Early triage of critically ill COVID-19 patients using deep learning

Abstract

The sudden deterioration of patients with novel coronavirus disease 2019 (COVID-19) into critical illness is of major concern. It is imperative to identify these patients early. We show that a deep learning-based survival model can predict the risk of COVID-19 patients developing critical illness based on clinical characteristics at admission. We develop this model using a cohort of 1590 patients from 575 medical centers, with internal validation performance of concordance index 0.894 We further validate the model on three separate cohorts from Wuhan, Hubei and Guangdong provinces consisting of 1393 patients with concordance indexes of 0.890, 0.852 and 0.967 respectively. This model is used to create an online calculation tool designed for patient triage at admission to identify patients at risk of severe illness, ensuring that patients at greatest risk of severe illness receive appropriate care as early as possible and allow for effective allocation of health resources.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1. Model performance comparison.

a Comparison of ROC curves for the Deep Learning Survival Cox model and the Cox proportional hazards model on the training-validation set. b The Kaplan–Meier curves for developing critical illness among patients in different risk groups in the training set. Shaded areas indicate 95% confidence interval. c ROC curves for the three external validation cohorts using the entire datasets. d ROC curves for the three independent external validation cohorts, excluding patients that were missing more than three values.

Fig. 2. Trend of 30-days critically ill risk probability in the follow-up visit after admission.

Red lines with triangle markers are critically ill patients. Green lines with circle markers are other patients. a Visualization of trend of each individual. Each marker indicates a follow-up exam. For better visualization, line color has been slightly disturbed for each patient. b Average trend for different groups of patients. Colored area corresponds to the 25% and 75% of the risk probability.

Fig. 3. Nomogram of the Deep Learning Survival Cox model to triage COVID-19 patients.

The patient’s total nomogram point is 209, overall critical illness probabilities are 0.58, 0.62, and 0.69 within 5, 10, and 30 days, respectively. The patient is triaged as high-risk.