Risk Factor Analysis And Prediction Model Construction For Necrotizing Enterocolitis(Bell’s Sage Ⅱ/Ⅲ) In Neonates

Objective: Necrotizing enterocolitis(NEC)is a serious gastrointestinal disorder in the neonatal intensive care unit.patients with Bell’s stage II/III NEC may have more severe clinical presentations,higher mortality,and more long-term complications than patients with Bell’s stage I NEC.Therefore,in this study,we constructed a predictive model for the onset of Bell’s stage II/III NEC to identify more severe NEC early so that timely measures can be taken to reduce mortality and adverse outcome rates in infants with NEC.Methods: Data were retrospectively collected on all NEC neonates admitted to the neonatal ward from January 1,2015,to January 1,2021.813 patients with NEC were initially screened,83 cases were removed according to the exclusion criteria,and 730 patients with NEC who met the requirements were finally included.The included population was divided into a moderate-severe NEC group(Bell’s stage II/III NEC)and a mild NEC group(Bell’s stage I NEC)according to Bell’s staging criteria.All children were randomly divided into a training group(n=530)and a validation group(n=200)in a7:3 ratio.SPSS 26.0 and Stata 16.0 statistical software were applied to conduct a singlefactor analysis of risk factors,and the results of the single-factor analysis(P<0.05)were subjected to multi-factor logistic regression analysis.And the independent predictors of moderate to severe NEC were finally screened out by adopting the stepwise backward method and according to the principle of AIC minimization.Then the nomogram was constructed in the software.The area under the subject’s working curve(AUC)was used to verify the discrimination of the model,the calibration curve and Hosmer-Lemeshow test were used to verify the calibration and clinical consistency of the model,and the clinical decision curve(DCA)was used to verify the value of the clinical application.Results:(1)Results of univariate analysis in the training group: gestational age(GA),body weight(BW),gestational hypertension,feeding method,red blood cell transfusion,asphyxia,neonatal anemia,coagulation abnormalities,cerebral hemorrhage,sepsis,hypoglycemia,hypoalbuminemia,patent ductus arteriosus(PDA)and hyperbilirubinemia were statistically different between the two groups(P<0.05).(2)Results of multivariate logistic regression analysis in the training group: sepsis(OR:5.563,95%CI = 3.194-9.690),hypoglycemia(OR: 4.279,95%CI = 2.464-7.431),BW<1500 g(OR: 4.057,95%CI = 1.390-11.837),GA<32w(OR: 3.457,95%CI =1.275-9.469),PDA(OR: 2.863,95%CI = 1.492-5.388)and asphyxia(OR.2.037,95%CI = 1.051-3.945)were independent risk factors for the development of moderate-severe NEC.(3)The above six risk factors were applied to construct a columnar plot of the risk of developing moderate-to-severe NEC: the area under the curve was 0.853(95%CI,0.82-0.89)and 0.846(95%CI,0.79-0.90)for the training and validation groups,respectively,indicating that the model had good discrimination.The HosmerLemeshow goodness-of-fit test P>0.05 and calibration curves for both groups were close to the y=x axis,indicating that the model has good calibration and clinical consistency.Moreover,our study showed that the nomogram had a high level of net clinical benefit when the prediction probability threshold for the training group was between 0.12 and 0.93,indicating that the prediction model has good clinical utility.Conclusion: The individualized prediction model composed of six predictors,sepsis,hypoglycemia,BW,GA,PDA and asphyxia has good discrimination,calibration,and clinical practicability,which is helpful for pediatricians to identify moderate-severe NEC early and reduce the incidence of NEC mortality and adverse outcomes.