BACKGROUND:

Substantial variability exists in the care of febrile, well-appearing infants. We aimed to assess the impact of a national quality initiative on appropriate hospitalization and length of stay (LOS) in this population.

METHODS:

The initiative, entitled Reducing Variability in the Infant Sepsis Evaluation (REVISE), was designed to standardize care for well-appearing infants ages 7 to 60 days evaluated for fever without an obvious source. Twelve months of baseline and 12 months of implementation data were collected from emergency departments and inpatient units. Ill-appearing infants and those with comorbid conditions were excluded. Participating sites received change tools, run charts, a mobile application, live webinars, coaching, and a LISTSERV. Analyses were performed via statistical process control charts and interrupted time series regression. The 2 outcome measures were the percentage of hospitalized infants who were evaluated and hospitalized appropriately and the percentage of hospitalized infants who were discharged with an appropriate LOS.

RESULTS:

In total, 124 hospitals from 38 states provided data on 20 570 infants. The median site improvement in percentages of infants who were evaluated and hospitalized appropriately and in those with appropriate LOS was 5.3% (interquartile range = −2.5% to 13.7%) and 15.5% (interquartile range = 2.9 to 31.3), respectively. Special cause variation toward the target was identified for both measures. There was no change in delayed treatment or missed bacterial infections (slope difference 0.1; 95% confidence interval, −8.3 to 9.1).

CONCLUSIONS:

Reducing Variability in the Infant Sepsis Evaluation noted improvement in key aspects of febrile infant management. Similar projects may be used to improve care in other clinical conditions.

Fever without an obvious source in well-appearing infants 7 to 60 days of age presents a common conundrum for health care providers. The most common cause of fever in these infants is viral13 ; however, fever is also a common initial symptom of bacterial infections such as meningitis, bacteremia, and urinary tract infection (UTI).4  Because these bacterial infections may be life-threatening, thousands of well-appearing, febrile infants 7 to 60 days old undergo evaluation for suspected bacterial infection in the United States annually, which often leads to overuse and unnecessary hospitalization,5,6  despite the fact that the vast majority of these infants do not have bacterial infections.4 

There is substantial practice variability among clinicians in terms of evaluation and treatment, including blood tests, microbiologic cultures, viral tests, cerebrospinal fluid analysis, empirical antibiotics, hospitalization, and length of stay (LOS).7,8  Additionally, the harms of testing and empirical treatment are challenging to quantify, physicians have widely varying risk tolerances, and commonly used algorithms have low positive predictive value for detecting bacterial infections.9 

In 2016, the American Academy of Pediatrics’ (AAP) Value in Inpatient Pediatrics (VIP) network, an established inpatient pediatric quality improvement (QI) network, designed and implemented a national QI initiative to standardize care for well-appearing infants ages 7 to 60 days evaluated for fever without an obvious source. This effort, entitled Reducing Excessive Variability in the Infant Sepsis Evaluation (REVISE), offered providers in the inpatient and emergency department (ED) settings evidence-based education, strategies for implementation, and tools to standardize care. The primary aim of REVISE was to standardize and improve appropriate hospitalization and LOS for our target population via a national QI collaborative.

The VIP network includes >150 hospitals in the United States and aims to improve health care value to pediatric patients in inpatient or ED settings through large-scale QI initiatives that are focused on standardizing processes, implementing evidence-based practices, eliminating harm, and reducing overuse. REVISE was sponsored by the VIP network and was approved by the AAP’s Institutional Review Board. Participating site teams obtained local approvals as necessary.

Population of Interest

REVISE included well-appearing infants ages 7 through 60 days evaluated for fever without an obvious source at the participating site’s ED or transferred to the inpatient unit from an outpatient setting and were discharged from the hospital from that site’s ED or inpatient unit. Infants were excluded if they (1) were not well appearing on presentation (as documented by terms such as “toxic,” “ill-appearing,” “lethargic,” etc), (2) had comorbid conditions predisposing to severe or recurrent bacterial illness such as central lines, metabolic disorders, or immunodeficiencies, (3) had an obvious source of infection (eg, bronchiolitis, cellulitis, etc), or (4) were transferred from another inpatient setting.

Planning the Intervention

During January and February 2016, an 8-person planning group was formed, consisting of university and community pediatric hospitalists, pediatric emergency medicine physicians, and pediatric infectious diseases specialists. Two consultants, selected for their decades of experience in the field and their national reputation, participated in an ad hoc manner on planning webinars, phone calls, and e-mails, specifically with regard to measure selection. Because no nationally promulgated evidence-based guideline existed, the planning group discussed available evidence and potential measures through webinars using a consensus-based approach.10  These discussions occurred over several months in early 2016. Approximately 9 potential measures were identified via literature review, and these were deliberated during webinars and agreed upon by consensus. After each measure was selected, they were discussed with a statistical consultant to ensure a reliable run chart could be produced. The discussion of each measure included (1) the strength of the evidence, (2) the feasibility of retrospective data collection, (3) whether it might be viewed as too controversial given the nature of the evidence base and historical precedent, (4) whether reasonable targets could be determined, and (5) whether the measure would be applicable to most hospitals. Measures that were considered but not selected included lumbar puncture use, herpes simplex viral testing, respiratory viral testing, blood culture use, and procalcitonin use.

