Utilizing Predictive Factors as a Screening Tool for Early-Onset Sepsis in Neonates

Abstract

Introduction: Neonatal early-onset sepsis (EOS) is a severe condition that affects newborns within the first three days of life, with high mortality rates, particularly in low- and middle-income countries (LMICs). In Vietnam, the diagnosis and management of EOS are challenged by ambiguous clinical signs and limited access to blood culture testing facilities. Early identification of at-risk neonates using a predictive risk factor model is crucial for improving neonatal care and reducing mortality. Objectives: This study aims to identify maternal and neonatal risk factors associated with EOS and develop a predictive screening tool to facilitate the early detection of at-risk neonates in Vietnam. Materials and methods: A nested case-control study was conducted on 225 neonates at the central neonatal unit in a principal tertiary hospital in southwestern Vietnam over a two-year period. Risk factors were identified using univariable and multivariable logistic regression analyses. A predictive nomogram was developed and evaluated for discrimination, calibration, and decision curve analysis (DCA). Results: The study identified eight significant risk factors for EOS, including maternal genital infections during the third trimester, urinary tract infections (UTIs) during pregnancy, hypertension during pregnancy, insufficient maternal weight gain, rupture of membranes (ROM) ≥18 hours, meconium-stained amniotic fluid, first-minute APGAR score <7, and preterm birth <34 weeks. The predictive model demonstrated excellent discrimination with an area under the curve (AUC) of 0.913 (95% CI: 0.876-0.95, p<0.001) and good calibration (Hosmer-Lemeshow test with χ²(df)=5.496 (5), p=0.358). The model-based nomogram showed high sensitivity (82.7%) and specificity (83.3%) at an optimal cutoff of 0.25. The DCA illustrates the model's good clinical utility, providing a higher net benefit across most threshold probability ranges (0.0-0.96). Conclusions: This study presents a robust predictive model for the early identification of neonates at risk of EOS in Vietnam, based on key maternal and neonatal risk factors. The model, with demonstrated accuracy and reliability, holds significant potential for improving neonatal outcomes through timely interventions. Future research should aim at external validation and inclusion of broader clinical data to enhance the model's applicability and generalizability.