The evaluation and management of young children with fever is one of the most anxiety-provoking and controversial issues faced daily by practitioners. The recognition that febrile infants without a recognizable source of infection often were bacteremic stimulated burgeoning literature and various strategies that continue to evolve.1 Hospitalization of young febrile infants was recommended in the late 1970s and early 1980s.2 More recently, we are making efforts to identify infants at low risk for bacterial infection, permitting safe outpatient management. The following communication is based around a 1993 practice guideline.3 This guideline represents a consensus of experts in pediatrics, infectious diseases, and emergency medicine. Significant variation in the proposed approach has been observed among practitioners, but at this point, this guideline is the most comprehensive we have for managing febrile young children.
This article reviews the proposed management strategies, emphasizing common points of departure among practitioners. Specific discussion will focus on how personal experience, the perspective of parents and their risk-taking behavior, limitations of clinical judgment, and physician thresholds in medical decision making may influence the approach. Finally, the need for appropriate documentation, careful followup, and an evidence-based approach to this controversial and changing area of medical knowledge will be emphasized.
FEVER IN PEDIATRIC PRACTICE
Fever is a common reason for infants to be evaluated by pediatricians and family physicians. Hoekelman et al4 reported an office-based study that found that 10.5% of 1068 children, aged 3 to 24 months, had temperatures >38.2°C. These children were common, occurring on average once every 1 to 2 days per practitioner's office. Numerous studies have documented that the prevalence of bacteremia in febrile infants ranges from 1.7% to 12%.5,8 The prevalence of any serious bacterial infection, including sepsis, meningitis, urinary tract infection, bacterial enteritis, bone and joint infection, and pneumonia, in febrile infants younger than 90 days ranges from 1.4% to 17.3%.9
Practitioners must be cautious in their interpretation of the medical literature in this area. Studies may examine different outcomes such as bacteremia and meningitis versus all serious bacterial infections. The literature also contains studies with different definitions of fever.1,10,11 The definition used in the recently published clinical practice guideline is a rectal temperature of ≥38°C in an unbundled infant who has not received recent antipyretic therapy.
EVALUATION OF THE FEBRILE INFANT AND THE CLINICAL PRACTICE GUIDELINE
The foundation of the clinical practice guideline is the performance of a careful history and physical examination and use of selected laboratory studies. Risk of infection is based on age, clinical appearance, and results of screening tests. Management options include hospitalization, outpatient treatment with or without antibiotics, and observation with careful follow-up. The guideline stratifies these options according to age: infants <28 days, 28 to 90 days, and 3 to 36 months of age. The guideline also separates children age 0 to 90 days into high and low risk. Low risk was defined as a nontoxic appearance in a previously healthy child plus a white blood cell count of 5000 to 15,000/mmp 3, an absolute band count of < 1500mmp 3 with 4500 bands, a normal urinalysis, (<5 white cells per high-power field or a negative gram-stained smear), and if diarrhea is present, <5 white blood count per high-power field on stool examination. High risk included children not meeting these criteria. Fever was defined as ≥38°C.
Infants Aged <28 Days
The guideline suggests two management options for infants <28 days of age:
* option I - treatment in the hospital with parenteral antibiotics while awaiting cultures of the blood, urine, and CSF (high and low risk), and
* option 2 - observation in the hospital without treatment while awaiting culture results (low risk only.)
Based on other recent data indicating that it is possible to identify low-risk infants with an impressive degree of diagnostic certainty, a more conservative option not demanding hospitalization also seems reasonable.5,12 Jaskiewicz et al5 in a study of 437 infants ^60 days who met low-risk criteria found a negative predictive value of 98.9% for any serious bacterial infection and a negative predictive value of 99.5% for bacteremia. This negative predictive value means that there was only about a 1% chance of serious bacterial infection and a less than 1% chance of bacteremia in those infants with negative screening tests. A similar study in Taiwan by Chiù et al12 evaluated 254 febrile neonates *^3l days and found 134 meet low-risk criteria. Among these very young infants, only one (0.7%) had bacteremia and meningitis and seven (5.3%) had urinary tract infections.
