To determine whether cerebrospinal fluid (CSF) enterovirus polymerase chain reaction (PCR) testing of febrile neonates is associated with a shorter hospital length of stay (LOS).
Retrospective cohort study.
Urban tertiary care children's hospital emergency department.
Febrile infants 56 days or younger evaluated by means of lumbar puncture.
Performance of CSF enterovirus PCR testing.
Main Outcome Measure
A CSF enterovirus PCR test was performed in 361 of 1231 eligible infants (29.3%); 89 of those tested (24.7%) were positive. The median LOS was 2 days. In multivariable analysis, CSF enterovirus PCR testing was associated with a 26.0% shorter LOS for infants with a positive test result (adjusted β coefficient, −0.305; 95% confidence interval, −0.457 to −0.153; P < .001) and an 8.0% longer LOS for those with a negative test result (0.075; −0.021 to 0.171; P = .12) compared with untested infants. In stratified analysis, LOS was shorter for all infants 28 days or younger who tested positive regardless of receipt of antibiotics before lumbar puncture. For infants 29 to 56 days old, a positive test result was associated with a shorter LOS only in those not previously receiving antibiotics. The median (interquartile range) turnaround time for CSF enterovirus PCR testing was 22.2 (15.1-27.4) hours, with no effect of turnaround time on LOS.
Among infants 56 days or younger, a positive CSF enterovirus PCR test result was associated with a shorter LOS compared with untested infants. The CSF enterovirus PCR test may improve the care of infants with fever.
Neonates and young infants routinely undergo lumbar puncture during the emergency department evaluation of fever.1,2 Although the prevalence of bacterial meningitis is low,3 these infants do not typically manifest the classic signs of meningitis found in older children and adults.4- 6 Furthermore, clinical prediction rules incorporating the results of lumbar puncture do not accurately distinguish young infants with aseptic meningitis from those with bacterial meningitis.7- 12 This differentiation is important because aseptic meningitis is usually self-limited, whereas bacterial meningitis requires prompt antibiotic therapy.4,11 Owing to uncertainty in diagnosis, these infants are often hospitalized to receive empirical antibiotic therapy while awaiting culture results.4,5,13- 16
Cerebrospinal fluid (CSF) enterovirus polymerase chain reaction (PCR) testing can help guide physicians in the management of young febrile infants because it provides for rapid and accurate diagnosis of infants with enterovirus meningitis,1,17- 21 the most common cause of aseptic meningitis.8,10,12 In older children, CSF enterovirus PCR testing is associated with decreased hospital length of stay (LOS) for those ultimately diagnosed as having enterovirus meningitis.11,12,20,22 However, these studies did not adjust for additional factors that may affect LOS and did not address the use of CSF enterovirus PCR testing in young infants, where the approach to treatment is often more conservative than that in older children.11,12,20 King et al23 found that a positive CSF enterovirus PCR test result during the enterovirus season was associated with a decrease in LOS in infants younger than 3 months. However, that study included only infants tested for enteroviral meningitis and, thus, could not compare LOS between tested and untested patients. Routine enterovirus PCR testing provides benefit only if positive test results decrease the LOS and negative test results do not increase the LOS compared with no testing.
The objective of this study was to evaluate the impact of CSF enterovirus PCR testing on hospital LOS for infants 56 days or younger. Specifically, we aimed to determine the effect of CSF enterovirus PCR testing on LOS for tested patients (with positive or negative results) compared with untested patients.
This retrospective cohort study was conducted at the Children's Hospital of Philadelphia, an urban tertiary care children's hospital. The Committees for the Protection of Human Subjects of the Children's Hospital of Philadelphia approved this study, with a waiver of informed consent.
Febrile neonates and young infants (birth through 56 days of age) who underwent a lumbar puncture in the emergency department between January 1, 2005, and December 31, 2007, were eligible for inclusion in the study. At the Children's Hospital of Philadelphia, infants 56 days or younger routinely undergo lumbar puncture during evaluation for fever.1,2,16 We excluded infants who had insufficient CSF for cell count determination because lack of CSF precluded further testing, including CSF enterovirus PCR testing.
The main exposure was CSF enterovirus PCR testing, classified into 3 categories: not performed, performed and results negative, and performed and results positive. The primary outcome was hospital LOS.
