Carl Galloway, MD
Clinical Question: What is the performance of the Step-by-Step approach to evaluate febrile infants, and how does it compare to other existing criteria?
Background: Multiple studies have been performed to find the best set of criteria to identify febrile infants at low risk for bacterial infection in order to manage them in a less invasive manner. Common criteria used in the United States include Rochester, Philadelphia, and Boston criteria, initially published in the early 1990s. Since that time, management has evolved with the introduction of newer biomarkers such as C-reactive protein (CRP) and procalcitonin (PCT), and epidemiology has changed with immunizations and improvement in intrapartum antibiotic prophylaxis.
A new algorithm, Step-by-Step, has been developed by a group of European pediatric emergency physicians and has been shown retrospectively to accurately identify groups of patients according to risk of noninvasive or invasive bacterial infection (IBI). This algorithm uses a sequential approach, evaluating the general appearance, age of the patient, urinalysis, and then other lab findings including CRP, PCT, and absolute neutrophil count (ANC). In this study, the authors sought to validate this algorithm prospectively in a larger multicenter population.
Study Design: Multicenter prospective study.
Setting: 11 European pediatric emergency departments in Spain, Italy, and Switzerland.
Synopsis: This study included infants ≤90 days of age presenting to the pediatric emergency department (PED) between September 2012 and August 2014 with fever without source (defined as temperature ≥38°C measured by thermometer at home or in the PED, with normal physical examination and no respiratory signs or symptoms or diarrheal process). Labs obtained for each patient included urinalysis, urine culture (obtained by bladder catheterization or suprapubic aspiration), white blood cell count, PCT, CRP, and blood culture. Further testing and management were determined by the treating physician and management protocols of each center.
Exclusion criteria included:
- Clear source of fever by history or physical examination.
- No fever in the PED and fever assessed only subjectively by parents prior to presentation without the use of a thermometer.
- Absence of one or more of the above lab tests.
- Refusal of parents to participate.
The study included 2,185 infants. Of these, 504 were diagnosed with bacterial infection, including 87 (3.9%) with IBI (defined as positive blood or cerebrospinal fluid culture) and 417 (19.1%) with non-IBI (409 of which were urinary tract infections). Following the first part of the Step-by-Step approach, which uses general appearance (well-appearing versus ill-appearing), age (older or younger than 21 days), and leukocyturia, identified 79.3% of patients with IBI and 98.5% of non-IBI. Adding the next steps in the approach, with PCT, CRP, and ANC, identified 991 low-risk patients (45.3% of the studied population). In this low-risk group, seven patients were subsequently identified as having IBI (0.7% of this group). Using the Step-by-Step approach led to a negative predictive value (NPV) of 99.3 for identifying IBI, with a negative likelihood ratio (LR) of 0.17.
In evaluation of the seven low-risk patients with IBI, three of these were noted to present to the PED within one hour of onset of fever, and three more patients had fever first detected on arrival in the PED. This short duration of fever, and the lack of time for a rise in biomarkers, is likely why these patients were missed in the initial assessment.
When the Rochester criteria were used for this group of 2,185 patients, 949 patients were identified as low risk, with 1.6% of the low-risk patients found to have IBI, leading to an NPV of 98.3 and negative LR of 0.41. The authors chose to compare their approach to the Rochester criteria because the other commonly used approaches (Boston, Philadelphia) recommend lumbar puncture in all febrile infants while Rochester does not, and more recent literature suggests an individualized approach rather than recommending the test systematically.