Identification of novel immune biomarkers to gauge the underlying pathology and severity of COVID-19 has been difficult due to the lack of longitudinal studies. Here, we analyzed serum collected upon COVID-19 admission (t1), 48 hours (t2), and seven days later (t3) using Olink proteomics and correlated to clinical, demographics, and therapeutic data. Older age positively correlated with decorin, pleiotrophin, and TNFRS21 but inversely correlated with chemokine (both C-C and C-X-C type) ligands, monocyte attractant proteins (MCP) and TNFRS14. The burden of pre-existing conditions was positively correlated with MCP-4, CAIX, TWEAK, TNFRS12A, and PD-L2 levels. Individuals with COVID-19 demonstrated increased expression of several chemokines, most notably from the C-C and C-X-C family, as well as MCP-1 and MCP-3 early in the course of the disease. Similarly, deceased individuals had elevated MCP-1 and MCP-3 as well as Gal-9 serum levels. LAMP3, GZMB, and LAG3 at admission correlated with mortality. Only CX3CL13 and MCP-4 correlated positively with APACHE score and length of stay, while decorin, MUC-16 and TNFRSF21 with being admitted to the ICU. We also identified several organ-failure-specific immunological markers, including those for respiratory (IL-18, IL-15, Gal-9) or kidney failure (CD28, VEGF). Treatment with hydroxychloroquine, remdesivir, convalescent plasma, and steroids had a very limited effect on the serum variation of biomarkers. Our study identified several potential targets related to COVID-19 heterogeneity (MCP-1, MCP-3, MCP-4, TNFR superfamily members, and programmed death-ligand), suggesting a potential role of these molecules in the pathology of COVID-19.