Objective. To evaluate selected characteristics of occult bacteremia in the post-Haemophilus influenzae type b (HIB) vaccine era.
Methods. A retrospective cohort study was performed involving 5901 children 2 to 24 months old with fever ≥39.0°C evaluated with a blood culture at an urban tertiary care children's hospital emergency department (ED) between February 1993 and June 1996. Patients were excluded if immune-suppressed, diagnosed with a focal infection, evaluated by lumbar puncture, or admitted to the hospital during initial evaluation. Prevalence of occult bacteremia, distribution of current pathogenic organisms, and time to positive culture in a continuously monitored system were determined. All patients with cultures positive for pathogenic bacteria were reevaluated and serious adverse outcomes were documented.
Results. The prevalence of occult bacteremia was 1.9% (95% confidence interval: 1.5%–2.3%). Streptococcus pneumoniae accounted for 82.9% of all pathogens and H influenzae was not a causative organism in this cohort.
The mean time to positive culture was significantly shorter for pathogens compared with contaminants (14.9 hours vs 31.1 hours). A culture that was positive in ≤18 hours was 13.0 (6.3–26.6) times more likely to contain a pathogen than a contaminant.
The average time from positive culture notification to reevaluation in the ED was 10.6 hours and over half of the patients recalled to the ED for positive cultures were admitted to the hospital. Of patients with occult pneumococcal bacteremia, 95.7% had resolution of their bacteremia without the use of parenteral antibiotics. Two patients had serious adverse outcomes. The rate of meningitis or death was .03% (.004%–.12%).
The contamination rate of blood cultures was 2.1% (1.7%–2.5%). Most (85%) of these patients were reevaluated in the ED and more than one third were admitted to the hospital before full identification of the organism.
Conclusions. Prevalence of occult bacteremia in the post-HIB vaccine era is lower than previously reported.Spneumoniae is the most common causative organism and resolves without parenteral antibiotics in the vast majority of cases. Continuously monitoring blood culture systems allow for early identification and can aid in differentiating contaminated from true pathogenic cultures by time to positive culture. Serious adverse outcome is an uncommon result of occult bacteremia. Updated epidemiology and microbiologic technology may impact the evaluation and treatment of children at risk for occult bacteremia.
Occult bacteremia, the presence of pathogenic bacteria in the blood of a well-appearing febrile child without an identifiable focus of infection, is a subject of controversy in pediatrics and emergency medicine. The evaluation and management of occult bacteremia is dependent on its prevalence in the population at risk, the characteristics of screening tests, the efficacy of treatment options, and the risk of adverse outcomes. Recent public health interventions and improvements in microbiology technology may have changed the prevalence and detection of occult bacteremia and the occurrence of adverse outcomes in febrile children.
Studies performed before the widespread use of the Haemophilus influenzae type b (HIB) vaccine found the prevalence of occult bacteremia to range from 2.8% to 11.6%.1–5 These studies identified Streptococcus pneumoniae as the cause in 60% to 85% of episodes of occult bacteremia, while HIB was responsible for 5% to 20%. Since the HIB vaccines were licensed in 1987, there has been a 96% decrease in the incidence of invasiveH influenzae infections in children <5 years old.6 A recent study estimated the current overall rate of occult bacteremia to be 1.6%.7 The risk of meningitis fuels the debate about the management of these patients. Although HIB is up to 12 times more likely than S pneumoniae to cause meningitis in patients with occult bacteremia,8,9 there has not been a determination of the risk of adverse outcomes since the near eradication of invasive HIB disease in young children.
In the past, laboratory identification of positive cultures by direct plating techniques, with a mean time to positive culture of 36 hours, resulted in a delay in follow-up of patients with occult bacteremia.10 However, the advent of continuously monitored blood culture systems allows for the majority of pathogens to be recovered within the first 24 hours.11,12 This decrease in laboratory detection time may significantly affect treatment decisions for patients at risk. Some authors in the past have advocated expectant wide-spectrum antibiotic treatment of children at risk for occult bacteremia.4,5,9,13–16 Careful evaluation of the current risk of adverse outcomes coupled with improved detection techniques may allow us to reassess this recommendation. Expectant antibiotic therapy carries with it considerable expenditures of time and resources, and the possible increased emergence of antibiotic resistance.17,18
The purpose of this study was to evaluate occult bacteremia in an urban emergency department population in the post-HIB vaccine era by: 1) determining the current prevalence and causative organisms of occult bacteremia; 2) demonstrating the time to positive culture and time to follow-up of children with culture-proven bacteremia; and 3) determining the incidence of severe adverse outcomes in children evaluated for occult bacteremia.
