Prediction of placenta accreta spectrum by a scoring system based on maternal characteristics combined with ultrasonographic features

ObjectiveIn order to create a comprehensive scoring system based on maternal characteristics and ultrasonographic features for predicting placenta accreta spectrum (PAS).Materials and methodsThis was a retrospective review of pregnant women who underwent routine ultrasound examination in the third trimester of pregnancy from January 2014 to November 2018 were used as a training set to establish the scoring system for PAS prediction while those who underwent examination from January 2019 to December 2019 served as a validation set.. Maternal characteristics including maternal age, parity, previous vaginal deliveries, previous curettage, previous cesarean section (CS), history of hypertension and diabetes mellitus, prenatal body mass index (BMI) were recorded. Ultrasonographic features including abnormal placental lacunae, subplacental hypervascularity, myometrial thinning, placental bulge, bladder wall interruption, location of placenta, placenta previa (yes or not) were recorded. Multivariate analysis was applied to analyze independent risk factors and assess the predictive power of selected parameters predicting PAS. Receiver operating characteristics (ROC) curve was used to evaluate the diagnosis power.ResultsParity, previous curettage and CS were independent risk factors. The best comprehensive scoring system was established as follow: the number of abnormal lacunae ≥3, 2 points; lacuna maximum dimension ≥2 cm, 5 points; subplacental hypervascularity (rich), 1 point; subplacental hypervascularity (extremely rich and disordered), 3 points; bladder wall interruption, 9 points; placental bulge, 9 points; placenta previa, 8 points; anterior placenta, 1 point; previous CS ≥ 1, 1 point; parity ≥ 4, 3 point; previous abortions ≥ 2, 1 point. The area under the ROC curve of the scoring system diagnosing PAS was 0.925. Sensitivity and specificity were 83.3% and 85.7%, respectively. Cross-validation for our model showed that sensitivity, specificity, positive predictive value and negative predictive value of the model in diagnosis of PAS were 82.6%, 81.8%, 82.6% and 81.8%, respectively. Diagnosis of 37 cases were consistent with the “gold standard”, and the coincidence rate was 82.2% (37/45).ConclusionThe comprehensive scoring system established in this study can effectively diagnose PAS.