Total white blood cell count (A), absolute neutrophil count (B), and platelet count (C) for identifying young febrile infants aged 0 to 28 days and 29 to 60 days with invasive bacterial infections. Black squares represent the optimal cutoffs; counts are in × 103 cells/µL. To convert absolute neutrophil count and white blood cell count to × 109 per liter, multiply by 0.001; to convert platelet count to × 109 per liter, multiply by 1.
eFigure. Flow Diagram of Enrollment.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Cruz AT, Mahajan P, Bonsu BK, et al. Accuracy of Complete Blood Cell Counts to Identify Febrile Infants 60 Days or Younger With Invasive Bacterial Infections. JAMA Pediatr. 2017;171(11):e172927. doi:10.1001/jamapediatrics.2017.2927
What is the accuracy of complete blood count parameters at routinely used thresholds in identifying young febrile infants with bacteremia or bacterial meningitis in the pneumococcal conjugate vaccine era?
In this cohort of 4313 febrile infants aged 0 to 60 days, 97 (2.2%) had bacteremia or bacterial meningitis. Sensitivities were low for white blood cell counts less than 5000/µL (sensitivity, 10%; specificity, 91%), white blood cell count ≥15 000/µL (sensitivity, 27%; specificity, 87%), and absolute neutrophil count ≥10 000/µL (sensitivity, 18%; specificity, 96%).
No complete blood cell count parameter at routinely used thresholds in isolation identified infants with bacteremia or bacterial meningitis with sufficient accuracy to substantially assist clinical decision making.
Clinicians often risk stratify young febrile infants for invasive bacterial infections (IBIs), defined as bacteremia and/or bacterial meningitis, using complete blood cell count parameters.
To estimate the accuracy of individual complete blood cell count parameters to identify febrile infants with IBIs.
Design, Setting, and Participants
Planned secondary analysis of a prospective observational cohort study comprising 26 emergency departments in the Pediatric Emergency Care Applied Research Network from 2008 to 2013. We included febrile (≥38°C), previously healthy, full-term infants younger than 60 days for whom blood cultures were obtained. All infants had either cerebrospinal fluid cultures or 7-day follow-up.
Main Outcomes and Measures
We tested the accuracy of the white blood cell count, absolute neutrophil count, and platelet count at commonly used thresholds for IBIs. We determined optimal thresholds using receiver operating characteristic curves.
Of 4313 enrolled infants, 1340 (31%; 95% CI, 30% to 32%) were aged 0 to 28 days, 2412 were boys (56%), and 2471 were white (57%). Ninety-seven (2.2%; 95% CI, 1.8% to 2.7%) had IBIs. Sensitivities were low for common complete blood cell count parameter thresholds: white blood cell count less than 5000/µL, 10% (95% CI, 4% to 16%) (to convert to 109 per liter, multiply by 0.001); white blood cell count ≥15 000/µL, 27% (95% CI, 18% to 36%); absolute neutrophil count ≥10 000/µL, 18% (95% CI, 10% to 25%) (to convert to × 109 per liter, multiply by 0.001); and platelets <100 × 103/µL, 7% (95% CI, 2% to 12%) (to convert to × 109 per liter, multiply by 1). Optimal thresholds for white blood cell count (11 600/µL), absolute neutrophil count (4100/µL), and platelet count (362 × 103/µL) were identified in models that had areas under the receiver operating characteristic curves of 0.57 (95% CI, 0.50-0.63), 0.70 (95% CI, 0.64-0.76), and 0.61 (95% CI, 0.55-0.67), respectively.
Conclusions and Relevance
No complete blood cell count parameter at commonly used or optimal thresholds identified febrile infants 60 days or younger with IBIs with high accuracy. Better diagnostic tools are needed to risk stratify young febrile infants for IBIs.
