Testing for Meningitis in Febrile Well-Appearing Young Infants With a Positive Urinalysis

We conducted a secondary analysis of data collected by the Reducing Excessive Variability in Infant Sepsis Evaluation (REVISE) quality improvement project led by the American Academy of Pediatrics (AAP) Value in Inpatient Pediatrics Network, which has led many national quality improvement projects.^15,16  The goal of REVISE was to increase the rate of appropriate evaluation, hospitalization, and length of stay (LOS) for well-appearing febrile infants aged 7 to 60 days presenting to the emergency department (ED) or inpatient setting.¹⁷ Of note, there were no specific recommendations on when to obtain CSF testing included as part of the intervention package for this project. A total of 124 university-affiliated and community hospitals participated, and deidentified data were collected retrospectively on eligible patients evaluated from September 2015 to November 2017. This study was considered exempt by the institutional review board of the AAP. Sites obtained institutional review board approval and data sharing agreements as required by their own site.

Well-appearing infants 7 to 60 days old were included if they were evaluated for fever without a source (temperature ≥38.0°C) in the site’s ED or admitted to the site’s inpatient setting and discharged from the site’s ED or inpatient unit. Patients were excluded if they were not well-appearing on presentation (defined as having documented terms in the chart such as “toxic,” “ill-appearing,” “lethargic,” “sick-appearing”), had comorbidities predisposing to severe or recurrent bacterial illness (including genetic, congenital, chromosomal, neuromuscular, or neurodevelopmental abnormalities), had a discharge diagnosis of bronchiolitis, or transferred to or from another inpatient setting. Eligible patients were identified by investigators at each site, and only patients who met inclusion criteria without any exclusion criteria based on manual chart review were entered into the database. The number of excluded patients and reasons for exclusion were not collected.

Age was recorded dichotomously as either 7 to 30 days or 31 to 60 days. A positive urinalysis result was defined as the presence of a positive leukocyte esterase result (including trace), a positive nitrite result, or >5 white blood cells (WBCs) per high-power field. Presence of abnormal inflammatory markers was recorded as a dichotomous variable and defined as the presence of ≥1 of the following: (1) WBC count <5000 or >15 000/mm³; (2) absolute band count ≥1500 cells per mm³; (3) bands-to-total neutrophil ratio ≥0.2; (4) elevated C-reactive protein or procalcitonin as defined by the institutional range; or (5) CSF WBC count >8/mm³ or positive Gram-stain result (if CSF obtained). Although these CSF studies do not necessarily constitute systemic inflammatory markers, these parameters were included to risk-stratify infants regarding appropriate hospitalization. Data on individual inflammatory marker results were not collected. Receipt of empiric antibiotics was defined as receipt of antibiotics within 24 hours of arrival to the ED or direct admission. LOS was defined as time between the first vital sign and placement of the discharge order (inclusive of the ED visit if at the same institution).

The REVISE data tool was designed to capture physician practice regarding treatment of infection; therefore, patients were categorized as receiving treatment for UTI if there was a “yes” answer to the question, “Did the urine culture grow an organism that was treated as a pathogen with a full course of antibiotics?” Similarly, patients were categorized as receiving treatment for meningitis if there was a “yes” answer to the question, “Did the CSF culture grow an organism that was treated as a pathogen with a full course of antibiotics?” A full course of antibiotics was defined as the patient receiving an antibiotic treatment course for the specified infection as determined by physician chart review. Definitions of bacterial infection were based on definitions used in previous multisite observational studies in this age group.^18,19  A delayed diagnosis of meningitis was 1 of the balancing measures of the quality improvement project and was defined as a “yes” answer to the question, “Did the patient return to the ED or get readmitted to the hospital for new diagnosis of bacterial meningitis within 7 days of the date of treat and release or hospital discharge?” Determination of a new diagnosis of bacterial meningitis was made through physician chart review. Detailed data on infecting organism and treatment regimen were not collected. Information on hospital characteristics was collected through a survey completed by the site lead at the beginning of the project. Team members were trained on data definitions and data entry into a standardized online tool via webinars.

