Impact of CSF Meningitis and Encephalitis Panel on Resource Use for Febrile Well-Appearing Infants

We performed a retrospective review of infants ≤60 days with fever presenting to the Pediatric Emergency Department (PED) of an urban, 300-bed tertiary-care children’s hospital and its affiliated community PED between July 2017 to April 2019. Infants were eligible if they had a chief complaint of fever or documented fever at triage and a CSF culture obtained during workup in the PED. Fever was defined as a temperature ≥38.0°C. Infants transferred from outside facilities that initiated workup and treatment (blood or urine testing, antimicrobial agents) were eligible if CSF was obtained in our PED. There were no direct admissions. Patients were included regardless of admission status.

Chart review was performed by a single author to confirm eligibility. Patients with bronchiolitis, upper respiratory infections, gastroenteritis, and viral meningitis were included because these infants are often the population of interest when deciding to discontinue antimicrobial agents. Patients were excluded if described as ill-appearing in initial ED provider documentation, admitted or transferred to an ICU, or hypothermic (<36.5°C)²²  at triage. Well-appearance was presumed unless documented as ill-appearing, sick-appearing, or lethargic.²³ 

Patients with a diagnosis of HSV or focal or systemic bacterial infection requiring antimicrobial treatment were excluded on the basis that LOS and duration of antimicrobial agents would be determined by what is needed for treatment. We identified patients with HSV and focal or systemic bacterial infections based on review of discharge summary diagnoses and International Classification of Diseases, 10 Revision (ICD-10) codes. Discharge summary diagnoses for focal and systemic infections consisted of urinary tract infection, pyelonephritis, bacteremia, bacterial meningitis, pneumonia, parotitis, skin and soft tissue infections, orbital cellulitis, and septic arthritis. A complete list of ICD-10 codes excluded in this study is available in Supplemental Table 3.

The primary outcome measure was hospital LOS. For patients presenting directly to our ED, LOS was defined as hours from ED registration to ED or inpatient discharge order. For patients transferred from outside facilities, LOS was defined as hours from registration time in our electronic health record when the patient was accepted for transfer to the ED or inpatient discharge order. Secondary outcome measures were antibiotic and acyclovir initiation and duration. Antibiotic or acyclovir initiation was defined as receiving a first dose of medication in either ED or inpatient setting. Antibiotic or acyclovir duration was defined as hours from first to final dose of any antibiotic or acyclovir administered. If a patient received 1 single dose of antibiotic or 2 separate antibiotics (eg, ampicillin and gentamicin) once within 1 hour, antibiotic duration was defined as 1 hour and will be referred to as receiving “1 dose” of antibiotics. If a patient received 1 single dose of acyclovir, acyclovir duration was defined as 1 hour. We identified ME panel testing as patients who had testing from CSF specimens sent from the ED, either ordered by an ED provider or later added on to the ED specimen by the inpatient team.

During the study period, there were no formal guidelines for febrile infant evaluation, ME panel testing or interpretation, initiation or cessation of antimicrobial agents, or discharge criteria from ED or inpatient units. There was no standardized training on ME panel use during site rollout.

We collected through electronic health record download patient demographics, laboratory results, antimicrobial agents administered, and antimicrobial duration. Antimicrobial administration and duration were checked with manual review. Charts were reviewed for past medical history (comorbidities, gestational age at birth, mode of delivery), admission status, admitting team, discharge date and time, outside facility transfer, outside facility laboratory results and antimicrobial administration, ME panel results, and discharge summary and ICD diagnoses. Chart review data were collected and managed using REDCap (Vanderbilt University; Nashville, TN) electronic data capture tools hosted at our study site.^24,25 

We describe our study population using standard descriptive statistics. We used the χ² test to evaluate associations between ME panel testing and independent categorical variables in bivariable analyses. Time in hours was treated as a continuous variable and presented as medians with interquartile ranges (IQR). Based on initial descriptive statistics, our continuous variables were not normally distributed; therefore, we used nonparametric tests (Mann-Whitney U and Kruskal-Wallis) to compare medians by ME panel testing, demographics, and laboratory variables. We performed subgroup analyses for LOS by age group and subset analyses removing groups that might impact LOS (viral diagnoses of bronchiolitis or gastroenteritis, discharge from ED, CSF pleocytosis). Associations between antimicrobial initiation and ME panel testing were analyzed using χ² tests and reported as odds ratios (OR) with 95% confidence intervals (CI). Statistical significance was determined a priori as a 2-tailed P value <.05.

We conducted multivariable analyses with covariates: age group, race and ethnicity, term status, complete blood count (CBC) white blood cell (WBC) count, and CSF pleocytosis. Age was analyzed as a categorical variable (<29 days and 29–60 days). CBC WBC count was defined as normal (5–15 K/mm³) or abnormal (<5 or >15 K/mm³).²⁶  CSF pleocytosis was defined as WBC count >15 cells/mm³ for infants <29 days and >9 cells/mm³ for infants 29 to 60 days.^27,28  Traumatic lumbar puncture cases (red blood cell count ≥10 000 cells/mm³)²⁹  were excluded from CSF analyses but were included in all other analyses. For multivariable analyses, we used logistic regression for antimicrobial initiation and nonparametric ANCOVA for LOS and antimicrobial duration.

Data were analyzed using IBM SPSS Statistics for Windows, version 28 (IBM Corp, Armonk, N.Y., USA). Our study site’s institutional review board certified this study as exempt.

We identified 337 infants ≤60 days with fever and CSF sampling during the study period. Triage temperatures ranged 36.4°C–40.2°C. Before applying exclusion criteria for bacterial infections, there were 3 false negative ME panel results for E.coli K1 because of the number of colony forming units per mL in each CSF sample being below the limit of detection. These patients all grew E.coli K1 from their CSF culture and were excluded from our analyses of LOS and antimicrobial administration. One of these patients with a false negative ME panel for E.coli was positive for EV on the same panel. There were no true or false positive bacterial pathogens by ME panel testing in this cohort of infants during the study period. There were 5 positive CSF culture bacterial pathogens (3 E.coli, 1 Streptococcus agalactiae, 1 Streptococcus bovis); these include the 3 false negative E.coli results and 2 infants that did not have ME panel testing.

We excluded 100 patients for HSV or focal or systemic bacterial infections, intensive care needs, ill-appearance, and/or hypothermia (Fig 1). There were 237 cases that met study criteria. Median age was 28 days (IQR: 20–40). Eighty-five infants (36%) had ME panel testing; 69 had testing in the ER, and 16 had testing added on while admitted. ME panel testing was performed on 41% of infants <29 days and 30% of infants 29 to 60 days of age (Table 1). Twenty (24%) had positive results, which included EV (n = 14), human parechovirus (n = 4), human herpesvirus 6 (n = 1), and cytomegalovirus (n = 1). There was no difference in median CSF WBC count between infants with and without ME panel testing (4, IQR: 2–17 vs 5, IQR: 2–27; P = .83).

There was no difference in median hospital LOS for infants with and without ME panel testing (42 hours, IQR: 36–52 vs 40 hours, IQR: 35–47; P = .09, adjusted P = .06) (Table 2). LOS was not associated with ME panel results (positive 48 hours, IQR: 37–55 vs negative 42 hours, IQR: 36–51; P = .14). In subgroup analyses, there was no difference in median LOS for infants not tested and those with a negative panel (P = .33). There was no association between ME panel testing and LOS in the subset of younger and older infants, admitted patients, or infants without CSF pleocytosis (Table 2). When we excluded infants with gastroenteritis or bronchiolitis, infants with ME panel testing had a significantly longer LOS.

More than 97% of infants with and without ME panel testing were initiated on antibiotics (Table 3). Two percent of infants with ME panel testing received 1 dose of antibiotics compared to 11% of infants without (Fig 2). Of patients receiving antibiotics, those with ME panel testing received antibiotics for a median of 32 (IQR: 25–40) compared to 30 hours (IQR: 23–35) for those without (P < .01). Antibiotic duration was not associated with ME panel results (positive 32 hours, IQR: 25–41 vs negative 32 hours, IQR: 25–40; P = .89).

The odds of acyclovir initiation for infants with ME panel testing was twice that of those without (33% vs 18%; OR: 2.2, 95% CI: 1.2–4.0; aOR: 2.1, 95% CI: 1.1–4.3). There was a difference in acyclovir initiation between infants with ME panel testing in the ER (22 of 69, 32%), ME panel testing added on while admitted (6 of 16, 38%), and those without ME panel testing (28 of 152, 18%) (P = .04). Among infants with ME panel testing, 25% received 1 dose of acyclovir compared with 9% for infants without testing (Supplemental Fig 3). For infants ages <29 days, 41% with ME panel testing were started on acyclovir compared to 16% of infants without (OR: 3.6, 95% CI: 1.5–8.2). For infants 29 to 60 days, ∼20% of infants in both groups were started on acyclovir (OR: 1.0, 95% CI: 0.4–2.8). Of infants started on acyclovir, ME panel testing in general was not associated with a shorter course of acyclovir use (Table 3). There was a difference in the median acyclovir durations for infants who had ME panel testing in the ER (1 hour, IQR: 1–1), ME panel testing added on while admitted (12 hours, IQR 1–19), and those without ME panel testing (4.2 hours, IQR: 1–21) (P = .04). There was no difference in acyclovir duration based on ME panel positivity (P = .97). For infants without bronchiolitis or gastroenteritis, those with ME panel testing similarly had longer antibiotic duration and more acyclovir initiation without a difference in acyclovir doses.

In this cohort, ME panel testing was not associated with decreases in hospital LOS or antimicrobial initiation and duration. This contrasts 4 recent studies that found statistically significant shorter LOS^18,19  and antimicrobial therapy for children with ME panel testing^17–20 ; however, only Posnakoglou et al in a prospective randomized cohort study focused analyses on neonates, specifically infants with CSF pleocytosis. This is the first study evaluating LOS and antimicrobial usage with implementation of the ME panel in febrile well-appearing neonates.