Febrile infants 60 days and younger are routinely evaluated in emergency departments (ED) for serious bacterial infections including urinary tract infections (UTIs), bacteremia, and bacterial meningitis.1 Urinary tract infections are common bacterial infections2,3 and the urinalysis is a highly sensitive, noninvasive, readily available, rapid turnaround test to facilitate diagnosis.4,5 Although novel serum biomarkers are being evaluated to aid the clinician to reliably detect the presence of invasive bacterial infections (IBIs, defined by bacteremia or bacterial meningitis),6-9 the lack of reliable physical examination findings and nonspecific symptoms add to the diagnostic uncertainty for IBI.3
Several algorithms and guidelines have been used to help risk stratify young febrile infants for IBIs.10-12 These algorithms were developed prior to the introduction of the pneumococcal conjugate vaccines, although this vaccine predominantly affects older infants. In addition, screening of pregnant women for group B streptococcus and provision of intrapartum antibiotic chemoprophylaxis has led to IBI becoming less common in the well-appearing febrile young infant,13,14 making risk stratification even more challenging. While there is substantial variation in the laboratory evaluation of young febrile infants,15,16 the most commonly obtained test is the complete blood cell count (CBC). Smaller studies have demonstrated suboptimal performance characteristics of CBC parameters including the peripheral white blood cell (WBC) count and absolute neutrophil count (ANC) for IBI.17-22 However, this needs further validation in larger prospective cohorts of young febrile infants. Serum inflammatory markers, such as procalcitonin and C-reactive protein, can more accurately predict which young infants have IBIs.6,7,19 However, these newer biomarkers are still being validated and are not readily available in all EDs.23-26 Therefore, the ease and tradition of obtaining CBCs has led clinicians to continue to use CBC parameters in algorithms to help risk stratify young febrile infants.
Given the changing epidemiology of IBIs, we conducted a prospective study in which we enrolled a large, geographically diverse cohort of febrile infants and conducted a planned subanalysis to determine the performance of the CBC to identify febrile infants 60 days and younger with IBIs.
Quiz Ref IDThis was a planned secondary analysis of a prospectively enrolled cohort of young febrile infants evaluated in 26 EDs in the Pediatric Emergency Care Applied Research Network between 2008 and 2013. In the parent study, we evaluated the association of host gene expression patterns with bacterial infections.26 Infants aged 60 days or younger were enrolled if they had documented temperatures of 38°C or higher and blood cultures were obtained. In addition to blood cultures, all infants had either cerebrospinal fluid (CSF) cultures obtained or telephone follow-up within 7 days of the ED visit to ascertain whether any patient had missed or developed bacterial meningitis. We excluded infants who were critically ill (eg, requiring emergent interventions such as intubation, cardiopulmonary resuscitation, or use of vasoactive medication), premature (completed gestational age <37 weeks), received antibiotics in the 4 days preceding the ED visit, or had major congenital malformations or comorbid medical conditions (eg, inborn errors of metabolism, congenital heart disease, chronic lung disease, immunosuppression or immunodeficiencies, or indwelling catheters or shunts). We also excluded previously enrolled infants and infants for whom the presence of bacteremia or meningitis was unknown. In addition, for this subanalysis, we excluded children whose CBC components were missing as well as febrile infants with urinary tract infections without IBIs. However, infants with urinary tract infections who also had either bacteremia and/or bacterial meningitis were included. Institutional review board approval was obtained at each site, and written consent was obtained from the guardian of each enrolled patient.
We defined bacteremia and bacterial meningitis as growth of a single pathogen in blood and CSF cultures, respectively. Organisms classified a priori as contaminants included Bacillus non-cereus/non-anthracis, diphtheroids, Lactobacillus, Micrococcus, coagulase-negative staphylococci, and viridans group streptococci.
Quiz Ref IDWe defined leukocytosis as a WBC count 15 000 cells/µL or greater and leukopenia as a WBC count less than 5000 cells/µL (to convert to × 109 per liter, multiply by 0.001). Neutrophilia was defined as an ANC greater than 10 000 cells/µL (to convert to × 109 per liter, multiply by 0.001). These thresholds were chosen a priori because these are the thresholds used by several existing algorithms to risk stratify young febrile infants.10-12,27 Thrombocytosis was defined as a platelet count 450 × 103 cells/µL or greater (to convert to × 109 per liter, multiply by 1). Thrombocytopenia was evaluated at 2 different platelet count thresholds: less than 100 × 103 cells/µL and less than 150 × 103 cells/µL.28 Band counts were not routinely available and therefore were not analyzed.
Quiz Ref IDThe primary outcome measures were the test characteristics of the WBC count, ANC, and platelet count to identify infants with IBIs.
We summarized categorical variables with counts and percentages and continuous measures with medians, interquartile ranges (IQRs), means, and standard deviations. For each CBC parameter analyzed, we reported sensitivity, specificity, positive predictive value, and negative predictive value for the commonly used thresholds. We also analyzed likelihood ratios with 95% confidence intervals. We constructed receiver operating characteristic (ROC) curves and determined an optimal cutoff value for each CBC parameter through minimizing the sensitivity, specificity coordinate pair from the point (1,1), also known as the Euclidean method.29,30 We calculated areas under the ROC curves (AUC), and based on prior work, we defined AUCs of less than 0.7 as having poor discriminatory value; AUCs of 0.7 to 0.8 as minimally accurate; AUCs of 0.8 to 0.9 as having good accuracy; and AUCs of greater than 0.9 as having excellent accuracy.17 We used SAS, version 9.4 software (SAS Institute) for analyses. P values less than .05 were considered significant, and all tests were 2-sided.
