Establishment And Validation Of Prediction Model Of Late-onset Sepsis In Preterm Infants With Birth Weight Less Than 2000g

Objective To establish a disease prediction model of late-onset sepsis in preterm infants with birth weight less than 2000g and verify its diagnostic value.Methods 360 premature infants with a birth weight of less than 2000g in the Department of Neonatology of the first people’s Hospital of Hangzhou from 2017 January to 2019 December were collected retrospectively for binary logistic regression analysis.Firstly,SPSS 24.0 was used for univariate analysis to preliminarily screen the risk factors with P<0.1.According to professional knowledge and col linearity diagnosis,the final variables were selected for multivariate analysis,and the nomogram was drawn by R software.38 cases of preterm infants with birth weight less than 2000g from 2020 January to June were prospectively collected,and the subject working curve（ROC curve）was drawn to verify the ability of the model to predict the risk of late-onset sepsis in preterm infants.Results Univariate analysis showed no significant difference in gender,delivery mode,antenatal hormone,premature rupture of membranes,maternal gestational diabetes mellitus,ecchymosis,anemia,thrombocytopenia,and CRP elevation in two groups（P>0.05）.Small gestational age,low birth weight,asphyxia,amniotic fluid pollution,maternal gestational hypertension,shortness of breath,apnea,tachycardia,feeding intolerance,poor response,abnormal leukocyte count,endotracheal intubation,peripheral central venous catheterization,fasting for≥3 days may increase the incidence of Los in preterm infants.Multivariate analysis finally included five study variables,namely birth weight（OR=0.999,95CI%:0.998-1.000,P=0.023）,feeding intolerance（OR=2.694,95CI%:1.139-6.694,P=0.027）,less response（OR=4.610,95%CI:1.971-10.860,P=0.000）,abnormal leukocyte count（or=3.106,95%CI:1.353-7.131,P=0.007）,fasting for≥3 days（OR=5.012,95%CI:2.265-11.702,P=0.000）.The external verification of the model in the verification set is c-index=0.921,that is,the area under the ROC curve=0.921,with high discrimination.Conclusions The results of this study show that low birth weight,feeding intolerance,poor response,abnormal leukocytes and fasting for≥3 days are independent risk factors for late-onset sepsis in preterm infants with birth weight less than 2000g.The nomogram model is consistent with the diagnostic criteria of neonatal late-onset sepsis,and has certain application value in the diagnosis of LOS.