A prediction model for placenta accreta spectrum: A multicentre external
validation study.
Abstract
Objective: To validate the Weiniger model, a multivariable prediction
model for placenta accreta spectrum (PAS). Design: Multicentre external
validation study. Setting: Two tertiary care hospitals in the United
States. Population: Cohort A included patients with risk factors (prior
caesarean delivery, placenta praevia) and/or ultrasound features of PAS
(variable risk) presenting to a tertiary care hospital. Cohort B
patients were referred to a tertiary care hospital specifically for
ultrasound features of PAS (higher risk). Methods: Weiniger model
variables (prior caesarean deliveries, placenta praevia and ultrasound
features of PAS) were retrospectively collected from both cohorts and
predictive performance of the model was evaluated. Main Outcome
Measures: Surgical and/or pathological diagnosis of PAS. Results: The
model c-statistic in cohorts A and B was 0.728 (95% CI: 0.662, 0.794)
and 0.866 (95% CI: 0.754, 0.977) signifying acceptable and excellent
discrimination, respectively. Based on calibration curves, the model
underestimated average PAS risk in both cohorts. In both cohorts, high
risk was overestimated and low risk underestimated. Use of this model
compared to a “treat all” strategy had greater net benefit at a
threshold probability of > 0.25 in cohort A, but no net
benefit in cohort B. Conclusions: This study provides multicentre
external validation of the Weiniger model for PAS prediction, making it
a useful triaging tool for management of this high-risk obstetric
condition. Clinical usefulness of this model is influenced by the
incidence of risk factors and PAS ultrasound features, with better
performance in a variable-risk population at threshold probability
>25%.