Association between maternal exposure to ambient air pollutants and preterm birth in a typical valley city of Northwest China: a retrospective cohort study

Abstract To investigate the association between maternal exposure to ambient air pollutants during pregnancy and preterm birth in Yan’an, a typical valley city in Northwest China, and to identify hierarchical exposure patterns among maternal characteristics and air pollutants using an exhaustive Chi-squared automatic interaction detection (CHAID) classification tree. A retrospective cohort study was conducted using data from 8298 singleton pregnancies delivered at two tertiary hospitals in Yan’an between January 2018 and December 2019, including 316 preterm births and 7982 term births. Maternal exposure to ambient air pollutants was estimated using the nearest air monitoring station. Binary logistic regression was performed after adjustment for epidemiologically relevant maternal covariates. An Exhaustive CHAID classification tree was established using a balanced training dataset generated by random under-sampling. Model performance was evaluated by ten-fold cross-validation and evaluated using the original cohort. Logistic regression analysis showed that pregnancy time ≥ 3, number of previous cesarean deliveries = 1, pregnancy complications, comorbid diseases, hypertension syndrome during pregnancy, SO 2 in the whole pregnancy, NO 2 in the first trimester and NO 2 in the second trimester were risk factors for preterm birth. Exhaustive CHAID identified whole pregnancy SO₂ as the primary splitting variable, followed by whole pregnancy O₃, whole pregnancy NO₂, and third trimester NO₂, with the proportion of preterm births ranging from 9.0 to 96.2% across terminal nodes. The model showed the area under the receiver operating characteristic curve (AUC) of 0.796 (95% CI 0.762–0.830) in the balanced training dataset and 0.793 (95% CI 0.769–0.818) when validated in the original cohort. Binary logistic regression identified associations of maternal characteristics and pregnancy period SO₂ and NO₂ exposure with preterm birth. Exhaustive CHAID revealed heterogeneous hierarchical exposure patterns with moderate discrimination and may serve as an exploratory risk stratification approach. Independent external validation is warranted.