A prediction nomogram based on machine learning methods for acute respiratory distress syndrome in patients with acute pancreatitis: A retrospective study

Abstract To construct a predictive nomograph for acute pancreatitis (AP) complicated with acute respiratory distress syndrome (ARDS) admitted to the intensive care unit (ICU) using machine learning methods. This study was designed as a retrospective investigation of data from patients enrolled with AP. These patients were divided into a training cohort and a testing cohort. Machine learning methods were used to select independent predictive factors and to establish a nomogram model, and the feasibility of the nomogram model was evaluated by the test set. The area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis (DCA) were used to assess the discriminatory ability, precision, and clinical applicability of the model. A total of 427 patients were included, 344 of whom were in the training cohort and 83 in the testing cohort. The support vector machine (SVM) model showed the best performance among the six machine-learning models. Age, sex, the sequential organ failure score (SOFA), C-reactive protein (CRP), platelet count (PLT), total serum bilirubin (TBIL), and direct bilirubin (DBIL) levels were the best predictors for patients with AP presenting ARDS through the SVM learning model. These seven variables were incorporated to construct a nomogram. The C-index of the model was 0.7977 in the training cohort and 0.8484 in the testing cohort. The calibration curve for predictive probability showed that the nomogram-based predictions were in good agreement with the actual observations. The DCA plot demonstrated a good net benefit for this model, and external validation confirmed its reliability. The prediction nomogram constructed based on the SVM model in this study can effectively predict the probability of AP complicated by ARDS.