Of the 4795 infants included in the main study, 4313 (90.7%) met inclusion criteria for this analysis (eFigure in the Supplement). Lumbar punctures were attempted in 3384 infants (78.5%) and were successful in 3324 (98.2%). The demographic characteristics of the study population are described in Table 1. The median age was 38 days (IQR, 25-48 days).
Ninety-seven infants (2.2%; 95% CI, 1.8% to 2.7%) had IBIs; 73 of 4313 (1.7%; 95% CI, 1.4% to 2.1%) with isolated bacteremia and 24 (0.6%; 95% CI, 0.4% to 0.8%) with bacterial meningitis. Of the 24 infants with meningitis, 11 also had documented bacteremia. Invasive bacterial infections were identified in 57 of 1340 infants 28 days and younger (4.3%; 95% CI, 3.2% to 5.3%) and in 40 of 2973 infants aged 29 to 60 days (1.4%; 95% CI, 1.0% to 1.8%). Pathogens isolated are listed in Table 2.
No CBC parameters reliably distinguished between infants with and without IBIs (Table 3). While children with IBIs did have higher WBC counts, ANCs, and lower platelet counts, there was not a threshold for these parameters at which IBI could be reliably predicted (Figure). The test characteristics of CBC parameters at various commonly used thresholds and optimal thresholds are presented in Table 4. Even at the optimal ANC threshold of 4100 cells/µL, one-third of infants with IBIs would have been missed. All CBC parameters had low sensitivity and high negative predictive values at commonly used thresholds. Low positive predictive values were in part owing to the low prevalence of IBIs. Using widely accepted normal ranges for WBC counts (5000 to 14 900 cells/µL) or an ANC of less than 10 × 103 cells/µL would have missed 61 (63%) and 80 (82%) of infants with IBIs, respectively.
We constructed ROC curves for each individual laboratory predictor variable for infants aged 0 to 28 days and infants aged 29 to 60 days separately (Figure). Complete blood cell count test characteristics were not improved when only infants older than 28 days were considered. No single CBC parameter at any threshold had both good sensitivity and good specificity. The WBC count, ANC, and platelet count each had poor discriminatory value.
Our analysis of a large, prospectively enrolled, geographically diverse cohort of febrile infants 60 days and younger likely represents the epidemiology of the bacterial pathogens responsible for bacteremia and bacterial meningitis. Of greater importance, our findings demonstrate that Quiz Ref IDindividual parameters of the CBC have poor discriminatory ability in identifying which young febrile infants have IBIs.
A minority of infants with IBIs had abnormal WBC counts. This finding is consistent with other studies evaluating young febrile infants, where peripheral leukocytosis was seen in fewer than one-half of infants with bacteremia and in a minority of infants with bacterial meningitis.31,32 We also evaluated the ability of the platelet count to identify infants with IBIs because platelets are an acute-phase reactant, and few data exist on the accuracy of platelet counts in risk stratifying young infants for IBIs.33 However, neither thrombocytosis nor thrombocytopenia, as can be seen with overwhelming infections, were sufficiently accurate. While we would expect the positive predictive value to decline and the negative predictive value to increase as the prevalence of bacteremia and bacterial meningitis decline, sensitivity and specificity of the CBC parameters should not be influenced by declining prevalence rates. One possible explanation for the CBC’s poor performance is the change of pathogens causing bacterial meningitis and bacteremia in young infants. In the pre–pneumococcal conjugate vaccine era, leukocytosis was commonly seen in older infants with Haemophilus and pneumococcal bacteremia.34,35 However, pathogens more commonly identified in the modern era may produce less of an inflammatory response by the host. A 2015 retrospective study of episodes of bacteremia in previously healthy febrile infants 90 days and younger receiving care in 17 children’s hospitals found that Escherichia coli, group B streptococcus, and Staphylococcus aureus were among the most common bacteria isolated,36 and that E coli and group B streptococcus accounted for almost two-thirds of bacterial meningitis cases in infants aged 0 to 90 days.37E coli bacteremia is less likely to be associated with peripheral leukocytosis than pneumococcal bacteremia in infants and toddlers,38-40 and 1 study of infants with late-onset group B streptococcus bacteremia found that only 45% had abnormal WBC counts.40,41
Another possible explanation for why CBC parameters poorly identified infants with IBIs is that infants in our cohort may have presented earlier in the natural histories of their infections than in previous studies. In 2001, investigators42 compared parental perceptions of childhood fever with a previous study conducted in 198043 and found that in the more contemporary era, parents were more apt to check their children’s temperatures more often, worry more about seizures as a complication of pyrexia, have bloodwork performed on their children during febrile illnesses, and seek care in an ED. These concerns regarding fever may well translate into seeking care earlier in the course of an infant’s illness. One 2016 study2 found that no single CBC or inflammatory parameter or combination of parameters had optimal sensitivity for identifying infants with IBIs who had pyrexia of short durations.2 Additionally, in this study, CBCs were evaluated at one point, and the parameters may have been more accurate had CBCs been obtained later into an infant’s illness.
