In this issue of Pediatrics,Drs Kramer and Shapiro review the practice guideline published inPediatrics in July 1993.1 The guideline addressed two patient populations with fever without source (FWS): 1) infants 0 to 90 days of age with a temperature ≥38.0°C, and 2) infants and children 3 to 36 months of age with a temperature ≥39.0°C. The first group was essentially stratified into two populations: 1a) infants <28 days, and 1b) infants 28 to 90 days of age. It was the panel's intention to follow the lead of Dr Keith Powell and his colleagues, and try to reduce the number of tests and unnecessary hospitalizations in young febrile infants by screening with “low-risk” clinical and laboratory criteria. Recent studies support this approach and suggest that the low-risk criteria may be used in younger children as well. Therefore, the definition of group 1a may need to be changed to 0 to 7 days of age.2
Kramer and Shapiro do not comment on the recommendations for young infants but take exception with the recommendations for children 3 to 36 months of age. They “believe that the use of the guideline may do more harm than good” but present no evidence to support their belief. Some of their statements are correct: 1) there is little data to suggest that practicing physicians are following the guideline; 2) the risk of occult bacteremia in children 3 to 36 months of age with FWS and white blood count (WBC) >15 000/mm3 is 8% to 15%; 3) Haemophilus influenzae bacteremia has been virtually eliminated by the use of Hib conjugate vaccines; 4) only a small proportion of children with Streptococcus pneumoniaebacteremia will go on to have a life-threatening infection. All of this was known to the panel in 1992–1993 and the guideline was based on these and other facts.
Other statements by Drs Kramer and Shapiro are incorrect. The use of the guideline will not result in unnecessary hospitalizations or repeat blood cultures. It is the failure to prescribe antibiotics to children at risk of occult bacteremia that leads to persistent bacteremia that will not only lead to hospitalization, but to death and serious morbidity. Children with blood cultures positive for S pneumoniae who have received parenteral antibiotics are more likely than untreated children to be afebrile (78% vs. 27%) and less likely to be admitted (14% vs. 50%)3. Furthermore, those treated are likely to have benefited from early antibiotic therapy. All 7252 children enrolled in the Fleisher and Bass studies were treated with antibiotics, as both investigators believed it inappropriate to not treat children at high risk of occult bacteremia. In these combined studies, 255 children had occult bacteremia, including 215 with S pneumoniae bacteremia and only 15 with H influenzaebacteremia. Presumably, if none had received antibiotics, approximately 11 children would have developed S pneumoniaemeningitis,4 and others might have developed other focal complications such as septic arthritis, pneumonia, or, in the case ofNeisseria meningitidis bacteremia, purpura fulminans. Children with otitis media or pneumonia are not comparable to children with FWS as these children are routinely treated with antibiotics, even though the diagnosis of otitis media is imprecise and most cases of pneumonia in children are viral in etiology. The panel considered oral antibiotic therapy for FWS but decided because of the emerging penicillin resistance among S pneumoniae to recommend ceftriaxone.
Only one third of children 3 to 36 months of age with FWS ≥39.0°C have a WBC count >15 000. Approximately 10% of these children (and of children with fever >40.0°C regardless of WBC count) have occult S pneumoniae bacteremia. If untreated, 5% of children with S pneumoniae bacteremia will go on to develop S pneumoniae meningitis.4 Those who question the guideline ultimately question the need of subjecting 600 children with FWS to a complete blood count, and those 200 with a WBC count ≥15 000 to a blood culture and empiric ceftriaxone therapy, to possibly prevent a single case of meningitis. As Fleisher and Bass did not have a placebo group in their studies, we do not know the morbidity and mortality of children with untreated pneumococcal bacteremia in a prospective randomized controlled trial with an adequate sample size. Let me propose the design of such a study to determine these outcomes. Using the assumptions stated at the beginning of this paragraph, and assuming that 40% of children with pneumococcal meningitis would suffer death or disability5 and that ceftriaxone therapy would prevent 90% of complications, the sample size required to detect a difference in these outcomes (α = .05, power = .8) would be 36 000 patients. A smaller sample size (15 000) would be needed to determine if antibiotic therapy prevents meningitis. In such a study, using the above assumptions, approximately 12 of 7500 controls and 1 of 7500 antibiotic-treated children would develop pneumococcal meningitis. Would such a study be feasible or ethical?
A physician's biases or beliefs, practice situation, and the reliability of the patient probably have more effect on practice patterns than the guideline. Those who would prefer to take a very small risk rather than subject an infant to “invasive” and time-consuming procedures are less likely to follow the guideline, especially if venapuncture and bladder catheterization are not part of usual office procedures. Those who routinely obtain blood by venapuncture and who do not wish to take risks, especially when follow-up is uncertain, are more likely to follow the guideline. Therefore, emergency physicians practicing in urban US emergency departments are justifiably more likely to test and treat than a private pediatrician with an established parent-physician relationship and good follow-up.
The issue regarding urine cultures is more straightforward. Urinary tract infections are a frequent (15% to 20%) cause of fever in children <3 months of age and occasional (2% to 5%) cause of fever in boys 3 to 6 months and girls 3 to 12 months of age. Urine collected by bag is not sterile and should not be sent for culture. Microscopic urinalysis fails to detect 20% of urinary tract infections in young infants and 10% in older infants and children. Multi-reagent strips are far less sensitive (50%).5,6 Therefore, we believe for all boys <6 months and girls <12 months, a urine obtained by bladder catheterization should be sent for urinalysis or Gram stain and culture.7 Infants with pyuria or bacturia should be treated immediately.
The guideline was never meant as a substitute for physician judgment and this was clearly stated. The guideline was based on the assumption that physicians should be informed of the evidence regarding FWS as a framework for rational medical care.
- Received February 8, 1997.
- Accepted March 10, 1997.
No reprints available.
- FWS =
- fever without source •
- WBC =
- white blood count
- Kramer MS,
- Shapiro ED
- Jaskiewicz JA,
- McCarthy CA,
- Richardson AC,
- et al.
- Baraff LJ,
- Oslund S,
- Prather M
- Copyright © 1997 American Academy of Pediatrics