PDF下载

[1]温美林颜涛徐飞.基于决策树算法的医疗费用超支风险预测与亚专业风险分级研究[J].中国卫生质量管理,2026,33(8):80-85.[doi:10.13912/j.cnki.chqm.2026.33.8.16]
　WEN Meilin,YAN Tao,XU Fei.Research on Medical Over-Expense Risk Prediction and Sub-Specialty Risk Stratification Based on Decision Tree Algorithm[J].Chinese Health Quality Management,2026,33(8):80-85.[doi:10.13912/j.cnki.chqm.2026.33.8.16]

点击复制

基于决策树算法的医疗费用超支风险预测与亚专业风险分级研究

《中国卫生质量管理》[ISSN:1006-7515/CN:CN 61-1283/R] 卷: 33 期数: 2026年8期页码: 80-85 栏目: 绩效管理与医院高质量发展出版日期: 2026-08-15

Title:: Research on Medical Over-Expense Risk Prediction and Sub-Specialty Risk Stratification Based on Decision Tree Algorithm

作者:: 温美林颜涛徐飞; 上海市第六人民医院

Author(s):: WEN Meilin; YAN Tao; XU Fei; Shanghai Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University

关键词:: 医疗费用超支; 风险预测; 决策树; 风险分级; 机器学习

Keywords:: Medical Over-Expense; Risk Prediction; Decision Tree; Risk Stratification; Machine Learning

分类号:: R197.323

DOI:: 10.13912/j.cnki.chqm.2026.33.8.16

文献标志码:: A

摘要:: 目的构建基于决策树算法的医疗费用超支风险预测模型，实现患者个体化风险识别与亚专业风险分层，为精细化管理提供数据依据。方法选取某三甲医院2024年骨科14个亚专业共69 423例住院患者数据，以“是否超支”作为因变量，采用决策树（CART）算法建模，以逻辑回归模型作为基准进行比较。根据预测结果计算各亚专业的预算偏离情况，利用卡方检验及事后多重比较确定不同风险等级。结果决策树模型在测试集上的曲线下面积（AUC）值为0.905，准确率为0.822，性能显著优于逻辑回归模型（AUC=0.883）。特征重要性显示“药耗占比”（63.1%）与“病种值”（19.6%）为核心影响因素。研究模型能精准识别出高危路径特征，根据风险程度将14个亚专业分为高、中高、中等、中低和低风险组，各风险等级之间有显著性差异（P<0.001）。结论所构建决策树模型兼具高预测性能与强可解释性，可应用于从个体预警到科室效率评价的多层次管理，为医院精准费用管理提供有效工具。

Abstract:: ObjectiveTo construct a medical over-expense risk prediction model based on the decision tree algorithm, realize individualized patient risk identification and sub-specialty risk stratification, and provide data support for refined management. MethodsData of 69 423 inpatients from 14 orthopedic sub-specialties in a tertiary hospital in 2024 were selected. Taking "whether there is over-expense" as the dependent variable, the classification and regression tree (CART) algorithm was adopted for modeling, with the logistic regression model as the benchmark for comparison. The budget deviation of each sub-specialty was calculated based on prediction results, and chi-square test with post-hoc multiple comparisons was used to determine different risk levels.ResultsThe decision tree model achieved an area under curve(AUC) of 0.905 and an accuracy of 0.822 on the test set, with significantly better performance than the logistic regression model (AUC=0.883). Feature importance analysis showed that "proportion of drug and consumable cost" (63.1%) and "disease case value" (19.6%) were the core influencing factors. The model could accurately identify high-risk path features, and the 14 sub-specialties were divided into high, moderately high, medium, moderately low and low risk groups according to risk degree, with significant differences among all risk levels (P < 0.001).Conclusion The constructed decision tree model has both high prediction performance and strong interpretability. It can be applied to multi-level management ranging from individual early warning to department efficiency evaluation, serving as an effective tool for precise hospital expense management.

参考文献/References:

［1］郑秉文.我国医保支付方式改革：现状、挑战与趋势［J］.人民论坛,2025(5):44-50.［2］王浩宇,周典,田帝,等.DRG改革背景下我国公立医院控费提质研究［J］.南京医科大学学报(社会科学版),2023(3):212-216,252.［3］裴晶,邰侠,汪茂艳.基于DRGs智能管理系统的医保精细化管理应用实践［J］.中国数字医学,2024，19(2):13-18.［4］王盟,许岩,刘珏,等.智慧医院一体化运营管理平台的构建与应用［J］.中国卫生质量管理,2025,32(2): 11-14.［5］SIDEY-GIBBONS C, PFOB A, ASAAD M, et al. Development of machine learning algorithms for the prediction of financial toxicity in localized breast cancer following surgical treatment ［J］. JCO Clin Cancer Info, 2021, 5: 338-347.［6］LAUFFENBURGER JC, MAHESRI M, CHOUDHRY NK. Not there yet: using data-driven methods to predict who becomes costly among low-cost patients with type 2 diabetes［J］. BMC Endocr Disord,2020,20(1):125.［7］朱萍，唐慧，张静，等. DRG支付方式改革下急性胰腺炎医疗费用结构分析：基于灰色—马尔科夫链预测模型［J］.中国卫生质量管理，2025,32(9):66-72.［8］刘萍,施辉,吴梦平,等.Logistic回归和决策树模型在手术患者住院超30天影响因素分析中的应用［J］.中国病案,2024，25(3):61-65.［9］张世冉. 医保支付方式改革背景下医疗住院费用影响因素研究--以两种病种为例［D］. 南昌：江西中医药大学,2023. ［10］张琪，魏丹，高玥珊，等. 940例老年慢病终末期患者住院费用影响因素分析［J］.中国卫生质量管理，2024,31(4):37-42.［11］任冠瑞.机器学习模型预测经椎间孔腰椎椎体间融合术后不良事件——基于控制住院费用角度［D］.南京:东南大学,2022.［12］郭禄义,张璇,张博雅,等.《中国医院质量安全管理》团体标准应用：医疗质量安全核心指标运行监测［J］. 中国卫生质量管理,2024,31(1): 6-9，18.［13］戴石元,聂钊,沈权平,等. DIP付费改革实施前后住院费用分析及医疗服务绩效综合评价［J］. 中国医院统计,2024,31(4):278-280.

相似文献/References:

[1]顾珍珍景欣范庆梅卫珂杭卓邱昌翠.人工智能在无痛胃肠镜麻醉风险预测中的应用进展[J].中国卫生质量管理,2026,33(8):64.[doi:10.13912/j.cnki.chqm.2026.33.8.13]
　GU Zhenzhen,JING Xin,FAN Qingmei.Research Progress on the Application of Artificial Intelligence in Anesthesia Risk Prediction for Painless Gastrointestinal Endoscopy[J].Chinese Health Quality Management,2026,33(8):64.[doi:10.13912/j.cnki.chqm.2026.33.8.13]

更新日期/Last Update: 2026-08-15