External Validation of the fullPIERS Risk Prediction Model in a U.S. Cohort of Women With Preeclampsia [ID 2683408

Abstract

INTRODUCTION: Hypertensive disorders of pregnancy are a leading cause of maternal morbidity and adverse pregnancy outcomes. The fullPIERS model was developed to predict severe adverse maternal outcomes within 48 hours of admission in patients with preeclampsia. This model has not been validated in the U.S. cohort. We sought to examine whether the fullPIERS model was predictive in the U.S. cohort. METHODS: This was a retrospective study of patients with preeclampsia who delivered at 23 weeks or greater from January 1, 2010, to December 31, 2020. Our primary outcome was a composite of maternal mortality or other serious complications of preeclampsia that occurred within 48 hours of admission. Using the prediction model that considers gestational age at admission, chest pain or dyspnea, serum creatinine, platelet, aspartate transaminase, and oxygen saturation, we calculated the probability of the composite outcome. We assessed performance using the area under the curve (AUC) of the receiver operating characteristic curve. Hosmer–Lemeshow test was used to assess the goodness-of-fit. RESULTS: Of 1,552 patients with preeclampsia, 77 (5.0%) had the composite outcome within 48 hours. The fullPIERS model had an AUC of 0.73 (95% CI, 0.66–0.80). However, the Hosmer–Lemeshow test showed poor calibration with a value of P<.001. The calibration plot showed the model underestimated when the predicted probability was less than 10% and overestimated when the predicted probability was over 28%. CONCLUSION: The fullPIERS model had good AUC in the U.S. cohort of patients with preeclampsia but poor calibration. The fullPIERS model may not be clinically useful in the United States.