ABSTRACT
BACKGROUND Accurately identifying COVID-19 patients at-risk to deteriorate remains challenging. Tools integrating host-protein expression have proven useful in determining infection etiology and hold potential for prognosticating disease severity. METHODS Adults with COVID-19 were recruited at medical centers in Israel, Germany, and the United States. Severe outcome was defined as intensive care unit admission, non-invasive or invasive ventilation, or death. Tumor necrosis factor related apoptosis inducing ligand (TRAIL) and interferon gamma inducible protein-10 (IP-10; also known as CXCL10) and C-reactive protein (CRP) were measured using an analyzer providing values within 15 minutes. A signature indicating the likelihood of severe outcome was derived generating a score (0-100). Patients were assigned to 4 score bins. RESULTS Between March and November 2020, 518 COVID-19 patients were enrolled, of whom 394 were eligible, 29% meeting a severe outcome. The area under the receiver operating characteristic curve (AUC) of the signature was 0.86 (95% confidence interval: 0.81-0.91). Performance was not confounded by age, sex, or comorbidities and superior to IL-6 (AUC 0.77; p = 0.033) and CRP (AUC 0.78; p < 0.001). Likelihood of severe outcome increased significantly (p < 0.001) with higher scores. The signature differentiated patients who further deteriorated after meeting a severe outcome from those who improved (p = 0.004) and projected 14-day survival probabilities (p < 0.001). CONCLUSION The derived immune-protein signature combined with a rapid measurement platform is an accurate predictive tool for early detection of COVID-19 patients at-risk for severe outcome, facilitating timely care escalation and de-escalation and appropriate resource allocation. FUNDING MeMed funded the study