Recruitment began in July 2016 via announcements at the pediatric hospital medicine annual meeting over applicable AAP LISTSERVS and via informational webinars. Sites were required to submit a written application describing a multidisciplinary improvement team consisting of at least 3 members, including 1 inpatient-based clinician and 1 ED-based clinician. Sites also had to demonstrate support for REVISE from hospital administration and pay a $750 enrollment fee.

The cornerstone of REVISE was the “change package” (Supplemental Information). This package, provided to all participating sites, included clinical algorithms, order sets, academic detailing (evidence-based education via live webinars), peer-to-peer sharing, and a freely available, smartphone-based mobile application (originally named “CMPeds,” currently named “PedsGuide”) sponsored by Children’s Mercy Hospital with input from REVISE leadership. The change package was designed to provide a comprehensive tool kit to small community-based sites that have limited resources and infrastructure for QI, whereas larger- or better-resourced sites may focus on using specific aspects of the package. The clinical algorithms and order sets were designed to standardize diagnostic and treatment options in the ED and the inpatient settings, thereby reducing unnecessary testing (eg, radiographs). The live webinars were used to provide education on QI methodology, beginning at a novice level and becoming more advanced as the project progressed. They were also used to report collaborative wide data to the sites, and individual, high-performing sites were invited to discuss their progress and local drivers of change. These webinars were recorded and could be accessed anytime by members of the collaborative. The PedsGuide mobile application remains freely available and is intended to walk providers through an evidence-based diagnostic and management plan based on patient data that the provider inputs into the application.

Communication with the collaborative was done through live webinars approximately every 2 months, designated coach support, and a project LISTSERV. Sites had access to real-time run charts for all project measures for use in benchmarking against the national collaborative. Institutions wishing to implement the REVISE change package can find additional information, order sets, and algorithms at www.aap.org/en-us/professional-resources/quality-improvement/quality-improvement-resources-and-tools/Documents/VIP%20Change%20Package%20(2016)%20Fever.pdf.

Physicians were eligible for 20 hours of Part 4 Maintenance of Certification (MOC) credit if they participated meaningfully at the local level, engaged in at least 6 webinars, and submitted 12 cycles of baseline data and 10 cycles of implementation.

Outcome

1. Proportion of Hospitalized Infants Who Were Hospitalized Appropriately

This primary measure was defined as the percentage of hospitalized infants who had an appropriate evaluation and who had an appropriate reason to be hospitalized.8,1113  An evaluation was deemed appropriate if it included at least a urinalysis and a marker of inflammation or infection (eg, white blood cell count, C-reactive protein, procalcitonin). These criteria were chosen because they provide the minimal amount of information necessary to apply standard risk stratification criteria.11  A lumbar puncture was not required because it is not required by at least 1 set of commonly used risk stratification criteria.12 

Hospitalization was considered appropriate if the infant was at “non–low risk” defined as (1) <31 days of age at the time of presentation, (2) abnormal findings on laboratory workup, (3) had a relevant previous medical history, or (4) required admission for social reasons. All other infants were considered “low” risk and therefore were considered inappropriate to hospitalize. The measure target was set at 90% compliance.

2. Appropriate LOS

Defined as the percentage of infants who had an appropriate LOS, this primary measure incorporated validated risk stratification criteria to determine if a hospitalized infant was discharged within an appropriate time period.8,14  For infants determined to be at non–low risk (as described above), appropriate LOS was defined as discharge within 42 hours of the first measured vital sign. For infants at low risk, appropriate LOS was defined as discharge within 30 hours of the first measured vital sign. These numbers represent a standard 36- or 24-hour period of observation for infants at non-low and low risk, respectively, with an additional 6-hour “window” period.8,14  The window period was built into the measure to allow, for example, infants whose 36-hour period of observation ended at 4 am to meet the measure if discharged by 10 am. Infants diagnosed with meningitis, bacteremia, or UTI were excluded from this measure. The target was set at 80% compliance.

Process

1. Urinalysis Use

This measure was defined as the percentage of infants who had a urinalysis performed by any method on the first day of encounter. The target was set at 95% compliance.15 

2. Chest Radiograph Use

This measure was defined as the proportion of infants without documented respiratory symptoms who received a chest radiograph within 24 hours of presentation.16,17  The target was set at <10%.

