A Prediction-scoring Model For Late-onset Neonatal Sepsis

Objective:A prediction-scoring model for late-onset neonatal sepsis was made on the basis of the risk factors,clinical characteristics and laboratory examination,and to validate the diagnostic performance of the model and evaluate clinical significance,so as to provide a reasonable basis for early detection and treatment.Methods:128 neonates with LOS were retrospectively studied in the First Hospital of Jilin University from January 2016 to December 2017.400 neonates who were more than 3days old with pneumonia,infection,urinary tract infection were retrospectively studied as control group during the same period.SPSS 22.0 software was used to analyze and process data.Independent risk factors which were found by multivariate logistic regression analysis were used to make a prediction-scoring model.The total prediction score for LOS of each neonate was the sum of variables’scores,and receiver operating characteristic(ROC)curve analysis was then used to find the cutoff of the best diagnosis,and its sensitivity and specificity were calculated.To validate the diagnostic performance of the model,71 neonates with LOS were studied as the case group in our department from January 2018 to December 2018.186 neonates who were older than 3days with pneumonia,infection,urinary tract infection were studied as control group during the same period.Agreement index(kappa)was made to evaluate their predictive accuracy.P<0.05 was statistically significant.Results:(1)In general information:For 128 neonates,70 male neonates(54.7%),58 female neonates(45.3%).The average gestational age in the case group and control group were(34.78w±4.60d)and(36.79w±4.14d).(2)The clinical manifestations of LOS were not typical,with the main manifestations of abnormal skin color(85.9%,110/128),poor dynamic response(77.3%,99/128),feeding intolerance(57.0%,73/128),fever(43.0%,55/128),respiratory distress(41.4%,53/128),and heart rate fast(28.1%,36/128).(3)Univariate analysis showed that fever,heart rate increase,abnormal skin color,poor dynamic response,feeding intolerance,abnormal WBC count(<5×10~9/L or>20×10~9/L),Neutrophilic percentage increased(>70%),low platelet count(<100×10~9/L),anemia(Hb<145g/L),CRP increased(>8mg/L),abnormal blood glucos e(<2.6mmol/L or>7.0mmol/L),birth weight(BW)<2500g,timing of feedings m ore than or equal to 4 days(d),peripherally inserted central catheter(PICC),en dotracheal intubation,gestational age(GA),premature rupture of membranes,respira tory distress syndrome(RDS)were related to LOS(P<0.05).(4)Conditional Logistic regression analysis suggested that fever(OR=2.09,95%CI:1.10-3.98),poor dynamic response(OR=2.48,95%CI:1.31-4.71),Neutrophili c percentage increased(OR=2.90,95%CI:1.58-5.33),low platelet count(OR=13.56,95%CI:2.45-14.52),CRP increased(OR=4.10,95%CI:2.21-7.59)、timing of feedi ngs more than or equal to 4 days(d)(OR=13.56,95%CI:4.91-37.41),endotrach eal intubation(OR=4.01,95%CI:1.82-8.83)、BW<2500g(OR=2.73,95%CI:1.11-6.74)were independent risk factors for the incidence of LOS.(5)A prediction-scoring model was made on the basis of above eight indep endent factors,and scores from 0 to 11.The area under ROC curve was 0.952(95%CI:0.934-0.971).The biggest Youden index was 0.798 when the total scores are 3.5,and sensitivity and specificity was 90.6%and 89.2%respectively.257 ne onates as the case group and control group in our department from January 2018 to December 2018 were used to validate the diagnostic performance of the mo del.The area under ROC curve was 0.886(95%CI:0.846-0.926).Agreement ind ex(kappa)was 0.682 with statistical significance(P<0.01).A prediction-scoring model had good the diagnostic performance and clinical significance.Conclusions:(1)fever,poor dynamic response,Neutrophilic percentage increased,low platelet count,CRP increased,timing of feedings more than or equal to 4 days(d),endotracheal intubation,BW<2500g were independent risk factors of LOS.(2)A prediction-scoring model had good the diagnostic performance and clinical significance.