Objective To investigate high risk factors with severe 2019 novel coronavirus pneumonia(COVID-19) via retrospective case control study, so as to establish and evaluate a predictive model for severe COVID-19 based on data analysis. Methods The clinical data of 314 patients with COVID-19 in the Third People's Hospital of Hubei Province Affiliated to Jianghan University from January 1 to March 10, 2020 were retrospectively analyzed. The patients were divided into severe group (n=76) and non severe group (n=238) according to whether they developed into severe or not. The independent risk factors of severe COVID-19 were obtained by the logistic regression analysis of the different clinical indicators between the two groups. The regression coefficients of all independent risk factors were substituted into the equation to establish a new combined predictor (L) model, through which the occurrence of severe cases could be predicted. Results According to the results of multivariate logistic regression analysis, aging(OR=1.138, 95%CI=1.080~1.199), increase of BMI(OR=2.346, 95%CI=1.509~3.646), leukopenia (OR=0.519, 95%CI=0.357~0.754), decrease of plasma albumin (OR=0.692, 95%CI=0.588~0.815), increase of c-reactive protein(OR=1.029, 95%CI=1.007~1.050) and increase of D-dimer(OR=1.278, 95%CI=1.089~1.499) were independent risk factors for progression to severe COVID-19. A new variable called the joint predictor was fitted by the regression equation after logistic statistical analysis. ROC curve analysis showed that the areas under the ROC curves (AUC) of the joint predictor was 0.985 (95%CI=0.974~0.996), which was the highest of all variables; the cutoff point was 12.90, the sensitivity of the model was 93.4%, and the specificity was 95.0%. Conclusion Among the variables, the joint predictor has the highest predictive value for severe cases of COVID-19, which could be useful in early identification of severe patients. |