Joc 31813

8/13/2019 Joc 31813

1/10

ORIGINAL CONTRIBUTION

Management and Outcomesof Care of Fever in Early InfancyRobert H. Pantell, MD

Thomas B. Newman, MD, MPH

Jane Bernzweig, PhD

David A. Bergman, MD

John I. Takayama, MD, MPH

Mark Segal, PhD

Stacia A. Finch, MA

Richard C. Wasserman, MD, MPH

FEBRILE INFANTS OFTEN LACK SUG-gestive clinical symptoms orfindings, making it difficult todistinguish between a minor fe-

brile illness and one that is life-threatening. To avoid the conse-quences of failing to detect seriousbacterial illness (SBI), such as bacter-emia and bacterial meningitis, a vari-ety of clinical strategies have been de-veloped to identify infants at high andlow risk, including policies that re-quire extensive laboratory testing, hos-pitalization, and treatment with intra-venous antibiotics.1-10 Although thesestrategies guarantee treatment of all in-fants with SBI, the costs are high, in-cluding considerable iatrogenic mor-bidity for some infants.10

Many strategies were developedfrom infants cared for in inner-cityemergency departments. Performanceof such strategies in the general popu-lation has not been evaluated. Studiessurveying responses to case scenarios

suggest that a large proportion ofoffice-based physicians do not rou-tinely follow these guidelines.11 Littleis known about the actual manage-ment of cases of febrile infants inoffice practice.

The purposes of this study were to(1) characterize the management, spec-trum of diseases, and clinical out-

comes of febrile infants aged 3 months

or younger in pediatric practices in theUnitedStates; (2)develop a clinicalpre-diction model for the identification of

infants with bacteremia/bacterial men-

Author Affiliations: Divisionof General Pediatrics, De-partment of Pediatrics (Drs Pantell, Newman, Bern-zweig, andTakayama),and Department of Epidemiol-ogy and Biostatistics (Drs Newman and Segal), Schoolof Medicine, Universityof California,San Francisco;De-partment of Pediatrics,Lucile Salter Packard ChildrensHospital at Stanford Medical Center, Stanford, Calif (DrBergman);PediatricResearchin OfficeSettings(PROS),Departmentof Practiceand Research, Center for Child

Health Research, American Academy of Pediatrics, ElkGrove Village, Ill (Ms Finch and Dr Wasserman); andDepartment of Pediatrics, Vermont College of Medi-cine, Burlington (Dr Wasserman).Corresponding Author: Robert H. Pantell, MD, Uni-versity of California, San Francisco, School of Medi-cine, Department of Pediatrics,Divisionof General Pe-diatrics, Box 0503, San Francisco, CA 94143-0503([email protected]).

Context Fever in infants challenges clinicians in distinguishing between serious con-ditions, such as bacteremia or bacterial meningitis, and minor illnesses. To date, thepractice patterns of office-based pediatricians in treating febrile infants and the clini-cal outcomes resulting from their care have not been systematically studied.

Objectives To characterize the management and clinical outcomes of fever in in-fants, develop a clinical prediction model for the identification of bacteremia/bacterialmeningitis, and compare the accuracy of various strategies.

Design Prospective cohort study.

Setting Offices of 573 practitioners from the Pediatric Research in Office Settings(PROS) network of the American Academy of Pediatrics in 44 states, the District ofColumbia, and Puerto Rico.

Patients Consecutive sample of 3066 infants aged 3 months or younger with tem-peratures of at least 38C seen by PROS practitioners from February 28, 1995, throughApril 25, 1998.

Main Outcome Measures Management strategies, illness frequency, and rates andaccuracy of treating bacteremia/bacterial meningitis.

Results The PROS clinicians hospitalized 36% of the infants, performed laboratorytesting in 75%, and initially treated 57% with antibiotics. The majority (64%) weretreated exclusively outside of the hospital. Bacteremia was detected in 1.8% of in-fants (2.4% of those tested) and bacterial meningitis in 0.5%. Well-appearing infantsaged 25 days or older with fever of lessthan38.6C had a rate of 0.4% for bacteremia/bacterial meningitis. Frequency of other illnesses included urinary tract infection, 5.4%;

otitis media, 12.2%; upper respiratory tract infection, 25.6%; bronchiolitis, 7.8%; andgastroenteritis, 7.2%. Practitioners followed current guidelines in 42% of episodes.However, in the initial visit, they treated 61 of the 63 cases of bacteremia/bacterialmeningitis with antibiotics. Neither current guidelines nor the model developed in thisstudy performed with greater accuracy than observed practitioner management.

Conclusions Pediatric clinicians in the United States use individualized clinical judg-ment in treating febrile infants. In this study, relying on current clinical guidelines wouldnot have improved care but would have resulted in more hospitalizations and labo-ratory testing.

JAMA. 2004;291:1203-1212 www.jama.com

See also p 1261 and Patient Page.

2004 American Medical Association. All rights reserved. (Reprinted) JAMA, March 10, 2004Vol 291, No. 10 1203

wnloaded From: http://jama.jamanetwork.com/ on 12/12/2013

8/13/2019 Joc 31813

2/10

ingitis; and (3) compare the accuracyof practitioners management with ex-isting guidelines.

METHODS

Sites and Practitioner Participants

This study was conducted by the Pedi-atric Research in Office Settings(PROS) practice-based research net-work of the American Academy ofPediatrics (AAP). A total of 573 mem-bers of the PROS network (91% ofwhom were physicians) from 219practices submitted data on eligibleinfants. Data were received from 44states, the District of Columbia, andPuerto Rico. Participant and nonpar-ticipant characteristics are shown inTABLE1. Practitioners who declined toparticipate or who did not respond to

recruitment efforts varied little fromparticipating practitioners. In addition,compared with AAP members wholisted patient care as their primaryactivity in a 1995 periodic survey,PROS practitioners were similar in ageand sex, but fewer (7.3% vs 12%;P.001) practiced in urban inner-cityareas. Infants who were eligible butnot enrolled also closely resembledenrolled infants with respect to tem-perature and on average were slightlyolder (4 days); hospitalization rates

were similar.

Patient Participants

Infants were eligible for thestudyif theywere aged 3 months or younger, hadbeen discharged from the hospital as anewborn, hada temperature of 38C orgreater either at home or in the clini-cians office, andhadno other major co-morbidities (eg, congenital anoma-lies, extreme prematurity, conditionsassociated with organ system failure).Foranalysis, we usedthemaximum rec-

tal temperature taken in the office orreported by the parent in the past 24hours, after adding 0.5C for axillarytemperatures. Data were collected for3131 consecutive infants, 3066 ofwhom met eligibilityrequirements.Thestudy was approved by the Commit-tee on Human Research of the Univer-sity of California, San Francisco.

