An advanced artificial intelligence model demonstrated a 95% accuracy rate in identifying postpartum hemorrhage among patients, while also uncovering 47% more bleeding cases than conventional methods. The results were published in npj Digital Medicine.
Postpartum hemorrhage remains a serious condition that can be life threatening and, in some instances, may necessitate a hysterectomy.
Researchers employed the Flan-T5 framework to process information from a cohort of 131 thousand women who gave birth over a 17-year span from 1998 to 2015. Their data included clinicians’ diagnoses of postpartum hemorrhage, which were encoded with a dedicated clinical code for accuracy and consistency.
The AI system identified 47% additional instances of bleeding compared with human assessment and achieved a 95% overall accuracy. These findings open the door for future studies to pinpoint factors that elevate the risk of postpartum hemorrhage and improve early intervention strategies.
Looking ahead, scientists plan to validate this approach further and explore real-time applications that could extend to other pregnancy-related complications, enhancing surveillance and response capabilities in obstetric care.
Earlier research highlighted a factor that increases the relative risk of vascular dementia by 78%, underscoring how predictive modeling can illuminate important health patterns across diverse areas.