Date: June 14th, 2020

Guest Skeptic: Dr. Dennis Ren is a Pediatric Emergency Medicine fellow at Children’s National Hospital in Washington, DC.

Reference: Kuppermann et al. A Clinical Prediction Rule to Identify Febrile Infants 60 Days and Younger at Low Risk for Serious Bacterial Infections. JAMA Pediatr. 2019.

CaseA 5-week-old full term female presents to the Emergency Department (ED) for fever with rectal temp of 100.6F (38.1C). Her mother states that she has been fussier today. She also seems “congested” and is not feeding as well. She continues to have the usual number of wet diapers. The mother is worried about her sick baby. She wants to know if they will need a spinal tap, be placed on antibiotics or will need to be admitted to the hospital?

Background: Fever without source in infants less than three months old represents a significant diagnostic dilemma for clinicians. Several criteria have been developed previously, including the Rochester (Jaskiewicz et al 1994), Boston (Baskin et al 1992) and Philadelphia (Baker et al 1993) criteria to help clinicians stratify the risk of serious bacterial infections (SBI).

Febrile infants commonly present to the emergency department. It is estimated 8-13% may have SBI that may include urinary tract infections, bacteremia, and bacterial meningitis. It is difficult to identify which infants have SBI by clinical examination alone. There are serious consequences from missed SBI. Workup for SBI may include lumbar puncture, antibiotics, and hospitalization.

These criteria (Rochester, Boston and Philadelphia) could be considered out of date in our current era of vaccinations. We covered a new protocol called the Step-by-Step approach on SGEM#171. The “Step-by-Step”rule combined both clinical factors and laboratory factors in febrile infants aged 22 to 90 days. It had a sensitivity of 98.9% to detect all SBIs.

The SGEM Bottom Line #171:If you have availability of serum procalcitonin measurement in a clinically-relevant time frame, the Step-by-Step approach to fever without source in infants 90 days old or younger is better than using the Rochester criteria or Lab-score methods. With the caveat that you should be careful with infants between 22-28 days old or those who present within two hours of fever onset.”

It is important to balance the consequences of missing an SBI with performing unnecessary procedures (lumbar punctures), exposing infants to antibiotics, and prolonging hospital stay. The new study proposes a novel way of identifying low risk febrile infants 29-60 days based on three objective lab criteria.

Clinical Question: Can a clinical prediction rule (tool) using laboratory data identify febrile infants under 60 days of age who are at low risk for serious bacterial infection (urinary tract infection, bacteremia, and bacterial meningitis) and reduce unnecessary lumbar punctures, antibiotic exposure, and hospitalization?

Reference: Kuppermann et al. A Clinical Prediction Rule to Identify Febrile Infants 60 Days and Younger at Low Risk for Serious Bacterial Infections. JAMA Pediatr. 2019.

  • Population: Febrile infants <60 days of age who look good and whose blood cultures were obtained to rule out SBI (fever was a rectal temperature of at least 38C)
    • ExcludedInfants who looked critically ill, had antibiotics in the previous 48 hours, history of prematurity (≤36 weeks’ gestation), pre-existing medical conditions, indwelling devices or soft tissue infections.
  • Intervention: Derivation and validation of accurate clinical prediction rule (tool) for infants at low risk of SBI using a negative urinalysis, ANC <4,090/uL, and procalcitonin 1.71 ng/ml or less
  • Comparison: Pre-existing algorithms combining subjective clinical findings and lab markers

  • Outcome: Accuracy of the prediction rule to identify infants at low risk for SBI (sensitivity, specificity, negative prediction value and negative likelihood ratio).
    • SBI was defined as bacterial meningitis, bacteremia or UTI.
      • UTI was defined as growth of a single urine pathogen with at least 1,000 cfu/ml on culture obtained by suprapubic aspiration, at least 50,000 cfu/ml from catheterized specimens or 10,000-50,000 cfu/ml from catheterized specimens in association with an abnormal urinalysis (presences of leukocytes esterase, nitrite or pyuria).

Authors’ Conclusions:We derived and validated an accurate prediction rule to identify febrile infants 60 days and younger at low risk for SBIs using the urinalysis, ANC, and procalcitonin levels. Once further validated on an independent cohort, clinical application of the rule has the potential to decrease unnecessary lumbar punctures, antibiotic administration, and hospitalizations.”