Table 1.Practitioner and Patient Characteristics*

Characteristics Participating Nonparticipating

Practitioners

No. of practitioners 573 781

SexMale 300 (52) 378 (49)

Female 273 (48) 389 (51)

Age, y45 307 (54) 433 (59)

45 266 (46) 303 (41)

RaceWhite 519 (91) 632 (81)

African American 3 ( 1) 17 (2)

Asian/Pacific Islander 36 (3) 50 (6)

Other/missing 15 (3) 82 (11)

Practice structureGroup 386 (67) 493 (64)

Solo 46 (8) 46 (6)

University 32 (6) 91 (12)

Health maintenance organization 22 (4) 40 (5)

Other 87 (15) 105 (13)

Practice regionNortheast 197 (34) 306 (39)

South 141 (25) 180 (23)

Central/Midwest 91 (16) 163 (21)

West 144 (25) 132 (17)

Practice settingUrban, inner city 42 (7) 108 (14)

Urban, not inner city 133 (23) 202 (26)

Suburban 259 (45) 329 (42)

Rural 131 (23) 101 (13)

Other/missing 8 (1) 41 (5)

Patients (n = 3066)

Demographic characteristicsAge

Mean (SD), wk 7.0 (3.4)

1-30 d 775 (25)

31-60 d 1220 (40)

60 d 1071 (35)

Female 1436 (47)

Race/ethnicityWhite, non-Hispanic 2150 (70)

Black 246 (8)

Asian 67 (2)

Hispanic 453 (15)

Other/missing 150 (5)

Medicaid insured 1074 (35)

Clinical characteristicsTemperature, C

Mean (SD) 38.7 (0.50)

Maximum (home or office)38.5 1361 (44)

38.5-38.9 1049 (34)

39.0-39.4 458 (5)

39.5 198 (7)

AppearanceVery ill 50 (2)

Moderately ill 767 (25)

Well or minimally ill 2206 (73)

*Data are expressed as No. (%) unless otherwise noted.Includes 18 participating Hispanic practitioners and 23 nonparticipating Hispanic practitioners.

TREATMENT AND OUTCOMES OF FEBRILE INFANTS

1204 JAMA, March 10, 2004Vol 291, No. 10 (Reprinted) 2004 American Medical Association. All rights reserved.


8/13/2019 Joc 31813

3/10

Procedures

We used a prospectivecohort study de-sign to follow theepisode of care for in-fants seen by PROS practitioners fromFebruary 28, 1995, through April 25,1998. Demographic and clinical data

were recorded by office staff and clini-cians on standard forms. Practitionersrecorded clinical signs and symptomsandan overall assessment of clinical ap-pearance before ordering laboratorytests. They also answered questionsabout clinical appearance similar tothose of the Yale Observation Scale,12,13

with the addition of an item on respi-ratory distress. For clinical appear-ance, practitioners indicated whetherthe infant appeared well/minimally ill,moderately ill, or very ill. Initial man-agement, changes in treatment strate-

gies, and subsequent medical contactswere decided by individual cliniciansand documented until resolution of ill-ness, when they recorded final diagno-sis. A variety of techniques were usedto maximize patient enrollment andminimizeselection bias and ensure dataquality.

Laboratory testing was performed atthe discretion of the clinician accord-ing to their usual and customary prac-tice. Testing wasperformed at the usuallaboratories of the practicesites; all clin-

cians were supplied with and asked touse urine dipsticks (Ames-Multistix,Miles Inc, Elkhart, Ind).

Clinical Diagnoses

A study manual was developed con-taining definitions of all clinical con-ditions andvariables. While other stud-ies have addressed SBI (ie, bacteremia,bacterial meningitis, urinary tract in-fection, and bacterial gastroenteritis) asthe main outcome variable, this reportfocuses on occult infections that have

generated the most uncertainty in de-veloping clinical strategies; ie, bacter-emia with pathogenic organisms andbacterial meningitis. We have re-ported elsewhere on urinary tract in-fections14 andalso report herein the fre-quency of other, less common seriousbacterial illnesses, such as cellulitis andosteomylelitis. A pediatric infectious

disease specialist reviewed all cases ofbacteremia and excludedthose not con-sidered to be caused by pathologic bac-teria. Similarly, all cases of meningitiswere reviewed to ascertain the accu-racy of the diagnosis and whether the

meningitis was likely bacterial, par-tially treated bacterial, or viral inorigin.

Statistical Analyses

STATA software, version 6 (CRC Inc,College Station, Tex) was used for sta-tistical analyses. For multivariate analy-ses of bacteremia/bacterial meningitisthat included laboratory data as pre-dictors, we included only patients forwhom a blood culture was obtained.For these models, the white blood cell(WBC) count was considered to be ab-

normal if it was less than 5000/L orat least15000/L andthe urinalysis wasconsidered abnormal if the dipstick testwas positive for leukocyte esterase ornitrite or if more than 5 WBCsper high-powered field were reported on micro-scopic examination. We used a back-wardstepwise logistic regression model,with aPvalue of .02 to predict order-ing of WBC counts and blood culturesand aP value of .10 to predict bacter-emia/bacterial meningitis. Standarder-rors for all logistic models were ad-

justed for clustering by practitioner andmodels tested for goodness of fit by themethod of Hosmer and Lemeshow.Tree-structured analyses were con-ducted using S-PLUS, version 3.4(MathSoft, Seattle, Wash).15,16

Clinical Prediction Models

To compare the accuracy of variousclinical prediction models, we ana-lyzed several alternative scenarios. Formodel 1, we used patient appearancealone as a predictor of bacteremia/

bacterial meningitis. Two additionalmodels were created by adding WBCcount and WBC count with urinalysis,respectively. For the fourth model, werelied on current guidelines. Currentguidelines6,17 and protocols5,7,18,19 differslightly in specific lower limits of ageto be considered low risk (28-30days), definition of abnormal WBC

count (15000/L to 20000/L),urinalysis performance, whether stoolWBC count should be performed ininfants with diarrhea, whether a lum-bar puncture is required in all infantsregardless of appearance and screen-

ing test results, and whether screeningshould be done in all febrile infants oronly in those without a source ofinfection. We blended these guidelinesso that i nfants ag ed 3 0 day s o ryounger and ill-appearing infantsrequired a WBC count, blood culture,urinalysis, urine culture, cerebrospinalfluid analysis and culture, hospitaliza-tion, and antibiotics; well-appearinginfants aged 31 days or older requireda WBC count and urinalysis as thebasis for proceeding with further man-agement. For the fifth model, we used

the decision tree derived from apply-ing tree-structured analysis to oursample. We compared the sensitivityand specificity of the 5 clinical predic-tion models described herein with theactual management and outcomes ofthe PROS practitioners. Sensitivity wascalculated as the number of infantswith bacteremia/bacterial meningitiswho would have been treated in eachstrategy or, for study infants, wereactually treated with antibiotics at theinitial visit divided by the number

with bacteremia/bacterial meningitis(n= 63). The concept of specificity,defined as the percentage of patientswithout the condition (bacteremia/bacterial meningitis) who were nottreated, does not strictly apply, sinceother diseases require antibiotics.However, to reflect the trade-offbetween sensitivity and specificity, wecalculated a specificity in which thedenominator used represents allinfants without bacteremia/bacterialmeningitis and other conditions

requiring antibiotics (ie, otitis media,urinary tract infection, pneumonia),while the numerator represents chil-dren not treated at the initial encoun-ter with an antibiotic. Specificity inthis context does not infer unneces-sary treatment but is used as an indi-cator of the relative frequency of anti-biotic use in different strategies.




