Variation in Care of the Febrile Young Infant <90 Days in US Pediatric Emergency Departments
BACKGROUND AND OBJECTIVES: Variation in patient care or outcomes may indicate an opportunity to improve quality of care. We evaluated the variation in testing, treatment, hospitalization rates, and outcomes of febrile young infants in US pediatric emergency departments (EDs).
METHODS: Retrospective cohort study of infants <90 days of age with a diagnosis code of fever who were evaluated in 1 of 37 pediatric EDs between July 1, 2011 and June 30, 2013. We assessed patient- and hospital-level variation in testing, treatment, and disposition for patients in 3 distinct age groups: ≤28, 29 to 56, and 57 to 89 days. We also compared interhospital variation for 3-day revisits and revisits resulting in hospitalization.
RESULTS: We identified 35 070 ED visits that met inclusion criteria. The proportion of patients who underwent comprehensive evaluation, defined as urine, serum, and cerebrospinal fluid testing, decreased with increasing patient age: 72.0% (95% confidence interval [CI], 71.0–73.0) of neonates ≤28 days, 49.0% (95% CI, 48.2–49.8) of infants 29 to 56 days, and 13.1% (95% CI, 12.5–13.6) of infants 57 to 89 days. Significant interhospital variation was demonstrated in testing, treatment, and hospitalization rates overall and across all 3 age groups, with little interhospital variation in outcomes. Hospitalization rate in the overall cohort did not correlate with 3-day revisits (R2 = 0.10, P = .06) or revisits resulting in hospitalization (R2 = 0.08, P = .09).
CONCLUSIONS: Substantial patient- and hospital-level variation was observed in the ED management of the febrile young infant, without concomitant differences in outcomes. Strategies to understand and address the modifiable sources of variation are needed.
- CBC —
- complete blood count
- CI —
- confidence interval
- CSF —
- cerebrospinal fluid
- ED —
- emergency department
- HSV —
- herpes simplex virus
- ICD-9 —
- International Classification of Diseases, Ninth Revision
- PHIS —
- Pediatric Health Information System
- SBI —
- serious bacterial infection
- UTI —
- urinary tract infection
What’s Known on This Subject:
Various low-risk criteria have been developed to guide management of the febrile young infant (<90 days), but they differ in age criteria, recommendations, and implementation. Therefore, variation in care is likely but has not been previously studied.
What This Study Adds:
There is wide variation in testing, treatment, and overall resource utilization in management of the febrile young infant across all 3 age groups: ≤28, 29 to 56, and 57 to 89 days. There may be opportunities to improve care variation without compromising outcomes.
Fever is one of the most common reasons for emergency department (ED) evaluation of young infants.1 Because of an immature immune system,2 vaccine naivety,3 and exposure to unique pathogens in the perinatal period,4 the febrile young infant is at high risk for serious bacterial infection (SBI), such as urinary tract infection (UTI),5 bacteremia, and bacterial meningitis. The prevalence of SBI in febrile young infants ranges from 8% to 12.5%,6 higher (up to 20%) in neonates ≤28 days old.7 Clinical scoring systems based on historical and physical examination findings8 do not reliably differentiate infants with and without SBI.9 As a result, febrile infants often undergo additional diagnostic evaluation consisting of a combination of urine, blood, and cerebrospinal fluid (CSF) testing.10–12 Criteria have been developed that incorporate the results of this diagnostic testing to identify febrile infants at low risk for SBI, but these criteria were developed >20 years ago and differ in age cutoffs and recommendations on extent of testing, antibiotic therapy, and need for hospitalization.10–12
Despite guidelines for the evaluation and management of the febrile young infant,13 there is variable implementation of these guidelines in practice, which leads to substantial variations in care.14–16 When surveyed, less than half of pediatric emergency medicine fellowship directors reported compliance with the published guidelines for management of febrile infants <8 weeks old.15 Similarly, significant variation was observed in rates of lumbar puncture, chest radiograph, and antibiotic use in febrile infants ≤90 days of age managed in 6 Canadian pediatric EDs.16 Even more striking, only two-thirds of febrile neonates ≤28 days old, the age group at highest risk for SBI, received recommended ED management in a recent multicenter study.14
Identifying practice variation patterns and subsequent effects on outcomes will allow hospitals and national organizations to tailor efforts to increase adherence to evidence and improve the delivery of care.17 Our objectives were to describe the variation in patient- and hospital-level ED testing, treatment, hospitalization rates, and outcomes of febrile young infants <90 days old among a large sample of pediatric EDs, evaluate patterns of variation across 3 age groups (≤28 days of age, 29–56 days, and 57–89 days), and determine the relationship between hospital-specific admission rate and the rate of return visits to the ED.
