1、中国病案2023 年第 24 卷第 2 期 67 基于分位数回归和决策树模型的AMI 住院费用影响因素分析 杨海琴 摘要 目的 回顾性分析重庆市某三甲医院 AMI 患者的病历资料，了解该类患者住院费用结构和影响因素，为科学控费提供依据。方法 采用横断面研究方法收集 2017 年 1 月 1 日-2020 年 12 月 31 日因急性心肌梗死住院的 1593 例患者的基本信息和费用信息，对患者基本信息采用描述性分析，对住院费用进行单因素分析，在此基础上对住院费用进行多因素线性回归和分位数回归，对比两者结果。根据对比结果选择更全面的影响因素纳入决策树模型构建住院费用的分组模型。结果 急性心肌梗死患
2、者住院费用构成中占比最大的是耗材费(51.58%)，最小的是治疗费(7.17%)。分位数回归分析中住院费用的影响因素包括了多因素线性回归分析结果中的影响因素，显示年龄、住院天数、年份、性别、住院次数、入院途径、转科、冠状动脉造影、PTCA、PCI、置入支架数量、治疗血管数量、冠状动脉溶栓、冠状动脉血栓抽吸、临时起搏器置入、IABP、CRRT、有创呼吸机、输血、抢救、心源性休克、肺炎、脓毒血症和治疗结局在住院费用的不同分位点产生影响(P0.05)。AMI 患者住院费用决策树分组模型显示住院花费最少的是没有实施PTCA、没有置入支架、没有合并肺炎且住院天数4 天的患者，花费最大的是实施 PTCA、
3、置入 2 及以上个支架且住院天数10 天的患者。结论 AMI 患者住院费用构成不合理，耗材费占比过大。优化诊疗流程，减少住院时间；规范诊疗行为，严格把控实施 PTCA 和置入支架的指征，减轻患者经济负担。关键词 分位数回归；决策树模型；急性心肌梗死；住院费用；影响因素 Analysis of Influencing Factors of AMI Hospitalization Expenses Based on Quantile Regression and Decision Tree Analysis of Influencing Factors of AMI Hospitalization
4、 Expenses Based on Quantile Regression and Decision Tree ModelModel Yang Haiqin Abstract Abstract ObjectivesObjectives The medical records of AMI patients in a Three A and Tertiary Hospital in Chongqing were retrospectively analyzed to understand the structure and influencing factors of hospitalizat
5、ion expenses for these patients,and provide a basis for scientific control of expenses.MethodsMethods A cross-sectional study method was used to collect the basic information and cost information of 1593 patients hospitalized for acute myocardial infarction from January 1st,2017 to December 31st,202
6、0,descriptive analysis was used for the basic information of the patients,and a univariate analysis was performed on the hospitalization costs of AMI.On this basis,multivariate linear regression and quantile regression were performed on hospitalization expenses,and the results were compared.Accordin
7、g to the comparison results,more comprehensive influencing factors were selected and incorporated into the decision tree model to construct a grouping model of hospitalization expenses.ResultsResults Consumables were the largest proportion of hospitalization expenses for AMI patients and the smalles
8、t was treatment expenses.The influencing factors of hospitalization expenses in the quantile regression analysis include those in the results of multivariate linear regression analysis,showing age,hospitalization days,year,gender,number of hospitalizations,admission route,transfer to department,coro
9、nary angiography,PTCA,PCI,number of stents placed,number of treated vessels,coronary thrombolysis,coronary thrombus aspiration,temporary pacemaker placement,IABP,CRRT,invasive ventilator,blood transfusion,rescue,cardiogenic shock,pneumonia,sepsis blood and treatment outcomes were affected in differe
10、nt quantiles of hospitalization costs(P0.05).The decision tree grouping model of hospitalization cost of AMI patients showed that the patients with the least hospitalization cost were those who did not perform PTCA,did not have stents,did not have pneumonia and were hospitalized for 4 days;showed th
11、at the patients with the largest hospitalization cost were patients who performed PTCA,placed 2 or more stents,and were hospitalized for more than 10 days.Conclusions Conclusions The inpatient cost of AMI patients was unreasonable,and the proportion of consumables was too large.Optimize the process
12、of diagnosis and treatment,reduce hospitalization time;standardize diagnosis and treatment behavior,strictly control the indications for PTCA and stent placement,and reduce the economic burden of patients.Key words Key words Quantile regression；Decision tree model；AMI；Hospitalization expenses；Influe
13、ncing factors FirstFirst-authorauthors addresss address Department of Medical Record Statistics,Chongqing Fourth Peoples Hospital,Chongqing 400014,China 1急性心肌梗死(acute myocardial infarction，AMI)是临床诊疗中一种常见的心血管疾患，也是中老年人群中的常见疾病之一。近年来，我国 AMI 的发病例数呈现上涨趋势，且患病人群越来越广泛，中国心血管病报告 2018 指出心血管病死亡率居于我国城乡居民死因的首位，AMI
14、死亡率呈现上升趋势。由于 AMI 疾病本身发病急、病情较重且迁延不愈，需要进行长期维持性的治疗，使患者承受巨大的疾病经济负担。AMI 住院费用自 2004 年以来每年以29.15%的速度增长，到 2016 年这一费用达到 重庆市第四人民医院病案统计科，重庆市，400014 190.85 亿元1，控制 AMI 医疗费用增长亟待解决。国内虽有与 AMI 患者住院费用的相关研究2-4，但大多数研究多采用传统的多因素线性回归模型，传统的多因素线性回归模型对偏态分布的资料具有局限性。基于此，本研究采用传统多因素线性回归模型和分位数回归模型分析 AMI 患者住院费用的影响因素，并对比两种模型的结果，选择更
15、可信更全面的回归模型结果中的影响因素构建 AMI 患者住院费用的决策树分组模型，探讨住院费用的关键影响因素，为科学高效控制 AMI 患者住院费用提供 参考。中国病案2023 年第 24 卷第 2 期 68 1 资料和方法 1.1 资料来源 该研究选取 2017 年 1 月 1 日-2020年 12 月 31 日重庆市某三甲医院出院所有主要诊断为AMI(依据国际疾病编码ICD-10临床版2.0 编码为 I21)并排除住院时间为 1 天的患者作为研究对象，共 1593 例。采用横断面研究从该院病案管理系统提取研究对象的基本信息、所有疾病诊断信息、所有手术操作信息以及费用信息。1.2 方法 将病案管
16、理系统提取的信息导入WPS13.0 软件建立数据库。为了费用数据对比具有可比性，依据中华人民共和国国家统计局公布的2017-2020 年重庆市居民消费价格指数修正住院费用。2017-2020 年重庆市居民消费价格指数分别为101%、102%、102.7%、102.3%10。1.3 统计学方法 运用SPSS26.0软件进行统计分析。住院费用呈偏态分布，用M(P25,P75)描述。以住院总费用为因变量，根据自变量类型采用Wilcoxon rank sum test、Kruskal-Wallis H、Spearman秩相关检验进行单因素分析。以单因素分析结果有影响的因素为自变量，以住院费用经 e 为底转换后的数据为因变量，进行多因素线性回归分析和分位数回归分析。对比两种回归分析的结果，纳入更全面的影响因素进入决策树模型建立住院费用分组模型。分位数回归模型：分位数回归是一种基于因变量的条件分布来拟合自变量的线性函数的回归方法。分位数回归可以任选任一分位数点进行参数估计，能提供更全面的条件分布信息，估计结果更加稳健5。该研究为了分析 AMI 患者住院费用低、中、高不同水平的影响因素，根据既往研究