Study Design and Setting
This retrospective cohort study included all children 2 to 24 months old with fever ≥39.0°C who had blood cultures drawn in an urban tertiary care children's hospital emergency department (ED) between February 1993 and June 1996. During the study period, approximately 54 000 children were seen in the ED annually. The Institutional Review Board of Children's Hospital of Philadelphia approved this study.
Standard practice during the study period was to obtain blood cultures on children 2 to 24 months old with temperatures ≥39.0°C but did not include routine complete blood counts (CBCs) or use of empiric antibiotic treatment for children evaluated for occult bacteremia. Lumbar puncture was performed based on clinical assessment by the examining physician and was not part of a standard protocol of children within this age range. During a subset of the study time period that accounts for one-third of the enrollment time, 82% of eligible children were documented to have blood cultures obtained.19 Blood cultures were obtained by ED nurses using sterile techniques and inoculated into pediatric blood culture bottles (Pedi-Bac T, Organon Teknika Corp, Durham, NC). A single bottle containing supplemented brain heart infusion broth with .02% sodium polyanethol sulfonate was inoculated for each blood culture ordered. A minimum of .3 and maximum of 4 mL are required by the manufacturer. Standard procedure in the ED was to inoculate .5–1.0 mL. Through a pneumatic tube delivery system, blood cultures were routinely received in the laboratory within an hour of when they were taken and were immediately loaded into the blood culture instrument. The microbiology laboratory used the BacT/Alert Microbial Detection System (Organon Teknika Corp, Durham, NC) to process all blood cultures. The BacT/Alert system monitored carbon dioxide (CO2) production within each bottle every 10 minutes 24 hours per day. Bottles identified as positive were immediately removed from the instrument, 24 hours per day, and an aliquot was taken for Gram stain and subculture. The ED was notified immediately of the positive culture and given information from the Gram stain. Bacterial isolates were identified by conventional procedures. Only information from the Gram stain, however, was available at the time of initial report of positive culture to the ED. Routine protocol included contacting families of all children with positive blood cultures for reevaluation.
Patients were excluded if during initial ED evaluation they were: a) known to have an underlying disease that predisposed them to bacteremia (eg, sickle cell anemia, oncologic disease, immunodeficiency, indwelling central catheter); b) diagnosed with definitive focal infection (eg, pneumonia, meningitis, cellulitis, varicella); c) underwent lumbar puncture; d) had an illness requiring hospitalization; or e) died during initial ED evaluation. Presumptive diagnoses of ear infection, gastroenteritis, or bronchiolitis were not considered proven sources of infection and those children were included.
Measured Outcomes and Protocol
Occult bacteremia was defined as a blood culture positive for pathogenic bacteria from patient fulfilling inclusion criteria. Bacteria that were considered pathogenic included: S pneumoniae, Staphylococcus aureus, group A Streptococcus, Enterococcus species, Neisseria meningitidis, Enterobacteriaciae, Salmonella species,Moraxella catarrhalis, Pseudomonas species,H influenzae, Campylobacter species, andEscherichia coli. Bacteria that were considered contaminants were: coagulase-negative Staphylococcus species, α-hemolytic Streptococcus, nonpathogenic Streptococcus species,Micrococcus, Clostridium species,Corynebacterium, Gram-positive rods, and nonpathogenicNeisseria species. Time to positive culture was measured in hours and tenths of hours and automatically recorded by a continuously monitored blood culture system. Serious adverse outcome was defined as meningitis or death within 2 weeks of the date that the blood culture was obtained.
All children with blood cultures obtained during the study months were identified using microbiology data from the BacT/Alert system (Organon Teknika Corp, Durham, NC). This data indicated the patient's name, medical record number, age, date of blood culture collection, disposition (admission or discharge), and blood culture final result (including time in hours to positive culture and identification of bacteria). Charts of identified patients were then abstracted for medical history, previous antibiotic use, ED discharge diagnoses, maximum fever documented in the ED, and antibiotic treatment or prescription. Charts of patients with blood cultures positive for pathogenic or contaminant bacteria were abstracted for additional data including: follow-up visit site, date, and time; diagnosis and disposition at follow-up visit; maximum temperature documented at follow-up visit; result of repeat blood culture, CBC, chest radiograph, spinal fluid assessment, urinalysis, and urine culture; and antibiotic treatment or prescription.
