Lab-based prediction model can rule out serious bacterial infections in febrile infants

Reference

Kuppermann N, Dayan PS, Levine DA, et al, for the Febrile Infant Working Group of the Pediatric Emergency Care Applied Research Network (PECARN). A clinical prediction rule to identify febrile infants 60 days and younger at low risk for serious bacterial infections. JAMA Pediatr 2019;173(4):342-351.

Synopsis

Most of us believe that clinical signs are unreliable in identifying serious illness in febrile infants, which results in extensive and invasive septic work-ups. These authors recruited a convenience sample of febrile infants (rectal temperature at least 38 degrees Centigrade or 100.4 degrees Fahrenheit) 60 days and younger who showed up in emergency departments during times when research staff were available. They excluded infants who appeared critically ill, those born prematurely, and those with chronic conditions. In addition to standardized clinical assessments, all infants had blood and urine cultures, and lumbar punctures were done at the discretion of the treating physician. Of 3230 eligible infants, 1821 had a procalcitonin sample drawn. The main outcome, the presence of a serious bacterial infection as defined by bacterial meningitis, bacteremia, or a urinary tract infection, was detected in 170 (9%). The researchers performed a variety of statistical gymnastics to derive a prediction model on a split sample of the infants and then validated the model on the rest. Using the validation sample, the combination of a negative urinalysis, an absolute neutrophil count less than 4090/mcL, and a procalcitonin level of less than 1.71 ng/mL was accurate at ruling out serious infections: 97.7% sensitivity (95% CI 91.3 - 99.6) and 60.0% specificity (56.6 - 63.3), with a positive likelihood ratio of 2.4 (2.1 - 2.7) and a negative likelihood ratio of 0.04 (0.006 - 0.15). Interestingly, the clinicians were asked to predict the likelihood of a serious infection and they were not particularly accurate.