Mount Sinai uses machine learning to improve the prediction of postpartum hemorrhage risk


What you should know:

Sema4, a Powered by AI The genomic and clinical data intelligence platform company, has recently published two studies demonstrating the usefulness of machine learning in predicting the clinical outcomes of postpartum hemorrhage (PPH). The studies, which will appear in a special print issue of “Computer Science for Sex and Gender Health” in the Journal of the American Medical Informatics Association (JAMIA), were conducted in collaboration with physicians from the system of hello Mount Sinai.

– Sema4 chose to focus its advanced machine learning methods on PPH, as the disease is the leading cause of maternal mortality worldwide. HPP accounts for about one-third of maternal deaths and often occurs in patients without known risk factors for bleeding. In addition, limitations in diagnostic guidelines and risk assessment tools may make it difficult for health care providers to properly identify and treat PPH, especially in patients without obvious symptoms.

First study

He first study It leveraged data from the unidentified longitudinal electronic medical record (EMR) of more than 70,000 parts of pregnancy at five hospitals in the Mount Sinai health system to develop and validate a complete digital phenotyping algorithm for PPH. The new algorithm incorporates not only accumulated blood loss, but also other critical characteristics related to the diagnosis and treatment indicated by PPH.

“HPP is a devastating condition that occurs with little warning. Current guidelines are based primarily on accumulated blood loss as the main diagnostic marker of HPP,” he said. Li Li, MD, SVP of Clinical Informatics at Sema4 and corresponding co-author. “We identified additional clinical features from the EMR data, which allowed us to identify PPH with 89% accuracy, while the standard definition based on blood loss was only 67% accurate. Therefore, we predict that our digital phenotyping algorithm will be of significant use in monitoring results and clinical research to develop better preventive interventions for PPH. “

Second study

The same cohort of patients was used in the second study to build, train, and validate a predictive model of PPH risk using advanced machine learning methods. The model uses 24 predictive markers, including five new potential PPH risk factors. These risk factors are easily obtained from routine laboratory tests, including complete blood count panels and vital signs, but are not currently used in standard risk assessment tools. The research also identified turning points for laboratory values ​​and vital signs where PPH risk increased substantially, which could serve as a guideline to control intrapartum risk. In comparison tests, the new Sema4 tool outperformed three existing models and clinical risk assessment tools.

“In our studies, we established a complete and physician-validated digital phenotyping tool for PPH using EMR data from a large U.S. health system. We then used this validated phenotype to evaluate the data. longitudinal anterior and intrapartum to build a robust predictive model for PPH risk after hospital admission, ”Drs. Li. “Our model identified thresholds of clinical features that can guide intrapartum follow-up with near-immediate clinical utility. Compared to current clinical standards of abnormal vital signs and laboratory parameters in the prepartum and intrapartum periods, we detected values ​​for to these 24 markers that could provide early warning signals to health care providers to monitor at-risk patients.After additional evaluation, our predictive tool can enable reproductive health care providers to predict and treat possible symptoms of high risk before they occur, allocating resources appropriately, and ultimately reducing morbidity and mortality from PPH. “

Both studies leveraged the knowledge of Centrellis®, Sema4’s health intelligence platform that drives Sema4’s ongoing research to understand and improve reproductive health. Key areas of research include predictive modeling to determine the optimal time for a pregnant patient to receive noninvasive prenatal testing to achieve more accurate results and generate models to better predict the potential risk of preeclampsia, another leading cause of mortality. maternal.

The Icahn School of Mount Sinai Medicine (Icahn Mount Sinai) has its share capital in Sema4. Dr. Schadt is the CEO of Sema4 and owns the company. In addition, Dr. Schadt is a part-time teacher at Icahn Mount Sinai.

Source link