Continuous variables were described using means, standard deviations, and 95% confidence intervals (CIs). Discrete variables were described using counts and percentages, with binomial exact 95% CIs. Continuous variables were analyzed with thet test. Categorical variables were analyzed with the χ2 test or Fisher's exact test. Relative risks with exact 95% CIs were calculated. Statistical significance was determined a priori as P value <.05.
The 5901 children who met inclusion criteria had a mean age of 12.4 (±5.5) months and 55% were male. The mean maximum presenting temperature was 39.9°C (± .5°C). Eight percent of the patients reported use of oral antibiotics before initial ED evaluation. The most common discharge diagnoses from the initial ED visit included viral syndrome or upper respiratory infection (41%), acute otitis media (38%), gastroenteritis (7%), and febrile seizure (4%).
The prevalence of occult bacteremia in the study cohort was 1.9% (95% CI: 1.5%–2.3%). Of note, the majority (82.9%) of pathogens wereS pneumoniae and none of the pathogens were H influenzae. Table 1 presents the distribution of all pathogenic bacteria recovered from study patients. The rate of occult bacteremia in children 2 to 6 months old was 1.0% (95% CI: .4%–2.0%) and the rate for children in the 6- to 24-month-old age group was 2.0% (95% CI: 1.6%–2.4%;P = .057).
The rate of contamination was 2.1% (95% CI: 1.7%–2.5%). Table 2 provides the rate of occult bacteremia and contaminated blood culture by age group. In children 6 to 24 months old, the rates of occult bacteremia and contaminated blood cultures were equal (P = .618). Children <6 months were actually more likely to have a contaminated culture than a true pathogen (P = .001).
Patients with occult bacteremia had higher mean temperature (40.2°C ± .5°C) when compared with patients with negative or contaminated cultures (39.9°C ± .5°C; P value <.001). Children with a temperature ≥40°C were 2.6 (95% CI: 1.8–3.9) times more likely to have occult bacteremia than if they had a lower fever. There was no statistical difference in the use of oral antibiotic before initial ED evaluation between those patients with occult bacteremia (3.6%) and those with negative or contaminated cultures (7.8%; relative risk [RR] = .4; 95% CI: .2–1.2). Amoxicillin was the most frequently (58%) reported antibiotic for those few children who were on antibiotics before evaluation.
The frequency of obtaining chest radiographs was not statistically different for patients who were later identified to have occult bacteremia (40.9%) and those with negative cultures (32.5%;P = .065). Patients with otitis media diagnosed during the initial ED visit were more likely to have occult bacteremia (RR = 2.2: 95% CI: 1.5–3.3). However, the rate of occult bacteremia in children diagnosed with bronchiolitis (.7%) was not statistically different from the rate in children with other diagnoses (1.9%; P = .176).
The mean time to positive culture for pathogenic bacteria was significantly shorter (14.9 hours ± 5.8 hours) than the mean time to positive culture for contaminated blood cultures (31.1 hours ± 20.6 hours; P value <.001). When only blood cultures positive for S pneumoniae were examined, the mean time to positive culture was even more rapid at 14.0 hours ± 2.4 hours.
Figure 1 presents a receiver operator characteristic (ROC) curve to demonstrate the true-positive rate (sensitivity) versus false-positive rate (1-specificity) over a range of time cutpoints until positive culture result. The area under the ROC curve is .92 (±.02). A culture that becomes positive in ≤18 hours is 13.0 (95% CI: 6.3, 26.6) times more likely to be a pathogen than a contaminant. Most pathogenic cultures (93.7%) became positive in ≤18 hours. Figure 2 illustrates the distribution of pathogenic and contaminated cultures that became positive within each time interval.
