A machine learning model developed by Weill Cornell Medicine researchers could give clinicians early warning of complications that may occur later in pregnancy.
Preeclampsia is a sudden-onset condition associated with high blood pressure before childbirth. It affects approximately 2% to 8% of pregnancies worldwide and can have serious consequences for both parents and children. A new study published March 6 in JAMA Network Open describes a machine learning-based computer model that provides continuously updated predictions of preeclampsia risk based on electronic health record data recorded during the third trimester of pregnancy. The study was co-led by Fei Wang, Ph.D., associate dean for AI and data science and Francis L. Loeb Professor of Medical Informatics in the Department of Population Health Sciences at Weill Cornell Medical College, and Zhen Zhao, Ph.D., professor of clinical pathology and laboratory medicine at Weill Cornell Medical College and director of the Central Research Institute at NewYork-Presbyterian/Weill Cornell Medical Center. Obstetric clinical expertise was provided by Dr. Tracy Grossman, assistant professor of clinical obstetrics and gynecology at Weill Cornell Medical College and maternal-fetal medicine specialist at NewYork-Presbyterian Brooklyn Methodist Hospital.
Existing models that assess preeclampsia risk early in pregnancy are primarily used as early warnings, allowing clinicians to prescribe aspirin as a prophylactic agent early in pregnancy and provide additional monitoring throughout the at-risk pregnancy. Although these approaches may reduce the risk of early-onset preeclampsia, their predictive accuracy is limited for late-onset and term-onset cases, which account for the majority of preeclampsia diagnoses. As a result, there are few tools available to help predict short-term preeclampsia risk during the third trimester, when most cases occur. To fill this gap, co-lead authors Haoyang Li, Ph.D., a postdoctoral fellow in population health sciences, and Yaxin Li, Ph.D., a postdoctoral fellow in pathology and laboratory medicine, collaborated with Drs. Wang, Zhao, and Grossman developed and tested a preeclampsia modeling tool using anonymized electronic medical record data from approximately 59,000 pregnancies at three New York-Presbyterian hospitals. The research team created the model using data on 35,895 pregnancies among patients who delivered at NewYork-Presbyterian/Weill Cornell Medical Center from October 2020 to May 2025. The model most accurately predicts the likelihood of preeclampsia around the 34th week of pregnancy, potentially giving clinicians time to take preventive measures.
The team then validated the model using data from 8,664 pregnancies at NewYork-Presbyterian Lower Manhattan Hospital and 14,280 pregnancies at NewYork-Presbyterian Brooklyn Methodist Hospital. This model showed that blood pressure in pregnant patients was the strongest predictor of preeclampsia. However, early in the third trimester, routine blood tests of a patient may yield abnormal results, indicating potential risks. These test results may suggest that there is a new problem with the placenta, which supplies nutrients and oxygen to the fetus, and may be contributing to preeclampsia at this stage. Later in the third trimester, the patient’s age and white blood cell count became more important indicators, suggesting that inflammation may be involved at this point.
This model helps clinicians identify patients in the third trimester who are most likely to develop preeclampsia and provides additional lead time to take timely clinical actions, such as increased monitoring, blood pressure control, and decisions regarding timing of delivery. Unlike previous approaches that provide a single static risk estimate, this model continuously updates the risk of preeclampsia using current electronic medical record data as the pregnancy progresses, aligning predictions with actual clinical decision-making in late pregnancy. Further research is needed to determine whether preeclampsia at different stages of late pregnancy has distinct causes, such as placental insufficiency or systemic inflammation. However, if these patterns are identified, they may help clinicians develop more targeted preeclampsia interventions that address the underlying causes.
sauce:
Reference magazines:
Lee, H. Others. (2026). Machine learning for dynamic and short-term prediction of pre-eclampsia using routine clinical data. JAMA network open. DOI: 10.1001/jamanetworkopen.2026.0359. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2845997