We refer to cumulative incidence proportions as rates. When examining rates of CSF testing by individual hospital site, we limited the sample to sites with ≥10 patients with a positive urinalysis in the respective age groups. To determine the association between the receipt of CSF testing and patient and site characteristics, we included all patients who had a urinalysis performed before the decision to admit or discharge from the ED. We used multilevel mixed-effects logistic regression and included patient and site characteristics that were selected a priori. Hospital site was included as a random effect to adjust for clustering. We “marginalized” the estimated coefficients from our mixed-effects models so that our estimates reflected the average effect across hospitals rather than the effect for an average hospital.²⁰  The rate of delayed diagnosis of meningitis was calculated by using a Clopper-Pearson 1-sided exact 95% confidence interval (CI).²¹  Data were analyzed using Stata version 15 (Stata Corp, College Station, TX).²² 

There were 20 570 well-appearing febrile infants 7 to 60 days old included in the REVISE quality improvement project. The majority (18 379 of 20 570; 89.4%) underwent urinalysis before the decision to admit or discharge from the ED, and 3572 of 18 379 (19.4%) had a positive urinalysis (Fig 1). Urine cultures were obtained by straight catheterization and/or suprapubic aspirate in 89% and “other/unknown” method in 10% of 3572 infants (1% had missing data). Seventy percent of patients with a positive urinalysis had CSF testing performed, and the proportion of infants with CSF testing was higher in the group with a positive urinalysis when compared to the group with a negative urinalysis (70.0% vs 58.1%; P < .001). When comparing infants with a positive versus negative urinalysis who underwent CSF testing, there was no difference in the percentage of infants who received treatment for bacterial meningitis (0.7% vs 0.9%; P = .37; Supplemental Table 4). Among infants with a positive urinalysis, the group that underwent CSF testing had a higher proportion of infants who were male, 7 to 30 days of age, and who had abnormal inflammatory markers when compared with the group who did not receive CSF testing (Table 1).

There was substantial variation in the percentage of infants with a positive urinalysis who underwent CSF testing. Among the 62 hospital sites with ≥10 patients 7 to 30 days old with a positive urinalysis, the proportion of patients who underwent CSF testing ranged from 64% to 100% (Fig 2). Among the 77 hospital sites with ≥10 infants 31 to 60 days old with a positive urinalysis, CSF testing rates ranged from 10% to 100%.

In multivariable analysis, infants with a positive urinalysis were more likely to undergo CSF testing if they were 7 to 30 days of age (adjusted odds ratio [aOR]: 4.6; 95% CI: 3.8–5.5) and had abnormal inflammatory markers (aOR: 2.2; 95% CI: 1.8–2.5) (Table 2). CSF testing was also more likely to be done in patients seen at hospitals with an annual volume of >300 febrile infants (aOR: 1.8; 95% CI: 1.2–2.6) when compared to hospitals with an annual volume of ≤100 febrile infants and less likely to be done in the Midwest region (aOR: 0.6; 95% CI: 0.4–0.8) when compared with the Northeast. Sex, university affiliation, and urban setting were not associated with CSF testing. After accounting for patient and hospital characteristics (Table 2), CSF testing in patients with a positive urinalysis was affected by hospital site (P < .001) with an intraclass correlation coefficient of 11.2% (95% CI: 2.4%–16.9%).

