Evaluating the Impact of Implementing a Clinical Practice Guideline for Febrile Infants With Positive Respiratory Syncytial Virus or Enterovirus Testing Free

On February 1, 2011, at our institution, we implemented a CPG for febrile infants ≤60 days of age that recommended RSV testing of infants with clinical signs of bronchiolitis and EV testing of cerebrospinal fluid (CSF) from infants evaluated for meningitis. Among low-risk RSV-positive infants, avoidance of both LPs and antibiotics was recommended (Fig 1). For low-risk RSV-positive infants ≤28 days old, hospital admission for observation (without antibiotics) was recommended; for low-risk RSV-positive infants >29 days old, discharge from the hospital with close follow-up was recommended. Among low-risk infants with EV meningitis, discontinuation of antibiotics and consideration of discharge was recommended (Fig 2). This was our institution’s first febrile infant CPG, and no revisions were made to the CPG after initial implementation.

This CPG was developed to address variability in the care of febrile infants, and it was created and implemented by a team consisting of physician and nurse representatives from the clinical areas most impacted by the guideline, physician and nurse informatics experts, experts in literature review, and a family member asked to represent the voice of the family. The quality of evidence was assessed by using the Grading of Recommendations Assessment, Development and Education approach.¹²  The febrile infant CPG was completed in January 2011 and implemented in the Children’s Mercy Hospital emergency departments (EDs), urgent care centers, and hospital in February 2011. An electronic order set guiding practice was developed for febrile infants ≤28 days old and for febrile infants 29 to 60 days old (Supplemental Figures 1 and 2, respectively). The guideline in both written and algorithmic format was placed on the Children’s Mercy Hospital Evidence-Based Practice Web site. In addition, a supporting “app” with decision support was developed. Academic detailing with the ED, urgent care, hospital medicine, and general pediatric groups was completed.

A retrospective study was performed at Children’s Mercy Kansas City (CMKC), a pediatric health system that includes a tertiary-care freestanding children’s hospital, a community-based suburban freestanding children’s hospital (with its own ED), and 3 urgent care centers. Infants ≤60 days of life who presented to our health care system with caregiver report of fever or documented temperature ≥38.0°C (≥100.4°F) between June 25, 2008 and January 31, 2013 were identified from medical records for inclusion by using both the International Classification of Diseases, Ninth Revision discharge diagnosis codes of 780.60 (fever, unspecified) and 780.61 (fever presenting with conditions classified elsewhere), and through query of our electronic health record (EHR). EHR query was done by using Business Objects software (Web Intelligence version 3.×; SAP Business Objects, San Jose, CA), which identified any of the following in medical provider notes: reason for visit being fever, documented temperature ≥38°C, or discharge diagnosis of fever. Qualifying charts were manually reviewed to ensure the presence of a fever.

Febrile infants found to be RSV-positive or to have EV meningitis were divided into 2 separate low-risk cohorts. A subset of infants ≤28 days old were analyzed separately in each low-risk cohort. RSV-positive infants were included in the low-risk RSV cohort if they met our CPG’s criteria for infants at low risk for serious bacterial infection on the basis of risk factors for herpes simplex virus, urinalysis, and/or clinical examination (Fig 1). For purposes of the study, ill-appearing infants included any infant described in the EHR by a health care provider as “irritable,” “ill-appearing,” “lethargic,” in “moderate-severe respiratory distress,” or in any distress not specified as respiratory. Of note, infant age was not a factor in determining which infants met low-risk criteria for RSV-positive infants.

Infants diagnosed with EV meningitis were categorized as low-risk by our CPG if they were afebrile, well-appearing, >7 days of age, and had negative bacterial culture results at 24 hours of incubation (Fig 2). Infants ≤7 days old were considered to be at higher risk because of the risk of disseminated EV among young infants. Because of limitations in manual chart review, neither time to fever resolution nor culture results at 24 hours of incubation could be captured; thus, only infants ≤7 days were excluded from the low-risk EV cohort.

