Individual subject level data.

SARS-CoV-2 infections lead to a wide-range of outcomes from mild or asymptomatic illness to serious complications and death. While many studies have characterized hospitalized SARS-CoV-2 patient immune responses, we were interested in whether serious complications of SARS-CoV-2 infection could be predicted early in ambulatory subjects. To that end, we used samples from SARS-CoV-2-infected individuals from the placebo arm of the BLAZE-1 clinical trial who progressed to hospitalization or death compared to individuals in the same study who did not require medical intervention and investigated whether baseline serum cytokines and chemokines could predict severe outcome. High-risk demographic factors at baseline, including age, nasal pharyngeal viral load, duration from symptom onset, and BMI provide significant predictive capacity for a hospitalization or death with an AUC of ROC = 0.77. The predictive performance of our outcome modeling increased when baseline serum protein markers were included. In fact, the one-marker model indicated that there were 51 individual proteins (including known markers of inflammation like IL-6, MCP-3, CXCL10, IL-1Ra, and PTX3) that significantly increased the AUC of ROC beyond high-risk patient demographics alone to range between 0.78 to 0.88. Moreover, a two-marker model incorporating levels of both IL-6 and PTX3 further improved the prediction over the addition of a single protein marker to an AUC of ROC = 0.91. While the analytes identified in this study have been well-documented to be altered in SARS-CoV-2 infection, this analysis demonstrates the potential value of their use in predicting hospitalization or death in ambulatory participants infected with SARS-CoV-2 and could guide early treatment decisions.