All historical data, vital signs, and clinical findings were obtained from the initial evaluation. Patients not tested for enterovirus at initial presentation were classified as “CSF enterovirus PCR test not performed.” Fever was defined as a reported tactile temperature or a recorded temperature higher than 38.0°C at or before presentation. Other vital sign abnormalities were defined as follows: hypoxia, percutaneous oxygen saturation level less than 90% or receipt of supplemental oxygen; tachypnea, respiratory rate higher than 70/min24; hypothermia, temperature lower than 36.1°C; and hypotension, systolic blood pressure lower than 63 mm Hg.25 Prematurity was defined as gestational age younger than 37 weeks.
Serious bacterial infection was defined as bacterial meningitis, bacteremia, or urinary tract infection. Bacterial meningitis was defined as either the isolation of a known bacterial pathogen from the CSF or, in patients who received antibiotics before evaluation, the combination of CSF pleocytosis and bacteria reported on CSF Gram stain. Bacteremia was defined as the isolation of a known bacterial pathogen from blood cultures excluding commensal skin flora. Urinary tract infection was defined as growth of a single known pathogen meeting 1 of 3 criteria: (1) at least 1000 colony-forming units (CFUs) per milliliter for urine cultures obtained by means of suprapubic aspiration, (2) at least 50 000 CFUs/mL from a catheterized specimen, or (3) at least 10 000 CFUs/mL from a catheterized specimen in association with positive urinalysis results.26- 28
Hyponatremia was defined by serum sodium levels less than 131 mEq/L (to convert to millimoles per liter, multiply by 1.0). Thrombocytopenia was defined by platelet counts less than 150 ×103/μL (to convert to ×109 per liter, multiply by 1.0). Levels of CSF glucose less than 40 mg/dL (to convert to millimoles per liter, multiply by 0.0555) were considered low. Levels of CSF protein greater than 0.17 g/dL (to convert to grams per liter, multiply by 10.0) for infants 28 days or younger or greater than 0.08 g/dL for infants aged 29 to 56 days were considered elevated.29 Pleocytosis of CSF was defined by previously published criteria: white blood cell count greater than 22/μL (to convert to ×109 per liter, multiply by 0.001) for infants 4 weeks or younger and greater than 15/μL for infants older than 4 weeks.29 Traumatic lumbar puncture was defined by a red blood cell count greater than 500 ×106/μL3 (to convert to ×1012 per liter, multiply by 1.0).30
The enterovirus season was defined as June 1 through October 31 for each year.31,32 The CSF enterovirus PCR test turnaround time was defined as the difference, in hours, from test ordering to result availability.
As previously described,23 enterovirus RNA was extracted from the CSF using an automated MagNAPure LC instrument (Roche Diagnostics Corporation, Indianapolis, Indiana) and total nucleic acid isolation kit (Roche Diagnostics). Amplification and detection were completed using either a sequence detection system (ABI Prism 7000; Applied Biosystems, Foster, California) or a 7500 real-time PCR system (Applied Biosystems). Standard techniques used by the Clinical Virology Laboratory at The Children's Hospital of Philadelphia to minimize the likelihood of false-positive and false-negative results have also been described previously.23 Because we used automated, magnet bead–based extraction technology, blood contamination of the CSF from traumatic lumbar punctures would not be expected to affect the PCR yield.33
Data were abstracted from medical records, entered on a standardized data collection form, and analyzed using a statistical software program (STATA 10; Stata Corp, College Station, Texas). Continuous variables were compared using the Wilcoxon rank sum test and are described using mean, median, range, and interquartile range (IQR) values. Categorical variables were compared using the χ2 test and are described using counts and percentages. Univariable analyses were conducted to identify factors associated with LOS. Multivariable linear regression was performed to identify factors independently associated with LOS. Because the LOS variable had a skewed distribution (P < .001, Shapiro-Wilk test), we used the logarithmically transformed LOS value as the dependent variable. The resulting β coefficients were transformed to reflect the percentage difference in LOS.
