您好, 访客   登录/注册

基于Logistic回归与决策树模型的老年高血压患者血压控制影响因素分析

来源:用户上传      作者:林菲 朱晓云

  摘 要 目的:了解上海市金山^社区管理模式下的老年原发性高血压患者血压控制现状及其影响因素,为完善社区高血压患者管理提供参考依据。方法:采用随机整群抽样方法,抽取高血压患病超过2年,且被社区医生管理的患者1 053例进行面对面问卷调查,平均年龄为(70.65±6.93)岁;其中男性457例,女性596例。采用卡方检验、Logistic回归模型和决策树模型分析影响血压控制的因素。结果:1 053例老年高血压患者的血压控制理想率为72.93%,Logistic回归分析显示,文化程度、高血压分级、食盐摄入量和是否忘记服药是血压控制影响因素;决策树分析显示,文化程度、高血压分级、食盐摄入量是血压控制影响因素;2个模型的ROC曲线下面积差异无统计学意义(Z=-0.691,P=0.489)。结论:Logistic回归模型与决策树模型对血压控制影响因素的分析结果有较高的一致性,在今后社区管理高血压患者的工作中应重视对文化程度低、高血压分级较高、食盐摄入量多、不规律服药患者的综合干预。
  关键词 老年高血压;血压控制;Logistic回归;决策树
  中图分类号:R544.1 文献标志码:A 文章编号:1006-1533(2022)16-0051-05
  引用文本 林菲, 朱晓云. 基于Logistic回归与决策树模型的老年高血压患者血压控制影响因素分析[J]. 上海医药, 2022, 43(16): 51-55.
  Analysis of influencing factors of blood pressure control in the elderly patients with hypertension based on Logistic regression and decision tree model
  LIN Fei, ZHU Xiaoyun(Department of Chronic Disease Prevention and Control of Jinshan District Center for Disease Prevention and Control, Shanghai 201599, China)
  ABSTRACT Objective: To understand the present situation and influencing factors of blood pressure control in the elderly people with essential hypertension in the community management mode in Jinshan District, Shanghai in order to provide the reference for the improvement of the management of community hypertension patients. Methods: One thousand and fiftythree patients with hypertension for more than two years and managed by the community doctors were selected by random cluster sampling to conduct the face-to-face questionnaire survey, and the average age was (70.65±6.93) years; among of them, there were 457 males and 596 females. The influencing factors of controlling blood pressure were analyzed with Chi square test, Logistic regression model and decision tree model. Results: The ideal blood pressure control rate of hypertension in 1 053 elderly patients was 72.93%, Logistic regression analysis indicated that education, hypertension classification, salt intake and taking medicine regularly or irregularly were the factors of the blood pressure control; the analysis of decision tree model indicated that education, hypertension classification and salt intake were the factors of impacting on the blood pressure control. There was no significant difference in the area under the ROC curve between the two models (Z=-0.691, P=0.489). Conclusion: Logistic regression model and decision tree model have high consistency in the analysis results of influencing factors of blood pressure control, in the future community management of the hypertension patients, more attention should be paid to comprehensive intervention for hypertension patients with low level of education, higher hypertension classification, more salt intake and irregular medication.

nlc202209131032

4(高中及以上),血压未控制的患者占该节点构成的13.4%;(2)文化程度≤2(小学以下),食用盐摄入适中,占该节点构成的33.7%;(3)文化程度≤2(小学以下),食用盐摄入偏淡,占该节点构成的18.3%;(4)文化程度≤2(小学以下),食用盐摄入偏咸,占该节点构成的47.3%;(5)文化程度为2-4(小学到初中),高血压分级为Ⅰ级,占该节点构成的18.4%;(6)文化程度为2~4(小学到初中),高血压分级≥Ⅱ级,占该节点构成的32.9%。
  2.4 两种模型分析结果比较
  绘制Logistic回归模型与决策树模型的受试者工作特征曲线(receiver operating characteristic curve,ROC)图。见图2。Logistic回归模型的ROC曲线下面积为0.647,决策树模型的ROC曲线下面积为0.631,2个模型的ROC曲线下面积差异无统计学意义,Z=-0.691,P>0.05,预测效果相近,见表4。
  3 论
  本研究通过Logistic回归模型与决策树模型筛选出的老年高血压患者血压控制影响因素有文化程度、血压分级和食盐摄入量。在Logistic回归模型中将是否忘记服药也纳入了影响因素,这与其他研究结果基本一致[5-7]。本研究结果显示,Logistic回归模型与决策树模型的ROC曲线下面积差异无统计学意义,提示2个模型的预测效果相近,且对血压控制影响因素的分析结果有较高的一致性。其中Logistic回归模型计算出了各个影响因素的OR值,反映了血压控制与文化程度、血压分级、食盐摄入量、是否忘记服药等4个因素之间的数量依存关系,但是无法展示各影响因素之间的交互作用,预测结果不直观[8]。而决策树则能很好展示各变量间的交互作用,如研究中文化程度是第1层的因素,说明文化程度是影响血压控制的最重要因素,第2层则展示了影响因素间的交互关系,食盐摄入偏多是文化程度最低的患者血压控制影响因素;而高血压分级偏高是文化程度稍高的患者血压控制影响因素。以上2种预测模型各具优势,在选择模型时将两者结合可更大程度地发挥两者的优势[9]。

nlc202209131032



转载注明来源:https://www.xzbu.com/6/view-15439564.htm

相关文章