8/13/2019 Joc 31813

4/10

RESULTSManagement Practices

of PROS Practitioners

Of the 3066 infants, 1975 (64%) weremanaged exclusively outside of the hos-pital. Only 125 episodes of care were

managed with a single office visit with-out other medical contacts (eg, tele-phone). A single visit was recorded for909 infants, while 761 infants had2 vis-its and 305 had 3 or more visits to thehospital. In addition, 1014 episodeswere accompanied by a single fol-low-up telephone call, 325 receivedmore than 1 telephone encounter. Ofinfants managed outside of the hospi-tal, 68 were seen in emergency depart-ments following the initial office visit.

Testing and management strategiesdid not vary by practitioner age, sex, or

region but varied with certain infantdemographicandclinical variables.Com-pared with older infants, those youngerthan 1 month were significantly morelikely to havea WBC count or blood cul-

ture (83.0 vs 71.4%; P.001), have alumbar puncture (54.8 vs 25.6%;P.001), begin antibiotic treatment atthetime oftheinitialexamination(68.2%vs 53.7%; P.001), and be hospitalized(60.1%vs27.3%;P001).Adjusted odds

ratiosforindependent predictors ofbloodtesting (WBC count and/or blood cul-ture) are documented inTABLE2. Pre-diction was fair, with the area underthe receiver operating characteristiccurve (AUROC)=0.735. Coding tem-perature and age as continuous vari-ables improved the fit slightly withAUROC= 0.746.Patients seen outside oftypical office hours were significantlymore likely to receive laboratory test-ingand to betreatedwith antibiotics andadmitted to the hospital. Infants receiv-ing Medicaid hadmore testing and hos-

pitalizations.Nearly one quarter of infants had no

blood, urine, or cerebrospinal fluidtestedduringtheir evaluation (n= 726),and slightly more than half had their

urine tested (n= 1666). Actual manage-ment strategies varied considerablyfromsuggested approaches to febrilein-fants.TABLE3compares PROS practi-tioners management with recom-mended strategies.

Final Diagnoses of Febrile Episodes

TABLE 4 lists the primary final diag-noses at the end of the illness epi-sodes. Most febrile episodes were dueto benign illnesses, while bacteremiawas present in 2.4% of infants withblood cultures and bacterial meningi-tis in 0.5% of the entire sample of in-fants. Other causes of serious bacte-rial illness were documented but many(eg, cellulitis) were not occult.

The 54 cases of bacteremia repre-sent cases reviewedby an infectious dis-

ease consultant and diagnosed as hav-ing pathogenic organisms; 18 othercases originally classified as bactere-mic by clinicians were recoded to eitherthe next listed diagnostic category orto an unidentified source. Of 16 in-fants originally classified as having bac-terial meningitis by clinicians, the ex-ternal reviewer confirmed 14 cases,including some cases in which the cul-ture was sterile because of prior anti-biotics but the cerebrospinal fluid andclinical findings were consistent with

bacterial meningitis. Five of the in-fants with bacterial meningitis also hadbacteremia. The bacterial organisms ininfants with bacteremia/bacterial men-ingitis are identified in TABLE 5. Muchof the bacteremia/bacterial meningitisoccurred in the first month after birth,when 4.1% of febrile infants had bac-teremia/bacterial meningitis com-pared with 1.9% in the second monthand 0.7% in thethird month (TABLE 6).

We also examined the frequency ofbacteremia/bacterial meningitis and

other bacterial illnesses by tempera-ture. Because fever of at least 38.0C athome was an eligibility criteria, someinfants included were afebrile at thetime of office visit. The frequency ofbacteremia/bacterial meningitis foundin infants who were afebrile in the of-fice as well as those who had highertemperatures is documented inTABLE 7.

Table 2.Multivariate Predictors of Blood Testing

Clinical Features Odds Ratio (95% CI) PValue

Age, d*31 2.86 (2.19-3.72) .001

31-60 2.31 (1.88-2.84) .001

Temperature, C38.5-38.9 1.40 (1.17-1.68) .001

39.0-39.4 1.60 (1.22-2.10) .00139.5 1.81 (1.22-2.71) .004

AppearanceModerately ill vs well 1.92 (1.55-2.38) .001

Very ill vs well 2.63 (0.99-7.00) .05

Unattentive 1.81 (1.36-2.48) .001

No smile 1.55 (1.24-1.93) .001

Decreased social interaction 1.35 (1.05-1.74) .02

Received care after hours 2.32 (1.70-3.18) .001

Medicaid insured 1.39 (1.12-1.72) .003

Source of fever known 0.63 (0.52-0.76) .001

Abbreviation: CI, confidence interval.

*Comparison group: aged 60 days.Comparison group: temperature 38.5C.

Table 3.PROS Practitioner Adherence to Guidelines

Age/Appearance RecommendationCases in Which Guideline

Was Followed, %

31 days Complete s epsis workup/hospital ization/antibiotics

45.7

31-90 days/moderatelyor very ill

Complete sepsis workup/hospitalization/antibiotics

35.8

31-90 days/minimally ill White blood cell count/ uri nalysis 41.6

Abbreviation: PROS, Pediatric Research in Office Settings.




8/13/2019 Joc 31813

5/10

Developing a Clinical

Prediction Model

We used logistic regression analyses toidentify the best clinical predictors ofbacteremia. The analysis inTABLE8isfor the entire sample of 3066 infants.

In this analysis, we exclude laboratoryresults to identify which clinical fea-tures are potentially useful in initiallyidentifying infants at high risk. We al-lowed for entry of variables with P.10because of the small number of casesof bacteremia. Age and very ill appear-ance emerged as the strongest predic-tors (AUROC= 0.820). To evaluate thepredictive value of laboratory testing,we first performed a logistic regres-sion on the 1746 infants who had bothWBC counts and blood cultures but ex-cluded WBC results. Without WBC

count, the AUROC was 0.767. By add-ingabnormal WBC count (asa dichoto-mous variable with abnormal countsc o n s i d e r e d t o b e 5 0 0 0 / L o r15000/L), the AUROC increased to0.803. The resulting model is summa-rized inTABLE9. By adding abnormalurinalysis, the AUROC increased to0.806, but the coefficient for an abnor-mal urinalysis was not statistically sig-nificant.

Finally, we used recursive partition-ing analysis with tree-structured analy-

sis to develop a classification tree toidentify low- and high-riskgroups. Theclassification procedure initially di-chotomizes on the single variable thatbest separates low- andhigh-risk groupsusing internal split sampling into tenths.It then continues to select the best vari-able for separating the remaining pa-tients into low- and high-risk groups.In the model (FIGURE), infants who aremoderately or severely ill appearinghave a bacteremia/bacterial meningi-tis rate of 4.4%, while those who ap-

peared well/minimally ill have a 1.2%occurrence. Of this latter group, 3.4%of those younger than 25 days had bac-teremia/bacterial meningitis while 0.8%of those aged 25 days or older had bac-teremia/bacterial meningitis. Finally, forinfantswho appeared well/minimally ill,were at least 25days old, and had a tem-perature of 38.6C or higher, bacter-

emia/bacterial meningitis was found in1.2%, while those with these charac-teristics but temperature of less than38.6C had a 0.4% chance of bacter-emia/bacterial meningitis, or a nega-tive predictive value of 99.6%.

