PDFlib PLOP: PDF Linearization, Optimization, Protection. Page inserted by evaluation version

Transcription

1 PDFlib PLOP: PDF Linearization, Optimization, Protection Page inserted by evaluation version

2 ACAD EMERG MED d December 2004, Vol. 11, No. 12 d A Low Peripheral Blood White Blood Cell Count in Infants Younger than 90 Days Increases the Odds of Acute Bacterial Meningitis Relative to Bacteremia Abstract In relying on the peripheral blood white blood cell (WBC) count to identify infants at high risk for acute bacterial meningitis and bacteremia, to the best of the authors knowledge, it has not been reported previously whether high and low values of the test have similar implications for predicting these separate infections. Objective: To analyze the relationship between the peripheral WBC count and the odds of acute bacterial meningitis relative to bacteremia among sick infants aged 3 to 89 days. Methods: Areas under the receiver operating characteristic curve (AUCs) and likelihood ratios at various intervals of the total peripheral blood WBC count were computed. Results: A pathogen was isolated from blood or cerebrospinal fluid (CSF) from 72 infants aged 3 to 89 days. Fifty-two infants had growth of a pathogen from the blood only, and 20 had growth from the CSF. The most common bacteria isolated were Escherichia Bema K. Bonsu, MB ChB, Marvin B. Harper, MD coli (32) and group B streptococci (32). The AUC for the peripheral WBC count when differentiating between acute bacterial meningitis and bacteremia was 0.75 (95% CI = 0.63 to 0.88). The odds of acute bacterial meningitis relative to bacteremia were sevenfold higher for a peripheral WBC cutoff below 5,000 cells/mm 3 and threefold lower for a peripheral WBC cutoff at or above 15,000 cells/mm 3. Conclusions: In young infants, the peripheral blood WBC count is useful for estimating the odds of acute bacterial meningitis relative to isolated bacteremia. A low peripheral blood WBC count should be considered a much more worrisome laboratory finding because it is associated with a relatively high risk for acute bacterial meningitis relative to the potential for bacteremia. Key words: bacterial meningitis; bacteremia; peripheral blood white blood cell count; infants. ACADEMIC EMERGENCY MEDICINE 2004; 11: In recent years, there has been an increased desire among clinicians to manage febrile young infants outside the hospital. Several protocols have been reported that attempt to achieve this goal. 1 3 Although useful, these protocols report cases of missed urinary tract infections, bacteremia, and acute bacterial meningitis in young infants discharged to home. Most common among these cases are young infants with unrecognized bacteremia and urinary tract infections. Fortunately, patients with these specific infections tend to have a good outcome if antibiotics are administered prior to discharge from the initial encounter. 1,2 However, those with unrecognized acute bacterial meningitis, if discharged to home, are at risk for serious complications. From the Department of Medicine, Division of Emergency Medicine, Children s Hospital (BKB), Columbus, OH; and the Department of Medicine, Divisions of Emergency Medicine and Infectious Diseases, Children s Hospital (MBH), Boston, MA. Received February 18, 2004; revision received June 2, 2004; accepted June 25, Address for correspondence and reprints: Bema K. Bonsu, MB ChB, Children s Hospital, Division of Emergency Medicine, 700 Children s Drive, Columbus, OH Fax: ; bonsub@pediatrics.ohio-state.edu. doi: /j.aem Among screening laboratory tests for detecting serious bacterial infections in infants, the most widely performed is the peripheral blood white blood cell (WBC) count. 4 6 Unfortunately, this test, within published norms, is unable to reliably exclude bacteremia, urinary tract infections, and acute bacterial meningitis in infants. 