Abstract Background Interleukin-6 (IL-6) was proposed to be associated with the severity of coronavirus disease 2019 (COVID-19). The present study aimed to explore the kinetics of IL-6 levels, validate this association in COVID-19 patients, and report preliminary data on the efficacy of IL-6 receptor blockade. Methods We conducted a retrospective single-institutional study of 901 consecutive confirmed cases. Serum IL-6 concentrations were tested on admission and/or during hospital stay. Tocilizumab was given to 16 patients with elevated IL-6 concentration. Results 366 patients were defined as common cases, 411 patients as severe, and 124 patients as critical according to the Chinese guideline on diagnosis and treatment of COVID-19. The median concentration of IL-6 was < 1.5 pg/ml (IQR < 1.50–2.15), 1.85 pg/ml (IQR < 1.50–5.21), and 21.55 pg/ml (IQR 6.47–94.66) for the common, severe, and critical groups respectively (P < 0.001). The follow-up kinetics revealed serum IL-6 remained high in critical patients even when cured. An IL-6 concentration higher than 37.65 pg/ml was predictive of in-hospital death (AUC 0.97 [95% CI 0.95–0.99], P < 0.001) with a sensitivity of 91.7% and a specificity of 95.7%. In the 16 patients who received tocilizumab, IL-6 concentrations were significantly increased after administration, and survival outcome was not significantly different from that of propensity-score matched counterparts (n = 53, P = 0.12). Conclusion Serum IL-6 should be included in diagnostic work-up to stratify disease severity, but the benefit of tocilizumab needs further confirmation. Trial registration retrospectively registered.