Validation of a predictive model for identifying febrile young infants with altered urinalysis at low risk of invasive bacterial infection.

Abstract

In 2015, a predictive model for invasive bacterial infection (IBI) in febrile young infants with altered urine dipstick was published. The aim of this study was to externally validate a previously published set of low risk criteria for invasive bacterial infection in febrile young infants with altered urine dipstick. Retrospective multicenter study including nine Spanish hospitals. Febrile infants ≤90days old with altered urinalysis (presence of leukocyturia and/or nitrituria) were included. According to our predictive model, an infant is classified as low-risk for IBI when meeting all the following: appearing well at arrival to the emergency department, being >21days old, having a procalcitonin value <0.5ng/mL and a C-reactive protein value <20mg/L. IBI was considered as secondary to urinary tract infection if the same pathogen was isolated in the urine culture and in the blood or cerebrospinal fluid culture. A total of 391 patients with altered urine dipstick were included. Thirty (7.7%) of them developed an IBI, with 26 (86.7%) of them secondary to UTI. Prevalence of IBI was 2/104 (1.9%; CI 95% 0.5-6.7) among low-risk patients vs 28/287 (9.7%; CI 95% 6.8-13.7) among high-risk patients (p < 0.05). Sensitivity of the model was 93.3% (CI 95% 78.7-98.2) and negative predictive value was 98.1% (93.3-99.4). Although our predictive model was shown to be less accurate in the validation cohort, it still showed a good discriminatory ability to detect IBI. Larger prospective external validation studies, taking into account fever duration as well as the role of ED observation, should be undertaken before its implementation into clinical practice.