Identification of infants with fever but at low risk of serious bacterial infection would allow outpatient management. The management of febrile illnesses in infants has been studied in infants older than one month of age but not in infants younger than that. Perhaps that is because of the perceived increased risk of serious bacterial illness when fever is present in infants younger than one month. Baker and Bell evaluated the spectrum of febrile illness in infants younger than one month of age to determine if out-patient management would be safe and effective in this age group.

The 254 infants in the three-year study were three to 28 days old and were initially evaluated in the emergency department of the Children's Hospital of Philadelphia. All of the infants presented with a rectal temperature of 38°C (100.4°F) or greater. A complete history was obtained and a physical examination was performed by an attending physician. Laboratory evaluation included a complete blood count with differential, microscopic urinalysis, chest radiographs, lumbar puncture and bacterial cultures of the blood, urine and cerebrospinal fluid. Stool specimens for white blood cells and stool cultures were obtained if the infant had a history of diarrhea. The neonates were admitted to the hospital and received empiric intravenous antibiotics. They were discharged after 72 hours if they were clinically well and all cultures were negative.

Laboratory results considered normal were as follows: white blood cell count, less than 15,000 cells per mm³ (15 × 10⁹ per L); band-to-neutrophil ratio, less than 0.2; urinalysis, less than 10 white blood cells per high-power field and no bacteria; cerebrospinal fluid, less than eight white blood cells per μL and negative Gram stain; and chest radiographs, no evidence of an infiltrate.

The investigators reviewed the medical records noting the results of all diagnostic studies, treatment plans and the final discharge diagnosis. They then applied the Philadelphia protocol to each case. The Philadelphia protocol was previously developed at their institution as a method of identifying febrile illnesses that carry a low risk of serious bacterial infection in infants 29 to 56 days of age. According to this protocol, the risk of serious bacterial illness is low in infants who appear well at initial presentation, have normal values on laboratory studies and have no evidence of bacterial infection on examination.

Of the 254 infants in the study, 145 (57.1 percent) were boys. A serious bacterial illness was found in 32 (12.6 percent) of the infants. A urinary tract infection was the most common infection, occurring in 17 of the 32 infants. Other infections included bacteremia in eight infants, meningitis in four and gastroenteritis in two. One patient each was found to have pertussis, peritonitis, osteomyelitis, cellulitis and bullous impetigo.

When the Philadelphia protocol was applied retrospectively to each infant in the entire group, 109 (42.9 percent) of the infants were identified as being at low risk for serious bacterial illness. However, five of these infants actually had a bacterial illness. These included two neonates with a urinary tract infection, two with bacteremia and one with Salmonella gastroenteritis. The negative predictive value of the Philadelphia protocol for detecting a bacterial illness in this group was 95.4 percent. The positive predictive value was only 18.6 percent; five children would have been incorrectly identified as not having serious bacterial illness.

The authors conclude that the Philadelphia protocol is not completely accurate in identifying febrile infants under one month of age who are at low risk of serious bacterial illness. In this age group, use of this protocol may actually miss up to 10 percent of febrile infants with a serious bacterial illness. The authors recommend that the management of febrile infants younger than one month continue to include a thorough diagnostic evaluation and the use of empiric antibiotic therapy until bacterial infection has been excluded.