Exposure to organophosphate esters (OPEs) of children (1-12y) in Guangxi, China: Predictors evaluation by machine learning

• Eleven OPEs were detected in a cohort of 84 children and their parents’ urine in Guangxi and TDCPP were dominated • Significant positive correlations were found among most target OPEs • The importance of 93 variables were analysed by machine learning method XGBoost coupled with SHAP • Daily use of fans affected the levels of five OPEs in children’ urine • Egg consumption and the number of family members were associated with the high levels of some OPEs Organophosphate esters (OPEs), widely employed as flame-retardants and plasticizers, have been linked to adverse health outcomes in children. This study evaluated the predictors of OPE exposure in children aged 1 to 12 years in Guangxi, China. Eleven OPEs were detected in a cohort of 84 children and their parents’ urine, with geometric mean (GM) levels ranging from 0.11 ng/g crea (tri-n-propyl phosphate, TPP) to 25.93 ng/g crea (tris(2,3-dichloropropyl) phosphate, TDCPP) in children, and from 0.04 ng/g crea (triphenyl phosphate, TPHP) to 18.71 ng/g crea (TDCPP) in parents. These findings are consistent with values reported in several large-scale national urinary biomonitoring studies on OPEs. Significant positive correlations among most target OPEs and no significant gender and age differences in children’s urine were found. The eXtreme Gradient Boosting (XGBoost) coupled with Shapley Additive explanations (SHAP), was employed to analyse the importance of 93 variables from the questionnaire and interpret the levels of OPEs in children’s urine. Models for six OPEs were constructed, revealing that daily use of fans affected the levels of five OPEs. Additionally, egg consumption and the number of family members were associated with higher levels of TPP/TCEP, and CDP/TEHP, respectively. There is a need to conduct more precise quantitative assessments of OPE exposure routes.