For this retrospective cohort study, we used the Pediatric Health Information System (PHIS), an administrative database that contains inpatient, ED, ambulatory surgery, and observation data from 44 children’s hospitals in the United States that are affiliated with the Children’s Hospital Association (Overland Park, KS). Contributing hospitals are located in 26 states and the District of Columbia and represent 85% of freestanding children’s hospitals in the United States.18 Data are deidentified before inclusion in the database; encrypted unique patient identifiers permit tracking of patients across ED revisits and subsequent hospitalizations.19 The Children’s Hospital Association and participating hospitals jointly ensure the quality and integrity of the data.20 For this study, we included data from 37 EDs after excluding 7 hospitals because of incomplete data or known data quality problems.21
The study was determined to not involve human subjects by the Yale University Human Investigations Committee and was exempt from review.
Infants <90 days of age were eligible for inclusion if they were discharged from a participating hospital between July 1, 2011 and June 30, 2013. Subjects were included if they were assigned 1 of the following International Classification of Diseases, Ninth Revision (ICD-9) admission or discharge diagnosis codes for fever: 780.6 (fever and other physiologic disturbances of temperature regulation), 778.4 (other disturbances of temperature regulation of newborn), 780.60 (fever, unspecified), or 780.61 (fever presenting with conditions classified elsewhere).14 We excluded infants with a complex chronic condition,22 because these infants may warrant nonstandard evaluation, and those transferred from another institution because we could not identify testing and treatment performed at nonparticipating hospitals. For infants with >1 ED visit for fever during the study period, repeat visits within 3 days of an index visit were considered revisits; visits >3 days after an index visit were considered as a new episode of illness (new index visit).
Demographic characteristics included gender, race or ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, Asian, or other), insurance status (commercial, government, or other), and region where the hospital was located, using US Census categories (Northeast, South, Midwest, West).
We determined whether the following diagnostic testing was performed in the ED using the associated billing codes: Urine testing included performance of urine dipstick, urinalysis, or urine culture; blood testing included complete blood count (CBC) or blood culture; and CSF testing included CSF cell count, culture, or a procedure code for performance of a lumbar puncture.23 Performance of chest radiograph was ascertained by using radiology billing codes. Antimicrobial treatment strategies were determined and assessed separately for hospitalized and discharged patients. For hospitalized patients, we considered any testing or treatment occurring on hospital days 0 or 1 to be performed in the ED, because billing code data in PHIS do not distinguish testing performed in the ED from the inpatient setting.14
An admission to an inpatient unit or observation status was considered a hospitalization.24,25 For patients discharged from the ED, we defined an ED revisit as a return visit within 3 days of ED discharge26,27; these patients were further categorized based on whether they were hospitalized or discharged on their return visit.