All of the patients with cultures that were identified to have pathogenic bacteria had documented follow-up in the ED, by telephone, or by their primary care physician. Most (97%) were reevaluated in the ED and 53% of these were admitted to the hospital. The average time from positive Gram stain notification to follow-up in the ED was 10.6 hours (±9.7 hours). Two patients returned to the ED for reevaluation before their blood culture becoming positive. Of those patients reevaluated in the ED, 33% were still febrile (≥38.5°C) and 96.3% had repeat blood cultures. The majority (93.3%) of the repeat cultures were negative. Only 4.8% had persistent bacteremia (4/92 with S pneumoniae and 1/6 with Salmonella) and 1.9% had contamination of repeat culture. This presents a 95.7% (95% CI: 89.2%–98.8%) rate of resolution without treatment with parenteral antibiotics for occult pneumococcal bacteremia.
When analyzing all patients with occult bacteremia who had repeat blood cultures on reevaluation, oral antibiotics prescribed at the initial visit did not affect the rate of persistent bacteremia (RR = .34; 95% CI: .06–1.93). Two of the 5 patients with persistent bacteremia on reevaluation and 67 of the 99 patients with resolved bacteremia had oral amoxicillin prescribed for them during initial evaluation in the ED before knowledge of the bacteremia.
A great majority of patients (98%) with cultures ultimately determined to be contaminated had documented follow-up. The majority (85%) were reevaluated in the ED before full identification of the organism and 35% of these patients were both febrile (≥38.5°C) and were admitted to the hospital on reevaluation. Of those patients reevaluated in the ED, 88% had repeat cultures and 2.2% of them were contaminated. None of these patients had repeat cultures that grew pathogens.
The focal bacterial infections that were identified on reevaluation of patients with occult bacteremia were: 8 patients with pneumonia, 4 patients with cellulitis, 2 patients with osteomyelitis, and 1 patient each with urinary tract infection, septic joint, and abscess. Therefore, the rate of focal infection per patient evaluated for occult bacteremia was .3% (95% CI: .2%–.5%). One patient was diagnosed with meningitis and 1 patient died of sepsis on reevaluation after identification of bacteremia. The rate of serious adverse outcome per patient evaluated for occult bacteremia was .03% (95% CI: .004%–.12%). The rates of outcome measures for patients evaluated for occult bacteremia are summarized in Table 3.
The patient who died was 5.7 months old and on autopsy was discovered to have congenital asplenia without any other associated anomalies. His temperature during initial ED evaluation was 40.1°C and he was diagnosed with otitis media and upper respiratory infection by an attending pediatric emergency medicine physician. At his initial ED visit he was evaluated and found to have a white blood cell count (WBC) of 9800 cells/mm3 and a chest radiograph with peribronchial thickening. Oral amoxicillin was prescribed for treatment of the otitis media. His initial blood culture grew S pneumoniae in 6.8 hours. He returned 6.4 hours after initial ED visit triage time in septic shock. Despite resuscitation efforts, he died in the ED.
The patient ultimately diagnosed with meningitis was 10.5 months old. During the initial ED visit, he had a temperature of 40.7°C, he was diagnosed with otitis media, and oral amoxicillin was prescribed. His blood culture grew S pneumoniae in 9.5 hours. He returned to the ED 6.9 hours after positive blood culture notification. On return, his fever was 38.4°C, WBC 14 300 cells/mm3 and lumbar puncture had 6200 WBC/mm3, 88 red blood cell count (RBC)/mm3, 235 mg/dL protein, 55 mg/dL glucose. The cerebrospinal fluid (CSF) Gram stain showed Gram-positive cocci but CSF culture and repeat blood culture remained negative. He recovered without sequelae.
Our study updates and redefines the risks associated with occult bacteremia by documenting the current low prevalence of bacteremia, the changing epidemiology of causative organisms, the ability of continuously monitored blood cultures to identify bacteremia quickly, and the very low incidence of severe adverse outcomes in children.
The prevalence of bacteremia (1.9%) in our study is lower than rates reported before the use of the HIB vaccine (2.8%–11.6%), but equivalent to that reported by Lee7 in the first post-HIB vaccine era cohort published.1–5,20,21 The noted decrease in prevalence may be attributable to both an absence ofH influenzae type b as a pathogen in occult bacteremia as well our study's strictly defined patient population. By excluding patients who had lumbar puncture or hospital admission at initial evaluation, we selected patients who fulfilled criteria for occult bacteremia and focused the numerator and denominator in our estimate of prevalence in children at risk for occult bacteremia. In an attempt to provide a truer estimate of the prevalence than previous studies, we eliminated patients with focal infections, hospitalization, or nonoccult bacteremia.1,2,4,5,20,21 Our participation rate of 82% is at least consistent with if not greater than previous studies' enrollment rates.1,4,7,19,21,22 As our study was a retrospective evaluation of a cohort treated by usual practice, we might expect that physicians would have obtained blood cultures on those children they considered to be at highest risk for occult bacteremia. This may have led to a prevalence rate that overestimated that of all-comers but closely estimates that in clinical practice.