Number of Physicians Selecting Outpatient Treatment Choices (N=53)*
These studies support the more conservative approach practiced by pediatricians surveyed in 1993.13 In this survey, practitioners were presented with the clinical scenario of a 3-week-old infant with a temperature of 38.5°C and no identifiable source of infection. When asked to choose a management strategy, 39% selected option 1 from the clinical practice guideline and 2% selected option 2. The remaining 59% chose outpatient strategies that deviated from the guideline as seen in the Table.
In a similar study in which pediatricians, family physicians, and emergency medicine physicians were asked to respond to the clinical scenario of a 3 week old with a fever of 40.30C and no source of infection, 98.1% of the pediatricians said they would admit the infant and 94.8% would begin empiric antibiotics while awaiting the results of cultures.14 The height of the temperature may have provoked a different response on the part of these pediatricians.
Is there a "best" course of action? Clearly, there is variability in the approach to these infants, and the clinical practice guideline represents a conservative strategy. If the practitioner manages a febrile infant <28 days of age as an outpatient, there must be assurance of close follow-up, timely access to medical care, and appropriate education of die family about potentially worrisome signs and symptoms. Although there are significant limitations in the ability of any practitioner or family to accurately assess an infant in the first 28 days of age by clinical appearance, the low incidence of serious bacterial infection in this age group may support occasional departure from the clinical guideline. The report by Jaskiewicz et al5 suggested that at the very least, the family should have a telephone, transportation, and ability to return to the health-care facility within 30 minutes of any perceived change in health status. Future trends in management will await further evidence and confirmation of these data.
Infants Aged 28 to 90 Days
The 1993 guidelines use the same clinical criteria to define low risk at 28 to 90 days as for 0 to 27 days except the child who has otitis media who may be treated like the low-risk group. According to the practice guideline, the low-risk group may be considered for two outpatient management options3:
* option 1 - treatment with parenteral ceftriaxone (50 mg/kg/day) as an outpatient pending results of cultures of blood, urine, and CSF, and
* option 2 - observation as an outpatient after urine culture is obtained.
These recommendations are based on the utility of clinical and laboratory criteria for defining risk among febrile infants5 and the observation that the probability of serious bacterial infection among low-risk infants was 0.2%.15 Infants who do not meet low-risk criteria should be hospitalized.3
However, practitioners have argued that the use of these laboratory tests is not practical or feasible given the variety of practice locations and the limited availability of diagnostic testing.16 It has been suggested that a clinician who knows the infant and the family has greater clinical certainty in evaluating a febrile infant. Furthermore, when otitis media is found in a febrile infant in this age group, practitioners feel comfortable in managing these infants with oral antibiotics, which represents a departure from the guideline. In a survey of practitioners presented with the clinical scenario of a 2-month-old nontoxic infant with a temperature of 38.7CC and otitis media, 87% of these practitioners would treat with oral antibiotics if screening laboratory tests were negative.13
Infants Aged 3 to 36 Months
During the first 2 years of life, 65% of children can be expected to visit a physician because of a febrile illness and 4.3% of patients with temperatures ≥39°C without an evident source can be bacteremic.3,17 Toxic infants in this age group are admitted to the hospital for a sepsis evaluation and are treated with parenteral antibiotics. The practice guideline suggests specific options for testing and treating nontoxic infants with temperatures ≥39°C in this age group to optimize clinical outcome.3
Because urinary tract infections are common and the urinalysis is not a sensitive screen in infants, urine cultures should be obtained by catheter or suprapubic aspirate from all febrile males <6 months and females <2 years. If signs of invasive enteritis are present (blood and mucous or >5 white blood cells per HPF on microscopic examination), a stool culture should be obtained and antibiotics started. A chest radiograph is recommended only if there are signs of tachypnea, rales, cough, or thonchi. The practice guideline offers two options for obtaining blood cultures and initiating antibiotic therapy: either obtain cultures and treat all children with temperatures ≥39°C or only those with temperature >39°C who have a white blood cell count ≥15,000.
