BACKGROUND:The effect of blood eosinophils (EOSs) on mortality in acute respiratory distress syndrome (ARDS) patients and whether corticosteroids affect this effect are unclear. METHODS:The Medical Information Mart for Intensive Care III database (version 1.4) was used to extract data. Patients with ARDS were selected for inclusion. Cox regression models using the backward stepwise method and propensity score matching (PSM) were used to assess the relationship between blood EOS counts and 28-day mortality. RESULTS:A total of 2,567 patients with ARDS were included, and the 28-day mortality rate was 24.19％. The crude 28-day mortality was significantly lower in patients with EOS counts ≥2％ (18.60％[85/457] vs. 25.40％ [536/2,110], P=0.002) than in those with EOS counts <2％. In the Cox regression model, the EOS counts ≥2％ showed a significant association with the decreased 28-day mortality (hazard ratio [HR] 0.731; 95％ confidence interval [95％ CI] 0.581–0.921, P=0.008). In the corticosteroid non-use subgroup, EOS counts ≥2％ was significantly related to decreased 28-day mortality (HR 0.697, 95％ CI 0.535–0.909, P=0.008), but the result was not significant in the corticosteroid non-use subgroup model (P=0.860). A total of 457 well-matched pairs were obtained by a 1:1 matching algorithm after PSM. The 28-day mortality remained significantly lower in the EOS counts ≥2％ group (18.60％ [85/457] vs. 26.70％ [122/457], P=0.003). CONCLUSIONS Higher EOS counts are related to lower 28-day mortality in ARDS patients, and this relationship can be counteracted by using corticosteroids.