3. Appropriate Antibiotic Use

This measure was defined as the percentage of infants who received recommended empirical antibiotic regimens within 24 hours of presentation.1820  Recommended antibiotics were considered to be (alone or in combination) none administered, ampicillin, aminoglycosides, or third-generation cephalosporins. All other antibiotics or combinations of antibiotics were considered inappropriate. The target was set at >90%.

Balance

1. Delayed Diagnosis for UTI, Bacteremia, and/or Meningitis

This measure assessed the percentage of infants who returned to the ED or were readmitted for a new diagnosis of meningitis, bacteremia, or UTI within 7 days of discharge from the ED or inpatient setting. The target was set at <2% on the basis of estimated prevalence of bacterial meningitis and bacteremia.8,11,21,22 

For the purposes of REVISE, meningitis, bacteremia, and UTI were defined as a cerebrospinal fluid, blood, or urine culture positive for an organism that was treated as a pathogen, respectively.

Sites collected 12 monthly cycles of baseline data from September 2015 through August 2016 retrospectively and 12 monthly cycles of implementation data from December 2016 through November 2017 on a rolling monthly basis. The project launched in September of 2016; thus, a 3-cycle window period wherein no data were collected allowed sites to obtain approvals and plan efforts. Data were deidentified and entered into the AAP’s Quality Improvement Data Aggregator, a Web-based data collection tool wherein sites could also view real-time local and national run charts.

Guidance was provided to each site to identify charts via manual review in the following manner. First, teams were asked to obtain a list of all infants between the ages of 7 to 60 days evaluated in the ED or admitted to the inpatient unit and their respective discharge diagnoses. The teams first excluded infants who did not have any discharge diagnoses suggestive of a potential need for evaluation for fever without an obvious source (eg, pyloric stenosis, constipation, etc). The remaining charts were then manually reviewed for the aforementioned inclusion or exclusion criteria. Data from the first 20 charts meeting criteria (or all eligible charts if <20) for each cycle were collected. If a site had >20 charts, they were asked to select either 20 charts at random or the first 20 charts of the cycle. This was done, in consultation with our statistician, because of the differences in sample sizes of large and small hospitals identified during a cursory evaluation while planning the project. By limiting large sites to 20 charts, we were attempting to limit the degree to which change at large sites overwhelmed change at small sites and thus drove change across the collaborative.

For our 2 outcome measures (our primary measures), we analyzed aggregate project performance using statistical process control run charting and Western Electric rules for detection of special cause variation. We used interrupted time series (ITS) modeling to estimate population-level changes in each measure that occurred during the course of the project. The ITS approach controls for secular trends and allows for the evaluation of clinical outcomes by using population-level data. Specifically, we estimated the overall time trend in the event rate, the postimplementation time trend in the event rate, the change in time trend associated with implementation, and the immediate effect of implementation on the event rate (equivalent to the average difference between prerollout and postrollout event rates, controlling for time trends). Odds of event were modeled by using logistic mixed models fit using the GLIMMIX procedure in SAS 9.4 (SAS Institute, Inc, Cary, NC). Each model included explanatory variables (the overall time trend, the postimplementation time trend, and the effect of implementation) and a random site intercept to adjust for clustering. For their data to be included, a site must have provided data for at least 1 baseline cycle and at least 1 implementation cycle. Statistical significance was set at P < .05.

In a follow-up exploratory subgroup analysis, we compared the effect of time period (pre- versus postimplementation) on odds of appropriate admission and odds of appropriate LOS for community versus university hospitals and for freestanding versus nonfreestanding hospitals by adding interaction terms to the models for these measures. These variables were selected for exploratory analysis given the potential for relevant differences in administrative infrastructure, value placed on QI, and available resources between subgroups.

A total of 133 hospitals were accepted into REVISE. Of these, 7 dropped out or did not provide data. In total, data on 20 702 infants were collected. Two sites provided data on a total of 132 infants but did not provide at least 1 cycle of baseline data and at least 1 cycle of implementation data, leaving 20 570 infants from the remaining 124 hospitals (from 38 states) for analysis (Table 1).