We identified the presence of an SBI using the following ICD-9 diagnosis codes: UTI or pyelonephritis (590.1, 590.2, 590.8, 599.0, and 771.82),28 bacteremia or sepsis (771.81, 771.83, 790.7, 995.91, 995.92, 785.52, and 038.xx), bacterial meningitis (036.0, 036.1, 320.0, 320.1, 320.2, 320.3, 320.7, 320.81, 320.82, 320.89, or 320.9),20 pneumonia (480.0–480.2, 480.8, 480.9, 481, 482.0, 482.30–482.32, 482.41, 482.42, 482.83, 482.89, 482.90, 483.8, 484.3, 485, 486, and 487.0),29 or bacterial enteritis (003.0, 004.xx, 008.0x–008.4x, and 008.5). Herpes simplex virus (HSV) infection was identified by the ICD-9 discharge diagnosis code 054.xx.30
We evaluated patient- and hospital-level variation in testing, treatment, and disposition for all infants and for 3 distinct age groups: ≤ 28 days, 29–56 days, and 57–89 days of age. We also assessed 3-day revisits and 3-day revisits that resulted in hospitalization.
Continuous variables were described using median and interquartile range or range values. Categorical variables were described using frequencies with 95% confidence intervals (CIs) and compared using the Kruskal–Wallis 1-way analysis of variance test because of the nonnormal distribution. The degree of hospital-level variation by age category was assessed using χ2 analysis; the magnitude of the variation is represented by the absolute χ2 value (higher χ2 value indicates greater variation). To address potential documentation bias related to the omission of fever as a diagnosis among infants with SBI, we repeated the analyses and included any infant with an ICD-9 diagnosis of SBI regardless of whether he or she had an ICD-9 code for fever.
To assess overall hospital-level variation, rates of utilization for laboratory testing (ie, urine, blood, and CSF) and hospitalization were calculated for each institution and ranked into quintiles. We assessed overall comparative utilization by hospital by rank-ordering the summation of quintile responses into tertiles, classifying utilization as high, moderate, or low.31 Bivariate analysis was used to evaluate associations between index visit admission rate and 3-day revisits and revisits that resulted in hospitalization. The analysis is reported as the coefficient of determination (R2) with P values. We determined statistical significance as a 2-tailed P < .05.
During the 2-year study period, there were 261 696 ED visits among infants <90 days old at the 37 participating hospitals; 37 907 (14.5%) had a diagnosis code of fever. Of these, we excluded 1892 (5.0%) infants who had a complex chronic condition and 1067 (2.8%) who were transferred from another institution (including 122 with a complex chronic condition), forming our study cohort of 35 070 patients. Approximately 22.0% of infants were ≤28 days of age, 42.9% were 29 to 56 days, and 35.1% were 57 to 89 days (Table 1).
Patient-level variation in management and outcomes of febrile young infants is listed in Table 2. The proportion of infants undergoing urine, blood, and CSF testing was inversely proportional to patient age, decreasing from 72.0% (95% CI, 71.0–73.0) among infants ≤28 days of age to 13.1% (95% CI, 12.5–13.6) among infants 57 to 89 days (P < .001). Admission rate also decreased as patient age increased (Fig 1). Among hospitalized patients, ampicillin use was highest among infants ≤28 days of age (94.9%; 95% CI, 94.4–95.5), decreased to 54.8% (95% CI, 53.6–56.0) in infants 29 to 56 days, and was lowest in infants 57 to 89 days (20.4%; 95% CI, 18.7–22.3) (P < .001). Acyclovir was administered to 19.6% (95% CI, 18.9–20.2) of infants overall, highest among infants ≤28 days old. For discharged patients, infants aged 29 to 56 days had the highest rate of ceftriaxone use (P < .001). The SBI rate (8.4% overall) was highest among infants ≤28 days of age (11.1%) and peaked in the third week of life at 12.9% (Fig 1). Among all infants, 1854 (5.3%) were diagnosed with UTI, 830 (2.4%) with bacteremia or sepsis, and 122 (0.3%) with meningitis. Twenty-two (0.06%; 95% CI, 0.04–0.09) infants had a diagnosis of HSV infection. Six infants in the cohort died; 3 of whom were ≤28 days of age (including 1 with HSV), and 3 were 29–56 days of age.
