Combination of Leukocyte Cell Surface Biomarkers Measured by Cytometry, to Differentiate Bacterial From Viral Infections in Emergency Department: a Multicentric Cohort for the Validation of Diagnostic Performances

Patients attending the ED of one of the participant centers for a suspicion of infection will be informed and asked to participate. After obtaining a non-opposition to participate, during the routine blood sampling performed in the ER, an additional volume of 30 ml of whole blood will be collected, aliquoted and stored at -80°C, comprising notably 12 ml of whole blood to which 1 ml of Cytodelics® Stabiliser will be added and incubated à room temperature for 10 mn before being aliquoted and stored at -80°C. The remaining whole blood will be collected on EDTA and Paxgen tubes, centrifugated, aliquoted and stored at -80°C for blood collection. Clinical data at admission (past medical history, vital parameters, infectious source (if any), treatments delivered in the ER) will be recorded into an eCRF. The participants will be followed up until leaving the hospital and no longer than day-28. All the microbiology tests performed during the hospital stay will be also recorded into eCRF. The diagnostic performance of the combinations of cell surface markers will be evaluated against a diagnostic reference on the bacterial of viral qualification of each infectious event. This diagnostic reference will be established by an independant expert comitee after reviewing all the clinical, and biological data recorded (excepted the results of flow cytometry), in order to adjudicate between bacterial, viral or no infection, and among infected patients to classify into infection, sepsis or septic shock (sepsis 3.0). After completing the recruitement of participants, a panel of cell surface markers will be measured by batch on a spectral cytometer, comprising notably the biomarkers of interest already published : HLA-DR, CD169 and CX3CR1 on monocytes, and MerTk, CD64 and CD24 on neutrophils. The performances of the combinations of cell surface markers already identified in the littérature will be tested prioritarily. However, in order to refine the best combinations of biomarkers to discriminate bacterial from viral infection, machine learning algorithms like gradient boosting tree and support vector machine tools will be applied on the entire results of cell surface markers measured. The diagnostic performance will be evaluated calculating the sensitivity, specificity, area under the ROC curve of the biomarkers combinations selected.