An Analysis of Predictive Factors for Severe Neonatal Infection and the Construction of a Prediction Model.

Abstract

To investigate theprimarypredictive factorsforthe occurrence ofsevereneonatal infection,constructaprediction model and assessitseffectiveness. A total of 160 neonateshospitalisedin theDepartment of NeonatologyatSuixi County Hospital from January 2019 to June 2022wereretrospectively analysed.Clinical data was analyzedto determine the primary predictive factors for the occurrence of severe neonatal infection. Predictive efficacy was evaluated using a receiver operating characteristic curve, and a nomogrammodel was constructed according to the predictors. A bootstrap technique was used to verify the accuracy of the model. The neonates weredivided,based on the degree of infection,intoamild infection group (n =80) andasevere infection group (n= 80)according to a 1:1 ratio. Multivariatelogistic regression analysis showed thatcompared with therecovery stage,white blood cell count (WBC) and platelet count (PLT) in the two groups were significantly decreased inthe early stage of infection,andthe ratio of mean platelet volume to PLT, as well asC-reactive protein (CRP) and procalcitoninlevels,waselevated(P<0.05). Thearea under the curves (AUCs) of decreased WBC, decreased PLTand elevatedCRP levels,and the combination of these three indicators,were 0.881, 0.798, 0.523 and 0.914, respectively.According to the filtered indicators, two models (a dichotomous variable equation model and a nomogram model) of continuous numerical variables were constructed, and their AUCs were 0.958 and 0.914, respectively.The calibration curve ofthenomogram model was validated with a consistencyindex of 0.908(95%confidence interval [0.862,0.954]). Decreased WBC and PLTlevels andan elevatedCRP levelweretheprimaryindependent predictors ofsevereneonatal infection.