Abstract: Objective：To validate the prediction value of adverse pregnancy outcomes of fullPIERS model in women with pre-eclampsia (PE). Methods: This was a retrospective study. 398 pregnant women with PE hospitalized in Suzhou Municipal Hospital from January 2014 to December 2015 were included to analyze. The fullPIERS was applied to predict the adverse pregnancy outcomes within 48 h and 2-7 days after hospitalization. The discrimination (ROC AUC), risk stratification and calibration (H-L test and calibration plot) of the model were assessed. The comprehensive external validation of fullPIERS model was accomplished. Results: Among 398 pregnant women with PE, 67 (16.8%) of whom had adverse outcomes, including 41 (10.3%) within 48 h and 21 (5.3%) within 2-7 days. The area under the curve of fullPIERS model to predict adverse outcomes was 0.855 within 48 h，and 0.826 with in 2-7 days after hospitalization. The negative prediction value (NPV) of model to predict adverse outcomes within 48 h and 2-7 days were 96.9% (286/295) and 98.7% (312/316), while the positive prediction value (PPV) were 31.1% (32/103) and 20.7% (17/82). The calibration of fullPIERS model was good ( χ2=6.431, P=0.599) within 48 h, but was poor ( χ2=36.778, P=0.000) within 2-7 days. Conclusions: The fullPIERS model performed well to predict the adverse pregnancy outcomes in women with PE in single center of east China, especially within 48 h after hospitalization. The model might be used as a routine test to predict the adverse pregnancy outcomes in women with PE.

Key words: Pre-eclampsia, Pregnancy outcome , Prediction model, Risk adjustment, External validation