As elevated levels of the soluble CXCL16 (sCXCL16) chemokine have been reported in severe coronavirus disease 2019 patients, this study examined whether sCXCL16 concentration on the first day of hospitalization predicted death in COVID-19 patients. A total of 76 patients with COVID-19 were admitted to the Military Hospital of Tunis, Tunisia, between October 2020 and April 2021, and later classified as survivors or nonsurvivors based on their outcomes. At admission, the groups were matched by age, gender, comorbidities, and the percentage of patients with moderate conditions. On the first day of admission, serum's sCXCL16 concentrations were measured using a magnetic-bead assay. There was an eightfold increase in serum sCXCL16 levels in the nonsurvivors' group (3661.51 ± 2464.87 pg/mL vs. 454.3 ± 338.07 pg/mL, p < 0.0001). For the optimal cutoff value of sCXCL16 at 2095 pg/mL, we found a 94.6% sensitivity and a 97.4% specificity, with an area under curve of 0.981 (p = 5.03E−08; 95% confidence interval [95% CI]: 0.951-1.0114). Considering the risk of death at a concentration above the threshold, the unadjusted odds ratio was 36 (p < 0.0001). The adjusted odd ratio was estimated at 1.003 (p < 0.0001; 95% CI: 1.002-1.004). Finally, there was a significant difference between survival and nonsurvival groups in leukocyte numbers (p = 0.006), lymphocytes (p = 0.001), polymorphonuclear neutrophils (p = 0.001), and C-reactive protein levels (p = 0.007), except for monocytes (p = 0.881).Based on these results, sCXCL16 level could be used for detecting nonsurvival COVID-19 patients. Therefore, we recommend assessing this marker in hospitalized COVID-19 patients.