S pneumoniae accounted for 83% of the cases of occult bacteremia in our study population. This proportion is lower than reported by Lee7 (92%), but similar to previously reported distributions.2–4,9,23 Notably, H influenzae was completely absent as a pathogen in our study. Rates of HIB occult bacteremia in early reports were 4.6% to 19.4%.1–4,9,23 During the years of our study, almost all (92%) of eligible children in the United States were immunized with 3 or more HIB doses and in our community the rate for complete HIB immunization was 88% (±4.2%).24 This information, in addition to the absence of HIB as a pathogen in Lee's study7 conducted in a different community, suggests that HIB is no longer isolated as a pathogenic bacterium in children evaluated for occult bacteremia.
The rate of contamination in our study (2.1%) is less than rates (3.6%–.6%) published previously.1,2,25 This rate, however, gains significance as the rate of occult bacteremia decreases. The youngest children (<6 months old) in our study had the highest contamination rate. The difficulty in technique of obtaining sterile blood cultures in young infants has also been noted in a previous study.2 A past study has demonstrated that a dedicated blood culture team or continued education of personnel obtaining blood cultures will decrease the contamination rate.26 A third of patients with cultures ultimately determined to be contaminated were admitted to the hospital and treated with antibiotics until full identification of the organism could be completed. We postulate that this unnecessary hospitalization rate results in a large expenditure of funds.
Risk factors including age and temperature have been evaluated in children with occult bacteremia.3,5,7,21,27 In this study, children with occult bacteremia were distributed equally between 6 and 24 months of age. We found, however, that children <6 months old had a prevalence of occult bacteremia equal to 1.0%. This low rate is similar to that reported in previous studies.1,7 It is possible that a larger percent of children in the youngest age group presenting with fever may have been excluded from the study because of a higher rate of invasive assessments, including lumbar puncture or hospitalization, to fully evaluate these young children who are more difficult to assess.
A higher documented fever was also associated with a higher risk of occult bacteremia. Children with a temperature ≥40°C were 2.6 times more likely to have occult bacteremia than children with lower temperatures. However, as the mean temperature difference between those children with occult bacteremia and those without was only .3°C, it is difficult to determine the clinical significance of this finding.
Importantly, this study indicates that the use of a continuously monitoring blood culture system and round-the-clock clinician notification allows for the identification of most cases of occult bacteremia within 18 hours. Follow-up was achieved within an average of <11 hours from the time of identification of positive culture to revisit ED triage time. This is significantly less than the 40 to 50 hours between evaluations cited in previous literature.10,31 As indicated by the ROC curve (Fig 1), knowledge of the time to positive Gram stain can guide the physician in determining if the cultured organism is more likely a pathogen or contaminant before full microbiologic identification. The ability to quickly identify cases of occult bacteremia and allow for reevaluation with pertinent data available may be useful in assessing treatment options for patients with suspected occult bacteremia. The implications of decreasing the cost and exposure to broad-spectrum antibiotics by using this information needs to be investigated further.
