基于决策曲线分析的临床预后重要影响因素研究  

Study on Important Influential Factors in Clinical Prognosis Based on Decision Curve Analysis

作　　者：孙磊[1] 李禄全[2] 杨晓慧[1] Sun Lei;Li Luquan;Yang Xiaohui(Data Analysis Technology Lab,Institute of Applied Mathematics,Henan University(475004), Kaifeng)

机构地区：[1]河南大学数据分析技术实验室,应用数学研究所,475004 [2]重庆医科大学附属儿童医院新生儿诊治中心

出　　处：《中国卫生统计》2018年第6期846-849,共4页Chinese Journal of Health Statistics

基　　金：河南省科技厅基础与前沿项目(162300410061);河南省教育厅科技攻关重点项目(16A120002);河南大学新兴交叉及特色学科建设项目(xxjc20170003).

摘　　要：目的针对临床预后重要影响因素自动筛选和重要性分析问题,探讨一种基于决策曲线分析(DCA)的研究方法。方法回顾性分析1998-2013年收治的食管闭锁患儿的临床资料,包括一般情况、实验室检查等临床资料,根据统计决策理论建立简明易用的决策曲线分析模型,对临床数据进行分析。结果结合临床实际意义,自动筛选出13个重要影响因素,并对这些影响因素做独立分析,并给出重要性排序。结论本文方法明显优于分布检验分析结果和ROC分析结果,主要体现在对样本数据要求不高,既能避免庞杂的哑变量处理,简易直观地筛选出重要影响因素,又能避免多重共线性问题导致筛选结果的不确定性。Objective In view of the problems of automatic screening and importance analysis of important influential factors,a research method based on decision curve analysis(DCA)is discussed.Methods Clinical data of children with esophageal atresia from 1998 to 2013,which includes those in general conditions and laboratory tests,was retrospectively analyzed.Based on the statistical decision theory,a simple and easy-to-use decision curve analysis model was established to analyze the clinical data.Results Thirteen important influential factors were automatically screened according to clinical practical significance,and an independent analysis of these influential factors was given to rank the importance.Conclusion This method is obviously better than the distribution test results and ROC analysis,mainly because it is less demanding on sample data,avoids complicated dummy variable processing,and can screen important influential factors easily and intuitively,avoids the uncertainty of screening results caused by multiple collinearity problems.

关 键 词：重要影响因素 DCA方法 食管闭锁 

分 类 号：R4[医药卫生—临床医学] O212.1[理学—概率论与数理统计]