Unsuccessful validation of 2004 model for predicting academic or behavioural limitations after childhood bacterial meningitis

In 2004, a model identifying children at risk of academic or behavioural limitations after bacterial meningitis (BM) was presented. Risk factors were male gender, low birthweight, lower educational level of the father, Streptococcus pneumoniae, lower cerebrospinal fluid (CSF) leucocyte count, delay between admission and start of antibiotics, dexamethasone <2 days, seizures and prolonged fever. The aim of this study was to validate that prediction model in an independent cohort. Academic or behavioural limitations were determined in 93 Dutch school-age BM survivors. Risk factors for limitations were obtained from medical files. Validation was performed by applying the model in the cohort, then assessing discrimination and goodness of fit. Multiple imputation techniques were used to deal with missing values. Although fit of the model appeared good when it came to similarity of expected and observed cases (p-value of the Hosmer-Lemeshow test 0.24-0.57), discrimination was poor. Area under the curve (AUC) of the receiver operated characteristics (ROC) curve of the model was 0.83 (95% CI: 0.77-0.89) in the development cohort and 0.53 (95% CI: 0.41-0.65) in the validation cohort. External validation of the model was unsuccessful. It is not suitable for implementation in practice