Professional Documents
Culture Documents
The online version of this article, along with updated information and services, is
located on the World Wide Web at:
http://www.pediatrics.org/cgi/content/full/126/6/1074
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
Febrile Infants With Urinary Tract Infections at Very
Low Risk for Adverse Events and Bacteremia
WHAT’S KNOWN ON THIS SUBJECT: Febrile infants aged 29 to 60 AUTHORS: David Schnadower, MD, MPH,a Nathan
days who have urinary tract infections are typically hospitalized Kuppermann, MD, MPH,b Charles G. Macias, MD, MPH,c
for ⱖ48 hours. Previous study results have indicated that most Stephen B. Freedman, MD,d Marc N. Baskin, MD,e Paul
patients have benign clinical courses, but small sample sizes Ishimine, MD,f Camille Scribner, MD,g Pamela Okada, MD,h
Heather Beach, MD,i Blake Bulloch, MD,j Dewesh Agrawal,
have limited the ability to identify those at near-zero risk of
MD,k Mary Saunders, MD,l Donna M. Sutherland, MD,m
adverse events. Mercedes M. Blackstone, MD,n Amit Sarnaik, MD,o Julie
McManemy, MD,p Alison Brent, MD,q Jonathan Bennett,
WHAT THIS STUDY ADDS: We derived prediction models in a MD,r Jennifer M. Plymale, MDb Patrick Solari, MD,s
large sample of patients that identify infants at very low risk for Deborah J. Mann, MD,t and Peter S. Dayan, MD, MSca for
adverse events and at low risk for bacteremia. Shorter the American Academy of Pediatrics Pediatric Emergency
hospitalization or outpatient treatment with close follow-up may Medicine Collaborative Research Committee
be feasible for selected patients. aPediatric Emergency Medicine, Morgan Stanley Children’s
1074 SCHNADOWER et al
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
ARTICLES
Urinary tract infection (UTI) is the most ment (ED), and to derive clinical pre- ciation with positive urinalysis
common serious bacterial illness in fe- diction models for infants at very low results.14
brile infants younger than 60 days of risk of adverse events and bacteremia. We defined positive urinalysis results
age, occurring in 4% to 10% of these These data may help clinicians make as those that met any of the following:
infants.1– 6 In 1999, an American Acad- more informed decisions regarding (a) any organisms visualized on Gram-
emy of Pediatrics guideline recom- the need for, or length of hospitaliza- stain; (b) trace or greater result for
mended outpatient management with tion of these infants. leukocyte esterase or nitrite on ED dip-
oral or parenteral antibiotics for chil- stick or laboratory-based urinalysis;
dren older than 2 months who have METHODS or (c) ⱖ5 white blood cells (WBCs) per
UTIs unless the child seemed “toxic, de- Study Design and Setting high-power field (standard micros-
hydrated, or unable to take oral in- copy) or per L (hemocytometer) on a
take,” in which case hospitalization We performed a retrospective chart
centrifuged or uncentrifuged urine
review at 20 medical centers that par-
was recommended.7 There is little in- specimen.15–17
ticipated in the Pediatric Emergency
formation however, to guide the man-
Medicine Collaborative Research Com-
agement of infants aged 29 to 60 Exclusion Criteria
mittee of the American Academy of Pe-
days.8,9 Although a recent clinical trial We excluded patients for any of the fol-
diatrics. The participating centers in-
found it feasible to treat febrile infants lowing: (a) transfer from other hospi-
cluded 16 tertiary care pediatric EDs
with UTIs as outpatients with paren- tals with previously obtained labora-
and 3 general EDs in the United States
teral antibiotics administered in an in- tory results; (b) urine specimens
and 1 Canadian tertiary care ED. Ap-
fusion center,8 most clinicians hospi- obtained by techniques other than su-
proval for the study with waiver of in-
talize these infants for ⱖ48 hours of prapubic aspiration or transurethral
formed consent and for data sharing
parenteral therapy.10–12 catheterization; (c) urine cultures that
with the coordinating institution and
There are several reasons why clini- with the centralized data center was grew multiple organisms; or (d) no
cians hospitalize infants aged 29 to 60 granted by the institutional review measured temperature of ⱖ38.0°C in
days of age who have UTIs, including board at each participating institution. the ED or at home within 24 hours of ED
the unclear risk of long-term renal in- presentation. We excluded patients
jury, and concern regarding follow-up. Patient Identification without a measured fever because we
The most important reasons, however, were interested in patients with a high
We performed case ascertainment by
are the concerns of acute adverse likelihood of true UTIs, not those with
querying laboratory databases for all
events and for missing concomitant asymptomatic bacteriuria.