The multivariable model initially incorporated age, CSF white blood cell count, diagnosis of serious bacterial infection, presence of seizures, enterovirus season, and the administration of antibiotics before lumbar puncture based on the a priori hypotheses of their effect on LOS. Additional variables were considered for inclusion in the multivariable model if they were associated with LOS in univariable analyses (P < .20).34 These variables were included in the final multivariable model if they remained statistically significant after adjustment for other factors or if their inclusion in the model resulted in a greater than 10% change in the effect size of the primary association of interest, enterovirus PCR testing and LOS, regardless of statistical significance.35 Statistical significance was determined a priori as a 2-tailed P < .05.
The possibility of important subgroup differences (ie, differences in outcome based on enterovirus season, age, antibiotic use before lumbar puncture, or CSF pleocytosis) was explored using interaction terms. The significance of these subgroups in multivariable analysis was evaluated using the likelihood ratio test. Because there was a significant interaction between age and the administration of antibiotics before lumbar puncture (P < .001, likelihood ratio test), the analysis was subsequently stratified according to these variables.
We examined the association between CSF enterovirus PCR test turnaround time and LOS in the subset of infants for whom CSF enterovirus PCR testing had been performed using linear regression with logarithmically transformed LOS values. The goal of this analysis was to determine whether differences in LOS could be due to test result availability. In univariable analysis, we estimated that 3 years of data collection would provide 80% power (α = .05) to detect a 17% difference in LOS between tested and untested infants based on a mean (SD) LOS of 2.5 (4.6) days.
During the study, 1256 febrile infants 56 days or younger underwent lumbar puncture in the emergency department. Twenty-five infants (2.0%) were excluded owing to insufficient CSF to perform a cell count, and the remaining 1231 infants were included in the subsequent analyses. The median (IQR) patient age was 32 (16-44) days (Table 1). The CSF enterovirus PCR test result was positive in 89 of 361 patients (24.7%) who were tested; most testing (77.6%) occurred during the enterovirus season, although only 535 infants (43.5%) presented during this period. Results of CSF enterovirus PCR testing were positive in 80 of 281 patients (28.5%) during the enterovirus season and in 9 of 81 patients (11.1%) outside the enterovirus season. Results of blood enterovirus PCR testing were positive in 13 of 42 tested infants (31.0%); 11 of these infants (84.6%) also had a positive CSF enterovirus PCR test result, 1 had a negative CSF enterovirus PCR test result, and 1 did not undergo CSF enterovirus PCR testing. Overall, 277 patients (22.5%) received antibiotics before lumbar puncture; 65.3% of infants previously receiving antibiotics were 28 days or younger. Results of initial laboratory studies are given in Table 2. Pleocytosis of the CSF was present in 34.4% of infants undergoing CSF enterovirus PCR testing and in 16.9% of those not tested (P < .001, χ2 test).
The median (IQR) LOS for all patients was 2 (1-3) days. For patients 28 days or younger, the median LOS was 3 (2-5) days, whereas for patients 29 to 56 days of age, the median LOS was 2 (1-2) days (Figure). In unadjusted analysis, there was no difference in LOS for those tested vs untested (unadjusted β coefficient, 0.077; 95% confidence interval [CI], −0.015 to 0.170; P = .10). However, compared with untested patients, the unadjusted LOS was 22.3% shorter in patients with a positive CSF enterovirus PCR test result (β coefficient, −0.252; 95% CI, −0.412 to −0.092; P = .002) and 20.6% longer in those with a negative result (0.187; 0.086 to 0.290; P < .001). The median LOS was longer for infants who received antibiotics before lumbar puncture (median [IQR] LOS, 3 [1-3] days) compared with those who did not (2 [1-3] days) (P < .001, Wilcoxon rank sum test).
Boxplot of the hospital length of stay stratified by age and results of cerebrospinal fluid enterovirus polymerase chain reaction testing. The top and bottom borders of each box mark the 25th and 75th percentiles, respectively, and the whiskers extend to 1.5 times the interquartile range. The black squares denote the median values.
Factors associated with a longer LOS in univariable analysis included enterovirus season, younger age, birth by cesarean delivery, and antibiotic use before lumbar puncture (Table 3). Abnormalities in certain laboratory test results, including hyponatremia, an elevated peripheral white blood cell count, thrombocytopenia, a low CSF glucose level, and CSF pleocytosis, were also associated with a longer LOS.