Testing, treatment, and hospitalization varied widely by study institution across all 3 age groups (Table 3). Despite this variation, the prevalence of SBIs and rate of 3-day revisits did not substantially differ across institutions. Infants aged 29 to 56 days and 57 to 89 days experienced similar variation in laboratory testing, greater than among neonates ≤28 days old, who had the least variation (Supplemental Table 4). The frequency and the variation in administration of antibiotic and acyclovir therapy for hospitalized infants were inversely associated with age. Hospitalization rates across institutions ranged from 53.0% to 97.3% in neonates ≤28 days of age, 17.2% to 80.4% in infants 29 to 56 days of age, and 3.0% to 65.3% in patients 57 to 89 days of age. The hospital-level variation in testing and treatment persisted when we additionally included patients with SBI diagnosis codes (Supplemental Table 5).
We examined the individual hospital-level utilization ranking for urine, blood, CSF testing, and admission rate, as well as the overall and age-stratified utilization tertiles calculated by summing the individual ranks (Fig 2). High-utilization hospitals were often high utilizers of multiple individual parameters, as indicated by the darkest-shaded boxes. Although hospitals varied in utilization across age groups, hospitals often remained consistent in their level of utilization: 12 hospitals (32.4%) remained within the same tertile in all 3 age groups, and 25 hospitals (67.6%) stayed within the same utilization tertile for 2 age groups. Of these 25 hospitals, 11 (44.0%) were in the same tertile for the ≤28-day and 29- to 56-day age groups, 10 (40.0%) for the 29- to 56-day and 57- to 89-day age groups, and 4 (16.0%) for the ≤28-day and 57- to 89-day age groups.
For the 20 392 infants discharged from the ED (58.1% of study population), the average 3-day revisit rate was 6.0% (95% CI, 5.7–6.3), and the average rate of revisits that resulted in hospitalization was 2.0% (95% CI, 1.8–2.1). One hundred-seventeen (0.6%; 95% CI, 0.5–0.7) returned within 3 days and were diagnosed with an SBI resulting in hospitalization. In the overall cohort, hospital-level admission rate at the index visit did not correlate with 3-day revisits (Supplemental Fig 4) or revisits that resulted in hospitalization (Fig 3). Hospitals with lower admission rates in the ≤28-day age group had higher 3-day revisits (R2 = 0.21, P = .004) and revisits that resulted in hospitalization (R2 = 0.16, P = .01), although the correlations were not strong. Index admission rate in the 29–56 day age group did not correlate with 3-day revisits or revisits with hospitalization. Conversely, a higher admission rate in the 57- to 89-day age group was associated with higher return visits that resulted in hospitalization (R2 = 0.28, P < .001).
We examined care variation among >35 000 febrile young infants presenting to EDs across a large sample of US children’s hospitals. We observed significant variation across hospitals in testing, treatment, and hospitalization rates, even among the youngest infants (≤28 days), who are at highest risk for SBI.7 There were also hospital differences in testing, treatment, and admission utilization patterns by age group, but hospitals tended to consistently remain high or low utilizers across multiple age groups. Despite this hospital-level variation in resource utilization, hospital-level outcomes were similar, particularly in infants >28 days of age. Targeting care variation may present an opportunity to better direct resources for the ED management of the febrile young infant.