Quality Checklist for Clinical Decision Tools:

  1. The study population included or focused on those in the ED. Yes
  2. The patients were representative of those with the problem. Yes
  3. All important predictor variables and outcomes were explicitly specified. Yes
  4. This is a prospective, multicenter study including a broad spectrum of patients and clinicians (level II). Yes
  5. Clinicians interpret individual predictor variables and score the clinical decision rule reliably and accurately. Yes
  6. This is an impact analysis of a previously validated CDR (level I). No
  7. For Level I studies, impact on clinician behavior and patient-centric outcomes is reported. N/A
  8. The follow-up was sufficiently long and complete. Unsure
  9. The effect was large enough and precise enough to be clinically significant. Unsure.

Key Results: The study included 1,821 febrile infants <60 days of age who had blood cultures collected. The mean age was 36 days old, 42% female, 2/3 had a fever of <12 hours prior to the ED visit, median YOS was 6.0 and SBI positive in 9.3% (7.7% from UTI alone).

Low risk prediction rule was derived based on three variables:

  1. Normal urinalysis
  2. Absolute Neutrophil Count (ANC) ≤4,090/µL
  3. Serum procalcitonin ≤1.71 ng/ml

Sensitivity of 97.7%, specificity of 60%, NPV of 99.6% and LR- of 0.04.

No infants with bacterial meningitis were missed.

There were 1,266 infants >28 days of age. The clinical prediction rule stratified 776/1,266 (61.3%) as low risk for SBI. Of that low risk subgroup, 523/776 (67.4%) had lumbar punctures performed. This is the number of lumbar punctures that could possibly be avoided in this age group for low-risk patients.

Overall, this is a very well-executed study with practice changing potential. The PECARN group does some great research. We covered the very important paper by Kupperman et al (NEJM 2018) looking at fluid infusion rates for children with diabetic ketoacidosis on SGEM#255.

1) Procalcitonin: This study lost a portion of eligible participants due to procalcitonin sample issues (41%). It is difficult to say whether or not this would have changed the results or analysis. The authors state that this group was similar to those with the procalcitonin measurements. There is a slight difference in SBI positive percentage in those who have procalcitonin test results available compared to those who did not (9.3% vs. 12%).

Laboratory tests besides procalcitonin like CRP, band counts and viral studies were not included. Past studies demonstrate concurrent viral infections may decrease risk of SBI but does not exclude SBI.

The procalcitonin samples were centrifuged and frozen at -80C, batched, and all sent to a central laboratory. This can increase the precision of the results by decreasing variability. Having 26 different laboratories running the procalcitonin levels could introduce more variability into the results. This would potentially decrease the precision by increasing the 95% confidence interval around the point estimate for the diagnostic accuracy of the clinical prediction rule.

To avoid over-fitting the data, they rounded off the ANC to 4,000/uL but more significantly decreased the procalcitonin level from 1.71 ng/ml to 0.5 ng/ml to see what would happen to the diagnostic accuracy.

The sensitivity remained identical and the specificity decreased a little with rounding off the ANC and decreasing the procalcitonin level.

2) Urinary Tract Infection: The definition used in this study followed the AAP guidelines that we detailed in the PICO. However, they did use a lower threshold for colony forming units.The reason for including this lower threshold was to account for lower colonies of bacteria sometimes found in urine of younger infants. This definition makes the prediction rule a bit more conservative.

3) Younger vs Older Infants: There was a difference between the two subgroup of infants identified a priori. Overall, 9.3% had SBIs but it was 13% in younger infants (≤28 days of age) vs. 7.7% in the older population (infants >28 days of age). Most of the SBI were UTI. Five infants had bacteremia and meningitis. Ten infants had UTI and bacteremia. One poor infant in the ≤28-day age group had UTI, bacteremia, and meningitis.

4) Number of Cases: We need to be careful about just considering negative predictive value (NPV). NPV is dependent on prevalence of disease while likelihood ratios are not dependant on prevalence of disease. They identified 170 infants with SBI (9.3%) which gives a fairly tight 95% CI for NPV. However, there were only four cases of bacterial meningitis alone (0.2%) which can make the NPV look pretty good (100%) but with very wide 95% CI.