TABLE 1

Demographic Data for Project Hospitals and Included Infants

VariableHospitals (N = 124), n (%)Infants (N = 20 570), n (%)
Hospital type   
 Community 49 (40) 6335 (31) 
 University 73 (59) 14 085 (68) 
Non-ICU beds   
 <10 6 (5) 262 (1) 
 11–30 44 (35) 3918 (19) 
 31–50 19 (15) 2958 (14) 
 >50 55 (44) 13 432 (65) 
Annual casesa   
 <50 15 (12) 1241 (6) 
 51–100 40 (32) 4113 (20) 
 101–200 26 (21) 4249 (21) 
 201–300 17 (14) 3308 (16) 
 >300 26 (21) 7659 (37) 
Board-certified PEM   
 No 24 (19) 1622 (8) 
 Yes 98 (79) 18 798 (91) 
Area   
 Urban (inner city) 39 (31) 6482 (32) 
 Urban (not inner city) 44 (35) 8977 (44) 
 Suburban 33 (27) 4634 (23) 
 Rural 6 (5) 327 (2) 
Census region   
 Midwest 39 (31) 5765 (28) 
 Northeast 27 (22) 3244 (16) 
 South 39 (31) 7944 (39) 
 West 19 (15) 3617 (18) 
VariableHospitals (N = 124), n (%)Infants (N = 20 570), n (%)
Hospital type   
 Community 49 (40) 6335 (31) 
 University 73 (59) 14 085 (68) 
Non-ICU beds   
 <10 6 (5) 262 (1) 
 11–30 44 (35) 3918 (19) 
 31–50 19 (15) 2958 (14) 
 >50 55 (44) 13 432 (65) 
Annual casesa   
 <50 15 (12) 1241 (6) 
 51–100 40 (32) 4113 (20) 
 101–200 26 (21) 4249 (21) 
 201–300 17 (14) 3308 (16) 
 >300 26 (21) 7659 (37) 
Board-certified PEM   
 No 24 (19) 1622 (8) 
 Yes 98 (79) 18 798 (91) 
Area   
 Urban (inner city) 39 (31) 6482 (32) 
 Urban (not inner city) 44 (35) 8977 (44) 
 Suburban 33 (27) 4634 (23) 
 Rural 6 (5) 327 (2) 
Census region   
 Midwest 39 (31) 5765 (28) 
 Northeast 27 (22) 3244 (16) 
 South 39 (31) 7944 (39) 
 West 19 (15) 3617 (18) 

PEM, pediatric emergency medicine.

a

Self-reported data on preproject survey.

There were 10 431 (51%) infants included in the 12 baseline cycles (mean per month ± SD; 869 ± 85 infants) and 10 139 (49%) infants included in the 12 implementation cycles (845 ± 110 infants).

The statistical process control chart (Fig 1A) for appropriate workup and hospitalization demonstrated special cause variation. The gains were sustained through the remaining 8 implementation cycles. The median site improvement was 5.3% (interquartile range [IQR] = −2.5% to 13.7%).

FIGURE 1

Statistical process control charts. A, Change in proportion of patients who received an appropriate workup and hospitalization. B, Change in proportion of patients who met appropriate LOS criteria.

FIGURE 1

Statistical process control charts. A, Change in proportion of patients who received an appropriate workup and hospitalization. B, Change in proportion of patients who met appropriate LOS criteria.

Close modal

The statistical process control chart (Fig 1B) for appropriate LOS demonstrated special cause variation. The gains were sustained through the remaining 11 implementation cycles. The median site improvement was 15.5% (IQR = 2.9 to 31.3).

Site-level performance in reference to baseline performance and amount of change achieved is presented for appropriate hospitalization in Fig 2A and for LOS in Fig 2B.

FIGURE 2

Site implementation performance by baseline performance. A, Appropriate hospitalization; B, Appropriate LOS. Markers represent individual sites. Markers above the “No change” line represent improvement, and markers below the line represent deterioration. Size is based on the number of enrolled patients during the baseline period.

FIGURE 2

Site implementation performance by baseline performance. A, Appropriate hospitalization; B, Appropriate LOS. Markers represent individual sites. Markers above the “No change” line represent improvement, and markers below the line represent deterioration. Size is based on the number of enrolled patients during the baseline period.

Close modal

ITS regression results are provided (Table 2). As the rate of appropriate admission changed in trend, it approached significance (P = .06). Odds of appropriate LOS increased by 70% with implementation (P < .001). Median (IQR) LOS dropped from 50 (41 to 64) hours preimplementation to 46 (36 to 59) hours postimplementation. There were no statistically significant changes in rates of delayed diagnosis of bacterial infection.

TABLE 2

ITS Regression With P Values Representing Slope Difference

Preslope (95% CI)Postslope (95% CI)Difference (95% CI)PImplementation Effect (95% CI)
Appropriate admission 1.8 (0.3 to 3.3) 4.2 (2.3 to 6.2) 2.4 (0 to 4.9) .06 −7.4 (−23.4 to 12.0) 
Appropriate LOS 1.1 (−0.9 to 3.2) 2.9 (0.7 to 5.1) 1.7 (−1.2 to 4.8) .25 70.3 (33.4 to 117.5) 
Urinalysis 2 (0.1 to 3.9) 2.7 (0.6 to 4.9) 0.7 (−2.1 to 3.6) .61 −22.8 (−38.6 to −3) 
Chest radiograph use −3 (−4.9 to −1.2) −2.3 (−4.7 to 0.1) 0.7 (−2.4 to 3.9) .65 0.1 (−21.9 to 28.4) 
Appropriate antibiotics 3.1 (−0.9 to 7.2) −4 (−8 to 0.2) −6.9 (−12.1 to −1.3) .02 2.8 (−37.3 to 68.5) 
Delayed treatment −0.7 (−6.2 to 5.3) −0.6 (−6.8 to 6.1) 0.1 (−8.3 to 9.1) .99 1.9 (−49.7 to 106.4) 
Preslope (95% CI)Postslope (95% CI)Difference (95% CI)PImplementation Effect (95% CI)
Appropriate admission 1.8 (0.3 to 3.3) 4.2 (2.3 to 6.2) 2.4 (0 to 4.9) .06 −7.4 (−23.4 to 12.0) 
Appropriate LOS 1.1 (−0.9 to 3.2) 2.9 (0.7 to 5.1) 1.7 (−1.2 to 4.8) .25 70.3 (33.4 to 117.5) 
Urinalysis 2 (0.1 to 3.9) 2.7 (0.6 to 4.9) 0.7 (−2.1 to 3.6) .61 −22.8 (−38.6 to −3) 
Chest radiograph use −3 (−4.9 to −1.2) −2.3 (−4.7 to 0.1) 0.7 (−2.4 to 3.9) .65 0.1 (−21.9 to 28.4) 
Appropriate antibiotics 3.1 (−0.9 to 7.2) −4 (−8 to 0.2) −6.9 (−12.1 to −1.3) .02 2.8 (−37.3 to 68.5) 
Delayed treatment −0.7 (−6.2 to 5.3) −0.6 (−6.8 to 6.1) 0.1 (−8.3 to 9.1) .99 1.9 (−49.7 to 106.4) 

Subgroup analysis by hospital type (community versus academic and freestanding versus nonfreestanding) revealed no statistically significant differences in the effect of time period on odds of inappropriate admission but did show higher odds of appropriate LOS associated with nonfreestanding versus freestanding hospitals (odds ratio 1.81; 95% confidence interval [CI] 1.46 to 2.23; P < .001).

Regarding our process and balancing measures, we noted no significant differences. The pre-post project-wide performances were 90.2% vs 90.9% (urinalysis), 16.9% vs 12.5% (chest radiograph use), 98.0% vs 98.1% (appropriate empirical antibiotics), and 0.9% vs 0.8% (delayed diagnosis). Adjusted estimates are presented in Table 2. To date, 737 physicians have received MOC credit for their participation in this project.

This QI collaborative demonstrated project-wide improvement in the proportion of appropriately hospitalized infants and the proportion of infants who had an appropriate LOS, with the largest increase in median performance noted in the latter. Many participating sites began the project at or near our measure target for meeting appropriateness criteria for hospitalization, thus limited improvement was possible as seen in the clustering of individual site performance in Fig 2A. For the LOS measure, there was greater variation in performance with greater dispersion of performance in Fig 2B. We also noted that nonfreestanding status seemed to confer an advantage for the LOS measure, which may be related to challenges achieving change across larger institutional settings.

The change noted in LOS also occurred more quickly, as evidenced by the control charts and the significant implementation effect identified in our regression. This may be due to several factors. First, LOS provides a specific goal that can be planned for in advance. Second, the vast majority of sites started farther from the target, and thus the collaborative as a whole had more room for improvement. Third, inherent in the appropriate hospitalization measure was some degree of subjectivity, whereas our LOS measure provided a more objective target. Lastly, there may be differences in the improvement processes between the ED and inpatient units, which may have resulted in different patterns of special cause variation for an inpatient-focused measure such as LOS and an ED-focused measure such as appropriate initial workup and hospitalization.

We did not demonstrate significant improvement in our 3 process measures. This may have been due to the fact that the project-wide baseline performances were close to (or met) the targets, limiting the degree of change needed. However, during the project, some sites reported extensive work on these measures and certain sites did make improvements. For this reason, we suggest assessing process measures in this style of large collaborative provided that doing so does not detract effort from the outcome measures.

Many aspects of well-appearing, febrile infant management are controversial, and there is no gold standard approach. Current management strategies are based on guidelines developed decades ago11,13,23  and data from regionally limited health systems.8  RNA biosignatures hold promise,24  but cost-effective, easily disseminated strategies are not currently available. We embarked on this project with the goal to improve and standardize care and decrease variability rather than derive optimal management strategies. We opted toward “conservative” approaches (eg, hospitalizing infants <31 days was considered appropriate)23 ; however, we also selected measures that would not penalize more “progressive” sites when we felt the evidence supported a variety of decisions.13,21  Our measures do not address clinical questions for which we could not arrive at consensus (eg, lumbar puncture).

