Development and Validation of a Multivariable Predictive nomogram for Diagnosis of Neonatal Bacterial Meningitis

Abstract

Purpose To develop and validate a multivariable predictive nomogram to diagnose neonatal bacterial meningitis (NBM).Methods A retrospective study was conducted and neonates with positive blood culture were included. Infants were divided into training and validation set. A predictive nomogram was established according to finding of univariate analysis and multivariate logistic regression. The diagnostic value was judged by receiver operating characteristic curves (ROC). ResultsAmong the 587 infants included finally, 177 of them developed NBM. Among training set (N=410), the incidence of abnormal temperature, seizure, bulging anterior fontanel were more common in NBM group (P<0.05). Higher of protein, WBC and lower glucose in cerebrospinal fluid (CSF) were found in NBM group (P<0.05). Bulging anterior fontanel, seizure, abnormal temperature, abnormal protein and WBC in CSF were considered as the independent predictors of NBM. A predictive nomogram was established, the C index was 0.89 and the sensitivity, specificity, positive predictive value, and negative predictive value were 63.4%, 94.3%, 83.3%, and 84.6% in internal validation.The model was validated in validation set (N=177). And the C index was 0.93, and the sensitivity, specificity, positive predictive value, and negative predictive value were 58.7%, 97.7%, 90.0%, and 87.1% in the validation set respectively in external validation. High score would increase the probability of NBM.ConclusionThe model including clinical manifestations had high potential diagnostic value in the prediction of NBM.