We cannot just consider the number of cases identified but also the number of missed cases. In this study there were three missed cases of SBI using the clinical prediction rule.  Two of the missed patients had culture positive urine without pyuria (E. coli and Pseudomonas) were in validation set. One patient with positive blood culture for Enterobacter cloacae was in derivation set but repeat blood cultures prior to antibiotics were never positive. No cases of bacterial meningitis were missed.

Four patients had herpes simplex virus (HSV) infections. Three with HSV in central nervous system (CNS) and one in nasopharynx. This is important because HSV most often presents in the first month of life.

5) Bias: This is something that systematically moves us away from the “truth” (best point estimate of effect) and is not random noise in the data. There could have been a number of sources of bias in this study.

One potential bias would be selection bias. To be included in the cohort they had to have blood cultures drawn. What were the differences in characteristics in those with blood cultures obtained and those infants without? Both groups presented to the ED with fever. They also only enrolled patients when a research coordinator was available. These two things could have introduced some selection bias.  

Another potential bias would be differential verification bias (double gold standard). This occurs when the test results influence the choice of the reference standard. So, a positive index test gets an immediate/gold standard test whereas the patients with a negative index test get clinical follow-up for disease. In this study, only 77% of febrile infants got a lumbar puncture to rule out meningitis. Those who did not get an LP had their families contacted by telephone 8 to 14 days after the ED visit and/or reviewed their medical records.  This type of bias can raise or lower sensitivity/specificity

Comment on Authors’ Conclusion Compared to SGEM Conclusion: We generally agree with the authors’ conclusions.

SGEM Bottom Line: This new clinical prediction rule has the potential to decrease unnecessary lumbar punctures, antibiotic administration, and hospitalizations in febrile infants 60 days and younger at low risk for serious  bacterial infections but needs to be externally validated.

Case Resolution: You explain to the mother that given her baby’s age (35 days), she is at risk of SBI. You need to check the blood and urine for signs of infection. Pending the results, you may also need to do an LP. Urine and blood samples are obtained and sent for culture. The urinalysis is unremarkable. The blood tests demonstrate an ANC and procalcitonin below their respective cut offs. The patient has taken a bottle in the ED and is now afebrile with stable vital signs. You do not administer antibiotics or admit the infant to the hospital. You reassure the mother it is OK to go home. She is told to follow up in the next 1-2 days with her pediatrician or return to the ED if she has any concerns.

Dr. Dennis Ren

Clinically Application: This is a well-designed and executed study that offers a novel way of identifying low risk febrile infants age 29-60 days based on objective lab criteria. The clinical prediction rule with three objective lab findings can help identify infants at low risk for SBI and may spare patients the need for lumbar puncture, empiric antibiotics, and hospitalization. I would NOT apply this rule to infants ≤28 days as they have higher risk of infection including HSV which the rule does not account for and the procalcitonin of ≤1.71 ng/mL is cut off when urinalysis and ANC are normal. As a reminder, clinical prediction rules do NOT replace clinical judgement. Prediction rules should help guide clinical judgement not dictate clinical care. Clinical predication tools often have lower diagnostic accuracy and wider confidence intervals when external validation is performed. This tool should be externally validated before recommending its general use.

What Do I Tell My Patient (parents/caregivers)? In babies under the age of 60 days with fever, we always think about the possibility of infection. This could be an infection in the urine, blood, or spinal fluid which coats the brain. I would like to start by testing your infant’s urine and blood for signs of infection. Depending on those results, we may need to perform a lumbar puncture to obtain spinal fluid, give antibiotics, or have your child stay in the hospital. We can talk more after the initial results come back.

Keener Kontest: There was no winner last week. The correct answer was YEARS is the pulmonary embolism risk stratification algorithms has also been validated in pregnancy.

Listen to the SGEM podcast to hear this weeks’ question. Send your answer to with “keener” in the subject line. The first correct answer will receive a cool skeptical prize.

 Other FOAMed Resources:

Remember to be skeptical of anything you learn, even if you heard it on the Skeptics’ Guide to Emergency Medicine