Cardio-Obstetrics in a Changing Climate: Environmental Exposures, Artificial Intelligence and Maternal Cardiovascular Risk Prediction

Abstract

Climate change is altering pregnant women's exposure to heat, air pollution, wildfire smoke and extreme weather, while artificial intelligence (AI) and machine learning (ML) are increasingly applied to predict maternal cardiovascular complications such as preeclampsia; whether these evidence streams have actually been integrated is unclear. This review synthesizes evidence across three levels: environmental/climate-sensitive exposure and maternal cardiovascular outcomes; AI/ML approaches to maternal cardiovascular prediction; and the extent of environmental-variable incorporation into AI/ML models, then proposes a research agenda. A structured search across four concept groups (maternal population, cardiovascular outcomes, environmental exposures, AI/ML methods) was conducted in PubMed/MEDLINE and peer-reviewed sources following PRISMA 2020 principles; given substantial heterogeneity, findings are reported as narrative synthesis rather than pooled quantitatively. Ambient heat and fine particulate matter were associated with gestational hypertension, preeclampsia and eclampsia (odds ratios generally 1.1–1.6), with larger increases for wildfire-sourced than urban-source particulate matter. ML models for preeclampsia prediction reported a pooled AUC of 0.91 in a 2026 meta-analysis, reflecting development/internal-validation performance, with extreme heterogeneity and lower pooled sensitivity (0.68) among externally validated models. Two studies incorporated genuine environmental exposure into an AI/ML analysis of a maternal cardiovascular outcome, using chemical biomarkers with a discriminative model in one case and climate-sensitive exposure with causal machine learning in the other, but none combined climate-sensitive exposure with a discriminative model showing demonstrated incremental predictive value. This integration gap is the principal finding; a four-stage research agenda spanning data integration, model development, validation and clinical translation is proposed.