Predicting Gestational Hypertension Risk in High-Density Cities: A Time-Weighted Heat-Humidity Exposure and Built Environment Analysis Using XGBoost in Shanghai

The synergistic mechanisms of hot-humid climate and the built environment (BE) on Hypertensive Disorders of Pregnancy (HDP) in high-density cities remain unclear. This study integrated longitudinal data from 4,142 singleton pregnancies (including 227 HDP cases) monitored throughout pregnancy at a Shanghai hospital (2016–2024). We constructed a neighborhood-level, multi-source environmental exposure database (temperature/humidity/BE) and compared the performance of multiple models (Discrete-Time Logistic Regression (DTLR), Deep Neural Networks (DNN), eXtreme Gradient Boosting (XGBoost)) for HDP risk. We further investigated the association mechanisms between different exposure types and HDP. Key findings include: 1. The XGBoost model achieved the optimal predictive performance (Area Under the Curve, AUC=0.847), outperforming a baseline model that neglected heat-humidity synergy and time-weighting by ΔAUC=0.05 (p=0.013); 2. Each 0.1-unit increase in green view index (GVI) significantly reduced HDP risk by 26% (Odds Ratio, OR=0.74, 95% Confidence Interval, CI: 0.62–0.89), while the street canyon effect (Sky View Factor, SVF >0.6) amplified the risk from equivalent heat exposure by 37%; 3. Cumulative exposure to just 3 days of moderate heat (Moderate Heat Days Lag 4, MHDL4) during the first trimester (<12 weeks) doubled HDP risk (OR=1.79, 95% CI: 1.48–2.16), with extreme heat events (Extreme Heat Days Lag 4, EHDL4) exhibiting more pronounced lagged effects (4-day lag OR=2.03). This study elucidates the potential impacts of environmental factors on maternal health, providing a crucial reference for pregnancy health protection and policy formulation amid increasing extreme climate events. These findings provide support for heat-resilient urban planning targeting maternal health in humid-hot megacities.