Host-Protein Scores Could Reduce Error of Physicians’ Diagnosis of Fever Etiology 2-Fold

Conference | <b>ID Week</b>

A recent study demonstrates a tool to help physicians better diagnose the cause of fevers in children.

Classifying bacterial from viral causes of a patient’s fever in the emergency department (ED) in order to give proper care is crucial, considering antibacterial resistance is on the rise, the patient’s health could deteriorate, and time restraints of care in the ED.

In a recent study, to distinguish bacterial from viral infections, a host-protein score (BV) which combined the circulating levels of TNF-related apoptosis-induced ligand (TRAIL), interferon-γ-induced protein-10 (IP-10), and C-reactive protein (CRP), which can accurately differentiate bacterial from viral infections, was investigated. Comparing the BV to physician’s etiological diagnosis, the investigators evaluated the BV’s potential to impact decision-making in the ED.

The study, “Bacterial vs viral etiology of fever: a prospective study to estimate a host score’s potential to support ED physicians' etiologic accuracy,” was presented as a poster by Meirav Mor, MD, Head of hospital infection control and epidemiology and attending ED physician Schneider Children's Medical Center of Israel; Tel Aviv University Petach Tikva, Israel, at the IDWeek Conference in Washington, DC, on October 19-23, 2022.

“Overall, BV could potentially correct physician diagnosis and reduce error greater than 2-fold, from 16.1% to 8.1%.” the Mor, et al, noted on the poster.

The study demonstrated that a BV could conceivably assist the diagnostic process. However, the investigators noted that further studies are warranted to evaluate the practical effect of BV in the ED.

The study’s participants had a febrile respiratory tract infection or a fever without source, and they ranged from 3 months to 18 years, and were recruited from a tertiary care pediatric ED.Symptoms for fewer than 7 days and immunocompetence were the main eligibility criteria. Reasons for not qualifying were having antibiotics longer than the previous 48 hours, having a fever below the threshold, and a recent episode of infection. The “ED physician’s initial etiological suspicion (bacterial/viral/I don’t know) and degree of confidence (low/medium/high) were recorded in a questionnaire.”

Then the investigators interpreted the BV based on pre-defined score thresholds (viral/bacterial/equivocal).” To make the study clearer, the physician was unable to access the results. The reference standard for etiology was based on the diagnoses (bacterial/viral/indeterminate) of 3 independent pediatricians who were provided with all available patient data but not allowed to see the BV.

The investigators wrote, “BV’s potential to impact the physician’s etiologic accuracy was estimated according to alignment between BV and etiological suspicion as filled in the questionnaire (assuming full adoption of BV by the physician).”


Only 290 out of the 348 patients recruited, met eligibility criteria and had the necessary serum sample. The patients’ median age was 1.3 years (interquartile range 1.7), and 37.5% were female. Only 211 patients had questionnaires filled out. The physician’s label and reference standard matched the BV in 72% of the cases (151/211). However, only, 52% (78/151) of these were marked by the physician with low/medium confidence. In these cases, the BV could reinforce the physician’s proposal. Only 12% of the cases (26/211) demonstrated that the BV aligned with the reference standard but not with the physician’s suspicion, and, therefore, could correct the physician's bacterial vs viral diagnosis. Interestingly, the “BV did not align with the reference standard in 7% of the cases (14/211) and was equivocal in 9% of the cases (20/211),” the investigators noted.