial mask, and non-invasive ventilation. The conditions of patients in the high-severity group correspond to invasive ventilation, multi-organ failure with extracorporeal membrane oxygenation required, and death. For the AI model input, we used 37 medical records including basic patient information, physical index, initial examination findings, clinical findings, omorbidity disease and general blood test results at an early stage. Feature importance analysis was performed with AdaBoost, random forest and XGBoost; AI model for predicting severe COVID-19 patients was developed with 5-layer deep neural network with 20 most important features. The ranked feature importance values of the 37 medical records; sensitivity, specificity, accuracy, balanced accuracy, and area under receiver operating characteristic (AUROC) metrics of the AI model. RESULTS We found that age is the most important factor for predicting the disease severity, followed by lymphocyte level, platelet count, and shortness of breath/dyspnea. Our proposed 5-layer deep neural network with 20 most important features provided high sensitivity (90.2%), specificity (90.4%), accuracy (90.4%), balanced accuracy (90.3%), and area under the curve (0.96). CONCLUSIONS Our proposed AI model with the selected features was able to predict the severity of COVID-19 accurately. We also made a web application (http://kcovidnet.site/) for anyone to access the model. We believe that opening the AI model to the public is helpful to validate and improve its performance. CLINICALTRIAL