决策树模型与Logistic回归模型在产后尿潴留发生影响因素分析中的作用(1)
[摘要] 目的 探讨决策树模型与Logistic回归模型在产后尿潴留(PUR)发生影响因素分析中的作用。 方法 收集2014年1月1日～2017年12月31日在浙江省丽水市人民医院(以下简称“我院”)经阴道分娩后发生PUR的180例产妇作为病例组，随机选取同期于我院经阴道分娩且未发生PUR的200例产妇作为对照组。采用决策树模型与Logistic回归模型回顾性分析PUR发生的相关影响因素。 结果 决策树模型和多因素Logistic回归模型分析均显示分娩镇痛(P = 0.047)、产钳助产(P = 0.001)以及会阴侧切(P < 0.001)是发生PUR的独立危险因素。此外，Logistic回归模型分析结果还提示巨大儿(P = 0.023)是PUR的影响因素，而决策树模型中未提示其对PUR的发生有影响。 结论 PUR的影响因素众多，决策树模型和Logistic回归模型互为补充，可从不同方面描述PUR发生的影响因素，为进一步制订预防措施提供参考依据。
[关键词] 决策树模型;产后尿潴留;影响因素;Logistic回归模型
[中图分类号] R714.64 [文献标识码] A [文章编号] 1673-7210(2019)09(b)-0085-04
The role of decision tree model and Logistic regression model in postpartum urinary retention influencing factors analysis
GUO Xueqi YAN Guizhen ZHANG Caixia CHEN Jingjuan
Department of Obstetrics， Lishui People′s Hospital， Zhejiang Province， Lishui 323000， China
[Abstract] 0bjective To investigate the role of decision tree model and Logistic regression model in postpartum urinary retention (PUR) influencing factors analysis. Methods From January 1， 2014 to December 31， 2017， 180 puerpera with PUR after vaginal delivery in Lishui People′s Hospital of Zhejiang Province ("our hospital" for short) were selected as case group， and 200 puerpera who without PUR in our hospital at the same period were selected randomly as control group. Decision tree model and Logistic regression model were used to determine influential factors for PUR. Results Decision tree model and Logistic regression model indicated that delivery analgesia (P = 0.047)， forceps delivery (P = 0.001) and episiotomy (P < 0.001) were independent risk factors of PUR， and fetal macrosomia (P = 0.023) was also the influencing factor in the Logistic regression model， but the decision tree model did not indicate its influence. Conclusion There are many factors influencing PUR. Decision tree model and Logistic regression model are complementary to each other， which can describe the factors from different aspects， and provide basis and reference for the further formulation of preventive measures.
[Key words] Decision tree model; Postpartum urinary retention; Influencing factors; Logistic regression model
產后尿潴留(postpartum urinary retention，PUR)指产妇经阴道分娩后6 h无法自行排尿，或自行排尿后超声监测或导尿管导出膀胱内残余尿量(PVRV)>150 mL，或自行排尿后膀胱内残余尿量>150 mL[1]。PUR是产科常见的一种并发症，不仅使产妇膀胱长期处于过度充盈的状态，影响产后子宫恢复;还可能因子宫收缩欠佳而造成产后大出血[2-3]，危害不容忽视[4]。PUR可导致严重的后果，故发现与阐明PUR的高危因素，指导临床采取针对性措施，对避免PUR的发生至关重要。目前，国内外多采用Logistic回归模型分析发生PUR的影响因素[5]。该分析方法虽能有效地展现因变量和自变量之间的数量依存关系，但尚不足以处理某些非线性、高度交互作用及含大量缺失值等特征的资料，同时也无法直观地显示各因素对结果变量的重要程度[6-7]。决策树法(decision tree)作为数据挖掘领域的重要方法，已被广泛应用于医药生物领域，在很大程度上弥补了传统Logistic回归模型的缺陷和不足[8]。, 百拇医药(郭学齐 闫贵贞 张彩霞 陈景娟)
[关键词] 决策树模型;产后尿潴留;影响因素;Logistic回归模型
[中图分类号] R714.64 [文献标识码] A [文章编号] 1673-7210(2019)09(b)-0085-04
The role of decision tree model and Logistic regression model in postpartum urinary retention influencing factors analysis
GUO Xueqi YAN Guizhen ZHANG Caixia CHEN Jingjuan
Department of Obstetrics， Lishui People′s Hospital， Zhejiang Province， Lishui 323000， China
[Abstract] 0bjective To investigate the role of decision tree model and Logistic regression model in postpartum urinary retention (PUR) influencing factors analysis. Methods From January 1， 2014 to December 31， 2017， 180 puerpera with PUR after vaginal delivery in Lishui People′s Hospital of Zhejiang Province ("our hospital" for short) were selected as case group， and 200 puerpera who without PUR in our hospital at the same period were selected randomly as control group. Decision tree model and Logistic regression model were used to determine influential factors for PUR. Results Decision tree model and Logistic regression model indicated that delivery analgesia (P = 0.047)， forceps delivery (P = 0.001) and episiotomy (P < 0.001) were independent risk factors of PUR， and fetal macrosomia (P = 0.023) was also the influencing factor in the Logistic regression model， but the decision tree model did not indicate its influence. Conclusion There are many factors influencing PUR. Decision tree model and Logistic regression model are complementary to each other， which can describe the factors from different aspects， and provide basis and reference for the further formulation of preventive measures.
[Key words] Decision tree model; Postpartum urinary retention; Influencing factors; Logistic regression model
產后尿潴留(postpartum urinary retention，PUR)指产妇经阴道分娩后6 h无法自行排尿，或自行排尿后超声监测或导尿管导出膀胱内残余尿量(PVRV)>150 mL，或自行排尿后膀胱内残余尿量>150 mL[1]。PUR是产科常见的一种并发症，不仅使产妇膀胱长期处于过度充盈的状态，影响产后子宫恢复;还可能因子宫收缩欠佳而造成产后大出血[2-3]，危害不容忽视[4]。PUR可导致严重的后果，故发现与阐明PUR的高危因素，指导临床采取针对性措施，对避免PUR的发生至关重要。目前，国内外多采用Logistic回归模型分析发生PUR的影响因素[5]。该分析方法虽能有效地展现因变量和自变量之间的数量依存关系，但尚不足以处理某些非线性、高度交互作用及含大量缺失值等特征的资料，同时也无法直观地显示各因素对结果变量的重要程度[6-7]。决策树法(decision tree)作为数据挖掘领域的重要方法，已被广泛应用于医药生物领域，在很大程度上弥补了传统Logistic回归模型的缺陷和不足[8]。, 百拇医药(郭学齐 闫贵贞 张彩霞 陈景娟)