7 9 More accurate screening tests are commonly available and include urinalysis for urinary tract infections and cerebrospinal fluid (CSF) analysis for acute bacterial meningitis. CSF specimens, however, may be technically difficult to obtain in young infants and, if contaminated with blood, challenging to interpret. When this occurs, or if specimens are insufficient to perform a preliminary analysis, clinicians may be compelled to admit young infants to the hospital because of uncertainty concerning the clinical outcome of such cases. Some, however (inferring from the disproportionately high rate at which the peripheral WBC count is performed relative to other tests 4 6 ), may discharge these young infants to home if the infants appear to be well, relying on the peripheral blood WBC count as a final check for identifying those patients who are at high risk for clinically unrecognized bacterial meningitis. We have shown previously that the peripheral WBC count, irrespective of test threshold, has a high negative predictive value because of the low prevalence

3 1298 Bonsu and Harper d PERIPHERAL BLOOD WBC AND INFANT BACTEREMIA of serious bacterial infections in young infants. For this reason, a more useful measure of the usefulness of this test for diagnosis is the ability to positively identify infants at high risk for sepsis. In this more limited role, we have reported that the peripheral blood WBC count, at values that are extremely low or high, is able to identify infants at increased risk for bacteremia and bacterial meningitis. 7,8 A question that remains, however, is whether high and low values of the test have similar implications for predicting these separate infections. A correct interpretation of this test has pertinence for sepsis protocols that do and do not recommend dual thresholds of the peripheral blood WBC count or routine lumbar puncture 1 3,10 16 when gauging the potential consequences of unwittingly discharging to home a well-appearing young infant with unrecognized serious bacterial infection. Because bacterial meningitis is a more serious infection, an abnormal peripheral blood WBC count that indicates a substantially higher risk for bacterial meningitis relative to bacteremia, particularly when preliminary results of CSF are unavailable, should prompt a reevaluation of decisions to discharge to home. In contrast, because missed bacteremia has a relatively good outcome, a peripheral blood WBC count that indicates a substantially lower risk for bacterial meningitis relative to bacteremia may permit the discharge of a well-appearing infant to home if appropriate cultures are obtained, a parenteral antibiotic is administered, and close follow-up is arranged before discharge. In this study, as an aid to decisions concerning patient disposition from the emergency department (ED), we examined whether the peripheral blood WBC count is able to estimate the odds of bacterial meningitis relative to bacteremia when preliminary results of CSF are unavailable in young infants evaluated for sepsis. METHODS Study Design. In this retrospective study, we reviewed laboratory test findings of consecutive infants, 3 to 89 days of age, seen in the ED who had peripheral blood and CSF sent concurrently for bacterial culture and peripheral WBC count between January 1992 and July The institutional review board of Children s Hospital, Boston, approved this study. Study Setting and Population. The ED at which the study was done is located within an academic children s hospital and serves a varied socioeconomic population, of whom approximately 70% have private insurance. Thirty-five percent of patients using the ED are white, 25% are African American, 20% are Hispanic, and 20% are of other races or did not state their racial origin. At our institution, a full sepsis evaluation is strongly encouraged in infants younger than 90 days of age who present with fever defined as a temperature.38 C. About 75% successfully complete a full sepsis evaluation