Background: The optimal method of urine collection in febrile infants is debatable; catheterization, considered more accurate, is technically difficult and invasive.
Objectives: To determine predictors of urethral catheterization in febrile infants and to compare bag and catheterized urine test performance characteristics.
Design: Prospective analysis of infants enrolled in the Pediatric Research in Office Settings' Febrile Infant Study.
Setting: A total of 219 practices from within the Pediatric Research in Office Settings' network, including 44 states, the District of Columbia, and Puerto Rico.
Patients: A total of 3066 infants aged 0 to 3 months with temperatures of 38 degrees C or higher.
Main outcome measures: We calculated adjusted odds ratios for predictors of catheterization. Diagnostic test characteristics were compared between bag and catheterization. Urinary tract infection was defined as pure growth of 100 000 CFU/mL or more (bag) and 20 000 CFU/mL or more (catheterization).
Results: Seventy percent of urine samples were obtained by catheterization. Predictors of catheterization included female sex, practitioner older than 40 years, Medicaid, Hispanic ethnicity, nighttime evaluation, and severe dehydration. For leukocyte esterase levels, bag specimens demonstrated no difference in sensitivity but somewhat lower specificity (84% [bag] vs 94% [catheterization], P<.001) and a lower area under the receiver operating characteristic curve for white blood cells (0.71 [bag] vs 0.86 [catheterization], P = .01). Infection rates were similar in bag and catheterized specimens (8.5% vs 10.8%). Ambiguous cultures were more common in bag specimens (7.4% vs 2.7%, P<.001), but 21 catheterized specimens are needed to avoid each ambiguous bag result.
Conclusions: Most practitioners obtain urine from febrile infants via catheterization, but choice of method is not related to the risk of urinary tract infection. Although both urine cultures and urinalyses are more accurate in catheterized specimens, the magnitude of difference is small but should be factored into clinical decision making.