基于白蛋白衍生指标建立急性高脂血症型胰腺炎重症化的机器学习预测模型

Establishment of a Machine Learning Prediction Model for the Severity of Acute Hyperlipidemic Pancreatitis Based on Albumin-Derived Indicators

摘要: 背景：建立基于白蛋白衍生指标的急性高脂血症型胰腺炎(HLAP)重症化的机器学习预测模型。方法：回顾性收集HLAP患者临床数据，按7:3分为训练集和验证集。训练集中通过Lasso回归筛选出与入住ICU显著相关的特征变量，多种机器学习算法构建预测模型，并在验证集中通过ROC曲线评估模型性能。结果：最终筛选的特征变量包括APACHE II评分、是否发热、直接胆红素(DB)、是否胰周液体积聚及C反应蛋白/白蛋白(CAR)。逻辑回归模型表现最佳，验证集中ROC曲线下面积(AUC)为0.757。结论：基于CAR的预测模型能较准确地评估HLAP患者入住ICU的风险，为早期识别HLAP患者重症化提供了一种简便实用的工具。

Abstract: Background: Establish a machine learning prediction model for the severity of acute hyperlipidemic pancreatitis (HLAP) based on albumin-derived indicators. Method: Retrospective collection of clinical data from HLAP patients, divided into a training set and a validation set in a 7:3 ratio. In the training set, Lasso regression was used to screen out feature variables significantly correlated with ICU admission. Multiple machine learning algorithms were used to construct predictive models, and the model performance was evaluated in the validation set using ROC curves. Result: The final selected feature variables include APACHE II score, presence of fever, direct bilirubin (DB), presence of peripancreatic fluid accumulation, and C-reactive protein/albumin (CAR). The logistic regression model performed the best, with an area under the ROC curve (AUC) of 0.757 in the validation set. Conclusion: The CAR-based prediction model can accurately assess the risk of HLAP patients being admitted to the ICU, providing a simple and practical tool for the early identification of severe HLAP patients.