Preliminary results from the RECOVERY trial indicated that dexamethasone usage markedly reduced death MESHD rate in COVID-19 patients receiving invasive mechanical ventilation. However, the overall reduction for the entire patient cohort in that trial was much more modest, indicating highly variable effects of corticosteroid usage among COVID-19 patients. While steroid treatment is known to have both clinical efficacy and detrimental adverse-effects, defining a clinic parameter that could guide the beneficial corticosteroid usage for treating COVID-19 remains an elusive, urgent, and critical unmet need in COVID-19 therapy. Here, we undertook a multicentered retrospective study on a cohort of 12,862 confirmed COVID-19 cases from 21 hospitals in Hubei Province, China, including 3,254 received corticosteroid treatment and 9,608 received usual care without corticosteroid. We uncovered that the clinical benefits of corticosteroid use were closely associated with the neutrophil-to-lymphocyte ratio (NLR) measured at admission. Among participants with NLR > 6.12 at admission, corticosteroid treatment was significantly associated with a lower risk of 60-day all-cause mortality of COVID-19 based on both Cox model with time-varying exposure and Marginal Structural Model. However, in patients with NLR ≤ 6.12 at admission, corticosteroid treatment was no longer associated with reduced risk of all-cause death MESHD, but rather with increased risks of severe adverse effects, particularly in hyperglycemia MESHD hyperglycemia HP and infection MESHD. In diabetic patients with COVID-19, corticosteroid treatment was associated with increased glycemia, but not with a higher risk of 60-day mortality. Therefore, our study has uncovered NLR as a clinical indicator to stratify COVID-19 patients in their response to corticosteroid therapy. This finding may assist clinical evaluation and future randomized controlled trials to establish proper guidelines for corticosteroid therapy in COVID-19 patients.