///2019 Abstract Details
2019 Abstract Details2019-07-13T07:45:15-05:00

Validation of a risk model for women with suspected placenta accreta spectrum

Abstract Number: T1C-305
Abstract Type: Original Research

Carolyn Weiniger MB ChB1 ; Elana Reitman MD2; Laurence Ring MD3; Lisa Deutsch PhD4; Ruth Landau MD5

Background: Correct antenatal diagnosis of placenta accreta spectrum (PAS) is crucial because outcomes are optimized when delivery occurs with highest capabilities and interventions at a level III or IV maternal care facility. A mathematical model to predict actual PAS among women using antenatal information from 3 variables (ultrasound imaging, number of prior cesarean deliveries (CD) and placenta previa) was previously developed on a cohort of 92 women (1). Our aim was to evaluate the robustness of the risk score using a validation cohort of suspected PAS cases from another tertiary medical center.

Methods: The validation cohort was prospectively gathered in a single-site tertiary delivery center. Ultrasound imaging, number of prior CD and diagnosis of placenta previa were recorded, as well as confirmation of PAS by OB surgeons at the time of delivery and/or pathology. The previously derived risk score (range 0 to 1 from the model equation combining the 3 variables) (1) was applied to the validation data and a receiver operator characteristic (ROC) curve generated according to actual status of PAS diagnosis; the area under the curve (AUC) was calculated as well.

Results: In the validation cohort (N=99), 85 had PAS confirmed by OB/pathology and 11 did not. The risk score performance to identify women with and without PAS according to OB and/or pathology diagnosis is presented in the Figures. The training set optimum cut-point 0.208 was applied on the validation dataset calculated risk score. It provides a sensitivity 97.7% and a specificity 36.4%. The combination model (pooled training and validation sets) had an of AUC 0.8241, 95%CI 0.7471 to 0.9012. Using the same 0.208 cut-point from the training model, sensitivity for the actual PAS diagnosis in the pooled model was 96.4% and specificity was 49.0%.

Conclusion: In this validation study, the mathematical model risk score appears to continue to differentiate aptly between those who did versus did not have PAS. If applied antenatally according to the 3 clinical variables (ultrasound imaging, number of prior CD and placenta previa) the sensitivity and specificity of the PAS diagnosis can be evaluated. The sensitivity of this model is higher than reported for ultrasound imaging alone, 91% (95%CI 87 to 94%) (2). This can aid optimum multidisciplinary planning for women with suspected PAS.

1. Int J Obstet Anesth 2013;22(4):273-9

2. PAS Obstetric Care Consensus No. 7. ACOG Obstet Gynecol 2018

SOAP 2019