Application of a Machine Learning Predictive Model for Recurrent Acute Pancreatitis
Application of a Machine Learning Predictive Model for Recurrent Acute Pancreatitis
复制标题
DOI：
10.1097/mcg.0000000000001936
复制
发表时间：
2024-10-01
影响因子：
2.9
通讯作者：
Tang，Xiaowei
中科院分区：
文献类型：
--
作者：
Ren，Wensen;Zou，Kang;Tang，Xiaowei
方法：选择2018-01/2019年12在西南医科大学附属医院就诊的首发急性胰腺炎住院患者531例。我们通过电子病历系统和电话或微信随访，确认患者是否有第二次发作，直到2021年12月31日。收集患者的临床和随访数据，按7：3的比例随机分配到训练集和测试集，利用训练集选择最佳模型，并用测试集对所选模型进行检验。用受试者操作特征曲线下面积、灵敏度、特异度、阳性预测值、阴性预测值、准确性、决策曲线和校准图来评估模型的有效性。结果：综合考虑多个指标，XGBoost为最优模型。XG Boost模型的受试者工作特征曲线下面积、准确度、敏感度、特异度、阳性预测值和阴性预测值分别为0.779、0.763、0.883、0.647、0.341和0.922。根据Shapley加性解释值，饮酒、吸烟、高甘油三酯水平和ANC的发生与RAP相关。结论：XGBoost模型对RAP有较好的预测效果，有助于识别高危患者。
Methods:A total of 531 patients who were hospitalized for the first episode of acute pancreatitis at the Affiliated Hospital of Southwest Medical University from January 2018 to December 2019 were enrolled in the study. We confirmed whether the patients had a second episode until December 31, 2021, through an electronic medical record system and telephone or WeChat follow-up. Clinical and follow-up data of patients were collected and randomly allocated to the training and test sets at a ratio of 7: 3. The training set was used to select the best model, and the selected model was tested with the test set. The area under the receiver operating characteristic curves, sensitivity, specificity, positive predictive value, negative predictive value, accuracy, decision curve, and calibration plots were used to assess the efficacy of the models. Shapley additive explanation values were used to explain the model.Results:Considering multiple indices, XGBoost was the best model. The area under the receiver operating characteristic curves, accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the XGBoost model in the test set were 0.779, 0.763, 0.883, 0.647, 0.341, and 0.922, respectively. According to the Shapley additive explanation values, drinking, smoking, higher levels of triglyceride, and the occurrence of ANC are associated with RAP.Conclusion:The XGBoost model shows good performance in predicting RAP, which may help identify high-risk patients.