Complete blood cell count parameters also had suboptimal specificity for identifying infants with IBIs in our cohort. While most febrile neonates are hospitalized and administered empirical parenteral antibiotics while awaiting culture results, accurate identification of low-risk infants in the second month of life would enable reduction in resource use, costs to families, and unnecessary antibiotic exposure. However, given the low incidence of IBIs, most abnormal CBC results will not be associated with the presence of IBI. Our data add to previous literature on the poor accuracy of CBCs in identifying young febrile infants with IBIs17-22 and questions how CBC parameters, particularly the ANC, may be integrated into newer guidelines. Rather than using CBC parameters in isolation, an approach that combines certain CBC parameters (ie, the ANC) with other laboratory tests, such as the urinalysis, procalcitonin, and/or C-reactive protein (when available), may help risk stratify these febrile infants more effectively.2
Quiz Ref IDRecognition that CBC parameters in isolation have poor ability to discriminate young febrile infants with and without IBIs has important implications for ED practice. Some hospital algorithms recommend only sending blood cultures in febrile infants if the WBC or ANC parameters are abnormal. The rationale is to decrease the frequency of blood culture contamination, which has historically been noted in 2% to 11% of blood cultures obtained from children in the ED.44-46 However, this practice would also miss most young febrile infants with IBIs. Two studies15,16 have described practice variation in terms of cultures obtained in young febrile infants, finding that even among 0- to 28-day-old neonates, fewer than two-thirds to three-quarters have blood, urine, and CSF cultures obtained.15,16 Another belief is that most infants with bacterial meningitis will have concomitant bacteremia and will thus have positive blood cultures even if CSF is not obtained at the time of the initial visit. While this was true for Haemophilus influenzae, data from the early 2000s indicate that bacteremia is only documented in 46% to 60% of infants with bacterial meningitis,47,48 consistent with our findings in this study.
This study has some limitations. Data were collected from a convenience sample of febrile infants evaluated in the EDs of large academic children’s hospitals, and the results may thus not be generalizable to community EDs. However, the frequency of bacteremia and meningitis were similar to that described in previously published studies,31,32,39,40,48,49 and we do not have a biological hypothesis why the accuracy of the CBC would be different in different settings. Critically ill-appearing infants were not enrolled in this study, possibly leading to spectrum bias. However, most risk-stratification tools aim to identify infants at low risk for IBI, not those at high risk, and critical appearance is more important than any laboratory value for identifying those at high risk. There are some data on the utility of band counts in predicting bacteremia in this age group,50 but we were unable to assess the utility of relative or absolute bandemia for identifying those with IBIs given variation in band count availability across study EDs. Bacterial cultures were considered the reference standard for analyses, despite recognition that falsely negative cultures51 can occur owing to sporadic bacteremia or if low blood volumes are inoculated into blood culture bottles. One 2017 study found that for each additional 1 mL of blood culture volume collected, microbial yield increased by 0.5% in children with pneumonia.52
Complete blood cell count parameters had poor accuracy in distinguishing febrile infants 60 days and younger with and without invasive bacterial infections in the postpneumococcal conjugate vaccine era, although the ANC had the highest sensitivity. Physicians who use CBC thresholds in an attempt to risk stratify febrile young infants may be falsely reassured by normal CBC parameters. When used in isolation, either at commonly used thresholds or at the optimal thresholds identified here, CBC parameters have at best modest discriminatory ability. In an era where better screening tests exist to identify infants with IBIs, we need to question our continual reliance on a test whose greatest strength may simply be in its ready availability in clinical practice.