There is little guidance for best practice. An American College of Emergency Physicians policy statement endorsed that febrile infants ≤28 days of age “should be presumed to have a serious bacterial infection” but did not define the extent of evaluation required; guidance for management of febrile infants aged ≥29 days is even less clear.13 The lack of clarity regarding best practices and resultant care variation is not surprising because several low-risk criteria exist that offer differing recommendations. Furthermore, there has been only 1 prospective comparison of these low-risk criteria, a study that compared the Philadelphia and Rochester criteria in a small cohort of low-risk patients.32 Less than half of febrile infants 29 to 56 days of age underwent lumbar puncture, possibly because of wide application of the Rochester criteria12 or modification of practice due to the documented low prevalence of bacterial meningitis in well-appearing febrile young infants.33,34
Opportunities to improve resource utilization may exist in areas where there is variation in care without disparate outcomes. We observed that overall resource utilization, antimicrobial use, and hospitalization rates varied across hospitals, but little variation was observed in revisit outcomes. Ampicillin has traditionally been administered to neonates suspected of having Listeria monocytogenes, an uncommon cause of bacteremia and meningitis in the febrile young infant,34–37 especially beyond the first month of life,38 and Enterococcus, similarly an uncommon cause of SBI.34,37,38 Additionally, the prevalence of ampicillin resistance has been reported to be 36% to 53% in pathogens that cause SBI in the febrile young infant.34,39 Despite these findings, nearly two-thirds of hospitalized febrile infants in our study received ampicillin in combination with gentamicin or a third-generation cephalosporin, with significant interhospital variation. Similarly, wide variation was observed in acyclovir use. Although neonatal HSV infection is associated with high mortality and neurologic morbidity if it is untreated40 or treatment is delayed,30 the incidence is low,41 especially in infants who are clinically well, without vesicles or mononuclear-predominant CSF pleocytosis.42 Neonatal HSV is especially uncommon after 21 days of age,43 yet 12.4% of febrile infants 29 to 56 days of age received acyclovir in our cohort, with a rate of >17.1% for one-quarter of the hospitals in this age group. Additional study of empirical antimicrobial coverage and the relationship to clinical outcomes is needed.
Improving resource utilization must be balanced against the risk of missed SBI.44 For neonates ≤28 days old in whom the low-risk criteria demonstrate diminished performance,7,45 we observed that hospitals with lower hospitalization rates experienced higher rates of 3-day revisits and subsequent hospitalization. Additional study is needed to explore the consequences of hospital-specific variation in admission rates in this age group. For infants 29 to 56 days of age, hospitalization rate was not associated with higher 3-day revisits or revisits requiring hospitalization. In infants >28 days old, a lower admission rate may not result in higher adverse outcomes, and hospitals with high admission rates may therefore evaluate opportunities to reduce hospitalization without affecting outcomes. Intermountain Healthcare System instituted an evidence-based care process model in 2008 to decrease variation within its system, which resulted in a higher proportion of low-risk patients being discharged from the ED without increasing adverse outcomes.46 Institution of evidence-based guidelines for care of the febrile young infant may similarly improve resource utilization without compromising outcomes.
Our study has several limitations. First, the validity of ICD-9 diagnosis codes for identifying young infants with fever is not well established, and thus our study is subject to misclassification bias. To maximize sensitivity, we included children with either an ICD-9 admission or discharge diagnosis of fever. Additionally, our findings did not materially differ after we included patients with a diagnosis code for SBI (but no fever code) in our cohort to identify patients in whom a diagnosis code for fever may have been omitted. It is also possible that some patients who did not undergo testing were misclassified as having a fever. Similarly, we may have overestimated or underestimated the true prevalence of SBI in our cohort because of ICD-9 code misclassification or lack of full testing for SBI in all patients. However, our observed rate of SBI was similar to that of previous studies,10,47 providing face validity to our findings.
Second, our study is also subject to bias related to the lack of clinical data available in the PHIS database. For example, we were unable to assess severity of illness, which probably influences management decisions. However, we excluded patients with complex chronic conditions, as well as transferred patients, to avoid misclassification related to testing or treatment that may have occurred before arrival. We were also unable to know the number of failed procedure attempts, although we included procedure codes for lumbar puncture to maximize capture of CSF testing.
