OBJECTIVES: This study evaluates whether bacterial meningitis prevalence differs by urinalysis result and whether antibiotic treatment of presumed urinary tract infection without cerebrospinal fluid (CSF) culture produces adverse sequelae in febrile infants 29 to 60 days old.
METHODS: This retrospective cohort study identified febrile infants 29 to 60 days old presenting to Kaiser Permanente Northern California sites from 2007 to 2015 who underwent urinalysis and blood, urine, and CSF cultures, comparing the prevalence of meningitis among infants with positive versus negative urinalysis results using a two 1-sided test for equivalence. Additionally, febrile infants treated with antibiotics for positive urinalysis results without CSF culture were identified and their charts were reviewed for adverse sequelae.
RESULTS: Full evaluation was performed in 833 febrile infants (835 episodes). Three of 337 infants with positive urinalysis (0.9%; 95% confidence interval [CI]: 0.0%–1.9%) and 5 of 498 infants with negative urinalysis (1%; 95% CI: 0.1%–1.9%) had meningitis. These proportions were statistically equivalent within 1%, using two 1-sided test with a P value of .04. There were 341 febrile infants (345 episodes) with positive urinalysis treated with antibiotics without lumbar puncture. Zero cases of missed bacterial meningitis were identified (95% CI: 0%–1.1%). Zero cases of severe sequelae (sepsis, seizure, neurologic deficit, intubation, PICU admission, death) were identified (95% CI: 0%–1.1%).
CONCLUSIONS: The prevalence of bacterial meningitis does not differ by urinalysis in febrile infants 29 to 60 days old. Antibiotic treatment of infants with positive results for urinalysis without lumbar puncture may be safe in selected cases.
Febrile infants 29 to 60 days often undergo full sepsis evaluation (blood, urine, and cerebrospinal fluid [CSF] cultures with intravenous [IV] antibiotic treatment). Studies reveal the safety of withholding lumbar puncture (LP) and IV antibiotics in low-risk infants, but criteria such as those established in Rochester, Boston, and Philadelphia all recommend the full sepsis evaluation in infants with a positive urinalysis, who are considered to be at high risk for serious bacterial infection.1–3
LP and hospital admission have significant costs to infants, families, and the health care system. Authors of studies report significant risks of minor but high-frequency adverse events such as IV infiltration; social costs to families such as parental anxiety, cessation of breastfeeding, and financial burden; and a cost to the health care system of >$7000 per such admission.1,4–6
The authors of recent retrospective studies have challenged the need for LP and IV antibiotics in febrile infants >28 days old with suspected urinary tract infection (UTI). There is a low prevalence of bacterial meningitis in well-appearing infants with febrile UTI; in 2 large multicenter studies, only 1 of 1206 and 2 of 934 infants with febrile UTI also had bacterial meningitis.7,8 Authors of a smaller study identified 0 cases of bacterial meningitis among 304 infants with febrile UTI.9 Similarly, the prevalence of bacterial meningitis is low in febrile infants with positive urinalysis; in 2011, Paquette et al10 found that among 392 infants 30 to 90 days old who underwent LP for bacterial culture, 57 had an abnormal urinalysis, and, of these, only 1 ill-appearing infant with a low white blood cell (WBC) count had bacterial meningitis. Because well-appearing infants with febrile UTI in this age group may be safely treated with antibiotics on an outpatient basis,11 infants with presumed febrile UTI may not require LP or IV antibiotics. Studies reveal that many febrile infants in this age group are managed without the full sepsis evaluation and suffer no worse outcomes, and some are presumably treated with antibiotics for bacterial infections such as UTI without first undergoing LP.12,13 However, many clinicians may be hesitant to give antibiotics to a young infant without previous LP because of the risk of partially treating an undetected bacterial meningitis. To date, no study has examined the outcomes of infants treated with antibiotics for presumed UTI without previous CSF collection.