Corresponding Author: Andrea T. Cruz, MD, MPH, Department of Pediatrics, Baylor College of Medicine, 6621 Fannin St, Ste A2210, Houston, TX 77030 (firstname.lastname@example.org).
Accepted for Publication: July 6, 2017.
Correction: This article was corrected on January 8, 2018, to correct an error in the Key Points section of the text.
Published Online: September 11, 2017. doi:10.1001/jamapediatrics.2017.2927
Author Contributions: Dr VanBuren had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Mahajan, Ramilo, Kuppermann.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Cruz.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Cruz, VanBuren.
Obtained funding: Mahajan, Ramilo, Kuppermann.
Administrative, technical, or material support: Cruz, Mahajan, Ramilo, Kuppermann.
Supervision: Mahajan, Ramilo, Kuppermann.
Conflict of Interest Disclosures: Dr Ramilo reports personal fees from Abbvie, Janssen, Regeneron, and Pfizer, and grants from Janssen. All these fees and grants are not related to the current work. No other disclosures are reported.
Funding/Support: The research reported in this publication was supported in part by grant H34MCO8509 from Health Resources and Services Administration, Emergency Services for Children and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award R01HD062477. This project is also supported in part by the Health Resources and Services Administration, Maternal and Child Health Bureau, and Emergency Medical Services for ChildrenNetwork Development Demonstration Program under cooperative agreements U03MC00008, U03MC00001, U03MC00003, U03MC00006, U03MC00007, U03MC22684, and U03MC22685.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Group Information: The participating centers and investigators of the Pediatric Emergency Care Applied Research Network are:
Ann and Robert H. Lurie Children’s Hospital, Chicago, Illinois (Elizabeth C. Powell, MD, MPH); Bellevue Hospital Center, New York, New York (Deborah A. Levine, MD; Michael G. Tunik, MD); Boston Children’s Hospital, Boston, Massachusetts (Lise E. Nigrovic, MD, MPH); Children’s Hospital of Colorado (Genie Roosevelt, MD); Children’s Hospital of Michigan (Prashant Mahajan, MD, MPH, MBA); Children’s Hospital of Philadelphia, Pennsylvania (Elizabeth R. Alpern, MD, MSCE); Children’s Hospital of Pittsburgh, Pittsburgh, Pennsylvania (Melissa Vitale, MD); Children’s Hospital of Wisconsin (Lorin Browne, DO; Mary Saunders, MD); Children’s National Medical Center, Washington, DC (Shireen M. Atabaki, MD, MPH); Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio (Richard M. Ruddy, MD); Hasbro Children’s Hospital, Providence, Rhode Island (James G. Linakis, MD, PhD); Helen DeVos Children's Hospital, Grand Rapids, Michigan (John D. Hoyle, Jr., MD); Hurley Medical Center, Flint, Michigan (Dominic Borgialli, DO, MPH); Jacobi Medical Center, New York, New York (Stephen Blumberg, MD; Ellen F. Crain, MD, PhD); Johns Hopkins Children’s Center, Baltimore, Maryland (Jennifer Anders, MD); Nationwide Children's Hospital, Columbus, Ohio (Bema Bonsu, MD; Daniel M. Cohen, MD); Nemours/Alfred I. DuPont Hospital for Children, New Castle County, Delaware (Jonathan E. Bennett, MD); New York Presbyterian-Morgan Stanley Children’s Hospital (Peter S. Dayan, MD, MSc); Primary Children’s Medical Center, Salt Lake City, Utah (Richard Greenberg, MD); St Louis Children’s Hospital, St Louis, Missouri (David M. Jaffe, MD; Jared Muenzer, MD); Texas Children’s Hospital, Houston, Texxas (Andrea T. Cruz, MD, MPH); University of California, Davis Medical Center (Nathan Kuppermann, MD, MPH; Leah Tzimenatos, MD); University of Maryland, College Park, Maryland (Rajender Gattu, MD); University of Michigan (Alexander J. Rogers, MD); University of Rochester, Rochester, New York (Anne Brayer, MD); and Women and Children’s Hospital of Buffalo, Buffalo, New York (Kathleen Lillis, MD).
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This information or content and conclusions are those of the authors and should not be construed as the official position or policy of, nor should any endorsements be inferred by Health Resources and Services Administration, Health and Human Services, or the US government.
Additional Contributions: We thank the research coordinators in Pediatric Emergency Care Applied Research Network and the project staff at the Data Coordinating Center at the University of Utah. All were compensated by grants that supported the study.
Create a personal account or sign in to: