Diagnostic Accuracy of a Bacterial Versus Viral Host-protein Test for Children Hospitalized With Acute Infections

Abstract

BACKGROUND: A host-protein test's diagnostic accuracy for discriminating bacterial from viral infections [MeMed BV (MMBV)] was established at emergency departments and urgent care settings. We determined its performance in children postadmission and in subcohorts stratified according to timing of the blood draw. METHODS: We analyzed postadmission MMBV data from children (3 months to 17 years) with suspected acute infections recruited across 5 previous studies. MMBV results were interpreted as bacterial/viral/equivocal according to the manufacturer's instructions. Reference standard infection etiology was as assigned in the original studies, where adjudicators were provided comprehensive patient data but blinded to MMBV. We calculated diagnostic performance by comparing MMBV to the reference standard. RESULTS: The study population comprised 1059 children, encompassing 659 patients sampled on admission day (day = 0), 69 patients sampled on day ≥1 of hospital stay and 331 with blood drawn postadmission without recorded timing. Median age was 1.9 years (interquartile range 1.0, 4.0), with 51.5% males. The most prevalent discharge diagnoses were systemic viral infections (29.7%), upper respiratory tract infection (17.6%) and lower respiratory tract infection (14.4%). MMBV attained comparable area under the receiver operating characteristic curves ( P > 0.9) of 0.92 (95% confidence interval: 0.90-0.94) for the study population, 0.92 (0.89-0.94) for those sampled on day = 0, 0.92 (0.82-1.0) for those sampled on day ≥1 of hospital stay and 0.92 (0.88-0.96) for those with sampling time unknown. CONCLUSIONS: These data support MMBV's performance in hospitalized children. Real-world studies are warranted to establish MMBV's utility postadmission.