Abstract Background Diagnosing bacterial infections in patients with symptoms of respiratory tract infections or unexplained fever is challenging in the Emergency Department (ED). Providers typically rely on clinical histories, physical examinations, and laboratory tests, e.g., white blood cell counts, absolute neutrophil counts, C-reactive protein (CRP), and procalcitonin (PCT) to assess the etiology of infection. Currently, there are no well-established and universally accepted clinical algorithms to reliably differentiate bacterial from viral infections in the ED. The lack of standardized diagnostic protocols causes uncertainty in clinical decision-making, leading to sub-optimal patient care. One major concern is the overuse of antibiotics. The MeMed-BV test offers a potential solution for rapid bacterial-viral differentiation by generating a single numeric score based on the levels of three host proteins: TNF-related apoptosis-inducing ligand (TRAIL), interferon gamma-induced protein-10 (IP-10), and CRP. Despite FDA approval, the clinical adoption of MeMed-BV remains limited due to insufficient real-world comparative data and the natural hesitancy of clinicians to evaluate and adopt a new diagnostic assay. This study aims to evaluate the analytical and clinical performance of MeMed-BV, comparing it with PCT and assessing its potential to reduce unnecessary antibiotic use. Methods We collected remnant serum samples from ED patients presenting with respiratory infection symptoms and/or unexplained fever who had a positive respiratory pathogens panel or bacterial culture, or elevated PCT (0.25 ng/mL). MeMed-BV results were classified as: high-viral infection/other non-bacterial etiology (0-10), mod-viral infection/other non-bacterial etiology (11-34), equivocal (35-65), mod-bacterial/co-infection (66-89), and high-bacterial/co-infection (90-100). We also ran an additional PCT offline on all these samples. Adjudication of etiology was performed by pathology residents and an ED physician based on clinical and laboratory findings. In addition, analytical validation including linearity and precision was carried out. Results Analytical validation was satisfactory, with linearity regression exceeding 0.999, recoveries between 90%-110%, and intra-day and inter-day precision within 2% and 9%, respectively, for all three markers. A total of 176 ED patients were included, with adjudication confirming 97 (55.1%) viral infections, 70 (39.8%) bacterial/co-infections, and 9 (5.1%) non-infectious cases. MeMed-BV identified 86 (48.9%) bacterial/co-infections, 74 (42.0%) viral infections, and 16 (9.1%) equivocal cases. Therefore, MeMed-BV exhibited a sensitivity of 95.5% (95%CI: 90.4-99.5%), specificity of 75.5% (95%CI: 66.8-84.2%), positive predictive value (PPV) of 73.3% (95%CI: 63.9-82.6%), and negative predictive value (NPV) of 95.9% (95%CI: 91.5-100%) on differentiating bacterial infections. In comparison, PCT exhibits a higher specificity (86.2%, 95%CI: 79.2-93.2%, p = 0.06) and PPV (77.6%, 95%CI: 66.9-88.3%, p = 0.56), but lower sensitivity (71.4%, 95%CI: 60.3-82.6%, p0.001) and NPV (81.8%, 95%CI: 74.2-89.4%, p=0.005). Notably, 43 out of 106 (40.6%) of non-bacterial infection cases received antibiotics. Of those, the MeMed-BV assay identified 25 (58.1%) cases as viral infection. Thus, incorporating the MeMed-BV test in clinical workflow could potentially reduce antibiotic prescription. Conclusion MeMed-BV demonstrated a high sensitivity and NPV for ruling out bacterial infections in ED patients as well as the promising potential to improve antibiotic stewardship.
Li et al. (Wed,) studied this question.