Development a prediction model for identifying bacterial meningitis in young infants aged 29–90 days: a retrospective analysis

Abstract

Abstract Background The early diagnosis and treatment of bacterial meningitis (BM) in young infants was very critical. But, it was difficult to make a definite diagnosis in the early stage due to nonspecific clinical symptoms. Our objectives were to find the risk factors associated with BM and develop a prediction model of BM especially for young infants. Methods We retrospectively reviewed the clinical data of young infants with meningitis between January 2011 and December 2020 in Children’s Hospital of Soochow University. The independent risk factors of young infants with BM were screened using univariate and multivariate logistic regression analyses. The independent risk factors were used to construct a new scoring model and compared with Bacterial Meningitis Score (BMS) and Meningitis Score for Emergencies (MSE) models. Results Among the 102 young infants included, there were 44 cases of BM and 58 of aseptic meningitis. Group B Streptococcus (22, 50.0%) and Escherichia coli (14, 31.8%) were the main pathogens of BM in the young infants. Multivariate logistic regression analysis identified procalcitonin (PCT), cerebrospinal fluid (CSF) glucose, CSF protein as independent risk factors for young infants with BM. We assigned one point for CSF glucose ≤ 1.86 mmol/L, two points were assigned for PCT ≥ 3.80 ng/ml and CSF protein ≥ 1269 mg/L. Using the not low risk criterion (score ≥ 1) with our new prediction model, we identified the young infantile BM with 100% (95% CI 91.9%-100%) sensitivity and 60.3% (95% CI 46.4%-72.9%) specificity. Compared with BMS and MSE model, our prediction model had larger area under receiver operating characteristic curve and higher specificity, the differences were statistically significant. Conclusion Our new scoring model for young infants can facilitate early identification of BM and has a better performance than BMS and MSE models.