职业与健康 ›› 2026, Vol. 42 ›› Issue (18): 2460-2466.

• 论著——职业卫生与职业相关疾病 • 上一篇    下一篇

873名汽车制造业工人焦虑情绪及影响因素分析

杨洁, 张晋蔚, 张玉侠, 赵远, 王致, 周海林()   

  1. 广州市第十二人民医院广东 广州 510620
  • 收稿日期:2025-10-21 修回日期:2026-02-11 出版日期:2026-09-15 发布日期:2026-08-28
  • 通信作者: 周海林
  • 作者简介:周海林,E-mail:zhouhl1998@163.com
    杨洁,女,研究实习员,主要从事职业卫生工作。
  • 基金资助:
    广州市科学技术局重点研发计划项目(202206010061);广州市医学重点学科建设项目(2025-2027);广州市市校(院)企联合资助专题(2023A03J0502);广州市市校(院)企联合资助专题(2025A03J3540);广州市市校(院)企联合资助专题(2025A03J3591)

Analysis of anxiety status and influencing factors among 873 automobile manufacturing workers

YANG Jie, ZHANG Jinwei, ZHANG Yuxia, ZHAO Yuan, WANG Zhi, ZHOU Hailin()   

  1. Guangzhou Twelfth People's HospitalGuangzhouGuangdong 510620, China
  • Received:2025-10-21 Revised:2026-02-11 Online:2026-09-15 Published:2026-08-28
  • Contact: ZHOU Hailin

摘要:

目的 应用机器学习分析广州市汽车制造业工人焦虑情绪及其可能的影响因素,为改善汽车制造业工人身心健康状况,促进制造业可持续发展提供数据参考。方法 采用简单随机抽样方法,于2022年8月—2023年9月,在广州市汽车制造业领域内抽取8家单位的873名一线劳动者进行问卷调查。将受访者按7∶3的比例随机分为训练组(n=611)和验证组(n=262),应用单因素Logistic回归、Boruta回归、Lasso回归和随机森林递归特征消除(recursive feature elimination,RFE)进行特征筛选。建立6种机器学习模型,通过受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under curve,AUC)、校准曲线、决策曲线(decisioncurve analysis,DCA)对不同模型的性能进行比较,通过Shapley加性解释(shapley additive explanation,SHAP)算法对影响焦虑情绪的因素进行评估。结果 广州市汽车制造业873名一线劳动者中有284人(32.5%)存在焦虑情绪,进行特征指标筛选出性别、婚姻状况、入职岗位时间、单位性质、健康法律知识水平、健康工作方式和行为水平、职业紧张状况、睡眠状况8个特征变量纳入6种机器学习模型构建预测模型,其中LightGBM模型的 AUC 最高,为0.705,其他5个AUC由高至低为0.683、0.667、0.634、0.630和0.601,DCA 曲线中 LightGBM 模型的临床决策净获益也更大、预测效能最佳。SHAP 算法分析得出,存在睡眠障碍、存在职业紧张、单位性质为外资企业、入职岗位时间较长会增加产生焦虑的风险,高水平的健康法律知识及高水平健康工作方式和行为会降低焦虑情绪风险。结论 广州市汽车制造业工人焦虑情绪值得关注,企业可通过合理安排员工岗位和工间休息时间、改善工作环境、加强职业健康教育来降低其焦虑情绪风险,还可应用LightGBM模型协助识别其产生焦虑的危险因素,并开展早期干预。

关键词: 汽车制造业, 劳动者, 焦虑情绪, 影响因素, 机器学习模型, Shapley加性解释算法

Abstract:

Objective To apply machine learning to analyze the anxiety status and influencing factors of workers in Guangzhou's automobile manufacturing industry,and provide data reference for improving the physical and mental health status of automobile manufacturing workers and promoting sustainable development of the manufacturing industry. Methods From August 2022 to September 2023,a simple random sampling was used to select 8 sample units in the automobile manufacturing industry of Guangzhou,and a total of 873 front-line workers were surveyed. Respondents were randomly divided into a training group(n=611) and a validation group(n=262) at a ratio of 7∶3,and feature screening was performed using univariate Logistic regression,Boruta regression,Lasso regression and recursive feature elimination(RFE).Six kinds of machine learning models were established,and the performance of different models was compared by receiver operating characteristic(ROC) area under the curve(AUC),calibration curve and decision curve analysis(DCA).Shapley additive explanation(SHAP) algorithm was used to evaluate the factors affecting anxiety. Results Among the 873 respondents,284(32.5%) experienced anxiety.After screening for characteristic indicators,eight characteristic variables including gender,marital status,length of employment,nature of the company,level of health and legal knowledge,level of healthy work Methods and behaviors,occupational stress,and sleep status were selected and incorporated into six machine learning models to construct predictive models.Among them,the LightGBM model had the highest AUC of 0.705,while the AUCs of the other five models,from high to low,were 0.683,0.667,0.634,0.630,and 0.601.The LightGBM model also demonstrated greater clinical decision net benefit in the DCA curve,indicating the best overall predictive performance. The SHAP algorithm analysis revealed that having sleep disorders,experiencing occupational stress,working in a foreign-funded enterprise,and having a longer tenure in the current position would increase the risk of anxiety,while high levels of health and legal knowledge and healthy work Methods and behaviors would reduce the risk of anxiety. Conclusion The anxiety of workers in Guangzhou's automotive manufacturing industry warrants attention.Enterprises can mitigate anxiety risks by optimizing job assignments,rest periods,and working conditions,as well as enhancing occupational health education.Additionally,the LightGBM model can be utilized to identify risk factors for anxiety,enabling early intervention.

Key words: Automobile manufacturing industry, Worker, Anxiety, Influencing factors, Machine learning models, Shapley additive explanation algorithm

中图分类号: