In this study, we investigated immune responses in a prospective cohort of hospitalized COVID-19 patients (derivation cohort, DC; n = 126) during the spring of 2020. Plasma samples were collected within four days of admission to measure a panel of innate and complement system immune markers. From a subset of this cohort (n = 80), we performed plasma proteomics and developed a predictive survival model based on proteins differentially expressed between survivors (n = 60) and non-survivors (n = 20). Logistic regression analyses were employed to evaluate the predictive value of these biomarkers for 30-day mortality. The developed proteomic prediction model and the regression analyses were subsequently validated using an independent cohort of hospitalized COVID-19 patients (validation cohort, VC) from the autumn/winter of 2020, who received treatment with remdesivir and dexamethasone.