Eosinophil count combined with routine indicators enhances early risk prediction value for ARDS in ICU patients: a retrospective cohort study.

Objectives: This study aims to systematically evaluate the incremental value of admission eosinopenia EOS(-) in enhancing the predictive efficacy of early-stage acute respiratory distress syndrome (ARDS) risk models for ICU patients. Methods: This study employed a single-center retrospective cohort design, enrolling 482 adult ICU patients between 2020 and 2023. Using Cox proportional hazards regression, we constructed and compared three models: Model 1 (baseline clinical model) included age, pneumonia, sepsis, and SOFA score; Model 2 added lymphocytes and eosinophils to Model 1; Model 3 further adjusted for BMI and diabetes. Predictive performance was assessed using the C-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI), and validated internally. Results: C-statistics for Models 1, 2, and 3 were 0.683, 0.710, and 0.729, respectively. Model 3 demonstrated optimal performance (NRI=0.185, p=0.004), with an adjusted C-statistic of 0.695. Multivariate analysis identified age, pneumonia, and sepsis as independent ARDS risk factors, while EOS(-) showed non-independence. Results from competing risks analysis were consistent with the primary analysis conclusions. Conclusions: Although eosinopenia is not an independent predictor of ARDS, it provides incremental information for early risk stratification based on conventional clinical factors, thereby aiding in the identification of high-risk patients.