Accuracy of Diagnostic Strategies

We compared the sensitivity of differ-ent approaches to identify infants withbacteremia/bacterial meningitis. The re-sults are displayed inTABLE10for in-fants with WBC counts and blood cul-tures. Only 58.1% with bacteremia/bacterial meningitis appeared clinicallyill; the specificity as defined for this

analysis was 68.1% (model 1). Addingabnormal WBC count as a predictor in-creases sensitivity to 83.9% but de-creases specificity to 54.0% (model2). The addition of a urinalysis in-creases the sensitivity to 87.1%, with a

small decrement in specificity. Cur-rent guidelines (model 4) call for treat-ing all infants with high-risk criteriawhile giving options for treating low-risk infants. Using this approach, 3 of62 with bacteremia/bacterial meningi-tis would have been classified as lowr i sk (1 i nfant w i th b acter em i a/ bacterial meningitis did not have clini-cal appearance recorded initially and is

Table 4.Final Primary Diagnoses

Primary Diagnosis No. Percentage

Upper respiratory tract infection 785 25.6

Unidentified source 655 21.4

Otitis media 375 12.2

Bronchiolitis 239 7.8

Gastroenteritis 222 7.2

Urinary tract infectionTotal 167 (18)* 5.4/10.3

Girls 107 7.5/13.4

Boys, uncircumcised 41 11.5/20.8

Boys, circumcised 15 1.3/2.6

Boys, circumcision status not specified 4 4.0/9.1

Pneumonia 102 3.3

Viral syndrome, nonspecific 85 2.8

Viral meningitis (nonherpes) 82 2.7

Noninfectious 76 2.5

Well child 69 2.3

Bacteremia 54 1.8/2.4

Influenza 41 1.3

Exanthematous illness 39 1.3

Pharyngitis 29 1.0

Viral syndrome, specific 21 0.7

Bacterial meningitis 14 (5)* 0.5

Cellulitis 6 0.2

Sinusitis 5 0.2

Conjuctivitis 4 0.1

Candidiasis 4 0.1

Pertussis 3 0.1

Other infectious 2 0.1Peritonitis 2 (1)* 0.1

Cervical adenitis 2 0.1

Osteomyelitis, omphalitis, mastitis, salmonella enteritis,renal abscess, herpes meningitis (1 case each)

6 0.3

Total 3066 100

*Numbers in parentheses representcases of bacteremiaoccurring simultaneously. Theseare includedin the totalnum-ber with bacteremia.

Of the 167 cases of urinary tract infection, 18 were bacteremic (17 with same organism) and 1 had renal abscess.Percentage when denominator included only those who had blood cultures performed.




8/13/2019 Joc 31813

6/10

excluded from this analysis). There-fore, we estimate that 59of 62 with bac-teremia/bacterial meningitis wouldhavereceived appropriate antibiotics.

For model 5, tree-structured analy-sis (Figure), the sensitivity closely re-

sembles current guidelines but themodel is less specific. The PROS prac-titioners (model 6) initially treated 61of the 63 infants with bacteremia/bacterial meningitis on the initial visit;all but 1 received parenteral antibiot-ics; 36% initially received ceftriaxone,34% ampicillinpluscephalosporin, 22%ampicillin plus gentamycin, and 8%other antibiotic combinations. The

sensitivity of PROS practitioners intreating bacteremia/bacterial meningi-tis was 96.8%. The 2 infantsnot treatedinitially included a 26-day-old infantwho appeared well, had an initial WBCcount of 13000/L, and had a blood

culture positive for group B strepto-cocci on the following day; the infantwas treated and had an uneventful re-covery. The other child was a 4-week-old infant who appeared well and hada WBC count of 15300/L with 8%bands. The infant was sent home with-out antibiotics, became progressivelymore irritable the next day, and was di-agnosed as having pneumococcal men-

ingitis; reports over the following yearindicated that the infant achieved nor-mal developmental milestones.

Among children without bacteremia/bacterial meningitis, otitis media, uri-nary tract infection, or pneumonia, cli-

nicians treated 64.5%, for a specificityof 35.5%. Practitioners also hospital-ized 309 fewer infants younger than 1month of age and conducted fewer di-agnostictests(Table 3) than they wouldhave had they followed current guide-lines.

COMMENT

Fever in young infants has generatedsubstantial interest, research, and con-troversyover the past 30 years. A num-ber of factors in the 1970s contributedto concerns aboutthe appropriate treat-

ment of febrile infants, including emerg-ing awareness of late-onset group B-hemolytic streptococcal sepsis andoccult bacteremia due to Streptococ-cus pneumoniae.20-22 In response, manyacademic centers encouraged exten-sive diagnostic testing, hospitaliza-tion, and antibiotic treatment of all in-fants younger than 60 or 90 days. Thisapproach had considerable costs andmorbidities10 and was followed by ef-forts to identify methods to distin-guish infants at high and low risk ofse-

rious bacterial i l lness. Differentstrategies were developed and tested inemergency departments in urban cen-ters (Boston, Mass,18 Philadelphia, Pa,7

Rochester, NY,4,5 and Pittsburgh, Pa19),and guidelines were also promul-gated.5,17 Subsequent experience re-vealed good sensitivity and negativepredictive value but less so in infantsyounger than 1 month.23-26 The gener-alizability of these strategies to officepractice hasnotbeen studied, andthereare indications that these guidelinesdo

n o t s e e m t o h a v e b e e n w i d e l yadopted11,27 and were imperfectly ap-plied in institutions where they weregenerated.6,28

This is the first nationwide study offebrile infants treated in community-based pediatric practices in the UnitedStates. We have developed a portrait ofhow febrile infants are cared for and the

Table 5. Bacteria Identified

Organisms Bacter emia Cas es, No. Bacterial Meningitis Cas es, No.

Escherichia coli 16 2*

Group BStreptococcus 14 1*

Staphylococcus aureus 5 0

Enterococcus faecalis 3 0

Streptococcus pneumoniae 3 2

Enterobacter cloacae 2 0

Staphylococcus(other) 3 0

Group AStreptococcus 2 0

Streptococcus bovis 1 1*

Klebsiella pneumoniae 1 0

Listeria monocytogenes 1 1*

Yersinia enterocolitica 1 0

Proteus mirabilis 1 1*

Gram-negative rod 1 0

Pseudomonas stutzeri 0 1

Neisseria meningitides 0 1

*Cases with simultaneous bacteremia.

Table 6.Patients With Bacteremia/Bacterial Meningitis by Age

Age, moTotal No. of

Patients

Cases ofBacteremia

Only, No.Cases of Bacterial

Meningitis, No.

Total No. (%) WithBacteremia/Bacterial

Meningitis

0-1 775 23 9* 32(4.1)

1-2 1220 18 5 23(1.9)

2-3 1071 8 0 8(0.7)

Total 3066 49 14 63(2.1)

*Five of these 9 cases of bacterial meningitis also had bacteremia.

Table 7.Office Temperature of Patients With Bacteremia/Bacterial Meningitis

Temperature, CTotal No. (%)

of PatientsNo. (%) With Bacteremia/

Bacterial Meningitis

38.0 835 (27) 6 (0.7)

38.0-38.4 1141 (37) 18 (1.6)

38.5-38.9 604 (20) 27 (4.5)

39.0 304 (10) 10 (3.3)

Missing data 182 (6) 2 (1.1)




8/13/2019 Joc 31813

7/10

types of illnesses seen in office prac-tice. It is clear that the PROS practi-tioners in our sample do not follow ex-isting guidelines for treating febrileinfants, even for those younger than 1month. While use of less-invasive test-

ing such as WBC count and urinalysisoccurs in the majority of cases, moreinvasive testing, such as lumbar punc-tures and hospitalizations, occurrs lessfrequently, especially in older infants,than called for by published strategiesor documented in series from emer-gency departments.