This retrospective cohort study aims to evaluate the prevalence of bacterial meningitis among febrile infants 29 to 60 days old with abnormal urinalysis, with an exploratory aim to evaluate other clinical predictors of meningitis. This study also examines the outcomes of infants treated with antibiotics for presumed UTI without previous CSF collection.
We performed a retrospective chart review at all Kaiser Permanente Northern California (KPNC) sites providing pediatric care (clinics, emergency departments, and pediatric wards) between 2007 and 2015. Kaiser Permanente is the largest health care maintenance organization in the United States, with over 3 million members receiving care in over 40 pediatric clinics, 19 emergency departments, and 10 pediatric hospital wards in Northern California. Because it is a closed health care system, most members receive all their care within the system. In a 7-year study of term infants born at KPNC sites, 92.4% of those patients remained within the system for >90 days.14 The Kaiser Foundation Research Institute’s Institutional Review Board approved this data-only, no-patient-contact study with a waiver of consent.
Patient Identification and Exclusion Criteria
We identified patients by querying KPNC’s laboratory database for all urinalyses performed in infants 29 to 60 days old, then reviewing KPNC’s electronic medical record (EMR) system to identify all infants presenting with fever, either at home or in the medical setting. Chart review abstraction was used to collect the reported home temperature, as well as demographic (age, sex, race) and clinical information (preterm birth, birth anomaly, presenting symptoms and history, and blood, urine, and CSF studies). For our primary study outcome, we included only infants who had a CSF culture within 72 hours of the urinalysis.
For our secondary study outcome, we then queried KPNC’s laboratory and pharmacy databases for all infants 29 to 60 days old who had a positive urinalysis who did not have a CSF culture within 72 hours of the urinalysis and who nevertheless received either a parenteral dose of antibiotics or a prescription for oral antibiotics. A chart review of the EMR was performed to identify subsequent diagnoses of bacterial meningitis or other adverse outcome (sepsis, seizure, neurologic deficit, PICU admission, intubation, death) within 1 month of presentation. Notes in the EMR were reviewed up to 1 month after the date of antibiotic administration to identify these events.
Urinalyses and urine cultures obtained by bag, catheter, and clean catch were included. In cases in which multiple specimens were obtained by different methods before antibiotic administration (usually bag followed by catheter), results from the most sterile specimen were used.
Patients were excluded from the study if:
fever occurred during the birth admission; or
the first documented temperature of ≥38°C was >24 hours after initial presentation to the medical setting.
To capture the full range of febrile infants who present for medical care, patients were not excluded on the basis of comorbid medical conditions, prematurity, or technology dependence.
Fever was defined as a measured temperature of ≥38°C at home or within 24 hours of presentation to a clinical site. Infants presenting with >1 episode of fever were counted as separate cases when presentation occurred >72 hours apart and were determined by chart review to represent separate episodes of illness.
A positive urinalysis was defined as the presence of any leukocyte esterase or nitrites on urinalysis, or any bacteria of ≥5 WBC per high-power field (hpf) when microscopy was available.15 The type of urinalysis varied over time and across sites and included both laboratory-performed dipstick and automated urinalysis. When microscopy results spanning the inclusion number (eg, 3–5 WBC per hpf) were reported, we reported these as positive. In some instances, the laboratory applied an adult protocol to perform microscopy only when the dipstick or automated urinalysis revealed positive results. We included infants who underwent any form of urinalysis, including those without microscopy. UTI was defined as having ≥10 000 colony-forming units per mL of a single pathogenic organism from urine obtained by catheterization or having ≥50 000 colony-forming units per mL from urine obtained by urine bag,15,16 with few deviations from this clinical definition for clearly documented reasons (eg, a known contaminated sample or a low concentration of pathogenic growth with a blood culture result positive for the same organism). Bacterial meningitis was defined as pathogenic growth of a single organism from CSF or, if the culture was obtained after antibiotics, a clinical diagnosis by a pediatric infectious disease specialist on the basis of history and CSF indices. In cases of diagnostic uncertainty, charts were reviewed by a pediatric infectious disease specialist (T.G.) who determined whether a diagnosis of bacterial meningitis was probable.
