首页|基于机器学习的医院CHS-DRG分组费用分析——以"胆囊切除手术不伴并发症和合并症"组为例

基于机器学习的医院CHS-DRG分组费用分析——以"胆囊切除手术不伴并发症和合并症"组为例

扫码查看
目的:采用机器学习技术建立院内"胆囊切除手术不伴并发症和合并症"分组的费用权重模型,进而分析该分组中的合理费用组成方式,找到实际管控和降低住院费用的方法.方法:以重庆市2021年发布的DRG管理方案为基础,收集2021年1月至2024年4月HC35分组的病案数据,按照住院总费用与分组补偿标准比值将其分为全部数据组、30%~100%正常补偿组和90%~100%标准补偿组,然后以物价模块分类为基础建立各类的费用集合,再采用梯度下降技术分析各物价模块在费用中的权重并以此建立分析模型,接着使用统计假设检验分别验证运算数据,经过分析后得到最终采用的权重模型.确认模型后,以重庆市2024年发布的DRG管理方案对前期数据进行重新分组计算,得到新的分组结果并追加收集2024年4月以后HC25分组的病案数据,经过模型拟合得到最新的费用权重及占比.结果:采用的数据模型为90%~100%标准补偿组模型,排列靠前的模块权重分别为(耗材费1.343[28.09%]、治疗费1.170[24.47%]、西药费0.846[17.70%]、检验检查费0.765[16.00%]、医疗服务费0.459[9.69%]).之后对各项内涵展开针对性讨论,探讨合理的费用控制方法.结论:将信息化技术结合临床路径等控制方式建立DRG分组的费用管理模型,是践行理论结合实际的良性操作行为,在未来的医疗机构费用管理中可以得到广泛的推广.
Analysis of CHS-DRG Grouping Cost in Hospitals Based on Machine Learning——Taking the Group of "Cholecystectomy Without Complications and Comorbidities" as an Example
Objective:The paper established a cost weight model for the group of "cholecystectomy without complications and comorbidities" within a hospital based on machine learning technology,and analyzed the reasonable cost composition in this group to find methods for controlling and reducing hospitalization costs in practice. Methods:Based on the DRG management plan released by Chongqing City in 2021,medical record data of HC35 groups from January 2021 to April 2024 were collected. According to the ratio of total hospitalization expenses to group compensation standards,they were divided into all data group,30%~100% normal compensation group,and 90%~100% standard compensation group. Then,based on the classification of price modules,various cost sets were established. Gradient descent technique was used to analyze the weight of each price module in the cost and establish an analysis model. Statistical hypothesis testing was used to verify the operation data,and the final weight model adopted was obtained after analysis. After confirming the model,the previous data was regrouped and calculated using the DRG management plan released by Chongqing City in 2024 to obtain new grouping results. Additionally,medical record data for HC25 grouping after April 2024 was collected,and the latest cost weight and proportion were obtained through model fitting. Results:The data model used is a 90%~100% standard compensation group model,with the top module weight being (consumables cost 1.343[28.09%],treatment cost 1.170[24.47%],western medicine cost 0.846[17.70%],testing and examination cost 0.765[16.00%],medical service cost 0.459[9.69%]). Afterwards,discussions will be conducted accordingly on various connotations to explore reasonable cost control methods. Conclusions:Combining information technology with clinical pathways and other control methods to establish DRG grouping cost management model is a beneficial operational behavior that combines theory with practice. It can be widely promoted in medical institution cost management in the future.

machine learningCHS-DRGcost management

郭亮、杜国平

展开 >

江苏省人民医院重庆医院 重庆 401420

机器学习 CHS-DRG 费用管理

2024

中国医疗保险
中国医疗保险研究会

中国医疗保险

影响因子:0.492
ISSN:1674-3830
年,卷(期):2024.(12)