Young infants with fever routinely undergo laboratory evaluation, and many are treated with empirical antibiotics even when the infant seems well. The requirement of a lumbar puncture (LP) as part of a routine evaluation is debated; however, administration of antibiotics without an LP can cause concerns for partially treated bacterial meningitis and make subsequent evaluation of the cerebrospinal fluid (CSF) confusing. The ability to predict which febrile infants have a CSF pleocytosis would assist in the decision to perform LP in febrile infants.
To develop a model to predict which febrile infants have a CSF pleocytosis.
We conducted a retrospective review of febrile children aged 90 days or younger seen in the emergency department. Electronic data sources provided the age of the infant, the triage temperature, and all laboratory values. After univariate analysis, recursive partitioning analysis was performed to develop a decision tree to predict febrile infants at increased risk for CSF pleocytosis, defined as a CSF white blood cell (WBC) count of 25/microL or greater in infants 28 days old or younger and 10/microL or greater in those 29 to 90 days old.
Two thousand three febrile infants were studied; 176 (8.8%; 95% confidence interval [CI], 7.6%-10.1%) had a CSF pleocytosis. Presentation during the summer season increased the risk of pleocytosis from 5.0% during nonsummer months to 17.4% (95% CI, 14.6%-20.6%). During the nonsummer season, 7.3% (95% CI, 5.6%-9.5%) of febrile infants with a temperature of greater than 38.4 degrees C and a WBC count of greater than 6100/microL had a CSF pleocytosis, as opposed to 2.9% (95% CI, 1.9%-4.4%) of those with lower temperature or lower WBC count. The decision tree has an overall sensitivity of 89% (95% CI, 83%-92%) and a negative predictive value of 97% (95% CI, 96%-98%).
A significant number of well-appearing febrile infants will have a CSF pleocytosis. A simple decision tree based on objective clinical information can help identify those at greatest risk for CSF pleocytosis.