不同粗分组方法建立DRGs模型的差异对比研究  

Research on the Differences of DRGs Models Established by Different Crude Grouping Methods

作　　者：程亮亮[1] 李亚茜[1] 段占祺[2] 张娟[1] 潘惊萍[2] 杨珉[1,3] 

机构地区：[1]四川大学华西公共卫生学院流行病与卫生统计系,610041 [2]四川省卫生和计划生育委员会卫生信息中心 [3]四川大学西部农村卫生发展研究中心

出　　处：《中国卫生统计》2017年第3期404-407,共4页Chinese Journal of Health Statistics

基　　金：四川省省级科研院所成果转化项目(15010112);美国中华医学基金会CMB项目(12-106)

摘　　要：目的利用2015年四川省病案首页280717例呼吸系统疾病病例,基于不同粗分组方法建立并评价DRGs模型。方法以临床诊断结合统计分析分组(粗分组一)、单纯临床诊断分组(粗分组二)两种方法对病例按临床特征进行粗分组,利用E-CHAID决策树方法进行细分组,并用变异系数(CV)对模型效度的组内同质性、方差减少量(RIV)对组间异质性进行评价;用回代检验评价模型的信度。结果粗分组一与粗分组二分别建立DRGs模型一(158个组合)和DRGs模型二(86个组合)。DRG模型一基于对数住院费用的CV平均值为32.77%,RIV为18.67%;基于原始住院费用CV平均值为69.21%,RIV为34.80%。DRGs模型二基于对数住院费用的CV平均值为41.09%,RIV为18.84%,基于原始住院费用CV平均值为77.87%,RIV为28.49%。总体上模型一在组内同质性及组间异质性优于模型二。回代检验显示DRGs模型一信度相对低下,而对数转换住院费用有提升模型效度的效应。结论不同粗分组下建立两组模型各有优缺点,需要多角度评价组内同质性及组间异质性,且有待建立综合评价指标体系。Objective Using 280717 cases of respiratory system diseases from the first-page medical records of Sichuan Province in 2015,we established and evaluated DRGs models based on different crude grouping methods. Methods Two crude grouping methods were used for comparison purpose： clinical diagnosis grouping combined with statistical analysis （crude groupl, CG1 ）and purely clinical diagnosis based grouping （crude group2, CG2）. The E-CHAID decision tree was used for the fine grouping. Validity in terms of homogeneity within groups and heterogeneity between groups was measured by coefficient of variance（CV） and reduction of variance （RIV）. Reliability by internal sub-sample test was performed. Results CG1 and CG2 established DRGs model 1 （156 fine groups ）and DRGs model 2 （86 fine groups）respectively. Base on the logarithmic hospitalization expense ,the DRGs model 1 demonstrated the average coefficient of variance（CV） at 32.77% and the average reduction in variance（RIV） 18. 67% ,while as the DRGs model 2 showed the two measures at 41.09% and 18.84% respectively. Base on the actual hospitalization expense,the averages of CV and RIV of the model 1 was 69, 21% and 34. 80% ,and 77.87% and 28.49% of the model 2 respectively. Although the DRGs model 1 had between validity than DRGs model 2 in general, the model 2 showed better reliability than the model 1. The logarithmic transformation had an effect of improving validity of models. Conclusion Both two models had their own advantages and disadvantages. We need to assess validity and reliability by multiple perspectives, and establish comprehensive evaluation index system.

关 键 词：诊断相关分组 DRGS 粗分组 呼吸系统疾病 效度评价 信度评价 

分 类 号：R195.4[医药卫生—卫生统计学]