Our results suggest that clinical char-acteristicsof febrileinfantshave a strongassociation with both diagnostic evalu-ation and management strategies.Younger infants, those appearing moreill, andthose with higher fever were sig-

nificantly more likely to receive labo-ratory evaluations, receive antibiotics,or be hospitalized. A few patient char-acteristics unrelated to clinical appear-ance predicted management. Medicaid-insured infants were more likely toreceive laboratory tests and be hospi-talized after adjusting for other fac-tors, likely due to perceived barriers inobtaining adequate follow-up care. Be-cause testing was less selective in in-fants with Medicaid, they had a lowerrate of bacteremia.

The majority of febrile infants in thisstudy had benign and self-limited ill-nesses; our sample size permits reason-able estimates of the risk of bacteremia/bacterial meningitis as well as otherserious illnesses, including urinary tractinfection. The frequency of urinary tractinfections was 5.4% in the total sampleof 3066 infants and 9.7% in the 1666infants who had urine testing at theini-tialvisit. The detection of bacterial men-ingitis in 0.5% of patients is less thanthat in an analysis of 14 studies from

1972 to 1991, in which 0.8% of 1703infants had bacterial meningitis.29 Mostof these studies were conducted in ur-ban emergency departments. More re-cent studies found 2 cases of bacterialmeningitis in a series of 394 (0.5%),19

17 of 5279 (0.3%),25 and 5 of 422(1.2%).6 We detected bacteremia in1.8% of our population of 3066 in-

fants (2.4% of those who hadblood cul-tures performed) compared with.7%19

and 1.2%25 in other contemporary se-ries. An earlier meta-analysis docu-mented a bacteremia/bacterial menin-gitis ratein febrile infants younger than

1 month of 3.7% and 1.6% in infantsaged 1 to 3 months.30 The sample sizeand geographic diversity of this studyhelps provide accurate estimates of therisk of various illnesses. These esti-

mates may be helpful in discussionswith parents about management strat-egies commensurate with the esti-mated level of risk. Along with par-ents individual preferences,31 thesecould enhance collaborative decision

making.We documented a wide array of or-ganisms responsible for bacteremia/bacterial meningitis. Given the low fre-quency of pneumococcal disease,

Table 8.Multivariate Predictors of Bacteremia/Bacterial Meningitis Before Laboratory Testing(n = 3066)

Demographic/ClinicalFeatures

No. ofPatients

UnadjustedOdds Ratio

Adjusted OddsRatio (95% CI)

P

Value

Age, d*30 775 5.72 5.56 (2.50-12.36) .001

31-60 1220 2.55 3.03 (1.35-6.81) .007

Medicaid insured 1074 0.68 0.56 (0.32-0.99) .05

AppearanceModerately ill 767 2.89 1.79 (0.95-3.38) .07

Very ill 50 20.1 8.90 (3.34-23.69) .001

URTI diagnosed 785 0.18 0.27 (0.06-1.15) .08

Ill family member(s) 1512 0.47 0.51 (0.30-0.89) .02

Temperature, C38.5-38.9 1049 2.63 2.37 (1.22-4.63) .01

39.0-39.4 458 2.59 1.84 (0.84-4.37) .12

39.5 198 4.51 3.61 (1.40-9.25) .008

Inner-city clinic 42 1.82 2.23 (0.97-5.13) .06

Abnormal cry 251 5.16 2.23 (1.16-4.29) .02

Abbreviations: CI, confidence interval; URTI, upper respiratory tract infection.

*Comparison group: aged 60 days.Comparison group: well or minimally ill.Comparison group: temperature 38.5C.

Table 9.Multivariate Predictors of Bacteremia, Including Laboratory Data*

Demographic/Clinical Feature Adjusted Odds Ratio (95% CI) PValue

Age, d30 4.03 (1.74-9.37) .001

31-60 2.39 (1.00-5.71) .06

Medicaid insured 0.54 (0.30-0.97) .04

AppearanceModerately ill 1.31 (0.69-2.48) .41

Very ill 5.26 (1.89-14.63) .001

Ill family member(s) 0.61 (0.34-1.08) .09

Temperature, C38.5-38.9 2.03 (1.03-4.02) .04

39.0-39.4 1.79 (0.78-4.09) .17

39.5 2.90 (1.09-7.74) .03

Abnormal cry 2.40 (1.20-4.96) .01Abnormal white blood cellcount

3.62 (2.13-6.15) .001

Abnormal urinalysis result 1.67 (0.90-3.13) .10

Abbreviation: CI, confidence interval.*Includes only those with complete blood cell counts and blood cultures; n = 1746.Comparison group: aged 60 days.Comparison group: well or minimally ill.Comparison group: temperature 38.5C.Defined as 5000/L or 15000/L.Defined as positive or negative esterase or 5 white blood cells per high-powered field.




8/13/2019 Joc 31813

8/10

introduction of the pneumococcal vac-cine subsequent to this study will notappreciably change the epidemiologyreported. Of interest was the fre-quency ofEscherichia coli infections,which was greater than group B -he-

molytic streptococcal infections. Thegreater use of intrapartum antibioticsmay have decreased the risk of groupB -hemolyticStreptococcus and in-creased the risk ofE coli.

This study also provides data on howwell current guidelines would per-form in infants seen in clinicians of-fices. We found that the current guide-

lines are very sensitive in detectingbacteremia/bacterial meningitis. Prac-titioners relying on their clinical judg-mentswere at least as sensitive in treat-ing bacteremia and bacterial meningitis,missing only 2 cases of bacteremia/

bacterial meningitiswhile sparingmanyinfants unnecessary hospitalization andtests. (At least 1 protocol calls for lum-bar puncture in all children, while onlya third ofpatients in thisstudy had lum-bar punctures.) Using recursive parti-tioning analysis, we were able to ap-proximate the performance of currentguidelines without the need for labo-

ratory tests. While this model did notperform quite as well as practitionersin this study, it identified patients atvery low risk of bacteremia/bacterialmeningitis. Only 4 in 1056 infants aged25 days or older who appeared mini-

mally ill and had temperatures of lessthan 38.6C had bacteremia/bacterialmeningitis.Many such infants might bespared unnecessary laboratory test-ing. This is of practical importance insituations in which laboratory testingmight not be immediately available.

The variation of the strategies ofPROS practitioners from establishedguidelines mayreflect the changing fre-quency of serious illnesses accompa-nying febrile illness in infants. With bac-terial meningitis in 0.5% and nobacterial meningitis in more than 1000

infants aged2 to3 months, it isnot sur-prising that clinicians used fewer labo-ratory tests than suggested in pub-lished guidelines.

Also noteworthy is that a majority ofinfants had more than 1 office visit andfrequent telephone contacts. The abil-ity to achieve this level of follow-upwasan important element in this group ofprimary care patients and may alsoexplain the low frequency of hospital-ization. The cost of 2 to 3 follow-up vis-its is substantially less than a hospital-

ization and was safe in this population.Any new guidelines for the manage-ment of fever in infants should con-sider a strategy of watchful waiting withrepeat observations for infants who, inthe judgment of the clinician, can besafely observed at home and have con-tinuing access to care.