The following possible clinical predictors of bacterial meningitis were identified by chart review and by querying KPNC’s EMR and laboratory database: home temperature, presenting temperature, maximum recorded temperature, ill appearance, evident infection, symptoms of upper respiratory tract infection, fussiness, decreased activity, respiratory distress, diarrhea, vomiting, apnea, seizure, dehydration, decreased urine output (UOP), decreased oral intake, exposure to a sick contact, WBC count of <5000, WBC count of >15 000, absolute neutrophil count (ANC) of <1000, ANC of >5000, and percent bands of >3%.
To limit subjectivity and to capture all potentially ill-appearing infants, clinical ill appearance was defined by the inclusion of the words “sick,” “shock,” “lethargic,” “toxic,” “ill appearing,” “poorly perfused,” “nonresponsive,” “difficult to arouse,” “decreased mental status,” “listless,” “inconsolable,” “hypotonic,” or “floppy” in any physician’s physical examination or assessment and plan at the time of presentation to the medical setting, regardless of resident or attending status or specialty. For all other chart review elements, positive documentation of any historical or physical exam finding by a physician was considered present; in cases of discrepancy between physicians, the attending physician’s note was considered definitive.
All febrile infants 29 to 60 days who underwent urinalysis and blood, urine, and CSF cultures were identified and divided into 2 groups on the basis of positive versus negative urinalysis. The proportion of infants in each group with bacterial meningitis were calculated, and equivalence testing was performed by using a two 1-sided procedure (TOST). Using TOST, the 90% confidence interval (CI) of the difference in proportions must be contained within an equivalency margin of ±δ. Because of a low observed prevalence of meningitis and the inability for the equivalency margin to cross 0, δ = 1% was used in the final analysis.
Age stratified by urinalysis result was described by using median and interquartile range (IQR) and compared by using a nonparametric Kruskal-Wallis test. All other demographic and clinical variables were described by using proportion and percent and compared using a χ2 test or Fisher’s exact test.
Subsequently, a bivariate analysis (Kruskal-Wallis test for continuous variables, χ2 test for categorical variables) was performed to identify clinical predictors of meningitis (home temperature, presenting temperature, maximum recorded temperature, ill appearance, evident infection, symptoms of upper respiratory tract infection, fussiness, decreased activity, respiratory distress, diarrhea, vomiting, apnea, seizure, dehydration, decreased UOP, decreased oral intake, exposure to a sick contact, WBC count of <5000, WBC count of >15 000, ANC of <1000, ANC of >5000, and percent bands of >3%).
For our secondary study outcome, we examined the incidence of adverse events in infants with positive urinalysis treated with antibiotics without previous LP and calculated 95% CIs around the observed incidence.
Urinalysis Result and Bacterial Meningitis
There were 835 episodes of fever evaluated with urinalysis and blood, urine, and CSF cultures in 833 infants (Fig 1). Of these, 337 infants had positive urinalysis and 498 had negative urinalysis.
Out of 835 total urinalyses, 485 were combined urinalysis (urinalysis plus microscopy), and 350 were urinalysis only (urinalysis without microscopy). The prevalence of positive urine culture did not differ between infants with negative combined urinalysis results (3%) and negative urinalysis-only results (4%), with a χ2 P value of .65.
Described in Table 1 are the demographic and clinical information stratified by urinalysis result.
Prevalence of Bacterial Meningitis by Urinalysis Result
There were 8 cases of bacterial meningitis, as follows: 3 cases per 337 infants with positive urinalysis for a proportion of 0.9% (95% CI: 0.0%–1.9%) and 5 cases per 498 infants with negative urinalysis for a proportion of 1% (95% CI: 0.1%–1.9%). These proportions were determined by TOST to be equivalent within a margin of ±1% with a P value of .04.
