Prediction of maternal and infant outcomes with a Mother-Child AI agent (doi.org)

Researchers have introduced the Mother-Child AI Agent (MoChiAgent), an innovative AI clinical assistant designed to enhance the prediction of maternal and infant health outcomes by analyzing extensive electronic health record (EHR) data. Unlike traditional models, which often focus on limited outcomes and expensive tests, MoChiAgent utilizes its core predictive engine, MoChiFormer, to integrate longitudinal EHR data from over 4.4 million clinical visits. It effectively reconstructs missing laboratory values and identifies gestational conditions with impressive accuracy—achieving AUROCs of 0.89 for placental abruption and 0.91 for preterm labor. This accuracy allows for a nuanced understanding of risks to both mothers and infants throughout the pregnancy trajectory. The significance of MoChiAgent lies in its ability to provide actionable, evidence-based treatment recommendations, which can greatly improve risk-stratified care in obstetrics. By integrating maternal and infant records, it revealed critical transgenerational risk factors, such as a twofold increased risk of neonatal jaundice linked to specific maternal health conditions. This advancement not only strengthens the predictive capabilities of AI in maternal-infant health but also paves the way for more personalized and effective healthcare interventions. Overall, MoChiAgent represents a substantial leap in leveraging AI for better maternal and infant health outcomes, making it a notable development for the AI/ML community and healthcare professionals alike.