1330. Evaluation of Multiple Host Response-Based Strategies to Classify Acute Respiratory Illness

Background Host response-based diagnostics are an alternative to pathogen-based tests. Host response strategies include proteomic and transcriptomic approaches. Here, we compare three host response strategies for ARI diagnosis: Procalcitonin (PCT), a 3-protein panel, and an mRNA panel. Methods PCT, a 3-protein panel (CRP, IP-10, TRAIL), and a host gene expression mRNA panel were measured in a cohort of 286 participants presenting to one of the four Emergency Departments with ARI due to bacterial (n = 47), viral (n = 162), or noninfectious (n = 77) etiologies. Multinomial logistic regression and leave-one-out cross-validation were used to train and evaluate the protein and mRNA panels. Performance characteristics were calculated for each method, and their combination, for the ability to discriminate bacterial vs. non-bacterial infection and viral vs. nonviral infection. PCT was not evaluated for viral vs. nonviral discrimination since it does not discriminate viral and noninfectious etiologies. McNemar’s test was used to compare overall accuracy of mRNA and protein panels. Results For discriminating bacterial vs. non-bacterial etiologies, the mRNA panel had an AUC of 0.93 vs. 0.83 for both the protein panel and PCT. A model utilizing all three strategies was the same as mRNA alone. Using previously established cutoffs, overall accuracy was similar between mRNA and protein panels, but the protein panel had widely discordant sensitivity (43%) and specificity (92%). When selecting an optimal cutoff for the protein panel that balanced the two (82% and 73%, respectively), the mRNA panel had a significantly greater overall accuracy (P < 0.001). Similar results were found when discriminating viral vs. non-viral subjects: the mRNA panel (AUC = 0.93) outperformed the protein panel (AUC = 0.84). Combining the mRNA and protein panels was equivalent to the mRNA panel alone. Conclusion A host-based gene expression signature is the most effective platform for classifying subjects with bacterial, viral, or noninfectious ARI. A gene expression approach, when translated to a clinically available platform, may facilitate diagnosis and clinical management of acute infectious diseases, mitigating antibiotic overuse. Disclosures Ephraim L. Tsalik, MD, MHS, PhD, Immunexpress: Consultant; Predigen, Inc.: Officer or Board Member, Research Grant.