In Table 2, we present clinical information for the 8 bacterial meningitis cases. All 3 infants with bacterial meningitis and positive urinalysis had positive urine culture for Escherichia coli. Two of 3 were bacteremic with E coli, and 2 of 3 had CSF culture positive for E coli. The diagnosis of bacterial meningitis was controversial in 1 infant with E coli UTI and bacteremia but negative postantibiotic CSF. The consulting pediatric infectious disease specialist diagnosed bacterial meningitis on the basis of a slightly low CSF glucose (37, no simultaneous serum glucose). This infant had received a 100 mg/kg dose of ceftriaxone on admission, followed by a day of ceftriaxone at 50 mg/kg before LP. The ceftriaxone dose was subsequently increased after the diagnosis of bacterial meningitis was made.
All 5 infants with bacterial meningitis and negative urinalysis had negative urine cultures. Two infants had growth of Group B Streptococcus (GBS) in their blood and CSF, 1 infant had growth of GBS in CSF only, 1 infant had growth of E coli in CSF only, and 1 infant had Salmonella typhi bacteremia with negative postantibiotic CSF but considerable CSF pleocytosis, low glucose, and high protein.
Clinical Predictors of Bacterial Meningitis
In a bivariate analysis, clinical variables statistically associated with meningitis (P value < .05) included ill appearance, decreased UOP, vomiting, and higher maximum temperature (Table 3). Each of the 8 patients with bacterial meningitis displayed ≥1 of these variables.
Outcomes of Antibiotic Treatment of Positive Urinalysis Without CSF Culture
There were 345 episodes of fever with positive urinalysis treated with antibiotics without CSF culture in 341 infants (Table 4). Compared with infants with positive urinalysis who underwent blood, urine, and CSF cultures, infants in this cohort were slightly older (median age 48 vs 40 days; P < .001), less likely to have birth anomalies (8.4% vs 13.7%; P = .03), and less likely to be ill appearing (3.2% vs 11.0%; P < .001) (Supplemental Table 5). They were more likely to have a positive urine culture (63% vs 48.1%; P < .001) and less likely to have a positive blood culture (7% vs 12.5%; P = .03).
Chart review revealed no cases with subsequent diagnosis of bacterial meningitis within 30 days (0%; 95% CI: 0%–1.1%) and no cases of related sepsis, seizure, neurologic deficit, PICU admission, intubation, or death (0%; 95% CI: 0%–1.1%).
In this study, we examine the association between urinalysis result and the probability of bacterial meningitis in febrile infants 29 to 60 days old, and we are the first to examine outcomes among febrile infants with positive urinalysis treated with antibiotics without previous LP.
In 835 instances of fever in infants 29 to 60 days old who underwent blood, urine, and CSF cultures, patients with positive urinalysis had an equivalent prevalence of bacterial meningitis compared with patients with negative urinalysis. Bacterial meningitis occurred at a prevalence of 1.0%, comparable to the prevalence described in a 2011 study of 392 febrile infants 29 to 90 days old, of whom 4 were found to have bacterial meningitis (a prevalence of 1%).10 In a 2016 study, Greenhow et al12 used data collected at KPNC clinical sites from 2010 to 2013 to show that laboratory testing was deferred in most febrile infants in this age group with no missed cases of bacterial meningitis; only 25% underwent full evaluation, and 74.5% did not have CSF culture. By extrapolation, the overall prevalence of bacterial meningitis in all febrile infants 29 to 60 days presenting to KPNC clinical sites is ∼0.2%. These findings are consistent with previous evidence that bacteremic UTI may rarely be complicated by bacterial meningitis; however, the rarity of this complication, combined with the relatively low positive predictive value of the urinalysis (48% in this study), is such that positive urinalysis results are not statistically associated with this event. Researchers who examine the association of febrile UTI with bacterial meningitis confirm that these co-occur rarely (0.6% of 322 infants 0 to 90 days old in Vuillermin and Starr,17 0.3% of 1609 infants 29 to 60 days old in Schnadower et al,7 0% of 304 infants 1 to 2 months old in Tebruegge et al,9 and 0.2% of infants 29 to 60 days old in Thomson et al8).
