Machine Learning Predicts Preeclampsia Risk in Pregnancy | Weill Cornell Medicine

A new machine-learning model offers the potential for more timely assessment of preeclampsia risk during late pregnancy, a condition that affects an estimated 2% to 8% of pregnancies globally. Published today, March 6, in JAMA Network Open, the model analyzes electronic health record data to provide continually updated predictions, potentially offering clinicians a crucial early warning system for a complication that can pose serious risks to both parent and child.

Understanding Preeclampsia and the Challenge of Late-Onset Cases

Preeclampsia is characterized by a sudden onset of high blood pressure during pregnancy, typically occurring after 20 weeks. It can lead to severe complications, including organ damage, seizures, and even death for both the parent and the developing baby. While existing models can identify risk factors early in pregnancy – often leading to preventative measures like low-dose aspirin – their accuracy diminishes when it comes to predicting late-onset preeclampsia, which accounts for the majority of diagnoses. This is where the new model, developed by researchers at Weill Cornell Medicine, aims to fill a critical gap.

The study, co-led by Dr. Fei Wang and Dr. Zhen Zhao, utilized data from over 35,000 deliveries and evaluated the model’s performance across nearly 59,000 pregnancies at three NewYork-Presbyterian hospitals. As reported by LetsDataScience, the model demonstrates the strongest predictive capability around 34 weeks of gestation.

How the Model Works and What Sets it Apart

Unlike earlier risk assessment tools, this model doesn’t rely on a single, static evaluation. Instead, it continuously analyzes data from electronic health records, incorporating new information as it becomes available. This dynamic approach allows for a more nuanced and up-to-date assessment of an individual’s risk profile throughout the third trimester. The researchers emphasize that the model is designed to assist clinicians, not replace their judgment. It provides an additional layer of information to inform clinical decision-making.

Clinical expertise in obstetrics was provided by Dr. Tracy Grossman, a maternal-fetal medicine specialist at NewYork-Presbyterian Brooklyn Methodist Hospital, highlighting the collaborative nature of the research. The study itself was approved by the Weill Cornell Medicine institutional review board, with a waiver of informed consent granted for the secondary use of deidentified patient data. Details of the study’s ethical oversight are available in the JAMA Network Open publication.

The Importance of Late-Pregnancy Monitoring

The timing of this predictive capability is particularly significant. Late-onset preeclampsia often progresses rapidly, leaving limited time for intervention. Early identification allows clinicians to closely monitor at-risk patients, potentially intervening with timely delivery or other medical management strategies to mitigate the risks. However, it’s crucial to understand that a prediction of increased risk does not equate to a diagnosis. It simply signals the need for heightened vigilance and more frequent monitoring.

Limitations and Future Directions

As with any predictive model, it’s important to acknowledge its limitations. The model was trained and validated on data from a specific population – patients at NewYork-Presbyterian hospitals. Its performance may vary when applied to different populations or healthcare settings. Further research is needed to assess its generalizability and to refine its accuracy across diverse patient groups. The study authors also note that the model’s predictive power is strongest around 34 weeks, suggesting that its utility may decrease earlier or later in the third trimester.

The researchers are continuing to explore ways to improve the model’s performance and to integrate it into clinical workflows. Future studies may focus on incorporating additional data sources, such as genetic information or lifestyle factors, to further enhance its predictive capabilities. Weill Cornell Medicine News reports that the team is also investigating the potential of using the model to personalize treatment plans for patients at risk of preeclampsia.

What’s Next for Preeclampsia Prediction?

The development of this machine-learning model represents a significant step forward in the effort to predict and prevent preeclampsia. However, it’s just one piece of the puzzle. Ongoing research is essential to improve our understanding of the underlying causes of preeclampsia and to develop more effective prevention and treatment strategies. Clinicians should stay informed about the latest research and guidelines, and patients should discuss any concerns they have about preeclampsia with their healthcare providers. Public health surveillance will continue to play a vital role in monitoring the incidence of preeclampsia and identifying emerging risk factors.