Rapid and accurate identification of COVID-19 infection through machine learning based on clinical available blood test results

Jiangpeng Wu; Pengyi Zhang; Liting Zhang; Wenbo Meng; Junfeng Li; Chongxiang Tong; Yonghong Li; Jing Cai; Zengwei Yang; Jinhong Zhu; Meie Zhao; Huirong Huang; Xiaodong Xie; Shuyan Li

The above data raised the hope that the tool, which combined machine learning algorithm and laboratory parameters, would play a prominent part in screening patients with COVID-19. The test set was completely independent of the entire modeling process and was adopted to prove the stability and reliability of this method again. As shown in Figure 2 , the performance of the method in the test set was consistent with the training set with an AUC of 0· 9926, achieving a sensitivity of 1.0000 and a specificity of 0· 9444. It confirmed that the discrimination tool had the ability to deal with a large number of suspected patients with COVID-19, which might greatly simplify the laboratory blood test process and provide timely treatment for the confirmed patients.