urine cultures with bacterial growth in
bacteremia. Several small studies
patients aged 29 to 60 days that were Data Collection and Potential Risk
have assessed the course of febrile in-
obtained in the ED between January Factors
fants with UTIs and suggest that other-
1995 and May 2006. Lactobacillus, Mi-
wise well-appearing infants with or All coinvestigators were trained by the
crococcus, diptheroids, Bacillus spe-
without concomitant bacteremia have study’s principal investigator in per-
cies and Staphylococcus epidermidis
benign clinical courses when treated son or during conference calls. Each of
were considered contaminants.
with appropriate antibiotics.9–13 Be- the coinvestigators reviewed the med-
cause of small sample sizes, investiga- Inclusion Criteria ical charts for all study patients at
tors have neither been able to provide their site. Study data were entered
a precise estimate of the risk of ad- Children were classified as having a
onto computerized PDF documents
verse events or bacteremia nor derive UTI if urine cultures grew a single
(“teleforms”) (Adobe Acrobat 8 Profes-
potential multivariable models to iden- pathogen and colony counts met at
sional, San Jose, CA) that were pro-
least 1 of 3 criteria:
tify risk factors for adverse events or grammed to improve data accuracy
bacteremia in young, febrile infants 1. ⱖ1000 colony-forming units (CFU)/mL and completeness. The teleforms were
with UTIs. for urine cultures obtained by su- then uploaded electronically or by fax
prapubic aspiration; into a central database to avoid sec-
The aims of this study were to deter-
mine the risk of adverse events and 2. ⱖ50 000 CFU/mL from a catheter- ondary transcription errors.
bacteremia in a large sample of febrile ized specimen; or We collected data on patient demo-
infants aged 29 to 60 days with UTIs 3. ⱖ10 000 and ⬍50 000 CFU/mL from graphics, past medical history, pre-
who present to the emergency depart- a catheterized specimen in asso- senting symptoms in the ED (including
1076 SCHNADOWER et al
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
ARTICLES
1078 SCHNADOWER et al
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
ARTICLES
electrolyte abnormalities.
68.6%– 84.1%), and the negative
predictive value was 96.8% (95%
nosed immediately in the ED in 26 of 51 Prediction Models for Adverse CI: 95.3%–97.8%). Bacteremia was
(51.0%) patients, within 4 hours of ED Events and Bacteremia present in 28 of 862 patients (3.2%
presentation in 10 of 51 (19.6%), be- [95% CI: 2.2%– 4.6%]) who met all the
Adverse Events low risk criteria. None of these 28 low-
tween 4 and 23 hours in 11 of 51
(21.6%), between 24 and 47 hours in 2 Recursive partitioning analysis identi- risk infants had an adverse event.
of 51 (3.9%), and after ⱖ48 hours in 2 fied a group of patients at very low risk Finally, Tables 3 and 4 demonstrate the
of 51 (3.9%). of adverse events (Fig 2). Patients be- unadjusted and adjusted odds ratios
longed to the very low risk group if they for predictor variables in the 2 analy-
Risk of Bacteremia were not clinically ill on ED examina- ses. The presence of clinical illness in
Bacteremia was present in 123 of 1877 tion (well-appearing, not dehydrated, the ED was the variable most strongly
infants from whom blood cultures not in respiratory distress, and no con- associated with adverse events,
were obtained (6.5% [95% CI: 5.5% of comitant acute disease) and did not whereas total peripheral band count
7.7%]) (Table 1). The time from obtain- have a high-risk past medical history. ⱖ1250 cells per L and peripheral
ing the blood culture to the time that The prediction model sensitivity for ad- ANC of ⬍1500 cells per L were most
the blood culture was noted to be pos- verse events was 98.0% (95% CI: highly associated with bacteremia.
