基于决策树模型的急性胰腺炎患者住院费用及病例组合分析

Analysis of Hospitalization Costs and Case Mix in Patients with Acute Pancreatitis based on Decision Tree Model

目的探索急性胰腺患者住院费用影响因素和分组模型,为医院制定住院费用分组和精细化管理提供理论依据。方法采用整体抽样方法,以宜昌市某三甲医院2015年1月1日-2019年12月31日之间2099例急性胰腺炎病例为研究对象,采用多元线性回归分析确定住院费用的主要影响因素并作为分类指标,应用决策树模型E-CHAID法建立急性胰腺炎DRGs病例组合分组方案及费用标准。结果住院天数、是否手术、年龄、有无合并感染作为影响急性胰腺炎病例住院费用的分类节点,建立了6个DRGs组合和费用标准,其中手术治疗、年龄大于60岁且住院天数超过11天的DRGs组权重最高,为2.41。没有手术、无合并感染且住院天数小于11天的DRGs权重最低,为0.54。总体样本组内变异系数CV为0.36,组间方差减少量RIV=0.61,组内同质性和组间异质性较好。结论采用决策树模型E-CHAID法对急性胰腺炎住院费用进行分组科学合理,可计算出DRGs组的预测费用和权重,可为本地区住院费用标准的制定和医院精细化管理提供参考依据。

Objectives This study aims to explore the influencing factors and grouping model of hospitalization expenses in patients with acute pancreas in order to provide a theoretical basis for the hospital to formulate inpatient expense grouping and refined management.Methods This study used the overall sampling method and took 2099 cases of acute pancreatitis from January 1,2015 to December 31,2019 in a tertiary hospital in Yichang as the research objects.Multiple linear regression analysis was used to determine the main influencing factors of hospitalization expenses and serve as classification indicators and the decision tree model E-CHAID method was applied to establish a combination grouping scheme and cost standard of acute pancreatitis DRGs.Results Days of hospitalization,surgery,age,and co-infection were classification nodes that affect hospitalization costs for acute pancreatitis cases.6 DRGs and hospitalization expenses standards were established.Among them,the DRGs group with surgical treatment,older than 60 years and hospitalized for more than 11 days had the highest weight,which was 2.41.DRGs with no surgery,no co-infection and less than 11 days of hospitalization have the lowest weights,which was 0.54.The overall sample within-group coefficient of variation CV is 0.36.Between-group variance reduction RIV=0.61.Conclusions Using the decision tree model E-CHAID method to group hospitalization expenses for acute pancreatitis was scientific and reasonable.Predictive costs and weights of DRGs groups can be calculated.It provides a reference basis for the formulation of the standard of hospitalization expenses in the region and the refined management of hospitals.