AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Development of a Clinical Risk Model for Pregnancy-Related Cardiovascular Complications
By Jowell A, Dsouza V, Ribeiro P et al.
Background: CV complications are a major cause of maternal morbidity and mortality. Pregnancy-related CV complications are rising, particularly among women without established CVD. Risk tools are needed to identify women at risk for pregnancy-related CV complications regardless of baseline CVD status. Objectives: To develop and validate a clinical risk prediction model for pregnancy-related CV complications using routinely collected pregnancy care data. Methods: We trained and validated a clinical risk model for pregnancy-related CV complications in two independent electronic health record (EHR)-based pregnancy cohorts. Model derivation and internal validation were performed in 36,461 pregnancies delivered within Mass General Brigham (MGB). External validation was performed in 6,956 pregnancies delivered at Beth Israel Deaconess Medical Center (BIDMC). Pregnancy-related CV complications were defined as a composite of major adverse cardiac events, severe hypertensive disorders of pregnancy, or maternal death occurring between 13 weeks’ gestation-1 year postpartum. Candidate predictors included routinely collected demographic, pregnancy, clinical, and laboratory variables. AIC-based stepwise selection was used to identify predictors for inclusion in a multivariable logistic regression model. Model performance was evaluated in the internal and external validation sets using discrimination metrics, including area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC), and calibration metrics, including Brier score, and integrated calibration index (ICI). Results: The derivation cohort included 29,192 pregnancies (33 ± 5 years, 15% events). The internal MGB validation set comprised 7,269 pregnancies (33 ± 5 years, 16% events), and the external BIDMC validation set included 6,956 pregnancies (33 ± 5 years, 12% events). The final model included 12 predictors and demonstrated good discrimination, with AUROC of 0.73 (95% CI 0.71-0.74) and 0.73 (95% CI 0.71-0.75) and AUPRC of 0.35 (95% CI 0.33-0.38) and 0.30 (95% CI 0.28-0.33) in the MGB and BIDMC validation cohorts, respectively ( Figure A ). The model also demonstrated good calibration with ICI of 0.007 (95% CI 0.005-0.014) and 0.026 (95% CI 0.023-0.028) and Brier score of 0.115 (95% CI 0.113-0.117) and 0.095 (95% CI 0.093-0.097), respectively ( Figure B ). Conclusions: A clinical model derived from routinely collected pregnancy care data predicted pregnancy-related CV complications in two independent EHR-based pregnancy cohorts. These findings suggest that routine demographic and clinical data may help identify at-risk individuals regardless of baseline CVD status and support targeted preventive efforts, like intensified antenatal surveillance, risk-guided use of preventive therapies, and prioritized postpartum CV follow-up.