itive was available for 91 of 123 pa- 88.2%–99.9%), and the negative pre-
tients (74%). The median time to posi- dictive value was 99.9% (95% CI: DISCUSSION
tivity was 16 hours (interquartile 99.5%–100%). Of 1206 infants in the In this large multicenter study, we de-
range: 13–24 hours); 80 of 91 (88%) very low risk group (65.5% of those an- rived a clinically sensible and parsimo-
were positive within 24 hours. Ten of alyzed), only 1 infant had an adverse nious prediction model that identifies
123 patients with bacteremia had event (0.1% [95% CI: 0%– 0.4%]): a 46- a group of infants aged 29 to 60 days
adverse events (8.1% [95% CI: 4.2%– day-old boy who was well-appearing with fever and UTIs who are at very low
13.8%]), whereas 41 of 1754 patients on physical examination in the ED but risk of developing adverse events
without bacteremia had adverse whose initial CSF studies were lost. A when treated with antibiotics. The
events (2.3% [95% CI: 1.7%–3.1%). subsequent lumbar puncture 24 hours model accurately predicts with very
1080 SCHNADOWER et al
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
ARTICLES
All infants with UTIs with a accurately identify infants at very low
blood culture performed
N = 1877 risk for adverse events did decrease
the likelihood of bacteremia to 3.2%.
123/1877 bacteremia (6.6%) Previous research has noted an asso-
1754/1877 no bacteremia (93.4%)
ciation between high peripheral band
counts or low peripheral ANC with
Clinically ill in the ED?a bacteremia caused by Gram-negative
No Yes organisms.12,23,26,27
87/1488 bacteremia (5.8%) 36/389 bacteremia (9.3%) Our study had several limitations typi-
1401/1488 no bacteremia (94.2%) 353/389 no bacteremia (90.7%)
cal of medical chart reviews. These in-
cluded potential biases in data ab-
High-risk PMH?b straction, particularly of subjective
No Yes
clinical findings. We attempted to min-
65/1237 bacteremia (5.3%) 22/251 bacteremia (8.8%)
imize these biases by creating a de-
1172/1237 no bacteremia (94.7%) 229/251 no bacteremia (91.2%) tailed manual of operations that in-
cluded specific key words to interpret
subjective findings, by conducting in-
Bands >1250 cells per µL?
terrater reliability analyses of subjec-
No Yes tive variables, and by using more ob-
31/900 bacteremia (3.4%) 34/337 bacteremia (10.1%) jective criteria to define adverse
869/900 no bacteremia (96.6%) 303/337 no bacteremia (89.9%)
events. These efforts would not ac-
count for clinician documentation bi-
ANC <1500 cells per µL? ased by previous knowledge of labora-
No Yes tory results or clinical course.
28/862 bacteremia (3.2%) 3/38 bacteremia (7.9%) However, clinicians in the ED would be
834/862 no bacteremia (96.8%) 35/38 no bacteremia (92.1%) unlikely to know of future clinical dete-
rioration before initial documentation
Bacteremia No Bacteremia Total in the medical chart.
Any predictor 95 920 1015
present
We were also challenged to define bac-
No predictor present 28 834 862 terial meningitis for patients with neg-
Total 123 1754 1877 ative CSF cultures. We used conserva-
Test characteristics of prediction model tive definitions for probable bacterial
Sensitivity: 77.2% (95% CI: 68.6-84.1%) meningitis to avoid missing patients
Specificity: 47.6% (95% CI: 45.2-50.0%)
NPV: 96.8% (95% CI: 95.3-97.8%) with this important outcome. It should
PPV: 9.4% (95% Cl: 7.7-11.4%) also be recognized that we queried mi-
LR negative: 0.48 (95% Cl: 0.34, 0.66)
crobiology databases rather than at-
FIGURE 3
Prediction model to identify infants aged 29 to 60 days with febrile UTIs at low risk of bacteremia.