Follow-up in 24 to 48 hours is recommended for all patients with positive cultures. If the blood culture is positive for Streptococcus pneumoniae and the patient has persistent fever, hospitalization, evaluation for sepsis, and initiation of parenteral antibiotics is recommended. However, if the patient is afebrile and well, antibiotics may be administered as an outpatient. If the urine culture is positive for a pathogen and the patient is febrile and ill, the infant should be admitted and treated with parenteral antibiotics; if the patient is afebrile and well, outpatient antibiotics may be administered.5
In a 1994 survey of 194 primary care pediatricians from Utah, many practitioners deviated from the strategies suggested above.13 When confronted with a nontoxic 20-month-old infant who had a temperature of 400C and no source of infection, 75% of those surveyed would obtain a complete blood cell count, 33% would obtain a blood culture, and only 6% would administer antibiotics. Pediatricians in practice are likely to be influenced by the low probability of poor outcome associated with any management strategy. Considering all febrile infants <2 years of age, the majority do well whether or not management includes antibiotic treatment. In a decision analysis evaluating this clinical problem, it was estimated that a practicing pediatrician who does not prescribe antibiotics for these febrile infants would experience a poor patient outcome (death or permanent disability) approximately once in 6 years.18 The infrequency of bad outcome may explain why practitioners choose to manage infants without antibiotic treatment, despite evidence that treatment is associated with a reduction in fever, persistent bacteremia, and meningitis.3
Further refinements of the evaluation process as suggested in the guideline may be available in the future. In a recent abstract by Kupperman et al,19 two diagnostic tests were found to be independent predictors of occult pneumococcal bacteremia: the absolute neutrophil count and temperature. In their study, 0.7% of 2713 infants ages 3 to 36 months with an absolute neutrophil count of < 10,000 and a temperature of <40°C had pneumococcal bacteremia whereas 9.1% of 617 infants with an absolute neutrophil count ≥10,000 and a temperature of ≥40°C had pneumococcal bacteremia. They favor a selective approach in the use of diagnostic tests and suggested the design of management algorithms that may limit the use of empiric antibiotics. Such approaches have yet to be studied.
Figure. Example of a treatment threshold. In this example, the low-risk infant's probability is less than the physician's treatment threshold; therefore, the infant is not treated. The toxic infant's probability of infection is higher than the physician's treatment threshold; therefore, the infant is treated.
PERSPECTIVE AND PREFERENCES OF PARENTS
What management strategy is favored by parents of febrile infants? In a study conducted by Oppenheim et al20 in an outpatient clinic, parents were interviewed regarding their preferences about fever management. Clinical scenarios were presented to them about infants with temperatures of 103°, risks and complications were discussed, and options from the clinical practice guideline were offered. Parents often selected options associated with higher long-term risk if lower rates of short term consequences, such as painful tests and procedures and more prolonged waiting time, could be avoided. Parents also expressed concern about the use of unnecessary antibiotics and cost, and emphasized their ability and willingness to return if their child's condition deteriorated. This difference in preference may be understandable when one considers the likelihood of encountering serious bacterial infection for physicians and parents. In any year, a physician may see hundreds of febrile children thus making it very likely that a serious bacterial infection will be encountered. A family, on the other hand, confronts fever as a singular and rare event with low probability of adverse outcome.
Do practitioners incorporate parental preference into their clinical decisions? Understanding the point of view of the family may influence the decision and contribute to details of management. Whether practitioners are aware of these data or incorporate parental preference into their decisions remains unclear.