In a multivariable analysis, there was no overall difference in LOS between tested and untested infants (adjusted β coefficient, −0.004; 95% CI, −0.094 to 0.087; P = .94). However, a positive CSF enterovirus PCR test result was associated with a 26.0% shorter LOS (adjusted β coefficient, −0.305; 95% CI, −0.457 to −0.153; P < .001), and a negative test result was associated with an 8.0% longer LOS (0.075; −0.021 to 0.171; P = .12) compared with untested infants; the difference in LOS between untested infants and those with a negative enterovirus PCR test result was not statistically significant (P = .12). Stratified analysis demonstrated shorter LOS for all infants 28 days or younger who tested positive for enterovirus compared with untested infants regardless of receipt of antibiotics before lumbar puncture (Table 4). For infants aged 29 to 56 days, a positive test result was associated with a shorter LOS only in patients who did not receive antibiotics before lumbar puncture (Table 4 and eTable).
The median (IQR) CSF enterovirus PCR test turnaround time was 22.2 (15.1-27.4) hours; the results of CSF enterovirus PCR testing were available before discharge for 317 of the 361 infants tested (87.8%). Most CSF enterovirus PCR test results (90%) were available within 36 hours; 95% of results were available within 48 hours. The difference in median CSF enterovirus PCR test turnaround time between infants with positive (23.1 hours) and negative (22.0 hours) results was not significant (P = .23, Wilcoxon rank sum test).
There was no difference in turnaround time for the enterovirus PCR testing performed during (22.0 hours) or outside of (22.3 hours) the enterovirus season (P = .50, Wilcoxon rank sum test). Among infants with a positive CSF enterovirus PCR test result, no effect of turnaround time on LOS was observed when stratified by age (age ≤28 days: β coefficient, −0.002; 95% CI, −0.272 to 0.023; P = .86; and age 29-56 days: 0.006; −0.007 to 0.019; P = .33). No effect of turnaround time on LOS was observed when stratified by antibiotic use before lumbar puncture (antibiotic use before lumbar puncture: β coefficient, 0.001; 95% CI, −0.018 to 0.020; P = .91; and no antibiotic use before lumbar puncture: 0.005; −0.011 to 0.022; P = .53). Finally, there was no effect of turnaround time on LOS (β coefficient, 0.006; 95% CI, −0.002 to 0.014; P = .13) in the subset of patients whose positive enterovirus PCR test result was available before discharge.
We found that in a large cohort of febrile infants, a positive CSF enterovirus PCR test result was associated with a shorter hospital LOS. The LOS for patients who tested negative was not significantly different compared with the LOS for untested patients. We believe that these findings lend support to routine CSF enterovirus testing for young febrile infants during the enterovirus season.
Two important factors affected the association between LOS and CSF enterovirus PCR testing—age and the receipt of antibiotics before lumbar puncture. The magnitude of reduction in LOS was similar for patients 28 days or younger and those 29 to 56 days of age; however, infants 28 days or younger had a longer median LOS than did infants 29 to 56 days of age. Febrile infants in this younger age group have a higher prevalence of serious bacterial infection1,36- 41 compared with older infants and children.42- 44 Furthermore, the diagnosis of meningitis is difficult because of the poor sensitivity of clinical examination in identifying young infants with any serious bacterial infection, including bacterial meningitis.16 Thus, we hypothesize that physicians hospitalize infants 28 days or younger for longer periods than they do infants 29 to 56 days of age because of concern about serious bacterial illness without overt clinical signs to indicate bacterial infection. The magnitude of the decrease in LOS for infants 28 days or younger who were enterovirus PCR positive was similar to that for infants aged 29 to 56 days.
The administration of antibiotics before lumbar puncture also affected the association between LOS and CSF enterovirus PCR testing. It is likely that enterovirus testing may disproportionately decrease LOS for patients who received antibiotics before lumbar puncture. The magnitude of reduction in LOS for those with a positive enterovirus PCR test result compared with those not tested was greater in the subsets that received antibiotics before lumbar puncture. Physicians cannot rely on negative culture results to exclude the possibility of bacterial meningitis when antibiotics have been administered before lumbar puncture because CSF bacterial cultures may become sterile within 15 minutes after antibiotic administration.45 Thus, previous antibiotic therapy may lead to longer hospitalizations as physicians elect to “observe” pretreated infants longer than infants who were not pretreated. We hypothesize that because of the small risk of bacterial and viral meningitis co-infection,46 physicians feel confident in attributing fever to enteroviral meningitis in the face of a potentially unreliable culture result, thereby facilitating earlier hospital discharge.
