A gradient boosting machine model for predicting prognosis in patients with acute respiratory distress syndrome

OBJECTIVES: To develop a gradient boosting model for predicting the prognosis of patients with acute respiratory distress syndrome (ARDS), providing a data-driven reference for early identification of high-risk patients in clinical settings. METHODS: This retrospective study analyzed the 28-day mortality in 307 ARDS patients treated at Qinzhou First People's Hospital between July 2023 and June 2024. Patients were divided into a mortality group (n=92) and a survival group (n=215) based on in-hospital death. Demographic characteristics, clinical variables, and biochemical parameters were collected. Univariate and multivariate logistic regression analyses were performed to identify independent predictors, which were subsequently used to construct a gradient boosting machine (GBM) model and a nomogram model. Model performance was evaluated with calibration curves and the area under the receiver operating characteristic (ROC) curve (AUC). RESULTS: Logistic regression identified age, oxygenation index (OI), neutrophil-to-lymphocyte ratio (NLR), interleukin-8 (IL-8), and N-terminal pro-B-type natriuretic peptide (NT-proBNP) as independent prognostic factors for ARDS. In the GBM model, the relative importance of NT-proBNP, age, NLR, IL-8, and OI was ranked. The nomogram indicated that older age, lower OI, and higher levels of NLR, IL-8, and NT-proBNP were associated with poorer prognosis. The AUC values for the GBM model in the training and validation sets were 0.907 (95% CI: 0.866-0.947) and 0.887 (95% CI: 0.803-0.971), respectively, which surpassed the values of 0.866 (95% CI: 0.810-0.923) and 0.835 (95% CI: 0.733-0.937) for the Nomogram model. CONCLUSION: The 28-day mortality rate among ARDS patients was 29.97%, and was mainly associated with age, oxygenation index, NLR, IL-8, and NT-proBNP levels. A GBM model constructed using these factors showed good predictive performance, offering valuable data references for clinical identification of ARDS patients at a high-risk of poor prognosis.