that includes the collection of blood, urine, and CSF specimens for analysis, Gram stain, and bacterial culture. Infants were included for analysis if a bacterium recognized to cause disease in young infants was isolated from culture of blood or CSF. Study Protocol. We reviewed medical records from a computerized database that included age, blood culture result, CSF culture result, and the values of the WBC count in peripheral blood and CSF. Excluded from the database were infants younger than 90 days diagnosed as having acute leukemia and infants in the immediate postnatal period (first 48 hours of life), most of whom have not yet been discharged from the hospital postdelivery and, hence, are unlikely to be managed by an emergency physician. Measures. We classified acute bacterial meningitis and bacteremia to be present if standard bacterial culture of CSF or blood isolated Streptococcus pneumoniae, Neisseria meningitidis, group B streptococci, Haemophilus influenzae type b, Listeria monocytogenes, Escherichia coli, or Salmonella species from patients at any age. Additionally, bacteremia was also determined to be present if other gram-negative rods such as Acinetobacter, Citrobacter, and Klebsiella species were isolated from the blood of an infant who was less than 1 month of age at the time of diagnosis. Contaminants were predetermined to be present if blood or CSF grew Staphylococcus epidermidis, Streptococcus bovis, viridans streptococci, Enterobacter species, Flavobacterium species, Enterococcus species, Corynebacterium species, and Bacillus species. These contaminants were coded as negative bacterial cultures but were retained in the database. Data Analysis. A receiver operating characteristic (ROC) curve was constructed to determine the accuracy of the peripheral blood WBC count for differentiating acute bacterial meningitis from bacteremia. 17,18 This curve plots sensitivity vs. 1 2 specificity (likelihood ratio) for all observed results of the peripheral blood WBC count. The area under this curve (AUC) defines the accuracy of the test. For our study, an AUC threshold of 0.7 was accepted 19 to be the nominal threshold for concluding that the peripheral blood WBC count was accurate. The AUC was also calculated for the absolute neutrophil count. For differentiating between acute bacterial meningitis and bacteremia, the sensitivity, specificity, predictive values, and interval likelihood ratios 20,21 were calculated for the following intervals of the peripheral blood WBC count based on norms published in the medical literature: 0 4,999 cells/mm 3, 5,000 14,999 cells/mm 3, 15,000 19,999 cells/mm 3, and $20,000 cells/mm 3. Indices of test performance were also

4 ACAD EMERG MED d December 2004, Vol. 11, No. 12 d calculated for a modification of these thresholds defined as a peripheral blood WBC count of 5,000 19,999 cells/ mm 3. Generally, likelihood ratios greater than 2.0 or less than 0.5 suggest that there is a clinically important increase or decrease, respectively, in the relative odds of a disease. 20,21 Because these data were not normally distributed, the Wilcoxon rank sum test was used to compare the medians of the peripheral blood WBC count for young infants with acute bacterial meningitis and bacteremia. All data were analyzed with the Stata statistical package version 6.0 (StataCorp LP, College Station, TX). RESULTS For the period of study, bacterial pathogens were identified from culture of blood or CSF from 72 (1.2%) of 5,885 infants aged 0 to 89 days (20 from CSF) among whom blood and CSF specimens were obtained for sepsis evaluation. The rates of bacterial meningitis and bacteremia were 0.3% (20/5,885) and 0.9% (52/5,885), respectively. Thirty (42%) of the 72 pathogens were from infants aged less than 29 days (21 blood, 9 CSF), 26 (36%) from infants aged 29 to 56 days (19 blood, 7 CSF), and 16 (22%) from those aged 57 to 89 days (12 blood, 4 CSF). Bacterial pathogens