The debate in management options for children at risk for occult bacteremia is fueled by the risk of severe adverse outcome. We, therefore, evaluated the risk associated with the current pathogens of occult bacteremia in this rigorously selected population. Importantly, more than 95% of the patients with occult pneumococcal bacteremia had resolution of their bacteremia without use of expectant broad-spectrum antibiotics. Although the numbers of involved affected patients were small, there was not a significant relationship between treatment with antibiotics and rate of resolution of pneumococcal bacteremia on reevaluation. Also, the risk of meningitis or death was extremely small. The natural history of HIB bacteremia involves a greater risk of invasive disease than S pneumoniae bacteremia. Previous studies have indicated that 25% to 44% of children with HIB bacteremia developed invasive disease.4,21–23 These and other studies have also indicated that S pneumoniaebacteremia had a lower risk of invasive disease (0%–6%).4,8,9,21–23 An important recent study indicated that the HIB vaccine has greatly reduced the overall occurrence of meningitis in children by reducing the risk of invasiveH influenzae disease.32 In the present study, only 1.8% (2/111) of children with bacteremia developed serious adverse outcomes, both of them infected with S pneumoniae. Nonparenteral antibiotics were prescribed for a large percentage (51%) of patients in this cohort at initial ED visit, however, a prior randomized clinical trial demonstrated that oral antibiotics do not affect the incidence of serious focal infections.3 Two meta-analyses of prior data also question the efficacy of expectant oral or parenteral antibiotics in preventing meningitis in children with S pneumoniae occult bacteremia.33,34
It is important to evaluate these findings in the context of clinical practice. Determining the percentage of adverse outcomes based on a posthoc knowledge of bacteremia is unrealistic.35 Children are evaluated and treated during their initial visit as “at risk for occult bacteremia” not “known to have occult bacteremia.” This suggests that the risk of adverse outcomes should be determined by the risk of bacteremia, not the presence of bacteremia. Therefore, using the at-risk population as the denominator, the risk of adverse outcomes in our study is actually .03%. Using this data, one would treat >3000 patients at risk for occult bacteremia with broad-spectrum parenteral antibiotics to possibly prevent 1 adverse outcome.
Previously published recommendations for the evaluation and treatment of children at risk for occult bacteremia have varied greatly.13,15,16,33–39 We suggest that incorporation of this new data may help physicians in the difficult decision process of caring for the febrile young child. The documented low risk of occult bacteremia associated with a high rate of resolution without using with parenteral antibiotics and an extremely low risk of meningitis or death may allow for better discrimination between the risks and benefits of testing for occult bacteremia and expectant broad-spectrum antibiotic treatment.17,18,33,34,39 Despite a recent article that reported data to the contrary,40 2 other published studies have demonstrated that parents are more willing to forego invasive testing or empiric antibiotics in light of the low risk of adverse outcomes accompanying occult bacteremia.41,42 Automated, continuously monitored blood culture systems allow for early identification of bacteremia and directed antibiotic treatment of patients with true bacteremia, opposed to those at risk. Further information on the evaluation and treatment of children at risk for occult bacteremia may be provided by risk-benefit and cost-effectiveness evaluations of the current data. The HIB vaccine has drastically changed the risk of meningitis32 and occult bacteremia. Development of an effective conjugate vaccine for S pneumoniae may ultimately lay the issue to rest.
Grant support was provided by the University of Pennsylvania Research Foundation.
We thank Sarah Michelman for her instrumental assistance in data collection, Michele Scutti for manuscript preparation, and the technologists in the Children's Hospital of Philadelphia Microbiology Laboratory.
- Received September 22, 1999.
- Accepted January 11, 2000.
Reprint requests and correspondence to (E.R.A.) Division of Emergency Medicine, Children's Hospital of Philadelphia, 34th St and Civic Center Blvd, Philadelphia, PA 19104. E-mail:
- HIB =
- Haemophilus influenzae type b •
- ED =
- emergency department •
- CBC =
- complete blood count •
- CI =
- 95% confidence interval •
- RR =
- relative risk •
- ROC =
- receiver operating characteristic •
- WBC =
- white blood cell count •
- RBC =
- red blood cell count •
- CSF =
- cerebrospinal fluid
- Baraff LJ,
- Oslund S,
- Prather M
- Krisher KK,
- Whyburn DR,
- Koepnick FE
- Baraff LJ,
- Bass JW,
- Fleisher GR,
- et al.
- Kaplan SL,
- Mason EO,
- Barson WJ,
- et al.
- Campbell GD,
- Silberman R
- Shaw KN, Gorelick MH, Alpern ER, McGowan KL. Prevalence of occult bacteremia in febrile young children in the emergency department. Submitted for publication
- McCarthy PL,
- Grundy GW,
- Spiesel SZ,
- Dolan TF
- Schutzman SA,
- Petrycki S,
- Fleisher GR
- Weinbaum FI,
- Lavie S,
- Danek M,
- Sixsmith D,
- Heinrich GF,
- Mills SS
- Jaffe DM,
- Fleisher GR
- Rothrock SG,
- Harper MB,
- Green SM,
- et al.
- Kramer MS,
- Shapiro ED
- Kramer MS,
- Lane DA,
- Mills EL
- Kramer MS,
- Etezadi-Amoli J,
- Ciampi A,
- et al.
- Copyright © 2000 American Academy of Pediatrics