Introduction
Placenta accreta spectrum (PAS) refers to a group of placentation disorders that are characterized by trophoblastic invasion beyond the physiologic decidual–myometrial junction zone (1). PAS is identified as one of the most serious pregnancy-related disorders because it is associated with substantial risk of massive obstetric hemorrhage, blood transfusion, surgical injuries, and thereby high risk of maternal intensive care unit (ICU) admission, reoperation, and prolonged hospitalization (2). Unfortunately, burden of PAS morbidity has been significantly aggravated as a result of the rising trend of cesarean section delivery (CS) among contemporary population (2).
To date, the most widely supported approach in management is PAS is cesarean hysterectomy without trying to separate placenta (placenta in-situ) (3). Although this approach may be associated with improved maternal outcomes, uterine preservation is routinely offered as an alternative or even considered as the primary approach in several regions of the world (4). Interventional radiology (IR) is another option that may reduce peripartum bleeding regardless of management approach (5). Despite being widely adopted, uterine preserving procedures are generally not robustly supported by evidence and data on clinical outcomes of these procedures are limited (6). Given the seriousness of PAS and presence of several proposed interventions, calculation of individualized probability of intrapartum and postpartum serious morbidity based on patient demographics, disease characteristics, and different treatment options may facilitate treatment decision and proper use of resources.
Machine learning (ML) is a subset of artificial intelligence, where a computer gains cumulative experience from an existing database, to be capable of making accurate predictions of studied outcomes (7). Generally, ML may provide more accurate prediction, reveal more complex relations between features and outcomes, and provide a scalable and readily applicable clinical tool compared to traditional statistics (7). The current study presents an international multicenter center of women with PAS who were managed conservatively or by cesarean hysterectomy. The study aimed at creating antepartum and peripartum prediction models of peripartum clinical outcomes, using ML technology, to enhance decision making with regard to PAS.