LIMITATIONS OF CLINICAL JUDGMENT
Identifying a child as "toxic" is associated with a high likelihood of serious bacterial infection. Can physicians properly identify low-risk infants and with what degree of diagnostic certainty? Attempts to quantitate clinical judgment have yielded variable and disappointing results. Clinical assessments alone are poor in detecting serious bacterial infection; a certain proportion of infants with serious bacterial infection look well. The published sensitivities of clinical assessment range from 64% to 80%.21,22 In a recent study of 6611 infants, the Yale Observation Score's highest sensitivity was only 28.6% in detecting bacteremia and was of limited use in discriminating bacteremic from nonbacteremic febrile infants.23 Another study comparing pediatric residents' and private practitioners' ability to diagnose bacteremia found no difference in the two groups; practitioners were, however, better at predicting the absence of bacteremia.24 Given these observations, clinicians must understand the limitation of clinical judgment in their evaluations of febrile infants. As suggested in the practice guideline, selected diagnostic tests add to the reliability of the evaluation process.
TREATMENT THRESHOLDS, DOCUMENTATION, AND FUTURE EVIDENCE
The concept of the treatment threshold is useful in understanding how physicians follow or depart from a clinical practice guideline.25,26 Practitioners synthesize complex sets of data and choose a single course of action when managing febrile infants. The decision is unique for the physician and the patient. One's need for diagnostic certainty is based on the penalty for being wrong and the reward for being right. Before a physician evaluates a febrile infant, there is a probability or percentage chance that the patient will have a serious bacterial infection. It seems important to introduce the sentence: Probability can be expressed as a number between 0 and 1 or as a percentage from 0 to 100 and is depicted in the Figure. Prior probability, p1, is the likelihood of disease based on the prevalence of disease in the population. After the physician evaluates an infant, there is a new probability of illness based on clinical appearance and other attributes such as screening laboratory criteria. This new probability, pp 2, may be higher if the infant is judged to be toxic or lower if the infant is considered low risk.
A physician will take action based on his or her own treatment threshold and the probability of disease after evaluating a febrile infant. The treatment threshold, p*, is the probability at which one considers treating and not treating of equal value. "Treatment" can be considered any course of action including following a clinical practice guideline, hospitalizing an infant, or administering antibiotic therapy. If the probability of serious bacterial infection is above the treatment threshold, the physician will treat; if the probability of infection is below the treatment threshold, the physician will not treat. This concept is depicted in the Figure.
To use treatment thresholds in making medical decisions, it is important to understand limitations of this concept. Physicians may be biased by personal experience. The infrequent occurrence of poor outcome in febrile infants may influence a physician's treatment threshold. Therefore, it is important to consider published evidence about disease prevalence and outcomes. Evidence, however, may not exist ?t may not be applicable to an individual physician's practice population. We need data that are current and applicable to pediatric practice to answer many of the questions that remain. In the absence of such data, expert consensus as promulgated in clinical practice guidelines, or personal experience, adjusted for bias, should be used for medical decisions. As new information becomes available, it must be incorporated into treatment thresholds and applied to patient management decisions.
The management of die febrile infant is truly a complex and evolving topic in pediatrics. The clinical practice guideline provides a map for physician decision making in this important clinical issue. An individual physician may alter the course of action based on his or her treatment threshold but must recognize the limitations of that approach. Documentation of management decisions that deviate from published guidelines are important for patient care. The medical record must carefully represent the appearance of the patient, the actions taken, and the discussion with the family regarding management. Close follow-up includes phone contact with documentation and reexamination of the patient in a reasonable period of time.
All who deal with the common problem of fever in infants should follow this subject closely and be ready to adopt new management strategies when evidence from outcomes-based research emerges. In the mean time, the current practice guideline offers an excellent starting point in the management of this common and vexing clinical problem.
1. Teele DW, Pelton SI, Grant MJA, et al. Bacteremia in febrile children under 2 years of age: resulti of cultures of blood of 600 consecutive children seen in a 'walk-in' clinic. J Pediatr. 1975;87:227-250.
2. Long SS. Approach to the febrile patient with no obvious focus of infection. Pediatr Rev. 1984;5:305-315.
3. Beraff L], Bass JA, Fleischer GR, et al. Practice guideline for the management of infants and children 0 ro 36 months of age with fever without source. Pediatrics. 1993;92:1-12.