Of the 1061 infants with a positive urinalysis who did not undergo CSF testing, 734 (69.2%) received empiric antibiotics and 505 received a full course of treatment for UTI (Fig 3). Among the 505 infants treated for UTI without CSF testing, most (407 of 505; 80.5%) were 31 to 60 days of age, and median LOS was 49 hours (interquartile range [IQR]: 29–65) (Table 3). Of the 483 infants (95.6%) discharged from the ED or hospitalized for <14 days, 13 (2.7%) had bacteremia. Nine of the 505 infants treated for UTI (1.8%) had a LOS ≥14 days, and 6 of these infants had positive blood culture results and were treated for bacteremia. There were 0 of 505 cases of delayed diagnosis of meningitis (0%; 95% CI: 0%–0.6%) within 7 days of ED or hospital discharge. Infants 7 to 30 days old had 0 of 98 cases (0%; 95% CI: 0%–3.0%), and infants 31 to 60 days old had 0 of 407 cases (0%, 95% CI 0%–0.7%).

In a nationally representative cohort of >3500 well-appearing febrile infants ≤60 days of age with a positive urinalysis, nearly 30% did not receive CSF testing, although these patients are classified as high risk by multiple risk stratification algorithms in which CSF testing would be recommended. We demonstrate wide variation in CSF testing between hospital sites in this population. An age of 7 to 30 days, abnormal inflammatory markers, and high annual volume of febrile infants were associated with increased CSF testing. No cases of delayed bacterial meningitis within 7 days of discharge were documented in infants who were treated for UTI without CSF testing, most of whom were 31 to 60 days old.

Previous studies have reported that a substantial proportion of febrile infants in other high-risk groups (eg, ≤ age 28 days) often do not undergo CSF testing.^5,23  In our cohort of febrile infants with a positive urinalysis, both younger age and abnormal inflammatory markers were associated with increased CSF testing in our model, which would be consistent with criteria promoted by major risk stratification algorithms.^2,–4,24  High institutional volume of febrile infants (>300 infants per year) was also associated with CSF testing, which could reflect institutional guidelines that incorporate published algorithms at larger centers. Previous work has revealed that children’s hospitals with clinical practice guidelines were more likely to perform CSF testing in young febrile infants.⁶ 

Our finding of 0 of 505 cases of delayed meningitis within 7 days among infants who were treated for UTI without CSF testing is consistent with a recent study by Young et al¹⁴  using data from Kaiser Permanente Northern California. They reported 0 of 345 cases of delayed diagnosis of meningitis within 30 days of initial evaluation among infants with a positive urinalysis in the second month of life treated empirically with antibiotics for UTI without CSF testing (213 had a positive urine culture result). Our study demonstrates similar results using a larger, nationally representative sample that includes university-affiliated and community sites. These findings suggest that there may be a subset of well-appearing febrile young infants with a positive urinalysis for whom the risks of CSF testing (procedural complications, family anxiety, traumatic taps, prolonged hospitalizations, or empiric treatment due to indeterminate results) outweigh the benefits (early diagnosis of bacterial meningitis and/or avoidance of the potential difficulty in interpretation of CSF pleocytosis in a pretreated child).

The absence of any reported cases of delayed meningitis within 7 days has several potential explanations. First, it could reflect the overall low rate of concomitant UTI and meningitis, reported to be from 0.8% to 1.2% in the first month of life and 0% to 0.3% in the second month of life.^7–11  Second, these results may also reflect the ability of practitioners to use clinical and laboratory data to select patients who they believed were at higher risk of meningitis and required CSF testing.²⁵ Third, cases of delayed meningitis may not have been captured during the follow-up period if they presented to another institution. Fourth, a small percentage of patients may have been treated presumptively for meningitis despite not receiving CSF testing because of their clinical presentation, presence of bacteremia, or other factors. A lumbar puncture may have been unsuccessful or declined by the family. Nine infants (1.8%) treated for UTI without CSF testing had a LOS ≥14 days, and although detailed clinical information is not available, these infants likely received parenteral antibiotics during that time. Fifth, some infants who received several days of parenteral antibiotics may have had sufficient treatment for early, undiagnosed meningitis.

