The fight against coronavirus disease 2019 (COVID-19) continues. Since the pandemic's onset, several biomarkers have been proposed to assess the diagnosis and prognosis of this disease. This research aimed to identify potential disease severity biomarkers in serum samples of patients with COVID-19 during the disease course. Data were collected using untargeted and targeted mass spectrometry methods. The results were interpreted by performing univariate and multivariate analyses. Important metabolite classes were identified by qualitative untargeted metabolomics in 15 serum samples from survivors of COVID-19. Quantitative targeted metabolomics on a larger patient cohort including 15 non-survivors confirmed serum 3-sulfate bile acids (i.e. GLCA-3S) were significantly increased in non-survivors compared to survivors during the early disease stage (p-value < 0.0001). Notably, it was associated with a higher risk of mortality (odds ratio of 26). A principal component analysis showed the ability to discriminate between survivors and non-survivors using the BA concentrations. Furthermore, increased BA-S is highly correlated with known parameters altered in severe clinical conditions.