Data were collected on patient demographics, initial clinical presentation, diagnostic tests and results, medications administered, and the following patient outcomes: hospital admission, hospital LOS, subsequent ED visit, and subsequent hospital readmission. Positive bacterial cultures were independently reviewed by 2 authors (A.D. and R.M.) to assess growth of a true pathogen versus a contaminant based on bacteria isolated and urine colony counts.

A positive urinalysis was defined by our febrile infant CPG as having >5 white blood cells per high-power field or positive nitrites. Urinary tract infection was defined by ≥10 000 colony-forming units of a single species of bacterium isolated from a urinary catheter–obtained specimen. Bacteremia and bacterial meningitis were defined by isolation of pathogenic bacteria in the blood or CSF. Presence of coagulase-negative Staphylococcus in the blood or CSF was determined to be a contaminant and was not considered a true infection unless multiple blood cultures grew isolates of coagulase-negative Staphylococcus with identical speciation and antimicrobial susceptibility.

Hours of therapy (HOT) was defined as the aggregate number of hours of antibiotic exposure. The HOT for a patient on multiple antibiotics is the sum of the hours each antibiotic was administered to the patient. For example, the HOT for an infant who received 24 hours of ampicillin and 24 hours of cefotaxime is 48 hours. HOT is derived from antibiotic days of therapy, a well-described method to measure antibiotic exposure and stewardship.^13–15  Our data set provided sufficient detail to determine antibiotic exposure at the level of HOT based on the date and time antibiotics were administered. We defined a subsequent visit and readmission as presenting or being admitted to our health care system because of a febrile illness within 14 days of discharge.

Hospital LOS was defined as the hours between the infant’s admission time and date and the infant’s discharge time and date as recorded in our health system’s EMR. Time spent outside of an inpatient unit, including time spent in the ED or urgent care among those infants discharged from the hospital from EDs or urgent cares, was not included in LOS calculations.

RSV testing during the study period was performed either as part of a multiplex polymerase chain reaction (PCR) panel (Biofire FilmArray or Luminex xTAG RVP) or by rapid antigen testing (BD Veritor System, BD Directigen EZ RSV, or Binax NOW RSV). Turn-around time for RSV testing ranged from ∼1 to 15 hours depending on which testing modality was used and from which CMKC location the sample was collected.

EV testing during the study period was performed via real-time reverse transcription PCR, by an Cepheid Enterovirus analyte-specific reagent assay (Cepheid, Sunnyvale, CA), an Enterovirus R-gene (bio-Mérieux, Marcy-l'Étoile, France), or via an in-house assay as described previously,¹⁶  with the exception that during the nonpeak EV incidence months of 2008 to 2010, EV testing was performed at a reference laboratory. Reference laboratory EV testing time to results ranged from ∼24 to 48 hours. In-house EV testing time to results ranged from ∼8 to 43 hours depending at which CMKC location the specimen was collected and on the time of year.

Primary outcomes among low-risk RSV-positive infants included the proportion of infants undergoing LPs, the proportion of infants who received antibiotics, and HOT and LOS pre- versus post-CPG implementation. Primary outcomes among low-risk infants with EV meningitis included HOT and LOS pre- versus post-CPG implementation. A subanalysis of low-risk infants with EV meningitis that excluded those who had testing performed at a reference laboratory, which had a longer turn-around-time of 24 to 48 hours, was conducted to assess whether observed differences in LOS and HOT were attributable to this subset of infants. Secondary outcomes among both cohorts included the proportion of subsequent ED visits and hospital readmissions pre- versus post-CPG implementation. We also determined the rates of positive blood, urine, and CSF bacterial cultures among both cohorts.

Fisher’s exact or χ² tests were used for categorical variables and 2-sample Wilcoxon rank (Mann–Whitney U test) tests were performed for continuous variables, as appropriate. Statistical significance was regarded at P < .05 (2-sided). Statistical analyses were performed by using Stata 13.0 (StataCorp, College Station, TX). The hospital’s institutional review board approved the study.

