Research from Sema4, a patient-centered health intelligence company, sets out the potential for the use of artificial intelligence to identify high-risk new mothers and reduce the risks of postpartum hemorrhage (PPH). This condition is the leading cause of pregnancy-related deaths.
In many parts of the world an alarming number of mothers experiencing excessive bleeding following the birth of a baby.in the U.S., this is on the rise, currently accounting for 35 percent of maternal deaths.
According to the U.S. CDC, the reason “for the overall increase in pregnancy-related mortality are unclear.”
The Agency notes, however, that: “Identification of pregnancy-related deaths has improved over time due to the use of computerized data linkages between death records and birth and fetal death records by states, changes in the way causes of death are coded.” Hence improved assessment and reporting could explain the step-change.
The research sets out how to harness the power of large-scale, comprehensive real-world data to predict meaningful outcomes. For this, Sema4 conducted the research in collaboration with the Mount Sinai Health System using an artificial intelligence technique to analyze de-identified electronic medical records of more than 70,000 pregnancies to improve the prediction of PPH.
The outcome suggests that by implementing the resulting predictive model into the clinical standard of care, healthcare providers may be able to improve PPH risk assessment and medical management for their pregnant patients resulting in better health outcomes.
Current guidelines fail to identify women without risk factors or those with the condition who do not show any symptoms, negatively affecting healthcare providers’ ability to predict PPH risk and perform appropriate monitoring during labor.
The new artificial intelligence -driven research creates a novel means of identifying PPH risk factors before, during, and after delivery. The research identified additional risk features that are not currently used in standard guidelines, which allowed for 89% accuracy in detecting PPH compared to 67 percent accuracy with standard guidelines in practice
From this, a predictive model was developed that identified five new PPH risk factors not currently included in standard risk tools but that are readily available on routine blood tests and vital sign monitoring. This model also identified critical changes in laboratory markers and vital signs where PPH risk rose substantially, offering providers indicators for early intervention.
The research appears in the Journal of the American Medical Association, titled “Improving postpartum hemorrhage risk prediction using longitudinal electronic medical records”.
