E-CHAID决策模型下骨髓增生异常综合征DRGs分组实际运行效果分析
Practical impact of DRGs categorization under E-CHAID decision model on the hospitalization costs of myelodysplastic syndromes
华恃彬 1任晋文 1朱佳英1
作者信息
- 1. 浙江省人民医院医保物价办公室,310000浙江杭州
- 折叠
摘要
目的 分析E-CHAID决策模型下骨髓增生异常综合征DRGs分组对住院费用的实际影响.方法 回顾性搜集2021年2月至2023年2月浙江省某三甲医院HIS系统中病案首页出院诊断首项为骨髓增生异常综合征的患者资料.对病案首页全部指标进行组间多元线性回归检验、Mann-Whitney U检验及Kruskal-Wallis H检验,以VIF(方差膨胀因子)＜10、单因素筛选中P＜0.05的指标为自变量,以住院费用为因变量,采用E-CHAID穷举算法建立DRGs分组决策树模型.以各DRGs病例组合"住院费用中位数+1.5倍标准差"为费用上限,计算各DRGs分组上限费用与所有病例上限费用之比,分析住院费用分布特征.结果 共纳入2 223例骨髓异常综合征患者.是否使用单克隆抗体、是否ICU 入住、输血次数3个变量为E-CHAID决策树模型的分层分类变量节点,建立了 12个节点,共生成7个终端节点,变异系数 CV分别为 0.40、0.15、0.23、0.21、0.25、0.51、0.46,各 DRGs 分组间 Kruskal-Wallis H 检验具备统计学意义(H=2 816.568,P＜0.001).2 223例患者中超出费用上限76例(3.42％),超上限费用患者的住院总花费为662 241.71万元,占 DRGs 总住院费用的 1.40％;DRGs 1～7 组相对权重分别为 1.49、2.42、2.14、1.33、1.10、0.47、0.24.结论 E-CHAID决策模型下骨髓增生异常综合征DRGs分组异质性较强,组间分布合理,可为浙江省后续开展骨髓增生异常综合征等恶性血液疾病的地方性DRGs医疗付费改革提供一定依据.
Abstract
Objective To analyze the practical effect of the diagnosis-related groups(DRGs)of myelodysplastic syn-dromes(MDS)on hospitalization costs under the E-CHAID decision model.Methods Data of patients with myelodysplastic syn-drome as the first discharge diagnosis were retrospectively collected in the first page of medical records in the Hospital Information System(HIS)records from a tertiary hospital in Zhejiang Province from February 2021 to February 2023.All indicators of the medical record home page were tested via multiple linear regression tests between groups,Mann-Whitney U tests,and Kruskal-Wallis H-tests.With indicators of variance inflation factor(VIF)＜10 and P＜0.05 in the univariate screen as the independent variable,and hospitalization costs as the dependent variable,the DRGs grouping decision tree model was established by using the E-CHAID exhaustive algorithm.The maximum cost was determined as the"median hospitalization cost+1.5 times the standard deviation"for each DRGs case combination,and the ratio of maximum cost for each DRGs grouping to the maximum cost for all cases was calculated to analyze the distribution characteristics of hospitalization costs.Results A total of 2 223 MDS patients were included.The three variables of whether monoclonal antibodies were used,whether the patients were admitted to the ICU,and the number of transfusions were the hierarchical classification variable nodes of the E-CHAID decision tree model.Twelve nodes were established,generating 7 terminal nodes.Variance coefficient CVs were 0.40,0.15,0.23,0.21,0.25,0.51,and 0.46 respectively,and the Kruskal-Wallis H test between the DRGs groupings held statistical significance(H=2816.568,P＜0.001).Among the 2 223 patients,76(3.42％)exceeded the maximum cost.The total hospitalization cost of patients excee-ding the cost ceiling was 662 241.71 yuan,accounting for 1.40％of the total DRGs hospitalization cost.The relative weights of DRGs groups 1 to 7 were 1.49,2.42,2.14,1.33,1.10,0.47 and 0.24.Conclusion DRGs grouping of myelodysplastic syn-dromes under the E-CHAID decision model is more heterogeneous and the distribution among groups is reasonable,which can be used for the future development of DRGs in Zhejiang Province.It can provide some basis for the subsequent development of local DRGs medical payment reform for malignant hematological diseases such as myelodysplastic syndromes.
关键词
疾病诊断相关分组/骨髓增生异常综合征/穷举卡方自动交互检测/医疗保险/住院费用Key words
diagnosis related groups/myelodysplastic syndrome/exhaustive chi-squared automatic interaction detector/medical insurance/hospitalization cost引用本文复制引用
基金项目
浙江省科学技术厅项目基金(2022C35105)
出版年
2024
NETL
NSTL
维普
万方数据