Application of a prediction model with laboratory indexes in the risk stratification of patients with COVID‑19

Abstract

In the present study, a prediction model with combined laboratory indexes in risk stratification of patients with COVID-19 was established and tested. The data of 170 patients with COVID-19 who were divided into an asymptomatic-moderate group (141 cases) and severe or above group (29 cases) were retrospectively analyzed. The clinical characteristics and laboratory indexes of the two groups were compared. Multivariate logistic regression analysis was performed to construct the prediction model based on laboratory indexes. A receiver operating characteristic (ROC) curve analysis was used to compare the diagnostic efficacy of different indexes. Decision curve analysis (DCA) was performed to quantify and compare the clinical validity of the prediction models. There were significant differences in blood cell count, high-sensitivity C-reactive protein (hsCRP) and procalcitonin (PCT) levels between the severe or above group and the asymptomatic-moderate group (all P<0.05). Among all individual indexes, hsCRP had the highest diagnostic efficacy (area under the curve=0.870), with a sensitivity and specificity of 0.828 and 0.802, respectively. The red blood cell count, hsCRP and PCT were used to construct the prediction model. The AUC of the prediction model was higher than that of hsCRP (0.912 vs. 0.870) but the difference was not significant (P=0.307). DCA suggested that the net benefit of the prediction model was higher than that of hsCRP in most cases and significantly higher than that of PCT, lymphocytes and monocytes. The prediction model with combined laboratory indexes was able to more effectively predict the clinical classification of patients with COVID-19 and may be used as a tool for risk stratification of patients.