基于LASSO回归模型的制造业工人非致命性职业伤害影响因素分析  

Impact factor selection for non-fatal occupational injuries among manufacturing workers by LASSO regression

作　　者：肖颖衡 陆春花[3] 钱娟[4] 陈颖 谷一硕 杨泽云[3] 丁道正[4] 李丽萍[2] 朱晓俊 XIAO Yingheng;LU Chunhua;QIAN Juan;CHEN Ying;GU Yishuo;YANG Zeyun;DING Daozheng;LI Liping;ZHU Xiaojun(National Center for Occupational Safety and Health,National Health Commission of the People's Republic of China,NHC Key Laboratory for Engineering Control of Dust Hazard,Beijing 102308,China;School of Public Health,Shantou University,Shantou,Guangdong 515041,China;Nantong Center for Disease Control and Prevention,Nantong,Jiangsu 226007,China;Yixing Center for Disease Control and Prevention,Yixing,Jiangsu 214206,China)

机构地区：[1]国家卫生健康委职业安全卫生研究中心,国家卫生健康委粉尘危害工程防护重点实验室,北京102308 [2]汕头大学公共卫生学院,广东汕头515041 [3]南通市疾病预防控制中心,江苏南通226007 [4]宜兴市疾病预防控制中心,江苏宜兴214206

出　　处：《环境与职业医学》2025年第2期133-139,共7页Journal of Environmental and Occupational Medicine

摘　　要：[背景]制造业作为我国支柱产业,其非致命性职业伤害发生率较高。该行业中个体、设备、环境及管理等非致命性职业伤害各层面因素众多且关联紧密,使其影响因素分析存在复杂性。[目的]探讨制造业工人非致命性职业伤害的影响因素,为后续开展针对性干预及监测提供依据。[方法]选择电缆及船舶制造企业内2243名一线作业工人作为研究对象,调查过去1年内非致命性职业伤害发生率及个体、设备、管理及环境等4个层面的因素情况。利用重抽样进行数据平衡,使用LASSO回归模型分析非致命性职业伤害影响因素,参考各变量系数估计值大小判断变量的影响程度及类型,其中系数估计值>0的变量为危险因素,反之则为保护因素,利用受试者工作特征(ROC)曲线下面积(AUC)检验模型性能,当AUC值>0.7时,说明模型性能良好。[结果]被调查的2243名制造业一线工人中男性占77.7%(1742/2243),主要年龄范围为40~49岁,占29.5%(661/2243),82.7%的工人(1854/2243)已婚,文化程度为初中学历占55.6%(1248/2243),51.0%(1144/2243)的工人平均月收入情况为5000~6999元。该人群非致命性职业伤害发生率为8.4%(189/2243),共计发现22个因素与非致命性职业伤害的发生有关联性(P<0.05)。分别是个体层面的性别、同事关系、吸烟、饮酒、平均运动时间、职业倦怠情况、工作疲劳感、肌肉骨骼疾患、心血管疾病及神经与感觉器官疾病等10个因素,设备层面的设备操作性、存在危险工件及安全隐患情况等3个因素,环境层面的从事低温作业、从事特种作业、从事噪声作业、作业空间大小、环境脏乱等5个因素,管理层面的每天工作时长、每周工作天数、加班情况及岗前技术培训等4个因素。LASSO回归模型AUC值=0.704,模型共计保留10个变量,其中非致命性职业伤害的危险因素共7个(系数估计值>0),包括存在安全隐患情况、存在肌肉骨骼�[Background]As a pillar industry in China,the manufacturing sector has a high incidence of nonfatal occupational injuries.The factors influencing non-fatal occupational injuries in this industry are closely related at various levels,including individual,equipment,environment,and management,making the analysis of these influencing factors complex.[Objective]To identify influencing factors of non-fatal occupational injuries among manufacturing workers,providing a basis for targeted interventions and surveillance.[Methods]A total of 2243 frontline workers from cable and shipbuilding enterprises were selected as study subjects to investigate the incidence of non-fatal occupational injuries and collect information at four levels:individual,equipment,management,and environment in past 12 months.Data balancing was performed using resampling,and LASSO regression was used to select factors of non-fatal occupational injuries.The influence degree and type of variables were judged based on the magnitude of the estimated coefficients of each variable,where variables with estimated coefficients>0 are risk factors,and those<0 are protective factors.The area under the receiver operating characteristic(ROC)curve(AUC)was used to test the performance of the model,with an AUC value>0.7 indicating good model performance.[Results]Among the 2243 frontline workers,males accounted for 77.7%(1742 out of 2243),with the main age range being 40-49 years old,representing 29.5%(661 out of 2243),82.7%of the workers(1854 out of 2243)were married,and 55.6%(1248 out of 2243)had a junior middle school education level.The average monthly income for 51.0%(1144 out of 2243)of the workers was between 5000 and 6999 Chinese Yuan.The incidence of non-fatal occupational injuries among the manufacturing workers was 8.4%(189/2243)in the past 12 months.Among the 22 factors associated with the occurrence of non-fatal occupational injuries(P<0.05),10 were individual-level factors,including gender,smoking,alcohol consumption,colleague relationships,average exerc

关 键 词：非致命性职业伤害 影响因素 LASSO回归 机器学习 制造业 

分 类 号：R13[医药卫生—劳动卫生]