Thefollowinglimitations of ourstudyshouldbe noted. While not all febrile in-fants were enrolledduring thestudy pe-riod, infants eligible but not enrolledwere slightly older, suggesting that the

truefrequency of SBI, including bacter-emia/bacterial meningitis, may be lessthan that reported. The distribution ofillness found in the sample of infants islikely representative of infants seen incommunity-based practice but is notbroadly generalizable to infants seen inemergency department settings. Ourmethods differed from most studies by

Figure.Classification Tree for Detecting Bacteremia and Bacterial Meningitis in Infants

ClinicalAssessment

Moderately or Very Ill Well or Minimally Ill

Yes NoAge

8/13/2019 Joc 31813

9/10

including children with normal tem-peratures in the office but febrile athome. However, we documented 6 casesof bacteremia/bacterialmeningitis in thisgroup, suggesting that these infantsshouldnot beignored.Finally, this study

included few African American, His-panic, or inner-city infants. The PROSnetwork has subsequently made effortsto recruit practices to reflect the cur-rent demographic portrait of theUnitedStates.

Whilethis report focuses on the suc-cess of clinicians in addressing 2 of themost serious underlying causes of fe-ver in infancy, we are not suggestingthat the quality of care could not be fur-ther improved. As reported in detailelsewhere,14 urinary tract infections intested infants were documented in 19%

of uncircumcisedboys, 13%of girls,and17% of infants with prolonged illness.Yet only slightly more than 50% of in-fants had a urine test, including uncir-cumcised boys, among whom 41% didnot have urine evaluations. This dis-crepancy represents a potential oppor-tunity to improve practice. However,obtaining urine tests in all infants maynot be necessary; the selective ap-proach by PROS practitioners did notresult in detected adverse outcomes.

In summary, we have documented

strategies for managing fever in in-fants by community practitioners andthe frequency of illnesses diagnosed.The large sample size has allowed us toprecisely assess the frequency and fac-tors associated with high risk of bac-teremia/bacterial meningitis in infants(age30 days, higher temperatures, illappearance, abnormal cry, and abnor-mal WBC count); and we have identi-fied a group with a risk of bacteremia/bacterial meningitis of 0.4% (wellappearing, aged 25 days or older, and

temperature38.6C). Despite lack ofadherence to guidelines, PROS clini-cians detected as many cases of bacter-emia/bacterial meningitis while per-forming fewer tests and hospitalizingfewer infants than would have oc-curred if strictlyadhering to practicepa-rameters. The findings suggest that ifclose follow-up care is attainable, the

management of selected cases by ex-perienced clinicians usingclinical judg-ment may be more appropriate thanstrict adherence to published recom-mendations, with the potential ben-efit of reducing considerable costs and

iatrogenic morbidity. While guide-lines have an important role in ensur-ing the quality of care for many clini-calissues,their performancein complexclinical situations, such as the manage-ment of febrile illnesses, should be ana-lyzed to evaluate whether the guide-lines actually optimize care.

Author Contributions: Dr Pantell had full access toallof thedata in thestudyand takes responsibility forthe integrity of the data and the accuracy of the dataanalysis.Study concept and design:Pantell, Newman, Bernz-weig, Bergman, Wasserman.Acquisition of data:Pantell, Newman, Bernzweig, Berg-man, Takayama, Finch, Wasserman.Analysis and interpretation of data:Pantell, New-man, Takayama, Segal, Finch, Wasserman.Drafting of the manuscript: Pantell, Bernzweig, Berg-man, Segal.Critical revision of the manuscript for important in-tellectual content: Pantell, Newman, Bergman,Takayama, Segal, Finch, Wasserman.Statistical expertise:Newman, Segal.Obtained funding: Pantell, Newman, Bergman,Wasserman.Administrative, technical, or material support:Pan-tell, Bernzweig, Takayama, Finch, Wasserman.Supervision:Pantell, Bernzweig, Wasserman.PROS FebrileInfant Study Sites/Practitioners (byAAPChapter): Alabama: Drs Heilpern & Reynolds PC(Birmingham),GrowingUp PediatricsPC (Birmingham),University of Alabama (Birmingham); Alaska: Anchor-age Neighborhood Health Center (Anchorage), An-chorage Pediatric Group(Anchorage), La Touche Pe-

diatrics (Anchorage), Eielson Clinic(Eielson); Arizona:Mesa Pediatrics Professional Associates (Mesa),Orange GrovePediatrics (Tucson),Tanque VerdePe-diatrics (Tucson), Cigna HealthCare (Tucson); Cali-fornia 1: Palo Alto Medical Clinic (Palo Alto), UCSF-Laurel Heights (San Francisco), Palo Alto MedicalFoundation (Los Altos), Palo Alto Medical Clinic(Fremont), Shasta Community Health Center(Redding), Healthy Trails Pediatric Medical Group(Freedom), Anita Tolentino-Macaraeg, MD (Hollister),Eureka Pediatrics (Eureka), Drs Cantor, Giedt, Kama-chi, & Brennan (Salinas), Marin Community Clinic(Greenbrae); California2: Clinic Sierra Vista(Lamont),Rose City Pediatrics Medical Group(Pasadena), Tou-raj Shafai, MD (Riverside), Facey Medical Group(Northridge); California 4: Edinger Medical GroupInc(Fountain Valley), Southern Orange County Pediat-ric Associates (Lake Forest);Colorado: Rocky Moun-tain Youth Medical Providers PC (Thornton); Con-

necticut: St Francis Pediatric Primary Care Center(Hartford), Hemant Panchal, MD (Enfield), Uwe Ko-epke, MD (Danbury);Delaware: Pediatric Associates(Newark);District of Columbia: George WashingtonUniversity Health Plan; Florida: SawgrassPediatrics PA(Coral Springs), Jonathan Rubin, MD, PA (Margate),MacKoul Pediatric (Cape Coral), SW Florida PediatricNetwork(Fort Meyers), AtlanticCoast Pediatrics (Mer-rittIsland), Orlando Health CareGroup (Orlando), Sa-cred Heart Pediatric Care Center (Pensacola), Gian-greco & Scarano Pediatrics (Bradenton), Emilio DelValle,MD (Fort Myers),Arnold PalmerWomen& Chil-

drens Hospital (Orlando); Georgia: ThePediatric Cen-ter (Stone Mountain), Childrens Hospital at Memo-rial (Savannah);Hawaii: Jeffrey Lim, MD (Honolulu),Melinda Ashton, MD (Honolulu), University of Ha-waii (Honolulu); Illinois: Southwest Pediatrics SC (Or-land Park), SIU Physicians & Surgeons Auburn(Auburn), LaGrangePediatrics(WesternSprings), Sid-ney Smith, MD (Carbondale), Signature Medical As-sociates (Elgin); Indiana: Georgetown Pediatrics