Third, the performance of viral testing is not reliably available in the PHIS database, and we were therefore unable to assess variation in viral testing or the influence of viral testing on treatment decisions. Fourth, because we were unable to have a more granular assessment of morbidity, additional study is needed to evaluate the impact of various testing and treatment strategies on outcomes. Infants who returned to a different ED were not captured, which would result in an underestimation of our overall ED revisit rates. Last, although this study includes patients from 37 children’s hospitals in the United States, our results may not be generalizable to EDs at non–children’s hospitals or other geographic regions not adequately represented in our data.
Substantial patient-level variation exists in the ED evaluation and management of the febrile infant <90 days of age. Moreover, hospital-level variation in evaluation, treatment, and admission of febrile young infants is significant across age groups. Variation in admission does not correlate with return visits to the ED, indicating opportunities for hospitals to improve resource use for management of the febrile young infant. Additional investigation is necessary to evaluate the impact of utilization on outcomes in neonates.
The Febrile Young Infant Research Collaborative includes the following additional investigators, who are acknowledged for their work on this study: Whitney L. Browning, MD, Angela L. Myers, MD, MPH, Fran Balamuth, MD, PhD, and Katie L. Hayes, BS.
- Accepted July 1, 2014.
- Address correspondence to Paul L. Aronson, MD, Section of Pediatric Emergency Medicine, Yale School of Medicine, 100 York Street, Suite 1F, New Haven, CT 06511. E-mail:
Drs Aronson and Neuman conceptualized and designed the study, analyzed and interpreted the data, and drafted the initial manuscript; Dr Thurm acquired and analyzed the data and contributed to drafting of the manuscript; Drs Alpern, Alessandrini, Williams, and Shah contributed to design of the study, analysis and interpretation of the data, and drafting of the initial manuscript; Drs Nigrovic, McCulloh, Schondelmeyer, and Tieder contributed to analysis and interpretation of the data; and all authors revised the manuscript critically for important intellectual content and approved the final manuscript as submitted.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Huebner RE,
- Mbelle N,
- Forrest B,
- Madore DV,
- Klugman KP
- Roberts KB,
- Subcommittee on Urinary Tract Infection, Steering Committee on Quality Improvement and Management
- Huppler AR,
- Eickhoff JC,
- Wald ER
- Schwartz S,
- Raveh D,
- Toker O,
- Segal G,
- Godovitch N,
- Schlesinger Y
- McCarthy PL,
- Sharpe MR,
- Spiesel SZ,
- et al
- Baker MD,
- Avner JR,
- Bell LM
- Jaskiewicz JA,
- McCarthy CA,
- Richardson AC,
- et al.,
- Febrile Infant Collaborative Study Group
- American College of Emergency Physicians Clinical Policies Committee,
- American College of Emergency Physicians Clinical Policies Subcommittee on Pediatric Fever
- Jain S,
- Cheng J,
- Alpern ER,
- et al
- Goldman RD,
- Scolnik D,
- Chauvin-Kimoff L,
- et al.,
- Fever in Infants Group Research, Pediatric Emergency Research of Canada
- Institute of Medicine
- Wood JN,
- Feudtner C,
- Medina SP,
- Luan X,
- Localio R,
- Rubin DM
- Nigrovic LE,
- Fine AM,
- Monuteaux MC,
- Shah SS,
- Neuman MI
- Tieder JS,
- Hall M,
- Auger KA,
- et al
- Shah SS,
- Aronson PL,
- Mohamad Z,
- Lorch SA
- Florin TA,
- French B,
- Zorc JJ,
- Alpern ER,
- Shah SS
- Biondi E,
- Evans R,
- Mischler M,
- et al
- Byington CL,
- Rittichier KK,
- Bassett KE,
- et al
- Kimberlin DW,
- Lin CY,
- Jacobs RF,
- et al.,
- National Institute of Allergy and Infectious Diseases Collaborative Antiviral Study Group
- Copyright © 2014 by the American Academy of Pediatrics