tempting to identify patients who pre-
a Clinically ill in the ED, defined as ill-appearing, dehydrated, or in respiratory distress or presence of sented to the ED with positive
an acute concomitant disease diagnosed in the ED; b high-risk past medical history, defined as urinalyses, which was not feasible.
genitourinary abnormalities, previous UTI, bacteremia, meningitis, previous laboratory evaluation for
fever, prematurity (⬍37 weeks’ gestation), or severe systemic disease (complex heart, chronic lung, Therefore, our results are applicable
metabolic, or neurologic diseases). PMH indicates past medical history. to those patients for whom urine cul-
ture results are known and cannot
necessarily be extrapolated to those
for whom only preliminary screening
TABLE 3 Bivariate and Multivariate Adjusted Odds Ratios for Predictors of Adverse Events in the tests for UTI are known.
Recursive Partitioning Analysis
Variable Unadjusted Odds P Adjusted Odds Ratio P CONCLUSIONS
Ratio (95% CI) (95% CI)
Clinically ill in ED 15.5 (7.8–30.4) ⬍.001 15.6 (7.9–30.9) ⬍.001 We derived a highly accurate predic-
High-risk past medical 4.0 (2.3–7.1) ⬍.001 4.3 (2.4–7.8) ⬍.001 tion model that identifies a group of
history febrile infants aged 29 to 60 days with
UTIs at very low risk for adverse apy and brief hospitalization (eg, 24 ACKNOWLEDGMENTS
events. We attempted but were unsuc- hours), within which time frame most The authors would like to thank John
cessful in deriving a very low risk bacteremia will be identified, seems Kanegaye, MD, Ma Long and Achilles
model to identify infants who had bac- appropriate management for this Kalnoky, MD, for their participation
teremia. Initiating antimicrobial ther- group of infants. Outpatient manage- and efforts in this study.
REFERENCES
1. Baker MD, Bell LM, Avner JR. The efficacy of nary tract infections in young febrile chil- brospinal fluid profiles of children with
routine outpatient management without dren. Pediatrics. 1999;104(1 pt 1):79 – 86 bacterial meningitis. Pediatrics. 2008;
antibiotics of fever in selected infants. Pedi- 10. Dayan PS, Hanson E, Bennett JE, Langsam D, 122(4):726 –730
atrics. 1999;103(3):627– 631 Miller SZ. Clinical course of urinary tract 19. Nigrovic LE, Kuppermann N, Macias CG, et al.
2. Baker MD, Bell LM, Avner JR. Outpatient infections in infants younger than 60 days of Clinical prediction rule for identifying chil-
management without antibiotics of fever in age. Pediatr Emerg Care. 2004;20(2):85– 88 dren with cerebrospinal fluid pleocytosis at
selected infants. N Engl J Med. 1993; 11. Bachur R. Nonresponders: prolonged fever very low risk of bacterial meningitis. JAMA.
