Comparison of two rapid host-response tests for distinguishing bacterial and viral infection in adults with acute respiratory infection

Abstract

Objectives
Distinguishing bacterial from viral acute respiratory infection (ARI) is challenging, leading to inappropriate antimicrobial use and antimicrobial resistance. We evaluated the accuracy of two host-response tests to differentiate bacterial and viral infection.
Methods
This study used patient blood samples previously collected during a randomised controlled trial of adults hospitalised with ARI. The aetiology for each patient was clinically adjudicated. PAXgene blood RNA samples were tested using the TriVerity test (which measures 29 mRNAs) and serum samples were tested using the MeMed BV test (which measures 3 proteins). Diagnostic accuracy was calculated against adjudicated aetiology.
Results
169 patients were tested. Median age was 60 (45-74) years and 152 (90%) received antibiotics. 60 (36%) were adjudicated as bacterial, 54 (32%) as viral, 26 (15%) as viral/bacterial co-infection, and 29 (17%) as non-infected. For bacterial (including bacterial/viral co-infection) versus non-bacterial infection, the TriVerity bacterial score had a Positive Percentage Agreement (PPA) of 81% (95%CI 70-89) and a Negative Percentage Agreement (NPA) of 66% (95%CI 55-79) and the MeMed BV score had a PPA of 96% (95%CI 90-99) and NPA of 34% (95%CI 23-47). The AUROC for the two tests was 0.77 (95%CI 0.70-0.84) and 0.81 (95%CI 0.74-0.87) respectively, p=0.388.
Conclusions
Both tests demonstrated similar overall accuracy for distinguishing bacterial infection with the Triverity test missing some bacterial infections and MeMed BV misclassifying most viral infections as bacterial. Prospective impact studies evaluating antibiotic use, safety and cost effectiveness are now required.

More information

Identifiers

ORCID for Nathan James Brendish: orcid.org/0000-0002-9589-4937

ORCID for Tristan Clark: orcid.org/0000-0001-6026-5295

Catalogue record

Export record

Altmetrics