We explored the relationship between PCR test turnaround time and LOS and found no significant difference in turnaround time for patients with a positive compared with a negative CSF enterovirus PCR test result. This finding assures us that differences in LOS are not attributable to CSF enterovirus PCR test result availability. A previous study23 at The Children's Hospital of Philadelphia found that for each 24-hour increase in turnaround time for infants with a positive CSF enterovirus PCR test result, the LOS increased by 13.6%. However, no association was found between LOS and turnaround time in the present study. Over time, the Children's Hospital of Philadelphia has transitioned from limiting enterovirus PCR testing to weekdays to performing enterovirus PCR testing daily. In this study, almost all CSF enterovirus PCR test results were available within 36 hours. Thus, it is possible that we reached a “floor effect” in which physicians were not comfortable discharging patients sooner than 36 hours despite having a positive CSF enterovirus PCR test result.
This study has several limitations. First, as in all observational studies, confounding by indication is possible because infants tested for enterovirus infection may be different from untested infants. We minimized confounding by adjusting for measured confounders, but we could not account for unmeasured confounding. However, because the adjusted LOS was not statistically different for untested infants compared with those who tested negative for organisms, it is likely that such confounding, if present, was accounted for in the multivariable analysis. Second, this study includes data from a single hospital, so generalizability to other institutions may vary according to different standards of treatment for young infants and the use and logistics of enterovirus testing. Third, we believe that the availability of CSF enterovirus PCR test results in a clinically relevant time frame depends on the specimen's timely arrival at the laboratory and on the ability of the laboratory to process the specimen and perform the test quickly. The turnaround time in this study was within a relatively narrow range. Other hospitals with fewer laboratory resources or greater demand for testing may show greater variability in turnaround time and, thus, clinically meaningful associations between test delays and LOS. Specifically, physicians practicing in settings where CSF enterovirus PCR testing is performed by commercial reference laboratories are unlikely to achieve meaningful reductions in LOS by routinely testing for enterovirus. Fourth, although enterovirus infections may occur throughout the year, the benefit of enterovirus testing may be greater during the enterovirus season, when the disease is most prevalent. In this study, too few tests were performed outside of the peak enterovirus season to allow us to comment on the role of enterovirus testing during nonpeak months. Fifth, this study was underpowered to detect differences in certain subgroups. The reduction in LOS was statistically significant for all subgroups except patients aged 29 to 56 days who received antibiotics before lumbar puncture; there were relatively few patients in this subgroup, making this study inadequately powered to detect even modest reductions in LOS in this subgroup.
In conclusion, we found that a positive CSF enterovirus PCR test result decreases hospital LOS for febrile infants 56 days or younger. Further studies should assess this practice prospectively in other populations and explore subgroups that may be most likely to benefit from testing.
Correspondence: Samir S. Shah, MD, MSCE, Division of Infectious Diseases, Children's Hospital of Philadelphia, Room 1526 (North Campus), 34th Street and Civic Center Boulevard, Philadelphia, PA 19104 (email@example.com).
Accepted for Publication: February 15, 2010.
Author Contributions: Dr Shah had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Hodinka and Shah. Acquisition of data: Dewan, Zorc, Hodinka, and Shah. Analysis and interpretation of data: Dewan, Hodinka, and Shah. Drafting of the manuscript: Dewan, Zorc, and Hodinka. Critical revision of the manuscript for important intellectual content: Dewan, Zorc, Hodinka, and Shah. Statistical analysis: Dewan and Shah. Obtained funding: Shah. Study supervision: Zorc, Hodinka, and Shah.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant K01 AI73729 from the National Institute of Allergy and Infectious Diseases (Dr Shah) and by the Physician Faculty Scholar Program of the Robert Wood Johnson Foundation (Dr Shah).
Dewan M, Zorc JJ, Hodinka RL, Shah SS. Cerebrospinal Fluid Enterovirus Testing in Infants 56 Days or Younger. Arch Pediatr Adolesc Med. 2010;164(9):824-830. doi:10.1001/archpediatrics.2010.153