329(20):1437–1441 among infants with urinary tract infections. 2007;297(1):52– 60
3. Bonadio WA, Smith DS, Sabnis S. The clinical Pediatrics. 2000;105(5). Available at: www. 20. Shah SS, Zorc JJ, Levine DA, Platt SL, Kupper-
characteristics and infectious outcomes of pediatrics.org/cgi/content/full/105/6/e59 mann N. Sterile cerebrospinal fluid pleocy-
febrile infants aged 8 to 12 weeks. Clin Pe- 12. Bachur R, Caputo GL. Bacteremia and meningitis tosis in young infants with urinary tract in-
diatr (Phila). 1994;33(2):95–99 among infants with urinary tract infections. Pe- fections. J Pediatr. 2008;153(2):290 –292
4. Hoberman A, Chao HP, Keller DM, Hickey R, diatr Emerg Care. 1995;11(5):280–284 21. DeAngelis C, Joffe A, Wilson M, Willis E. Iatro-
Davis HW, Ellis D. Prevalence of urinary tract 13. Honkinen O, Jahnukainen T, Mertsola J, Es- genic risks and financial costs of hospitaliz-
infection in febrile infants. J Pediatr. 1993; kola J, Ruuskanen O. Bacteremic urinary ing febrile infants. Am J Dis Child. 1983;
123(1):17–23 tract infection in children. Pediatr Infect Dis 137(12):1146 –1149
5. Shaw KN, Gorelick M, McGowan KL, Yakscoe J. 2000;19(7):630 – 634 22. Crain EF, Gershel JC. Urinary tract infections in
NM, Schwartz JS. Prevalence of urinary 14. Zorc JJ, Levine DA, Platt SL, et al. Clinical and febrile infants younger than 8 weeks of age. Pe-
tract infection in febrile young children in demographic factors associated with uri- diatrics. 1990;86(3):363–367
the emergency department. Pediatrics. nary tract infection in young febrile infants. 23. Bonadio WA, Smith DS, Madagame E, Machi
1998;102(2). Available at: www.pediatrics. Pediatrics. 2005;116(3):644 – 648 J, Kini N. Escherichia coli bacteremia in
org/cgi/content/full/102/2/e16 15. Gorelick MH, Shaw KN. Screening tests for children: a review of 91 cases in 10 years.
6. ZorcJJ,KiddooDA,ShawKN.Diagnosisandman- urinary tract infection in children: a meta- Am J Dis Child. 1991;145(6):671– 674
agement of pediatric urinary tract infections. analysis. Pediatrics. 1999;104(5). Available 24. Ginsburg CM, McCracken GH Jr. Urinary
Clin Microbiol Rev. 2005;18(2):417–422 at: www.pediatrics.org/cgi/content/full/ tract infections in young infants. Pediatrics.
7. American Academy of Pediatrics, Commit- 104/5/e54 1982;69(4):409 – 412
tee on Quality Improvement and Subcom- 16. Shaw KN, McGowan KL, Gorelick MH, 25. Kaplan RL, Harper MB, Baskin MN, Macone
mittee on Urinary Tract Infection. Practice Schwartz JS. Screening for urinary tract in- AB, Mandl KD. Time to detection of positive
parameter: the diagnosis, treatment, and fection in infants in the emergency depart- cultures in 28- to 90-day-old febrile infants.
evaluation of the initial urinary tract infec- ment: which test is best? Pediatrics. 1998; Pediatrics. 2000;106(6). Available at: www.
tion in febrile infants and young children. 101(6). Available at: www.pediatrics.org/ pediatrics.org/cgi/content/full/106/6/e74
Pediatrics. 1999;103(4 pt 1):843– 852 cgi/content/full/101/6/e1 26. Bonadio WA, Smith DS, Mathews S, Rock A. Clini-
8. Doré-Bergeron MJ, Gauthier M, Chevalier I, 17. Herr SM, Wald ER, Pitetti RD, Choi SS. En- cal significance of newly documented neutrope-
McManus B, Tapiero B, Lebrun S. Urinary hanced urinalysis improves identification nia in febrile young infants evaluated for sepsis.
tract infections in 1- to 3-month-old infants: of febrile infants ages 60 days and younger Pediatr Infect Dis J. 1991;10(5):407–408
ambulatory treatment with intravenous an- at low risk for serious bacterial illness. Pe- 27. Paul M, Gafter-Gvili A, Leibovici L, et al. The epide-
tibiotics. Pediatrics. 2009;124(1):16 –22 diatrics. 2001;108(4):866 – 871 miology of bacteremia with febrile neutropenia:
9. Hoberman A, Wald ER, Hickey RW, et al. Oral 18. Nigrovic LE, Malley R, Macias CG, et al. Ef- experience from a single center, 1988–2004. Isr
versus initial intravenous therapy for uri- fect of antibiotic pretreatment on cere- Med Assoc J. 2007;9(6):424–429
1082 SCHNADOWER et al
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010
ARTICLES
Downloaded from www.pediatrics.org. Provided by University of Pennsylvania Library on December 29, 2010