Abstract
OBJECTIVE: To explore the clinical characteristics of neonatal late-onset sepsis (LOS) and analyze the independent risk factors for secondary neonatal purulent meningitis (NPM). METHODS: This retrospective case-control study included infants diagnosed with LOS at the Children's Hospital of Soochow University between January 2018 and December 2023. The study divided the patients into two groups: the NPM group and the non-NPM group, based on the presence of secondary purulent meningitis. Clinical characteristics, laboratory markers, pathogen distribution, and treatment regimens were compared between the two groups. Independent risk factors were identified through multivariable logistic regression analysis, and a receiver operating characteristic (ROC) curve was used to evaluate the predictive performance. RESULTS: A total of 453 LOS patients were included, with 98 (21.6%) cases in the NPM group. Compared to the non-NPM group, the NPM group exhibited a higher frequency of prolonged fever (>3 days), fever peak >39 °C, tachypnea, seizures, irritability, poor feeding, and bulging anterior fontanel (all P < 0.05). Laboratory tests showed elevated procalcitonin (PCT) in the NPM group, while albumin, cholinesterase, glycocholic acid, and creatine kinase (CK) levels were decreased (all P < 0.05). Blood culture results revealed that the NPM group had a significantly higher proportion of non-Group B Streptococcus and Enterobacter cloacae, but a lower proportion of Staphylococcus aureus (P < 0.05). Multivariable analysis identified prolonged fever (>3 days), fever peak >39 °C, tachypnea, PCT >10.50 ng/mL, and CK <200 U/L as independent risk factors for LOS complicated by NPM. ROC analysis showed that the combined prediction model had an AUC of 0.804 (95% CI: 0.751-0.856), with a sensitivity of 75.24% and specificity of 72.83%, which outperformed the individual predictors for predicting NPM. CONCLUSION: Prolonged high fever, abnormal respiration, elevated PCT, and decreased CK levels are important independent predictors of LOS complicated by NPM. The combined prediction model demonstrates high diagnostic efficacy, providing a useful reference for early identification of high-risk infants and the development of personalized intervention strategies.