Construction of a diagnostic prediction model of severe bacterial infection in febrile infants under 3 months old

Abstract

Abstract Introduction Fever is a common cause of paediatric admissions in emergency departments. An aetiological diagnosis is difficult to obtain in those less than 3 months of age, as they tend to have a higher rate of serious bacterial infection (SBI). The aim of this study is to find a predictor index of SBI in children under 3 months old with fever of unknown origin. Methods A study was conducted on all children under 3 months of age with fever admitted to hospital, with additional tests being performed according to the clinical protocol. Rochester criteria for identifying febrile infants at low risk for SBI were also analysed. A predictive model for SBI and positive cultures was designed, including the following variables in the maximum model: C-reactive protein (CRP), procalcitonin (PCT), and meeting not less than four of the Rochester criteria. Results A total of 702 subjects were included, of which 22.64% had an SBI and 20.65% had positive cultures. Children who had SBI and a positive culture showed higher values of white cells, total neutrophils, CRP and PCT. A statistical significance was observed with less than 4 Rochester criteria, CRP and PCT levels, an SBI (area under the curve [AUC] 0.877), or for positive cultures (AUC 0.888). Using regression analysis a predictive index was calculated for SBI or a positive culture, with a sensitivity of 87.7 and 91%, a specificity of 70.1 and 87.7%, an LR+ of 2.93 and 3.62, and a LR− of 0.17 and 0.10, respectively. Conclusions The predictive models are valid and slightly improve the validity of the Rochester criteria for positive culture in children less than 3 months admitted with fever.