isolated from blood and CSF included E. coli (32), group B streptococci (32), S. pneumoniae (4), H. influenzae type B (1), Salmonella spp. (2), and Citrobacter koseri (1). The 52 (72%) pathogens isolated from blood alone included E. coli (23), group B streptococci (23), S. pneumoniae (3), H. influenzae type b (1), and Salmonella spp. (2). Twenty pathogens identified from culture of CSF included E. coli (9), group B streptococci (9), S. pneumoniae (1), and C. koseri (1). Seven (35%) of the 20 patients with acute bacterial meningitis also grew the same pathogen from the blood. Because our analysis was limited to young infants who had blood and CSF obtained for culture, the total number of infants evaluated for sepsis (denominator for infection rate) was the same for cases with bacteremia and meningitis, canceling out when evaluating test performance. The absolute neutrophil count (AUC = 0.65, 95% CI = 0.51 to 0.78) did not perform as well as the peripheral blood WBC count (AUC = 0.75, 95% CI = 0.63 to 0.88) when differentiating between children with acute bacterial meningitis and those with isolated bacteremia. For this reason, the performance characteristics of the peripheral blood WBC count alone are presented below. The median peripheral blood WBC count for children with acute bacterial meningitis was 9,500 cells/mm 3 (interquartile range of 3,495 to 13,120 cells/ mm 3 ) and differed significantly (p = 0.001) from the median WBC count of 15,524 cells/mm 3 (interquartile range of 10,760 to 18,825 cells/mm 3 ) for children with bacteremia alone. The prevalence of both infections (positive predictive value) remained low at all thresholds of the test (Table 1). Bacterial meningitis (8/244 = 3.2%) was more prevalent than bacteremia (3/244 = 1.2%) at a peripheral WBC below 5,000 cells/ mm 3. Above this threshold, however, bacteremia was increasingly more prevalent with a rate overall of 0.9% (49/5,641) compared with a rate overall for bacterial meningitis of 0.2% (12/5,641). Because of the low prevalence of infection, the negative predictive value was high at all thresholds of the peripheral blood WBC count. No single cutoff or interval was highly sensitive and specific for identifying young infants with these bacterial infections (Table 2). Finally, interval likelihood ratios for adjusting the prior odds of acute bacterial meningitis relative to bacteremia at previously stated peripheral blood WBC count cutoffs are presented in Table 3. Infants with bacterial meningitis were seven times more likely than those with bacteremia to have peripheral WBC counts below 5,000 cells/mm 3. Conversely, infants with meningitis were about three times less likely than those with bacteremia to have a peripheral WBC count at or above 15,000/mm 3 (Table 2). Infants with bacteremia and those with acute bacterial meningitis were equally likely to have a peripheral TABLE 1. Positive and Negative Predictive Values at Various Intervals and Thresholds of the Peripheral White Blood Cell (WBC) Count for Detecting Serious Bacterial Infections (SBI) in Young Infants Peripheral Blood WBC Count (cells/mm 3 ) Positive Predictive Value: Meningitis % (N) Positive Predictive Value: Bacteremia Positive Predictive Value: SBI* % (N) Negative Predictive Value: Meningitis % (N) Negative Predictive Value: Bacteremia Negative Predictive Value: SBI* % (N) 0 to 4, (8/244) 1.2 (3/244) 4.5 (11/244) 99.8 (5,629/5,641) 99.1 (5,588/5,641) 98.9 (5,580/5,641) $15, (4/1,358) 2.0 (27/1,358) 2.3 (31/1,358) 99.6 (4,511/4,527) 99.4 (4,502/4,527) 99.1 (4,486/4,527) $20, (1/406) 3.0 (12/406) 3.2 (13/406) 99.7 (5,460/5,479) 99.3 (5,421/5,479) 98.9 (5,420/5,479),5,000 or $15, (12/1,602) 1.9 (30/1,602) 2.6 (42/1,602) 99.8 (1,590/4,283) 99.5 (4,261/4,283) 99.3 (4,253/4,283),5,000 or $20, (9/650) 2.3 (15/650) 3.7 (24/650) 99.8 (5,224/5,235) 99.3 (5,198/5,235) 99.1 (5,187/5,235) *Serious bacterial infections (acute bacterial meningitis and bacteremia).