Our work has some limitations. First, the data were collected retrospectively, and there were challenges in categorizing infants within certain criteria (eg, well-appearing). Second, there are limits to the internal validity of rolling out a change package to such a large collaborative. It is difficult to know the degree of uptake at individual sites; thus, we are unable to provide insight into which components of our change package elicited positive change. Similarly, we are unable to provide insight into the degree to which site-specific aspects evoked positive (or negative) change at the individual sites. Third, our balancing measure of delayed treatment included a 7-day window because we felt that this time period most accurately represented the potential for a missed or delayed treatment resulting directly from our project. It is possible that a longer time period would have identified additional cases or that infants returned to EDs or inpatient units of nonparticipating sites. Fourth, we cannot comment on the sustainability of the improvement beyond the implementation period. There may have been an unidentified Hawthorne effect in that sites focused more on improvement while they knew they were being monitored or incentivized by MOC credit. Finally, limiting this QI effort to 12 baseline and 12 intervention data points may provide excessive α risk for missing smaller positive or negative changes.

In a national QI initiative of 124 sites and >20 000 well-appearing, febrile infants, we noted improvement in key aspects of febrile infant management including the proportion of infants hospitalized appropriately and LOS, without a change in 7-day ED or inpatient revisits. Similar collaboratives may improve care in other clinical conditions.

We thank the teams at the following 124 participating sites for their effort and commitment to this patient population and for their dedication to standardize best practices for these young infants.

Abington Hospital – Jefferson Health; Adventist Hinsdale Hospital; Advocate Children’s Hospital; Akron Children’s Hospital; AnMed Health Women’s and Children’s Hospital; Ann and Robert H. Lurie Children’s Hospital of Chicago; Beaumont Children’s Hospital; Blank Children’s Hospital; Bridgeport Hospital; Bryn Mawr Hospital; Carilion Children’s Hospital; Carolinas Medical Center; Central DuPage Hospital; Children’s Medical Center Plano; Children’s Healthcare of Atlanta at Scottish Rite Hospital; Children’s Hospital & Medical Center; Children’s Hospital at Montefiore; Children’s Hospital Colorado; Children’s Hospital Los Angeles; Children’s Hospital of Georgia; Children’s Hospital of Greenville Health System; Children’s Hospital of Illinois; Children’s Hospital of Michigan; Children’s Hospital of Philadelphia at Virtua; Children’s Hospital of Richmond at Virginia Commonwealth University; Children’s Hospital of Wisconsin; Children’s Hospital of Wisconsin Fox Valley; Children’s Medical Center Dallas; Children’s Mercy Hospital Kansas City; Children’s National Health System; CHRISTUS St. Vincent; Cleveland Clinic Children’s Hospital; Community Medical Center, Missoula Montana Hospital; Concord Hospital; Cook Children’s Medical Center; Covenant Children’s Hospital; C. S. Mott Children’s Hospital; Dell Children’s Medical Center of Central Texas; Duke University Medical Center; Egleston Children’s Hospital; Elmhurst Hospital Center; Florida Hospital for Children; Golisano Children’s Hospital at the University of Rochester Medical Center; Hackensack University Medical Center; Holtz Children’s Hospital; Hughes Spalding Children’s Hospital, Children’s Healthcare of Atlanta; Joe DiMaggio Children’s Hospital; John Muir Medical Center of Stanford Children’s Health Network; Kaiser Permanente Women and Children’s Center – Roseville; Kaiser Permanente Walnut Creek Medical Center; Kapiolani Medical Center for Women and Children; WellStar Kennestone Regional Medical Center; Kings County Hospital; Lakeland Regional Health Medical Center; Le Bonheur Children’s Hospital; Lucile Packard Children’s Hospital at Stanford; MacNeal Hospital; Massachusetts General Hospital; Medical University of South Carolina; Memorial Children’s Hospital; Mercy Children’s Hospital; Mission Children’s Hospital; Monmouth Medical Center; New York-Presbyterian Morgan Stanley Children’s Hospital; Newark Beth Israel Medical Center, Children’s Hospital of New Jersey; Nicklaus Children’s Hospital; NorthShore University Health System; Northwestern Lake Forest Hospital; Norton Children’s Hospital; New York University Langone Medical Center; Our Lady of the Lake Children’s Hospital; Palmetto Health Children’s Hospital; Kaiser Permanente Panorama City Medical Center; PeaceHealth St Joseph Medical Center; Penn State Children’s Hospital; Peyton Manning Children’s Hospital at St Vincent Indianapolis; Phoenix Children’s Hospital; Presbyterian Hospital; Providence Tarzana Medical Center; Rady Children’s Hospital-San Diego; Rainbow Babies and Children’s Hospital; Randall Children’s Hospital at Legacy Emanuel; Reading Health Systems; Renown Children’s Hospital; Riley Hospital for Children; Rochester General Hospital; Rush University Children’s Hospital; Sacred Heart Medical Center & Children’s Hospital; St Louis Children’s Hospital, Washington University in St Louis; University of Florida Health Shands Children’s Hospital; Shore Medical Center; Silver Cross Hospital; Sparrow Hospital; SSM Health; Cardinal Glennon Children’s Hospital; St Luke’s University Health Network; St Christopher’s Hospital for Children; St Cloud Hospital; St David’s Children’s Hospital; St Joseph Mercy Hospital; St Mary’s Hospital; Steven and Alexandra Cohen Children’s Medical Center of New York; Stony Brook Children’s Hospital; Stormont Vail Health; State University of New York Downstate Medical Center; Swedish Covenant Hospital; Texas Children’s Hospital and Baylor College of Medicine; Floating Hospital for Children at Tufts Medical Center; New York-Presbyterian Hospital, Weill Cornell Medicine; The Studer Family Children’s Hospital at Sacred Heart; University of California, San Francisco Benioff Children’s Hospital Oakland; University of Florida Health Jacksonville; University of Chicago Medicine Comer Children’s Hospital; University of Illinois at Chicago; University of Iowa Children’s Hospital; University of Minnesota Masonic Children’s Hospital and Fairview Ridges; University of Mississippi Medical Center, Batson Children’s Hospital; University of New Mexico Hospital; University of Vermont Medical Center; University of Virginia Children’s Hospital; Virginia Mason Memorial (formerly Yakima Valley Memorial Hospital); WakeMed Health and Hospitals; Wolfson Children’s Hospital; Women’s and Children’s Hospital, University of Missouri Hospitals and Clinics; West Virginia University Medicine Children’s Hospital; Kaiser Permanente Panorama City Medical Center.

