Prediction for serious bacterial infection in febrile children aged 3 years or younger: comparison of inflammatory markers, the Laboratory-score, and a new laboratory combined model

Purpose To compare the efficacy of inflammatory markers, the Laboratory-score, and a new laboratory combined model for predicting serious bacterial infection (SBI) in young febrile children. Methods The presence of SBI was reviewed in previously healthy children aged 3 years or younger with fever (> 38℃) who visited the emergency department from 2017 through 2018. Areas under the curves (AUCs) of the receiver operating characteristic curve for SBI were compared with individual inflammatory markers (white blood cells [WBC] count, erythrocyte sedimentation rate [ESR], C-reactive protein [CRP], procalcitonin [PCT], and urine WBC count), the Laboratory-score, and a laboratory combined model. The latter model was developed using logistic regression analysis including ESR, CRP, and PCT. Results Of the 203 enrolled children, SBI was diagnosed in 58 (28.6%). For SBI prediction, the Laboratory-score showed 51.7% sensitivity (95% confidence interval [CI], 38.2%-65.0%) and 83.5% specificity (95% CI, 76.4%-89.1%). The AUC of the Laboratory-score (0.76) was significantly superior to the values of all individual inflammatory markers (WBC, 0.59 [P = 0.032]; ESR, 0.69; and CRP, 0.74 [P < 0.001]) except that of PCT (0.77, [P < 0.001]). The AUC of the laboratory combined model (0.80) was superior to that of the Laboratory-score (0.76) (P < 0.001). Conclusion In this study, the new laboratory combined model showed good predictability for SBI. This finding suggests the usefulness of combining ESR, CRP, and PCT in predicting SBI.