Artificial Intelligence for Early Prediction of Maternal Cardiopulmonary Deterioration in Pregnancy Using Multimodal Digital Biomarkers: A Systematic Review

Digital health technologies and artificial intelligence (AI) have expanded the capacity for continuous physiologic monitoring during pregnancy. While wearable-derived digital biomarkers are increasingly studied, evidence specific to the early detection of maternal cardiopulmonary deterioration remains fragmented, and no focused synthesis has addressed this domain. We systematically reviewed existing evidence on wearable-derived digital biomarkers and AI-based analytic approaches used to identify early cardiopulmonary physiologic instability during pregnancy and the postpartum period. This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. Major bibliographic databases and clinical trial registries were searched through the final search date. Studies involving pregnant or postpartum populations that used wearable or remote monitoring technologies to capture cardiopulmonary-related physiologic signals were eligible. Owing to substantial heterogeneity in study design, outcomes, sensor modalities, and analytic approaches, findings were synthesized qualitatively without meta-analysis. Of 1,320 records identified, 27 studies met the inclusion criteria. The literature encompassed continuous monitoring of heart rate variability, resting heart rate, respiratory parameters, oxygenation patterns, skin temperature, activity cycles, and multimodal biometric signatures. Across studies, physiologic trajectories followed consistent gestational patterns, while deviations from individualized baselines were associated with preterm birth risk, labor onset, or emerging cardiometabolic strain. Several investigations proposed AI-based models integrating multiple physiologic streams to generate personalized risk profiles; however, prospective validation in high-risk obstetric populations remains limited. Feasibility and adherence to wearable monitoring were generally high, although methodological variability limited cross-study comparability. Current evidence suggests that AI-enabled digital biomarkers derived from wearable sensors hold promise for the anticipatory detection of maternal cardiopulmonary deterioration. Advancing clinical translation will require standardized biomarker definitions, harmonized analytic pipelines, and rigorously designed prospective studies aligned with obstetric care pathways.