The strengths of this study are that it represents recent data from a large, nationally representative pediatric cohort including both university-affiliated and community hospitals. In addition, the diagnosis of fever without a source was confirmed by chart review, ensuring appropriate cohort identification and minimizing the probability of misclassification, which may be present in studies in which only administrative data were used.²⁶  Lastly, whereas most previous related studies were focused exclusively on the probability of meningitis in infants with UTI, to our knowledge, this study represents the largest study to date in which the probability of delayed diagnosis of meningitis in infants treated for UTI without initial CSF testing is reported.

There are several limitations to this study. First, the data were collected as part of a multisite quality improvement initiative; consequently, interrater reliability was not assessed, and automated quality checks were not able to be performed. However, sites received standard training on data definitions and data entry through multiple webinars and project materials.

Second, detailed data on symptoms, examination findings, laboratory studies, and treatment regimen were not collected. Thus, our regression model may not have all relevant patient-level characteristics that could influence the decision to perform CSF testing. Third, not all infants received inflammatory markers, and the measure of abnormal inflammatory markers combined tests with variable performance characteristics into a single surrogate, so we were unable to evaluate the relationship between the type of inflammatory marker used or the magnitude of the abnormality on CSF testing. In addition, inclusion of CSF parameters in the abnormal inflammatory marker variable may have overestimated the association of serum inflammatory markers with CSF testing if infants had pleocytosis without elevation of serum inflammatory markers. However, previous research in infants with UTIs has revealed that sterile pleocytosis is associated with higher WBC and band counts.¹⁰  Our data captured a provider’s clinical diagnosis rather than a disease diagnosis based solely on laboratory data, which might overestimate rates of infection because of potential misclassification. However, provider documentation of a treatment course for UTI (with or without laboratory confirmation) has been used as a “gold standard” reference for the accuracy of administrative codes to identify hospitalized infants with UTIs, with an 85% positive predictive value (in contrast to a positive predictive value of 50% for laboratory-confirmed UTI).²⁷  Therefore, our data reflect a higher proportion of provider-confirmed UTI (100%) when compared with the 85% that would be expected with studies using administrative codes.^28 ,29  Additionally, most of our cohort of infants treated for UTI without CSF testing were 31 to 60 days old, so the low rate of delayed meningitis may not apply to the 7 to 30 day age group. Lastly, it is possible that cases of delayed meningitis may have presented after the 7-day follow-up period. Data on timing of presentation of partially treated meningitis is limited; however, 1 retrospective study of 83 children ≤24 months of age pretreated with oral antibiotics before being diagnosed with bacterial meningitis indicated that they presented on average 2.4 days after initiation of antibiotics.³⁰  Those data suggest that most patients ≤24 months of age with delayed meningitis would present within the 7-day follow-up time period. Regardless, more data are likely needed to understand the time course of presentation of delayed meningitis in the ≤60-day age group.

There is significant variation in CSF testing among febrile infants ≤60 days of age with a positive urinalysis , and CSF testing is more likely in infants 7 to 30 days of age, those with abnormal inflammatory markers, and infants evaluated in sites with a high annual volume of febrile infants. There were no cases of delayed diagnosis of meningitis within 7 days of discharge in our large, multicenter cohort of well-appearing febrile young infants treated for UTI without CSF testing (most of whom were 31 to 60 days old), suggesting that CSF testing may not be needed for all infants with a positive urinalysis in the 31 to 60 day age group. Additional large studies are needed to further elucidate the risk/benefit profile for CSF testing in well-appearing febrile infants with a positive urinalysis.

We thank all of the local teams who participated in the REVISE project and Brittany Jennings, Naji Hattar, Faiza Wasif, and Vanessa Shorte at the AAP for their valuable leadership and contributions.

AAP

American Academy of Pediatrics

aOR

adjusted odds ratio

CI

confidence interval

CSF

cerebrospinal fluid

ED

emergency department

IQR

interquartile range

LOS

length of stay

REVISE

Reducing Excessive Variability in Infant Sepsis Evaluation

UTI

urinary tract infection

WBC

white blood cell