(Indianapolis), Pediatric Advocates (Peru), SouthernIndiana Pediatrics (Bedford), Bloomington Pediatric As-so cia te s P C (B lo o min g to n ), L y n n Ry a n , M D(Lawrenceburg), Marshall County Pediatrics(Plymouth), Jeffersonville Pediatrics (Jeffersonvill e),Childrens Health Care (Batesville), Northpoint Pedi-atrics Fischers (Fischers), Southern Indiana PediatricsL L C (B lo o min g to n ); Iowa: D a v i d K e l l y , M D(Marshalltown), West Des Moines Family Physicians(WestDes Moines);Kansas: Bethel Pediatrics(Newton),Ashley Clinic (Chanute);Kentucky: Pediatric & Ado-lescent Medicine (Lexington); Maine: John Salvato,MD (Waterville), Intermed Pediatrics (Portland);Mary-land: Clinical Associates PA (Towson), Drs Andorsky,Finkelstein, andCardin (Owings Mills), ChildrensMedi-cal Group (Cumberland), Steven Caplan, MD(Baltimore), Shore Pediatrics (Easton), ODonovan &Ahluwalia, MD, PA (Baltimore), Drs Wiczer, Koren-gold, and Mayol (Bethesda), DAlbora & Osha, MD,

PA (Rockville),Shady Side Medical Associates (ShadySide), TheChildrensDoctors(Westminster),Drs Cole-man, Sachs, & Thillairajah (Rockville), Potomac Phy-sicians (Severna Park);Massachusetts: FraminghamPe-diatric PC (Framingham), Garden City Pediatrics(Beverly), Burlington Pediatrics (Burlington), River-bendMedical Group(Chicopee), Holyoke Pediatric As-sociates (Holyoke), John Mulqueen, MD (Gardner),Pediatric Associates of Norwood (Norwood), CapeCodPediatrics (Forestdale), Winthrop Community HealthCenter (Winthrop); Michigan: Botsford Pediatrics(Farmington), H. M. Hildebrandt, MD (Ypsilanti), Es-sexville Medical Clinic (Bay City), Downriver Pediat-ric Associates PC (Lincoln Park), Child Health Associ-ates(Ann Arbor), Pediatric & Family Careof RochesterHills PC (Rochester Hills), Drs Lee & Kim Associates(Warren),OrchardPediatrics (West Bloomfield);Min-nesota: South LakeClinic (Minnetonka);Missouri: Pe-diatric Associates of SW Missouri (Joplin), Childrens

Clinic (Springfield), Doctors Clinic (Carruthersville);Montana: Stevensville Pediatrics (Stevensville);NewHampshire: ExeterPediatricAssociates(Exeter), Lahey-Hitchcock Clinic Concord (Concord), Dartmouth-Hitchcock Clinic (Lebanon), Laconia Clinic (Laconia),Pediatric & Adolescent Medicine (Kingston), Lahey-Hitchcock Clinic Keene (Keene);New Jersey: Kids CarePediatrics (Egg Harbor Township), Salem Road Pedi-atrics (Burlington), Coventry Family Practice (Phill-ipsburg);New Mexico: Albuquerque Pediatric Asso-ciates Ltd (Albuquerque), University of New MexicoHospital (Albuquerque); Nevada: Physicians CenterWest (Fallon), Jobs Peak Primary Care Specialists(Gardnerville); New York1: Panorama Pediatric Group(Rochester), ElmwoodPediatric Group (Rochester), Al-banyMedical CollegePediatricGroup (Albany), South-ern Tier Health Associates (Wayland), Parkway Pedi-atrics (Rochester), LewisPediatrics (Rochester), GayleBuckley, MD (Ballston Lake), Genesee HealthService

(Rochester), Pine Street Pediatric Associates PC(Kin g sto n ), No rth Co u n try Ch ild re n s Cl in ic(Watertown), SpringvillePediatrics (Springville); NewYork2 : Women & ChildrensHealthCenter (Long Is-landCity), GaryMirkin, MD (Great Neck), Southamp-ton Pediatric Associates (Southampton), SoniaVinas,MD (Brooklyn);New York 3: Saint Vincents Pediat-ric Associates (New York); North Carolina: NovantHealth, Eastover Pediatrics (Charlotte), Triangle Pe-diatric Center (Cary),Peace Haven Family Health Cen-ter (Winston-Salem); NorthDakota: Medical Arts Clinic(Minot), Altru Clinic (Grand Forks), Dakota Clinic Ltd




8/13/2019 Joc 31813

10/10

Jamestown( Jamestown); Ohio: Bryan MedicalGroup(Bryan), South Dayton Pediatrics Inc (Dayton), Ox-ford Pediatrics & Adolescents (Oxford), John DiTra-glia (Portsmouth), FamilyHealth Center(Idaho), Ober-lin Clinic (Oberlin), Childrens Hospital PhysiciansAssociates (Twinsburg), North Central Ohio FamilyCare Center Inc (Galion), Drs Harris & Rhodes (Lan-caster);Oklahoma: Pediatric & Adolescent Care LLP(Tulsa), Medical Care Associates of Tulsa (Tulsa), OU

Pediatric Clinic(Tulsa);Oregon: Eugene Clinic(Eugene),NBMC (Coos Bay);Pennsylvania: Pennridge Pediat-ric Associates (Sellersville), Praful Bhatt, MD (LockHaven), Reading Pediatrics Inc (Wyomissing), Chil-drens Health Care (Allentown), Erdenheim Pediat-rics (Flourtown), Yoon-Taek Chun, MD (EastStroudsburg), Pediatric Associates of Plymouth (Ply-mouth Meeting), Plum Pediatrics (Pittsburgh), Ein-stein Community Health Associates (Philadelphia), Ce-vallos andMoise PediatricAssociates PC (Quakertown),Pediatric Group Services (Philadelphia), VNA Kids-Care (Bethlehem), Laurel Health Center (Blossburg);PuertoRico: Edna Zayas,MD (San Juan),PediatricAm-bulatory Clinic (San Juan); Rhode Island: MarvinWasser,MD (Cranston);SouthCarolina: Anderson Pe-diatric Group (Anderson), Grand Strand Pediatrics &Adolescent Medicine (Surfside Beach), Barnwell Pe-diatrics PA (Barnwell); Tennessee: Johnson City Pe-

diatrics PC ( Johnson City); Texas: ThePediatricClinic(Greenville), Winnsboro Pediatrics (Winnsboro), Pe-diatrics (Sherman), Sarah Helfand, MD (Dallas), Cleve-land Pediatric & Adolescent Clinic (Cleveland), UTHealthCenterat Tyler Pediatrics(Tyler), Pediatric Clinic(Mineral Wells),WhiteRock PediatricsPA (Dallas), UN-THSC at Fort Worth Pediatric Clinic (Fort Worth),Fam-ily Medical Center (Big Spring); Utah: Utah ValleyPe-diatricsLC (AmericanFork), Mountain ViewPediatrics

(Sandy),Granger MedicalCenter (Salt Lake City), Wil-low Creek Pediatrics (Salt Lake City), John Weipert,MD (American Fork), University of Utah Health Sci-ences Center (SaltLake City); Vermont: University Pe-diatrics, UHC Campus (Burlington), Pediatric Medi-cine(SouthBurlington), Timber LanePediatrics(SouthBurlington), Hagan & Rinehart Pediatrics (SouthBurlington), BrattleboroPediatrics(Brattleboro), Uni-versity Pediatrics Williston Office (Williston), Re-becca Collman, MD (Colchester), Mousetrap Pediat-rics (Milton), Green Mountain Pediatrics PC(Bennington), St Johnsbury Pediatrics (St Johnsbury);Virginia: Pediatric Association of Richmond Inc(Richmond), Alexandria Lakeridge Pediatrics(Alexandria), Drs Casey, Goldman, Lischwe, Garrett& Kim (Arlington), Fishing Bay Family Practice(Deltaville), Stafford Pediatrics PC (Stafford), Pediat-ric Clinic (Arlington); Washington: Valley Childrens