Certain clinical variables revealed a statistically significant association with bacterial meningitis and are as follows: ill appearance, decreased UOP, vomiting, and maximum temperature. Each of the 8 patients with bacterial meningitis displayed ≥1 of these variables on presentation. Statistically, each of these variables was more strongly associated with bacterial meningitis than urinalysis result, suggesting they might be useful to risk stratify infants who would benefit from LP and empiric antibiotics. These findings in an exploratory analysis merit further large study to assess these factors for further risk stratification of febrile infants.
In the second part of our study, our retrospective review revealed that the exercise of clinical judgment by clinicians who chose to treat infants with positive urinalysis results with antibiotics (and no LP) did not lead to late presentation with partially treated bacterial meningitis or other adverse outcomes. Clinicians were selective in the infants they chose to manage this way; these infants were older, lower risk because of the absence of birth anomalies, and more often well appearing. The accuracy of their judgment is reflected in the higher prevalence of UTI, lower prevalence of bacteremia, and absence of cases of missed bacterial meningitis or other adverse outcomes in this cohort. Although a retrospective review cannot exclude the possibility that a patient presented to a different hospital with partially treated meningitis, the likelihood of such an event is mitigated by the structure of KPNC as a closed health care system; most members receive all their care within the system. In a 7-year study of term infants born at KPNC sites, 92.4% of those patients remained within the system for >90 days.14
Other limitations of this study include possible biases of chart abstraction, especially regarding subjective clinical data such as ill appearance; we attempted to mitigate bias by limiting “ill appearance” to the inclusion of key words in the physical examination or assessment and plan. We found that the laboratory inconsistently applied a protocol to perform microscopy on infant urinalyses, forgoing microscopy in 42% of cases (when the urinalysis results were negative, as per the adult protocol). We included infants who underwent any form of urinalysis, because excluding those without microscopy would disproportionately exclude infants with negative urinalysis. Recent data reveal that urine dipstick testing compares favorably to combined urinalysis with a negative predictive value of 98.7%,18 and, in our study, the prevalence of positive urine cultures did not differ between infants with negative combined urinalysis (3%) and negative urinalysis-only results (4%); therefore, we do not expect the partial absence of microscopy to affect validity. The predictive value of the urinalysis may have been affected by the inclusion of bag and clean catch specimens as well as catheterized specimens and our definition of positive microscopy as a WBC count of ≥5 cells per hpf (with the inclusion of those reported as 3–5 WBC per hpf). These choices were made to avoid missing cases where an equivocally positive urinalysis may have been associated with bacterial meningitis. Defining cases of bacterial meningitis was challenged by common confounding circumstances, such as traumatic LPs and LPs performed after antibiotic initiation. These cases were reviewed by a pediatric infectious disease specialist (T.G.).
In our cohort, the urinalysis result was not predictive 1 way or the other for bacterial meningitis, and no adverse outcomes were identified among febrile infants with positive urinalysis who were treated with antibiotics without LP. Previous studies have revealed the safety of treating well-appearing infants with febrile UTI in this age group with antibiotics on an outpatient basis.11 A more selective approach should be considered when deciding whether to perform LP in febrile infants 29 to 60 days with positive urinalysis and no other risk factors. Whether to perform LP and admit for IV antibiotics depends on each clinician’s tolerance for risk, because no protocol, short of obtaining CSF from every patient, can eliminate the risk of missing a case of bacterial meningitis. With these data, we suggest that the urinalysis should not be the deciding factor.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by a grant from the Kaiser Permanente Northern California Community Benefit Program.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Jaskiewicz JA,
- McCarthy CA,
- Richardson AC,
- et al
- Bachur RG,
- Harper MB
- Byington CL,
- Reynolds CC,
- Korgenski K,
- et al
- Schnadower D,
- Kuppermann N,
- Macias CG,
- et al
- Greenhow TL,
- Hung YY,
- Pantell RH
- Roberts KB
- Velasco R,
- Benito H,
- Mozun R,
- et al
- Glissmeyer EW,
- Korgenski EK,
- Wilkes J,
- et al
- Copyright © 2018 by the American Academy of Pediatrics