Demographic and clinical outcome data collected from a total of 2036 febrile infants are presented in Table 1. A flow diagram depicting the included 134 low-risk RSV-positive infants, of whom 131 were diagnosed with RSV through rapid antigen testing, and the 274 low-risk infants with EV meningitis is presented in Fig 3.

In the low-risk RSV cohort, the proportion of LPs performed, the proportion of infants receiving antibiotics, HOT, and LOS were unchanged post-CPG implementation (Table 2). In the low-risk EV cohort, HOT (median 79 hours pre-CPG versus median 46 hours post-CPG, P < .001) and LOS (median 47 hours pre-CPG versus median 43 hours post-CPG, P = .01) both decreased post-CPG implementation (Table 2). Among infants ≤28 days old included in the low-risk EV cohort, HOT decreased (median 90 hours pre-CPG versus median 75 hours post-CPG, P = .03) and LOS was unchanged (Table 2).

A subanalysis of infants with EV meningitis that excluded 27 infants whose EV testing was performed at a reference laboratory showed a similarly significant decrease in both LOS and HOT post-CPG implementation.

There were no differences in subsequent ED visits or hospitalizations post-CPG implementation in either cohort. In the low-risk RSV cohort, 8.2% of pre-CPG versus 12.3% of post-CPG (P = .57) infants had subsequent ED visits, and 4.9% of pre-CPG versus 9.6% of post-CPG (P = .35) infants had subsequent hospital readmissions. All revisits or readmissions were because of symptoms related to bronchiolitis, and there were no positive results for blood, urine, or CSF bacterial cultures noted on subsequent ED visits or readmissions. In the low-risk EV cohort, 2.9% of pre-CPG versus 0% of post-CPG (P = .34) infants had subsequent ED visits and 1% of pre-CPG versus 1.5% of post-CPG (P = .56) infants had subsequent hospital readmission. All revisits or readmissions were because of recurrence of fever, and there were no positive results for blood, urine, or CSF bacterial cultures noted on subsequent ED visits or readmissions.

No cases of bacterial meningitis or bacteremia were diagnosed in either low-risk RSV-positive infants or in low-risk infant with EV meningitis. Urine culture results were positive in 0% of low-risk RSV-positive infants and in 1.2% of low-risk infants with EV meningitis. As mentioned above, infants were excluded from our low-risk RSV cohort if they had a urinalysis that contained nitrites or >5 white blood cells per high-power field.

We noted significant differences in patient outcomes when comparing low-risk RSV-positive infants and low-risk infants with EV meningitis evaluated pre-CPG to all other febrile infants evaluated pre-CPG. Specifically, compared with all other febrile infants evaluated pre-CPG, low-risk RSV-positive infants evaluated pre-CPG were observed to have a decreased proportion of both LP obtainment (29% of RSV-positive infants versus 70% of all other febrile infants, P < .001), and antibiotic administration (31% of RSV-positive infants versus 70% of all other febrile infants, P < .001). Compared with all other febrile infants evaluated pre-CPG, low-risk infants with EV meningitis evaluated pre-CPG had decreased HOT (79 hours [interquartile range (IQR) 48–96] versus 90 hours [IQR 52.5–105], P = .001) and LOS (47 hours [IQR 38–56.5] versus 48 hours [IQR 41–63], P = .03).

We found that implementing a febrile infant CPG that gives explicit management recommendations based on RSV and EV test results was associated with no changes in the proportion of infants undergoing LP, antibiotic exposure, or LOS among low-risk RSV-positive infants but that HOT and LOS decreased among low-risk infants with EV meningitis. We also found a low rate of concomitant bacterial infections in both cohorts. CPG implementation was not associated with adverse patient outcomes, suggesting that adjusting clinical management of a well-defined low-risk subset of febrile infants based on RSV and EV test results is reasonably safe.

