赵亮, 刘明升, 张丹丹, 张丽, 张磊, 王睿, 符刚, 冯宝佳, 曾强. 应用分类树模型分析环境因素暴露对儿童急性淋巴细胞白血病的影响[J]. 环境与职业医学, 2017, 34(12): 1054-1059. DOI: 10.13213/j.cnki.jeom.2017.17435
引用本文: 赵亮, 刘明升, 张丹丹, 张丽, 张磊, 王睿, 符刚, 冯宝佳, 曾强. 应用分类树模型分析环境因素暴露对儿童急性淋巴细胞白血病的影响[J]. 环境与职业医学, 2017, 34(12): 1054-1059. DOI: 10.13213/j.cnki.jeom.2017.17435
ZHAO Liang, LIU Ming-sheng, ZHANG Dan-dan, ZHANG Li, ZHANG Lei, WANG Rui, FU Gang, FENG Bao-jia, ZENG Qiang. Application of classification tree model in analyzing environmental risks of childhood acute lymphoblastic leukemia[J]. Journal of Environmental and Occupational Medicine, 2017, 34(12): 1054-1059. DOI: 10.13213/j.cnki.jeom.2017.17435
Citation: ZHAO Liang, LIU Ming-sheng, ZHANG Dan-dan, ZHANG Li, ZHANG Lei, WANG Rui, FU Gang, FENG Bao-jia, ZENG Qiang. Application of classification tree model in analyzing environmental risks of childhood acute lymphoblastic leukemia[J]. Journal of Environmental and Occupational Medicine, 2017, 34(12): 1054-1059. DOI: 10.13213/j.cnki.jeom.2017.17435

应用分类树模型分析环境因素暴露对儿童急性淋巴细胞白血病的影响

Application of classification tree model in analyzing environmental risks of childhood acute lymphoblastic leukemia

  • 摘要: 目的 应用分类树模型构建儿童急性淋巴细胞白血病(ALL)的影响因素模型,筛选环境危险因素,为预防儿童ALL提供科学依据。

    方法 采用病例-对照研究的流行病学方法,通过问卷回顾性调查315名研究对象(儿童ALL患者为病例组,共179名;社会健康儿童为对照组,共136名)的相关信息,包含可能的儿童ALL影响因素30个。利用分类树模型卡方自动交互检测(CHAID)法建立儿童ALL影响因素模型,通过收益图、索引图及受试者征(ROC)曲线评价模型的应用价值。

    结果 分类树模型包括4层,共11个节点,筛选出5个儿童ALL的解释变量,分别是:儿童不爱吃蔬菜水果(χ2=47.070,P<0.001)、母亲孕期服用药物(χ2=13.638,P<0.001)、母亲孕期感冒(χ2=8.650,P=0.003)、儿童接触油漆涂料(χ2=8.403,P=0.004)、母亲孕期二手烟暴露(χ2=8.803,P=0.003)。模型的ROC曲线下面积为0.781,与曲线下面积0.5相比,差异有统计学意义(P<0.001),模型的拟合效果较好。

    结论 部分环境因素暴露可能与儿童ALL存在关联。

     

    Abstract: Objective To introduce classification tree model to analyze the effects of environmental risk factors on childhood acute lymphoblastic leukemia (ALL), screen the environmental risk factors, and provide scientific references for ALL prevention.

    Methods A case-control survey was conducted using questionnaires to retrospectively collect information on 30 substantial environmental risk factors among 315 children (179 ALL children as case group, 136 healthy children as control group).Classification tree model was built using chi-square automatic interaction detection (CHAID) method.The application value of final model was evaluated by gain diagram, index chart, and receiver operating characteristic (ROC) curve.

    Results The classification tree model contained four stratum and eleven nodes.Five explanatory variables for childhood ALL were screened out from the model including children not eating vegetables and fruits (χ2=47.070, P<0.001), children being exposed to paint (χ2=8.403, P=0.004), and maternal medication use (χ2=13.638, P<0.001), secondhand smoke exposure (χ2=8.803, P=0.003), and catching a cold during pregnant (χ2=8.650, P=0.003).The area under the ROC curve was 0.781, which was significantly different from 0.5 (P<0.001), indicating that the final classification tree model fit well.

    Conclusion Several environmental risk factors could be related to childhood ALL.

     

/

返回文章
返回