4. Hoekelman R, Lewin EB, Shapira MB, Sutherland SA. Potential bacteremia in pediatric practice. AmJ Dis Child. 1979;133:1017-1019.
5. Jaslciewicz JA, McCarthy CA. Richardson AC, it al. Febrile infants at low risk for serious bacterial infection - an appraisal of the Rochester Criteria and implications for management. Pediatrics. 1994:94:390-396.
6. Bass JW, Steele RW, Wittier RR, et al. Antimicrobial treatment of occult bacteremia: a muttcenter cooperative study. Pediatr Infect Dis J. 1993;12:466-473.
7. Fleisher GR, Rosenberg N, Vinci R, et al. Intramuscular versus oral antibiotic therapy for the prevention of meningitis and other bacterial sequelae in young, febrile children at risk for occult bacteremia. J Pediatr. 1994;124:504-512.
8. Jaffe DM, Tara RR, Davis AT, Henretig F, Fleisher G. Antibiotic administration to treat possible occult bacteremia in febrile children. N Engl J Med. 1987:317:11751180.
9. Banff L], Oslund S. Schriger DL, Stephen ML Probability of bacterial infection in infants less than 3 months of age: a meta-analysis. Pediatr Infect Dis J. 1992;11:257-265.
10. Baker MD, Bell LM, Avner JR. Outpatient management without antibiotics of fever in selected infants. N Engl J Med. 1993;329:1437-1441.
11. Bonadio WA. Evaluation and management of serious bacterial infections in the young febrile infant. Pediatr Infect Dis J. 1990;9:905-912.
12. Chiu CH, Lin TY, Bulloni MJ. Application of criteria identifying febrile outpatient neonates at low risk for bacterial infections. PeoW Infect Dis J. 1994;13:946-949.
13. Young PC. The management of febrile infants by primary care pediatricians in Utah: comparison with published practice guidelines. Pediatrics. 1995;95:623-627.
14. Jones RG, Bess JW. Febrile children with no focus of infection: a survey of their management by primary care physicians. Pediatr Infect Dis J. 1993;12:179-183.
15. Klassen TP, Rowe PC. Selecting diagnostic tests to identify febrile infants less than 3 months of age as being at low risk for serious bacterial infection: a scientific overview. J Pediatr. 1992;121:671-676.
16. Hyman D. Holland BC. Lockman BE. DeFelke PL. Belser NL. Letter to the Editor. Pediatria. 1994;93:344-345.
17. Soman M. Characteristics and management of febrile young children seen in a university family practice. J Fam Pract. 1985;21:117-122.
18. Downs SM, McNutt RA, Margolis PA. Management of infants at risk for occult bacteremia: a decision analysis. J Pediatr. 1991;118:11-20.
19. Kupperman N, Jaffe D, Fleisher G. Independent predictors of occult pneumococcal bacteremia in young febrile children. Pediatr Res. 1996;39:135A. Abstract.
20. Oppenheim Pl, Sotiropoulos C1 Baraff L]- Incorporating patient preferences into practice guidelines: management of children with fever without source. Amt Emerg Med. 1994;24:836-841.
21. Crain E, Shelov S. Febrile infants: predictors of bacteremia. J Pediatr. 1982;108:1-12.
22. BerkowiK CD, Uchiyama N. TuIIy SB, et al. Fever in infants less than 2 months of age: spectrum of disease and predictors of outcome. Pediatr Emerg Care. 1985;1:128135.
23. Teach SJ, Fleisher GR. Efficacy of an observation scale in detecting bacteremia in febrile children 3 to 36 months of age, treated as outpatients. J Pediatr. 1995;126:877881.
24. Dershewitz RA, Wigder HN, Widger CM, Nadelman DH. A comparative study of the prevalence, outcome, and prediction of bacteremia in children. J Pediatr. 1983;103:352-358.
25. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med. 1980;302:1109-1117.
26. Sox HC, Bergen MR, Young KM. Stanford Faculty Development Program: Threshold Model 1993.
Number of Physicians Selecting Outpatient Treatment Choices (N=53)*