Green space exposure and preterm birth risk in Wuhan, China: A machine learning study

Green space exposure, as measured by the Enhanced Vegetation Index (EVI), is a crucial component of the urban built environment. It has the potential to reduce the risk of preterm birth (PTB) by lowering heat and improving air quality. This study explores the effect of EVI on PTB risk in Wuhan, China, within a framework that considers individual, environmental, and built environment factors. We used birth data from the Hubei Provincial Health Commission, covering all live births in Wuhan (2014-2016), and environmental data, including daily PM₂.₅, O₃, temperature, humidity, and EVI from national meteorological and remote sensing databases. To optimize model performance, we employed Bayesian optimization and five-fold cross-validation to identify the optimal hyperparameters for five machine learning models. The best-performing Extreme Gradient Boosting (XGBoost) model was selected to explore the nonlinear relationship between EVI and PTB risk at different pregnancy stages. This model was further used to identify exposure thresholds and assess spatial variations in PTB risk. Our findings revealed that moderate levels of EVI were associated with lower PTB risk, with optimal exposure ranges of 0.15 and 0.50 in early pregnancy, 0.30 to 0.50 in mid-pregnancy, and 0.10 to 0.35 in late pregnancy. Higher PTB risks were observed in central urban areas characterized by low green space and high urban development intensity. Additionally, we found a significant interaction between EVI and temperature, where both low EVI and extreme temperatures increased PTB risk. Specifically, when EVI is low and temperatures fall below 15 °C or rise above 21 °C during early or late pregnancy, the risk of PTB increases significantly. This study provides empirical evidence on how green space affects PTB at different pregnancy stages and identifies regions where high-risk groups live. These findings have important implications for urban green planning and the development of targeted maternal health strategies.