We thank the members of the project planning group and those who provided expert guidance during development: Drs Jeffrey Bennett, Jennifer Light, Robert Pantell, and Kenneth Roberts. We also thank the current and former staff at the AAP and VIP network without whom this project would not have been possible: Brittany Jennings, Naji Hattar, Vanessa Shorte, and Faiza Wasif.

Dr Biondi designed the project, participated in data analysis, and drafted the manuscript; Drs McCulloh, Garber, Arana, Barsotti, Natt, Alan Schroeder, Lisa Schroeder, Wylie, and Ralston participated in project planning, study design, and key aspects of the quality improvement initiative and reviewed and edited drafts of the manuscript; Drs Staggs and Hall participated in study design, performed the analysis, and provided edits to the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

     
  • AAP

    American Academy of Pediatrics

  •  
  • CI

    confidence interval

  •  
  • ED

    emergency department

  •  
  • IQR

    interquartile range

  •  
  • ITS

    interrupted time series

  •  
  • LOS

    length of stay

  •  
  • MOC

    Maintenance of Certification

  •  
  • QI

    quality improvement

  •  
  • REVISE

    Reducing Excessive Variability in the Infant Sepsis Evaluation

  •  
  • UTI

    urinary tract infection

  •  
  • VIP

    Value in Inpatient Pediatrics

1
Byington
CL
,
Enriquez
FR
,
Hoff
C
, et al
.
Serious bacterial infections in febrile infants 1 to 90 days old with and without viral infections
.
Pediatrics
.
2004
;
113
(
6
):
1662
1666
2
Kimberlin
DW
,
Lin
CY
,
Jacobs
RF
, et al;
National Institute of Allergy and Infectious Diseases Collaborative Antiviral Study Group
.
Natural history of neonatal herpes simplex virus infections in the acyclovir era
.
Pediatrics
.
2001
;
108
(
2
):
223
229
3
Sharp
J
,
Harrison
CJ
,
Puckett
K
, et al
.
Characteristics of young infants in whom human parechovirus, enterovirus or neither were detected in cerebrospinal fluid during sepsis evaluations
.
Pediatr Infect Dis J
.
2013
;
32
(
3
):
213
216
4
Baskin
MN
.
The prevalence of serious bacterial infections by age in febrile infants during the first 3 months of life
.
Pediatr Ann
.
1993
;
22
(
8
):
462
466
5
Biondi
EA
,
Byington
CL
.
Evaluation and management of febrile, well-appearing young infants
.
Infect Dis Clin North Am
.
2015
;
29
(
3
):
575
585
6
Wynn
JL
,
Levy
O
.
Role of innate host defenses in susceptibility to early-onset neonatal sepsis
.
Clin Perinatol
.
2010
;
37
(
2
):
307
337
7
Aronson
PL
,
Thurm
C
,
Alpern
ER
, et al;
Febrile Young Infant Research Collaborative
.
Variation in care of the febrile young infant <90 days in US pediatric emergency departments
.
Pediatrics
.
2014
;
134
(
4
):
667
677
8
Byington
CL
,
Reynolds
CC
,
Korgenski
K
, et al
.
Costs and infant outcomes after implementation of a care process model for febrile infants
.
Pediatrics
.
2012
;
130
(
1
).
9
Hui
C
,
Neto
G
,
Tsertsvadze
A
, et al
.
Diagnosis and management of febrile infants (0-3 months)
.
Evid Rep Technol Assess (Full Rep)
.
2012
;(
205
):
1
297
10
Dalkey
N
,
Helmer
O
.
An experimental application of the DELPHI method to the use of experts
.
Manage Sci
