Abstract
BACKGROUND: The relationship between triglyceride-glucose (TyG) index and acute respiratory distress syndrome (ARDS) in acute pancreatitis (AP) patients is still lacking. This study aimed to explore the association between the TyG index and ARDS in AP patients using an 8-year retrospective dataset. METHODS: This study was performed in Changsha Central Hospital from January 2015 to December 2022. Univariate analysis was done to discuss the relationship between different characteristics and ARDS in AP. Multivariate regression analysis was employed to investigate the relationship between the TyG index and ARDS in AP. Eight machine learning models were employed to assess the in-hospital ARDS risk in AP patients. The SHapley Additive exPlanations (SHAP) method was utilized to verify the importance of TyG in the models. RESULTS: A total of 2,382 AP patients were finally enrolled, and ARDS occurred in 137 patients. With per-unit increment in TyG index, the risk of ARDS in AP increased by 133%(OR = 2.33, 95%CI: 1.51-3.60, p = 0.0001) after adjusting all potential confounders. The relationship between the TyG index and ARDS in AP was non-linear. The XGBoost (AUC = 0.857 ± 0.034) and Random Forest (AUC = 0.851 ± 0.045) algorithms were the best two performance methods. In the SHAP analysis, TyG was the second most important feature in the RF model and the seventh in the XGBoost model. CONCLUSION: TyG index was associated with in-hospital ARDS in AP. The XGBoost and Random Forest models based on the TyG index had the best performance for predicting ARDS in AP patients. The SHAP method further confirmed that the TyG index serves as a significant predictor for the development of ARDS in patients with acute pancreatitis.