Our study demonstrated higher odds of acyclovir initiation for patients with ME panel testing. Acyclovir use doubled between 1999 to 2012 from 7.6% to 15.6% in infants ≤60 days, predominantly in infants 30 to 60 days and those with milder disease severity.³⁰  With previous CSF HSV PCR testing, infants <29 days had a twofold longer LOS than infants without.¹⁰  There was concern that prolonged acyclovir use during an HSV rule out made physicians perceive patients as more ill than infants not started on acyclovir, thereby leading to prolonged hospitalizations despite negative HSV results.^10,30  In theory, for infants initiated on acyclovir without high-risk features,³¹  the fast turnaround time of the ME panel test should decrease infant exposure time to acyclovir and mitigate prolonged hospitalizations¹⁵ ; however, it may lead to more single doses given if providers initiate acyclovir expecting ME panel results before the second dose. The timing of ME panel results may affect provider antimicrobial decision-making. In our sample, we noted a difference in acyclovir duration when comparing infants with ME panel testing in the ED, added on while as an inpatient, and with no ME panel testing done. Although this would support the hypothesis that ME panel testing would shorten acyclovir duration, it should be interpreted with caution as it is a subgroup analysis of a secondary outcome. Future studies with a larger sample size of infants are needed to evaluate acyclovir use changes with ME panel testing.

The greatest potential benefit of a positive viral ME panel result may be decreasing LOS. In our sample with 24% positive ME panel results, we did not demonstrate this. However, in a larger cohort study, Aronson et al demonstrated a one-third shorter LOS with positive CSF enteroviral testing.³²  There is low risk for bacterial meningitis in children with a positive enteroviral PCR.³² Perhaps with a low-risk fever workup⁵  and a positive ME panel for a viral agent (eg, EV), providers might be comfortable observing off antimicrobial agents or consider discharge from the ED if the patient is stable. During the study period, 1 infant had a positive EV result on the ME panel and a CSF culture positive for E. coli; this may have impacted provider decision-making around a positive viral ME panel.

There is no consensus on when the ME panel should be performed in the evaluation of infants. Blaschke et al demonstrated that the ME panel is highly sensitive and outperformed conventional testing in infant CSF samples.²¹  Other studies have demonstrated its utility in detection of uncommon neonatal pathogens such as Listeria monocytogenes³³  or diagnosing pathogens missed by conventional methods.³⁴  It is unknown whether the ME panel is most useful in practice if it is universally performed on all infants undergoing CSF testing or only those with concerning CSF results (eg, positive Gram stain, pleocytosis). Posnakoglou et al demonstrated a shorter antimicrobial duration and hospital LOS in infants ≤3 months with CSF pleocytosis with ME panel testing.¹⁹  However, Byington et al showed that many infants with EV meningitis lack CSF pleocytosis; therefore, the yield of molecular diagnostic testing may also be high in children who lack a pleocytosis.³⁵  In our study, there was no difference between median CSF WBC counts between those with and without ME panel testing. Standardizing the indication for ME panel testing in febrile infants may help address utilization disparities and reduce differential antimicrobial use.

This study has several limitations. First, as a retrospective study there is risk for confounding by indication³⁶  for LOS, ME panel testing, and antimicrobial use. We attempted to ameliorate potential confounding through our multivariable and subgroup analyses. Second, during the study period, there were no formal guidelines at our institution for febrile infant evaluation and treatment. Decisions regarding LOS and antimicrobial use were provider dependent. We cannot postulate the extent to which our negative findings were driven by a concern for bacteremia and antibiotic decisions were based on pending blood cultures.³⁷  Third, we cannot determine why providers ordered the ME panel and how they interpreted results without formal guidelines and education for ordering physicians; as such, result interpretation and subsequent actions likely varied. Patel et al discusses that rapid diagnostics only improve clinical outcomes if they are coupled with stewardship teams that properly interpret results and apply them to treatment decisions.³⁸  The fact that the ME panel failed to detect 3 cases of E.coli meningitis during the study period might have affected provider perceptions of the test. Fourth, we used WBC count as a covariate in our multivariable analyses, which is not a sensitive indicator of invasive bacterial infections in febrile infants³⁹ ; however, during the study period, inflammatory markers had variable turnaround times and were inconsistently used by providers and hence were not incorporated in our multivariable analyses. Lastly, there is no standardized method to measure antimicrobial utilization; we defined a single dose of antimicrobial administration as 1 hour of duration of treatment, which may not accurately reflect antimicrobial utilization.⁴⁰ 

In our sample, ME panel testing was not associated with decreases in hospital LOS or antimicrobial use. It was associated with acyclovir initiation. Additional clinician guidance, longer use of the ME panel, and a larger cohort study may be required to see an impact of ME panel use on hospital LOS and clinician decision making. These factors should be considered in determining costs/benefits of adopting ME panel use in the evaluation of febrile well-appearing infants.