JMIR Publications

ABSTRACT

Background:

In late December 2019, a pneumonia caused by SARS-CoV-2 was first discovered in Wuhan, and it spread worldwide. Until now, no specific medicine has been used for the treatment of coronavirus infections.

Objective:

The aim of this study is to find a tool to predict the likelihood of critical patients, which help clinical physicians prevent COVID-19 progression.

Methods:

In this retrospective study, Clinical characteristics were collected and analyzed from 175 confirmed cases of COVID-19. Univariate analysis and least absolute shrinkage and selection operator (LASSO) regression were used to select variables. Multivariate analysis was applied to find independent risk factors in COVID-19 progression. We established a nomogram to evaluate the probability of COVID-19 progression to severe within three weeks after the disease onset, which was verified through the use of calibration curves and a receiver operating characteristic (ROC) curve.

Results:

Risk factors given by multivariate Cox regression were Age (1.035; 95%CI: 1.017-1.054; P<0.001), creatine kinase(CK)(1.002; 95%CI: 1.0003-1.0039; P=0.022), CD4(0.995; 95%CI: 0.992-0.998; P=0.002), CD8%(1.007; 95%CI: 1.004-1.012; P<0.001), CD8(0.881; 95%CI: 0.835-0.931; P<0.001), and C3(6.93; 95%CI: 1.945-24.691; P=0.003). The area under the curve (AUC) of the prediction model for 0.5-week, 1-week, 2-week and 3-week were 0.721, 0.742, 0.87 and 0.832 respectively, and the calibration curves showed that the model had a good ability to predict COVID-19 progression to severe within three weeks after the disease onset.

Conclusions:

This study presents a critical COVID-19 patient prediction nomogram based on LASSO and multivariate Cox regression. The clinical use of the nomogram may allow for the timely detection of potential critical COVID-19 patient and instruct clinicians to give these patients intervention timely to prevent the disease from worsening. Clinical Trial: none