Clinic (Renton), Rockwood Clinic (Spokane), YakimaValley FarmWorkersClinic (Toppenish), PaulousePe-diatrics (Pullman), Columbia Health Center (Seattle);West Military: Wilford Hall Medical Center (Lack-land Air Force Base); West Virginia: Grant MemorialPediatrics (Petersburg), Tess Alejo, MD (Martins-burg);Wisconsin: Beloit Clinic SC (Beloit), LutheranHospital (La Crosse), Waukesha Pediatric Associates(Waukesha), CHW Pediatric Clinic (Milwaukee), Au-

rora Health Center Waukesha (Waukesha); Wyo-ming: Jackson Pediatrics ( Jackson), Cheyenne Chil-drens Clinic (Cheyenne).Funding/Support: Thisstudy was supported by grantR01 HS06485 from the Agency for Healthcare Re-search and Quality, with additional support from theHealth Resources and Services Administration Mater-nal and Child Health Bureau (grant MCJ-177022).Role of theSponsor: The studysfunding agencies hadno role in the design and conduct of the study, in thecollection, analysis, andinterpretation of the data, andin the preparation, review, or approval of the manu-script.Acknowledgment: We thank G. Mark Spitalny andMichelle Mayer, PhD,for theircontributionsand dataanalyses, and Jay Tureen, MD, who provided infec-tious diseaseconsultationon thepathogenicity of or-ganisms.

REFERENCES1. RobertsKB, Borzy MS.Feverin thefirst eight weeksof life.Johns Hopkins Med J.1977;141:9-13.2. Anbar RD, Richardson de Corral V, OMalley PJ.Difficulties in universal application of criteria identi-fying infants at low risk for serious bacterial infec-tion.J Pediatr.1986;109:483-485.3. Dagan R, PowellKR, Hall CB,Menegus MA.Iden-tification of infants unlikely to have serious bacterialinfection although hospitalized for suspected sepsis.J Pediatr. 1985;107:855-860.4. Dagan R, Sofer S, Phillip M, Shachak E. Ambula-tory care of febrile infants younger than 2 months ofage classified as being at low risk for having seriousbacterial infections.J Pediatr.1988;112:355-360.5. BaraffLJ, Bass JW,Fleisher GR,et al.Practice guide-line for the management of infants and children 0 to36 months of age with fever without source. AnnEmerg Med.1993;22:1198-1210.

6. Baker MD, Bell LM, Avner JR. The efficacy of rou-tine outpatient management without antibiotics of fe-ver in selected infants.Pediatrics.1999;103:627-631.7. Avner JR, Baker MD. Management of fever in in-fants and children. Emerg MedClin North Am.2002;20:49-67.8. Baker MD,AvnerJR, Bell LM.Failure of infantob-servation scales in detecting serious illness in febrile,4- to 8-week-old infants. Pediatrics.1990;85:1040-1043.9. Kramer MS,ShapiroED. Management of theyoungfebrile child: a commentary on recent practice guide-lines.Pediatrics.1997;100:128-134.10. DeAngelis C, Joffe A, Wilson M, Willis E. Iatro-genic risks and financial costs of hospitalizing febrileinfants.AJDC.1983;137:1146-1149.11. YoungPC. Themanagementof febrileinfantsbyprimary-care pediatricians in Utah: comparison with

published practice guidelines.Pediatrics. 1995;95:623-627.12. McCarthy PL, Jekel JF, Stashwick CA, et al. Fur-therdefinition of history andobservation variablesin as-sessing febrile children.Pediatrics.1981;67:687-693.13. McCarthy PL, Lembo RM, Fink HD, Baron MA,Cicchetti DV. Observation, history, and physical ex-amination in diagnosis of seriousillnessesin febrilechil-dren less than or equal to 24 months.J Pediatr. 1987;110:26-30.14. Newman TB, Bernzweig JA, Takayama JI, FinchSA,Wasserman RC,Pantell RH.Urinetesting anduri-narytract infections in febrile infantsseen in office set-tings: the Pediatric Research in Office Settings Fe-brile Infant Study. Arch Pediatr Adolesc Med.2002;156:44-54.15. Breiman L, Friedman JH, Olshen RA, Stone CJ.Classification and Regression Trees. Belmont, Calif:

Wadsworth; 1984.16. Segal MR. Extending the elements of tree-structured regression.Stat Methods Med Res.1995;4:219-236.17. Baraff LJ. Management of fever without sourcein infants and children. Ann Emerg Med. 2000;36:602-614.18. Baskin MN, ORourke EJ, Fleisher GR. Outpa-tient treatmentof febrileinfants 28 to 89 days of agewith intramuscular administration of ceftriaxone.J Pe-diatr.1992;120:22-27.19. Herr SM, Wald ER, Pitetti RD, Choi SS. En-hanced urinalysisimproves identification of febrile in-fants ages 60 days and younger at low risk for seri-ous bacterial illness.Pediatrics.2001;108:866-871.20. Feder HM, Jr Occult pneumococcal bacteremiaand the febrile infant and young child. Clin Pediatr(Phila).1980;19:457-462.

21. Gotoff SP, Behrman RE. Neonatal septicemia.J Pediatr. 1970;76:142-153.22. Bratton L, Teele DW, Klein JO. Outcome of un-suspected pneumococcemia in childrennot initiallyad-mitted to the hospital.J Pediatr.1977;90:703-706.23. Baker MD, Bell LM. Unpredictability of seriousbacterial illness in febrile infants from birth to 1month of age. Arch Pediatr Adolesc Med.1999;153:508-511.24. Kadish HA, Loveridge B, Tobey J, Bolte RG, Cor-neli HM. Applying outpatient protocols in febrile in-fants 1-28 days of age: can the threshold be low-ered?Clin Pediatr (Phila).2000;39:81-88.25. Bachur RG, Harper MB. Predictive model for se-riousbacterialinfectionsamong infants younger than3 months of age. Pediatrics.2001;108:311-316.26. Kuppermann N. Diagnostic testing of the febrileneonate: it is time to collaborate. Arch Pediatr Ado-

lesc Med.2003;157:508-509.27. Jones RG, Bass JW. Febrile children with no fo-cusof infection: a surveyof their management by pri-mary care physicians.Pediatr Infect Dis J.1993;12:179-183.28. DeAngelis C, Joffe A, Willis E, Wilson M. Hospi-talization v outpatient treatmentof young,febrile in-fants.AJDC.1983;137:1150-1152.29. Baraff LJ, Oslund SA, Schriger DL, Stephen ML.Probability of bacterial infectionsin febrile infants lessthanthree months of age:a meta-analysis. Pediatr In-fect Dis J.1992;11:257-264.30. Baskin MN. The prevalence of serious bacterialinfections by age in febrile infants during the first 3months of life.Pediatr Ann.1993;22:462-466.31. Bennett JE,Sumner W 2nd,Downs SM, JaffeDM.Parents utilities for outcomes of occult bacteremia.Arch Pediatr Adolesc Med. 2000;154:43-48.



Documents

Joc 31813