Association between occupational lead exposure and multiple health indicators: A machine learning-based study
- VernacularTitle:基于机器学习的职业性铅暴露与多健康指标的关联
- Author:
Jiali QIAN
1
;
Boshen WANG
2
;
Qinheng ZHU
3
;
Xiaoru DAI
4
;
Baoli ZHU
5
Author Information
- Publication Type:Investigation
- Keywords: lead-acid battery enterprise; occupational health; machine learning; CatBoost; Naive Bayes model; random forest
- From: Journal of Environmental and Occupational Medicine 2026;43(5):621-629
- CountryChina
- Language:Chinese
- Abstract: Background Lead (Pb) is a highly toxic heavy metal that accumulates in the body, potentially leading to multi-systemic impairment. Compared with traditional statistical methods, machine learning techniques offer unique advantages, opening new avenues for occupational health risk assessment and the exploratory analysis of complex associations. Objective To examine the association between occupational lead exposure and multiple health indicators and to identify key risk factors for lead toxicity. Methods A cross-sectional study was conducted, integrating occupational hygiene investigation results from 16 lead-acid battery enterprises in Jiangsu Province with occupational health examination data from 1914 lead-exposed workers. Inter-group differences were analyzed using the χ2 test or Fisher's exact test. Binary logistic regression and machine learning algorithms [CatBoost, Naive Bayes model (NBM), and random forest (RF)] were employed to evaluate the association between blood lead (PbB), urine lead (PbU), and health indicators including blood pressure (BP), red blood cell count (RBC), and alanine aminotransferase (ALT). Results The prevalence of abnormal PbB and PbU were 14.52% and 9.35%, respectively. The risks of abnormal BP, RBC, and ALT were significantly increased in the population with high lead levels (P<0.05). PbB abnormalities were closely associated with gender, environmental lead concentration, wearing masks, smoking, and alcohol consumption (P<0.05). Regarding occupational hazards, workers exposed to lead dust had a 1.98-fold risk of PbU abnormality compared to those exposed to lead fumes. The plate coating and acid leaching process posed the highest risk for both PbB (OR=8.81) and PbU (OR=5.46) abnormalities compared with assembly process. Furthermore, the risks of PbB and PbU abnormalities were significantly elevated among workers with abnormal BP, RBC or ALT (P<0.05). Among the models, CatBoost performed best in predicting RBC abnormality (accuracy: 95.8%; precision: 44.9%; F1 score: 0.952; AUC: 0.981). Feature importance analysis identified PbB and PbU as the core factors affecting abnormal RBC and ALT, while RBC and ALT abnormalities as key features for predicting the risk of PbB and PbU abnormalities. Conclusion By integrating traditional statistical methods with machine learning, this study reveals a complex bidirectional association between occupational lead exposure and multiple health indicators, and identifies gender, job category, and environmental Pb concentration as the key factors influencing PbB abnormalities. These findings provide a scientific foundation for the implementation of precision occupational health management models.