5 1300 Bonsu and Harper d PERIPHERAL BLOOD WBC AND INFANT BACTEREMIA TABLE 2. Sensitivity and Specificity at Various Intervals and Cutoffs of the Peripheral Blood White Blood Cell (WBC) Count for detecting Serious Bacterial Infections (SBI) in Young Infants Peripheral Blood WBC Count (cells/mm 3 ) Sensitivity: SBI* Sensitivity: Meningitis Sensitivity: Bacteremia Specificity: No SBI* 0 to 4, (11) 40 (8) 6 (3) 4 (233) $15, (31) 20 (4) 52 (27) 77 (4,486) $20, (13) 5 (1) 23 (12) 93 (5,420),5,000 or $15, (42) 60 (12) 58 (30) 73 (4,253),5,000 or $20, (24) 45 (9) 29 (15) 89 (5,187) Total 100 (72) 100 (20) 100 (52) 100 (5,813) *Serious bacterial infections (acute bacterial meningitis and bacteremia). WBC count between 5,000 cells/mm 3 and 14,999 cells/mm 3. DISCUSSION This study suggests that there is an inverse relationship between the height of the peripheral WBC count and the likelihood of acute bacterial meningitis relative to bacteremia. Specifically, for purposes of adjusting the pretest odds of infection, bacterial meningitis compared to bacteremia is seven times more likely to be associated with a peripheral blood WBC count of less than 5,000 cells/mm 3 and three times less likely to be associated with a peripheral blood WBC count of greater than or equal to 15,000 cells/mm 3. These findings suggest that while both high and low values of the peripheral blood WBC count predict serious bacterial infections, a cutoff less than 5,000 cells/mm 3 is a more ominous finding because it portends a higher risk for bacterial meningitis. Because acute bacterial meningitis is more regularly associated with serious complications, a low peripheral blood WBC count should prompt clinicians to reevaluate plans to discharge to home, if such plans are made without the benefit of clearly interpretable CSF. The reasons for these observations are unclear, but the findings indicate that febrile young infants who fail to mount a vigorous leukocyte response in blood to the presence of bacteria in blood may be more susceptible to invasive infections of the CSF space. Further work is needed to elucidate the mechanism(s) underlying these observations. For now, however, our TABLE 3. Likelihood of Acute Bacterial Meningitis Relative to Bacteremia at Various Intervals of the Peripheral Blood White Blood Cell (WBC) Count in Young Infants Peripheral Blood WBC Count (cells/mm 3 ) Interval Likelihood Ratios 95% Confidence Intervals 0 to 4, , 24* $15, , 0.98* $20, , ,000 to 14, , ,000 to 19, , 1.18 *Statistically significant. findings have implications for screening strategies designed for detecting serious bacterial infections in infants. 1 3,10 16 These strategies combine several tests for diagnosis but, with regard to the peripheral blood WBC count, fall generally into two groups: those that utilize either a single low or a single high cutoff of the peripheral blood WBC count 1,2,15,16 and others that utilize both low and high thresholds of this test. 3,11,14 Our results show that both approaches to interpreting the peripheral WBC count have poor sensitivity for identifying infants at low risk for bacteremia and bacterial meningitis. Still, for identifying infants at high risk for these infections, screening strategies that utilize both high and low thresholds of the peripheral WBC count are preferable to those that utilize only a single threshold, and specific test results predict bacteremia or acute bacterial meningitis more strongly. Because acute bacterial meningitis is a more severe infection, protocols that utilize a single high peripheral WBC count cutoff should be amended to also include a lower threshold for this test. We recommend lumbar puncture with close scrutiny of preliminary tests of CSF fluid whenever sepsis is considered to be a possibility in infants presenting with fever. If, however, discharge to home from the ED is contemplated without the benefit of clearly interpretable CSF fluid results, a peripheral blood WBC count finding of less than 5,000 cells/mm 3 should prompt a period of further observation and the firming up of plans for even closer follow-up. Because these measures can fail, however, strong consideration should be given to hospitalizing infants, or to repeating the lumbar puncture after an appropriate interval. Finally, irrespective of peripheral blood WBC count, close follow-up after discharge from the hospital is necessary in all infants because of the poor sensitivity of the test. LIMITATIONS Our study has a number of limitations. It is retrospective, so it suffers from the common limitations of this study design. Still, we chose this approach because it was the only feasible way to answer the study question given the low rates of bacteremia and bacterial meningitis in young infants. Inadequate

6 ACAD EMERG MED d December 2004, Vol. 11, No. 12 d documentation, in particular, meant that we were unable to accurately identify all young infants pretreated with antibiotic agents. Fortunately, this limitation is unlikely to affect our conclusions because prior antibiotic treatment is highly unlikely in this age group. We were also unable to characterize accurately the clinical appearance of young infants in our study because full and consistent documentation of clinical status prior to testing was not always present in the medical record. Still, likelihood ratios reported for specific intervals of the peripheral blood WBC count are valid because testing by the laboratory is not biased by the clinical evaluation. CONCLUSIONS In young infants, no single cutoff of the peripheral blood WBC count is accurate for separating acute bacterial meningitis from bacteremia, and there is wide distribution of these infections in all strata of the test. The test, however, provides information that may alter clinical decisions to discharge to home when the initial CSF fluid specimen is insufficient for screening analysis or difficult to interpret. Specifically, a peripheral blood WBC count of less than 5,000 cells/mm 3 is associated with a higher likelihood of acute bacterial meningitis relative to bacteremia. Although not a component of a number of published protocols for detecting sepsis, this laboratory result in febrile young infants should be recognized as an even more ominous finding than a peripheral blood WBC count greater than or equal to 15,000 cells/mm 3, and should prompt practitioners to manage these patients more aggressively. References 1. Baskin MN, O Rourke EJ, Fleisher GR. Outpatient treatment of febrile infants 28 to 89 days of age with intramuscular administration of ceftriaxone. J Pediatr. 1992; 120(1): Baker MD, Bell LM, Avner JR. Outpatient management without antibiotics of fever in selected infants. N Engl J Med. 1993; 329: Jaskiewicz JA, McCarthy CA, Richardson AC, et al. Febrile infants at low risk for serious bacterial infection an appraisal of the Rochester criteria and implications for management. Febrile Infant Collaborative Study Group. Pediatrics. 1994; 94: Belfer RA, Gittelman MA, Muniz AE. Management of febrile infants and children by pediatric emergency medicine and emergency medicine: comparison with practice guidelines. Pediatr Emerg Care. 2001; 17(2): Wittler RR, Cain KK, Bass JW. A survey about management of febrile children without source by primary care physicians. Pediatr Infect Dis J. 1998; 17:271 7; discussion Zerr DM, Del Beccaro MA, Cummings P. Predictors of physician compliance with a published guideline on management of febrile infants. Pediatr Infect Dis J. 1999; 18: Bonsu BK, Harper MB. Utility of the peripheral blood white blood cell count for identifying sick young infants who need lumbar puncture. Ann Emerg Med. 2003; 41: Bonsu BK, Harper MB. Identifying febrile young infants with bacteremia: is the peripheral white blood cell count an accurate screen? Ann Emerg Med. 2003; 42: Lin DS, Huang SH, Lin CC, et al. Urinary tract infection in febrile infants younger than eight weeks of age. Pediatrics. 2000; 105(2):E King JC Jr, Berman ED, Wright PF. Evaluation of fever in infants less than 8 weeks old. South Med J. 1987; 80: Dagan R, Powell KR, Hall CB, Menegus MA. Identification of infants unlikely to have serious bacterial infection although hospitalized for suspected sepsis. J Pediatr. 1985; 107: Kadish HA, Loveridge B, Tobey J, Bolte RG, Corneli HM. Applying outpatient protocols in febrile infants 1 28 days of age: can the threshold be lowered? Clin Pediatr (Phila). 2000; 39(2): Baker MD, Bell LM, Avner JR. The efficacy of routine outpatient management without antibiotics of fever in selected infants. Pediatrics. 1999; 103: Chiu CH, Lin TY, Bullard MJ. Identification of febrile neonates unlikely to have bacterial infections. Pediatr Infect Dis J. 1997; 16(1): Crain EF, Shelov SP. Febrile infants: predictors of bacteremia. J Pediatr. 1982; 101: Caspe WB, Chamudes O, Louie B. The evaluation and treatment of the febrile infant. Pediatr Infect Dis. 1983; 2: Zweig MH. Apolipoproteins and lipids in coronary artery disease. Analysis of diagnostic accuracy using receiver operating characteristic plots and areas. Arch Pathol Lab Med. 1994; 118: Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982; 143(1): Swets JA. Measuring the accuracy of diagnostic systems. Science. 1988; 240: Jaeschke R, Guyatt G, Sackett DL. Users guides to the medical literature. III. How to use an article about a diagnostic test. A. Are the results of the study valid? Evidence-Based Medicine Working Group. JAMA. 1994; 271: Gallagher EJ. Clinical utility of likelihood ratios. Ann Emerg Med. 1998; 31:391 7.