高脂血症性急性胰腺炎患者疾病严重程度预测模型构建OA

Objective To construct a disease severity predictive model for patients with hyperlipidemia acute pancreatitis by using machine learning methods.Methods A retrospective study was conducted on 320 patients with hyperlipidemic acute pan-creatitis in a tertiary hospital in Binzhou City,Shandong Province,and the clinical examination indicators and the Atlanta severity score of pancreatitis were used as variables to construct the machine learning predictive model.Results The Logistic regression showed that body mass index(BMI),triglyceride(TG),glucose(Glu),amylase(AMY)and white blood cell(WBC)were in-dependent influencing factors for the severity of acute pancreatitis with hyperlipidemia,and serum calcium level and high density lipoprotein cholesterol(HDL-C)level were protective factors for the severity of acute pancreatitis with hyperlipidemia.The logis-tic regression model's receiver operating characteristic(ROC)curve area under the curve(AUC)value for the test set was 0.949,the Adaboost model AUC value was 0.969,the gradient boosting decision tree(GBDT)model AUC value was 0.938,and the random forest(RF)model AUC value was 0.951.Among the severity prediction models for hyperlipidemic acute pancre-atitis,the logistic regression test set calibration curve was closest to the ideal calibration curve in the 0-0.7 probability range.Conclusion The logistic multivariate regression model constructed with machine learning demonstrates superior overall predictive performance compared to other machine learning models in predicting the severity of patients with hyperlipidemic acute pancreatitis.