CIRPMC: An online model with simplified inflammatory signature to predict the occurrence of critical illness in patients with COVID‐19

• Published in: Clinical and Translational Medicine 2020-10, Vol.10 (6), p.e210-n/a
• Main Authors: Gao, Yue, Chen, Lingxi, Zeng, Shaoqing, Feng, Xikang, Chi, JianHua, Wang, Ya, Li, Huayi, Jiang, Tengping, Yu, Yang, Jiao, XiaoFei, Liu, Dan, Feng, XinXia, Wang, SiYuan, Yu, RuiDi, Yuan, Yuan, Xu, Sen, Cai, Guangyao, Xiong, Xiaoming, Chen, Pingbo, Mo, Qingqing, Jin, Xin, Wu, Yuan, Ma, Ding, Li, Chunrui, Li, Shuai Cheng, Gao, Qinglei
• Format: Article
• Language: English
• Subjects: Letter to Editor
• ISSN: 2001-1326; 2001-1326
• DOI: 10.1002/ctm2.210

Hyperinflammatory response induced by immune dysfunction is reported to underpin critical COVID‐19.1 Uncontrolled release of cytokines results in tissue damage and further leads to multiple organ failure, which is the major cause of death in patients with COVID‐19.2 As expected, the differences of multiple cytokines and immune features between critical ill and noncritical ill patients were observed in clinical practice. [...]LASSO analysis identified four features (CRP, TNF‐α, IL‐2R, and IL‐6) for the development of critical illness classifier. Performance metrics for critical illness risk prediction of models in cohorts AUC (95% CI) Accuracy (95% CI) SN (95% CI) SP (95% CI) PPV (95% CI) NPV (95% CI) Kappa F1 Brier SFV cohort SVM 0.946 (0.923‐0.969) 0.927 (0.904‐0.946) 0.653 (0.548‐0.747) 0.976 (0.959‐0.987) 0.827 (0.722‐0.904) 0.940 (0.917‐0.959) 0.688 0.729 0.057 LR 0.944 (0.921‐0.968) 0.916 (0.891‐0.936) 0.758 (0.659‐0.840) 0.944 (0.921‐0.962) 0.706 (0.608‐0.792) 0.956 (0.935‐0.972) 0.681 0.731 0.067 GBDT 0.928 (0.902‐0.954) 0.900 (0.874‐0.922) 0.663 (0.559‐0.757) 0.942 (0.919‐0.960) 0.670 (0.566‐0.764) 0.940 (0.917‐0.959) 0.608 0.667 0.077 KNN 0.894 (0.853‐0.934) 0.913 (0.888‐0.933) 0.579 (0.473‐0.680) 0.972 (0.954‐0.984) 0.786 (0.671‐0.875) 0.928 (0.904‐0.948) 0.618 0.667 0.068 NN 0.939 (0.915‐0.963) 0.849 (0.819‐0.876) 0.000 (0.000‐0.038) 1.000 (0.993‐1.000) NA (0.000‐1.000) 0.849 (0.819‐0.876) 0.000 NA 0.085 OV cohort SVM 0.969 (0.945‐0.992) 0.966 (0.951‐0.977) 0.667 (0.540‐0.778) 0.992 (0.983‐0.997) 0.880 (0.757‐0.955) 0.971 (0.957‐0.982) 0.741 0.759 0.028 LR 0.968 (0.946‐0.989) 0.963 (0.948‐0.975) 0.803 (0.687‐0.891) 0.977 (0.964‐0.987) 0.757 (0.640‐0.852) 0.983 (0.971‐0.991) 0.759 0.779 0.034 GBDT 0.964 (0.944‐0.9845) 0.958 (0.942‐0.971) 0.773 (0.653‐0.870) 0.975 (0.961‐0.985) 0.729 (0.609‐0.828) 0.980 (0.967‐0.989) 0.727 0.750 0.041 KNN 0.908 (0.861‐0.954) 0.957 (0.941‐0.970) 0.606 (0.478‐0.724) 0.988 (0.977‐0.995) 0.816 (0.680‐0.912) 0.966 (0.951‐0.978) 0.673 0.696 0.037 NN 0.963 (0.940‐0.986) 0.919 (0.899‐0.937) 0.000 (0.000‐0.054) 1.000 (0.995 – 1.000) NA (0.000‐1.000) 0.919 (0.899‐0.937) 0.000 NA 0.052 Abbreviations: AUC, area under the curve; SVM, supported vector machin; GBDT, gradient boosted decision tree; KNN, k‐nearest neighbour; LR, logistic regression; NN, neural network; NPV, negative predictive value; OV, Optical Valley Campus of Tongji Hospital; PPV, positive predictive value; SFV cohort, internal validation cohorts of Sino‐French