A Predictive Model For Poor Prognosis In Very Low Birth Weight Infants With Late-onset Sepsis

Objective: The study aimed to develop a predictive model to assess the probability of poor prognosis in very low birth weight infants(VLBWI)with late-onset sepsis(LOS).Methods:(1)Clinical data of all VLBWI with LOS admitted to the neonatal intensive care unit(NICU)of the First Affiliated Hospital of Anhui Medical University,Jixi Road Campus and Gaoxin Campus between January 1,2010,and November 1,2022,were collected retrospectively.In accordance with the condition of the infants at the at the hospital discharge,we divided them into four groups: death,survival with severe complications,survival with mild complications,and survival without any complications.Among them,we defined death and survival with severe complications as having a poor prognosis,and survival with mild complications and survival without complications were defined as having a good prognosis.Eligible patients were randomly assigned to the development and validation cohort according to 8:2.(2)Independent predictors of poor prognosis in VLBWI with LOS were identified based on logistic regression analysis.Finally,use the "rms" package in the R software to construct a nomogram containing these independent predictors.(3)The evaluation indicators of the prediction model mainly included discrimination and calibration.The area under the curve(AUC)of the receiver operating characteristic(ROC)is used to determine the discrimination of the prediction model,and the calibration plots determine the calibration of the prediction model.Results:The 309 cases of VLBWI were eventually included in this study and were randomized into the development cohort(n=247)and the validation cohort(n=62).(1)In the development cohort,univariate analysis showed that birth weight(BW),gestational age(GA),ionized calcium(i Ca),plateletcrit(PCT),platelet distribution width(PDW),albumin(ALB),alkaline phosphatase(ALP),estimated glomerular filtration rate(EGFR),bicarbonate ion concentration(HCO3),serum calcium(Ca),and serum phosphorus levels were lower in the poor prognosis group compared to the good prognosis group.Red blood cell distribution width-SD(RDW-SD),aspartate aminotransferase(AST),urea nitrogen(BUN),creatinine(CRE),uric acid(UA),and serum potassium(K)levels were higher in the poor prognosis group compared to the good prognosis group.The above indicators were statistically significant between the two groups(P < 0.05).(2)Include the variables with P<0.05 in the above univariate analysis in the multivariate analysis,and independent predictors of poor prognosis were screened out using the forward stepwise analysis.Ultimately,the results showed that HCO3,ALB,i Ca,BUN,GA,and BW were independent predictors of poor prognosis in VLBWI with LOS.(3)In the development cohort,nomogram(AUC=0.845,95%CI,0.795-0.894,P<0.05)had a better predictive value for poor prognosis of in VLBWI with LOS than BW(AUC=0.749,95%CI,0.687-0.810),i Ca(AUC=0.740,95%CI,0.676-0.803),GA(AUC=0.731,95% CI,0.670-0.793),BUN(AUC=0.661,95% CI,0.594-0.729),HCO3(AUC=0.643,95% CI,0.574-0.713)and ALB(AUC=0.588,95% CI,0.517-0.659).Among them,the sensitivity and specificity of nomogram and BW,GA,i Ca,BUN,HCO3,and ALB were 83.7% and 74.0%,55.8% and 87.0%,54.4% and 85.0%,56.5% and 84.0%,72.1% and 52.0%,44.2% and 82.0%,33.3% and 87.0%.In addition,the calibration plots fitted the ideal curves well in both the development and validation cohorts.The prediction model has good discrimination and calibration.Conclusion: This study shows that HCO3,ALB,i Ca,BUN,GA,and BW are independent predictors of poor prognosis of VLBWI with LOS in the development cohort.The predictive model constructed through them has a higher predictive value than individual indicators.The model can help clinicians to identify the high-risk group with poor prognosis early and provide important clinical reference information.