Association between occupational heat exposure and early renal dysfunction among Chinese petrochemical workers: A combined machine learning and WQS modeling study

Qingyu Li^1,2Chuancheng Wu^1,2Minhua Li^1,2Zhang YiLin^1,2Chen Yifeng^1,2Shanshan Du³Rong Xu⁴Zhihu Lv⁴Weimin Ye^5*Wei Zheng^6*Jianjun Xiang^1,2*

• ¹Department of Preventive Medicine, School of Public Health, Fujian Medical University, Fuzhou, China
• ²Key Laboratory of Environment and Health of Fujian Higher Education Institutes, School of Public Health, Fujian Medical University, Fuzhou, China
• ³Institute of Population Medicine, Fujian Medical University, Fuzhou, China
• ⁴Quangang Hospital, Quanzhou, China
• ⁵Department of Epidemiology and Health Statistics, School of Public Health, Fujian Medical University, Fuzhou, China
• ⁶The First Affiliated Hospital of Fujian Medical University, Fuzhou, China

Objective: To investigate the association between occupational heat exposure and hyperuricemia among petrochemical workers. Methods: We retrospectively analyzed the association between workplace heat exposure and hyperuricemia by using 10 years of occupational health examination records from 2312 petrochemical workers in Fujian Province, China. Generalized linear models (GLMs) were employed to estimate the effects of individual exposures. Weighted quantile sum (WQS) regression model was used to evaluate the combined effects of multiple occupational exposures and to identify the relative contribution of each exposure factor. A hyperuricemia risk prediction model was developed using the LightGBM machine-learning algorithm, with feature importance assessed using SHAP (SHapley Additive exPlanations) values. Results: Occupational heat exposure was significantly associated with an increased risk of hyperuricemia (OR=1.68, 95% CI: 1.28–2.20). In the GLM analysis, co-exposure to heat with benzene (OR=1.93, 95% CI 1.05–3.55), H2S (OR=3.38, 95% CI 1.94–5.88), gasoline (OR=2.58, 95% CI 1.49– 4.48), acid anhydride (OR=2.21, 95% CI 1.09–4.48) and CO (OR=2.14, 95% CI 1.16–3.97) further increased the risk (all P<0.05), suggesting synergistic effects. The WQS analysis indicated that in the mixed occupational hazards exposure, heat exposure (49.2%) contributing nearly half the effect to the overall effect. The LightGBM machine learning model identified length of service, age, BMI, gender, and heat exposure as the main predictors of hyperuricemia. The SHAP analysis confirmed heat exposure as a key independent contributor alongside length of service. Conclusion: Occupational heat exposure in petrochemical settings is significantly associated with hyperuricemia, suggesting potential early renal dysfunction risk. Integrating machine learning–based predictive models into workplace health surveillance may facilitate the early identification and management of high-risk workers. However, causal inference remains limited by the retrospective design and potential residual confounding, underscoring the need for prospective studies to validate and extend these findings.