.
1963
;
9
(
3
):
458
467
11
Huppler
AR
,
Eickhoff
JC
,
Wald
ER
.
Performance of low-risk criteria in the evaluation of young infants with fever: review of the literature
.
Pediatrics
.
2010
;
125
(
2
):
228
233
12
Jaskiewicz
JA
,
McCarthy
CA
,
Richardson
AC
, et al;
Febrile Infant Collaborative Study Group
.
Febrile infants at low risk for serious bacterial infection--an appraisal of the Rochester criteria and implications for management
.
Pediatrics
.
1994
;
94
(
3
):
390
396
13
Dagan
R
,
Powell
KR
,
Hall
CB
,
Menegus
MA
.
Identification of infants unlikely to have serious bacterial infection although hospitalized for suspected sepsis
.
J Pediatr
.
1985
;
107
(
6
):
855
860
14
Biondi
EA
,
Mischler
M
,
Jerardi
KE
, et al;
Pediatric Research in Inpatient Settings (PRIS) Network
.
Blood culture time to positivity in febrile infants with bacteremia
.
JAMA Pediatr
.
2014
;
168
(
9
):
844
849
15
Herr
SM
,
Wald
ER
,
Pitetti
RD
,
Choi
SS
.
Enhanced urinalysis improves identification of febrile infants ages 60 days and younger at low risk for serious bacterial illness
.
Pediatrics
.
2001
;
108
(
4
):
866
871
16
Heulitt
MJ
,
Ablow
RC
,
Santos
CC
,
O’Shea
TM
,
Hilfer
CL
.
Febrile infants less than 3 months old: value of chest radiography
.
Radiology
.
1988
;
167
(
1
):
135
137
17
Bramson
RT
,
Meyer
TL
,
Silbiger
ML
,
Blickman
JG
,
Halpern
E
.
The futility of the chest radiograph in the febrile infant without respiratory symptoms
.
Pediatrics
.
1993
;
92
(
4
):
524
526
18
Biondi
E
,
Evans
R
,
Mischler
M
, et al
.
Epidemiology of bacteremia in febrile infants in the United States [published correction appears in Pediatrics. 2014;133(4):754]
.
Pediatrics
.
2013
;
132
(
6
):
990
996
19
Mischler
M
,
Ryan
MS
,
Leyenaar
JK
, et al
.
Epidemiology of bacteremia in previously healthy febrile infants: a follow-up study
.
Hosp Pediatr
.
2015
;
5
(
6
):
293
300
20
Greenhow
TL
,
Hung
YY
,
Herz
AM
,
Losada
E
,
Pantell
RH
.
The changing epidemiology of serious bacterial infections in young infants
.
Pediatr Infect Dis J
.
2014
;
33
(
6
):
595
599
21
Pantell
RH
,
Newman
TB
,
Bernzweig
J
, et al
.
Management and outcomes of care of fever in early infancy
.
JAMA
.
2004
;
291
(
10
):
1203
1212
22
Biondi
E
,
Lee
B
,
Ralston
S
, et al
.
Prevalence of bacteremia and bacterial meningitis in febrile neonates and infants in the second month of life: a systematic review and meta-analysis
.
JAMA Network Open
.
2019
;
2
(
3
):
e190874
23
Baker
MD
,
Bell
LM
,
Avner
JR
.
Outpatient management without antibiotics of fever in selected infants
.
N Engl J Med
.
1993
;
329
(
20
):
1437
1441
24
Mahajan
P
,
Kuppermann
N
,
Mejias
A
, et al;
Pediatric Emergency Care Applied Research Network (PECARN)
.
Association of RNA biosignatures with bacterial infections in febrile infants aged 60 days or younger
.
JAMA
.
2016
;
316
(
8
):
846
857

Competing Interests

POTENTIAL CONFLICT OF INTEREST: Dr Biondi provides consultation for McKesson Corporation; the other authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: Dr Biondi provides consultation for McKesson Corporation and from time to time provides expert testimony; the other authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data