Efficient management strategy of COVID-19 patients based on cluster analysis and clinical decision tree classification

Abstract

Abstract Early classification and risk assessment for COVID-19 patients are critical for improving their terminal prognosis, and preventing the patients deteriorate into severe or critical situation. We performed a retrospective study on 222 COVID-19 patients in Wuhan treated between January 23rd and February 28th, 2020. A decision tree algorithm has been established including multiple factor logistic for cluster analyses that were performed to assess the predictive value of presumptive clinical diagnosis and features including characteristic signs and symptoms of COVID-19 patients. Therapeutic efficacy was evaluated by adopting Kaplan–Meier survival curve analysis and cox risk regression. The 222 patients were then clustered into two groups: cluster I (common type) and cluster II (high-risk type). High-risk cases can be judged from their clinical characteristics, including: age > 50 years, chest CT images with multiple ground glass or wetting shadows, etc. Based on the classification analysis and risk factor analysis, a decision tree algorithm and management flow chart were established, which can help well recognize individuals who needs hospitalization and improve the clinical prognosis of the COVID-19 patients. Our risk factor analysis and management process suggestions are useful for improving the overall clinical prognosis and optimize the utilization of public health resources during treatment of COVID-19 patients.