The lack of observed changes among low-risk RSV infants is likely multifactorial, but it could be attributed to preexisting clinician knowledge and skill in incorporating RSV testing into febrile infant management. Compared with all other febrile infants evaluated pre-CPG at our institution, low-risk RSV-positive infants evaluated pre-CPG were observed to have a decreased proportion of both LP obtainment and antibiotic administration. Thus, on the basis of these observations, we speculate that CPG recommendations did not reflect a novel way of managing these patients in our health system, and that clinicians may have already been incorporating RSV testing and interpretation into their management practices. LOS was unchanged among low-risk RSV-positive infants. This in our view is not surprising, given that RSV-positive infants likely had many confounding factors contributing to LOS, particularly symptoms of bronchiolitis. Additionally, although our CPG provided explicit discharge criteria for patients with EV, no similar criteria for discharge for RSV-positive infants were provided. No changes in management or outcomes were observed among low-risk RSV-positive infants ≤28 days of life, a population expected to be more conservatively managed pre-guidelines. This lack of change may be because of small sample size in this subset of infants, or health care providers may still have had uncertainty regarding optimal management of these young infants even after CPG implementation.

Our study highlights the positive role that EV testing, when combined with explicit management recommendations for infants with EV meningitis, can play in medical decision-making. We build on previous research performed at both our institution and elsewhere that describes an associated reduction in hospital LOS and antibiotics received among febrile infants with positive EV testing.^17,18  Although EV testing had been available at our institution for more than a decade and had been associated with substantial reductions in antibiotic exposure and LOS,¹⁷  we observed even further decreases in both post-CPG implementation. Specifically, compared with all other febrile infants evaluated pre-CPG, low-risk infants with EV meningitis had decreased HOT and LOS. Post-CPG implementation, we observed even further decreases in HOT and LOS for low-risk infants with EV meningitis (Table 2). Thus, we propose that guidelines that give explicit management recommendations based on test results are likely needed to optimize changes in practice patterns. We do note the discrepancy in the number of low-risk infants with EV meningitis evaluated pre-CPG versus post-CPG implementation. Because testing patterns were unchanged pre-CPG versus post-CPG (Fig 3), we attribute this discrepancy to natural temporal fluctuations that occur with EV.¹⁹ 

Our study has several limitations. This review was performed at a single pediatric health system and was not a multicenter review, decreasing its generalizability. Febrile infant encounters may have been misclassified by using our case identification strategy; however, this risk was reduced through manual confirmation of patient encounter eligibility. Because HOT is based on the aggregate number of antibiotics given, the use of HOT as a marker of antibiotic exposure could confound results if infants presenting post-CPG received a different number of antibiotics than infants evaluated pre-CPG. We find this unlikely to affect our results because there were no significant differences in the number of antibiotics given to either low-risk RSV-positive infants or low-risk infants with EV meningitis pre-CPG versus post-CPG. A subanalysis of infants within each low-risk cohort broken down by age ≤28 or >29 days similarly revealed no significant change in number of antibiotics given pre-CPG versus post- CPG. We cannot fully assess the impact that variations in test turnaround time could have had on clinical course or patient outcomes. Because of the retrospective nature of this study, changes seen in clinical practice patterns and patient outcomes cannot be attributed to CPG implementation with certainty. However, clinicians did use an electronic order set, which was linked to and based on the CPG in 80% of febrile infant encounters, suggesting that elements of the CPG were at least reviewed in most clinical encounters. Finally, patient and provider characteristics that may have driven viral testing were not analyzed. There may be a selection bias among infants who underwent viral testing, which we were unable to account for in a retrospective study. Specifically, although the proportion of infants tested for EV was unchanged, it is unknown if inherent clinical differences among infants tested for EV pre- versus post-CPG affected our results.

Integrating testing and clinical management recommendations for RSV and EV into CPGs for febrile infants may enhance medical decision-making and poses a minimal risk of missing concomitant bacterial infections. However, merely including recommendations regarding test interpretation into a CPG may be insufficient. Further studies are needed to determine specific barriers and facilitators to effectively incorporate diagnostic viral testing into medical decision-making for these infants.