Construction and validation of COVID-19 infection risk prediction model based on febrile seizures in children

Objective To construct the risk prediction model of COVID-19 infection in FS. Methods Ninety children with FS were included, and the clinical data were collected for univariate regression analysis. The predictive factors for the risk of infection of COVID-19 were screened by LASSO regression analysis. The predictive model was constructed by nomogram, and verified by the ROC curve. Results The median ages of children in the COVID-19 and non-COVID-19 group were 5.0 and 3.3 years, respectively. And there was males dominance in the COVID-19 group compared with non-COVID-19 group ( p  = 0.0174). The history of FS was more common in the non-COVID-19 group than in the COVID-19 group ( p  = 0.0146). The blood white blood cells levels ( p  < 0.0001), lymphocyte levels ( p  < 0.0001) and serum sodium levels of the COVID-19 group were lower than those of the non-COVID-19 group ( p  = 0.0098). The proportion of myelin sheath immature in the non-COVID-19 group was lower than that in the COVID-19 group ( p  = 0.0121). The LASSO regression analysis revealed gender (Coef.:-0.63586571), age (Coef.:0.60214948), lymphocyte levels (Coef.:-0.6587662), serum sodium levels (Coef.:-0.01968962), normal head MRI (Coef.:4.43661865), Other pathogen infection (Coef.:-3.2293900) and history of FS (Coef.:-3.18077790) were all predictive variables of COVID-19 infection risk. The predictive model was established by the above seven predictive variables, and the model discrimination was verified by AUC of the ROC curve. The traindata AUC was 0.965, testdata AUC was 0.864, and the slope of the correction curve was close to 1. Conclusion Male, elderly children, hyponatremia, and decreased lymphocyte levels indicated an increased probability of COVID-19 infection.