Prediction of criticality in patients with severe Covid-19 infection using three clinical features: a machine learning-based prognostic model with clinical data in Wuhan

Background: COVID-19 appeared in Wuhan, China in December 2019, and since then it has immediately become a serious public health problem worldwide. No specific medicine against COVID-19 has been found until now. However, mortality risk in patients could potentially be predicted before they transmit to critically ill. Methods: We screened the electronic records of 2,799 patients admitted in Tongji Hospital from January 10th to February 18th, 2020. There were 375 discharged patients including 201 survivors. We built a prognostic prediction model based on XGBoost machine learning algorithm and then tested 29 patients (included 3 patients from other hospital) who were cleared after February 19th. Results: The mean age of the 375 patients was 58.83 years old with 58.7% of males. Fever was the most common initial symptom (49.9%), followed by cough (13.9%), fatigue (3.7%), and dyspnea (2.1%). Our model identified three key clinical features, i.e., lactic dehydrogenase (LDH), lymphocyte and High-sensitivity C-reactive protein (hs-CRP), from a pool of more than 300 features. The clinical route is simple to check and can precisely and quickly assess the risk of death. Therefore, it is of great clinical significance. Conclusion: The three indices-based prognostic prediction model we built is able to predict the mortality risk, and present a clinical route to the recognition of critical cases from severe cases. It can help doctors with early identification and intervention, thus potentially reducing mortality.