Objective  To explore the application value of six prediction models reported at home and abroad for adverse outcomes of hypertensive disorders of pregnancy(HDP) in eastern and western China.

  Methods  For all patients who delivered in the First Affiliated Hospital of Soochow University and Sichuan Provincial Maternal and Child Health Care Hospital from May 1, 2011 to April 30, 2019 and were diagnosed with HDP, their clinical data were retrospectively analyzed. Six models, fullPIERS, miniPIERS, Zwertbroek, PREP, Ngwenya, and Ma Guojun, were used to predict the risk of adverse outcomes for the patients. The predictive performance of the models was evaluated in terms of discrimination and calibration.

  Results  A total of 2978 patients were eligible. Combined adverse outcomes occurred in 13.6% (405/2978) of women within 48 h of admission, and 22.0% (655/2978) at any time during admission. The delivery < 34 weeks (49.4%, 200/405), need for blood product transfusion (43.5%, 176/405), and placental abruption (23.5%, 95/405) were the most common adverse outcomes within 48 hours of admission. The area under of the curve of the six models for predicting adverse outcomes in the patients with HDP within 48 hours of admission/during hospitalization ranged from 0.600 to 0.897, the sensitivity ranged from 57.1% to 69.5%, and the specificity ranged from 60.1% to 76.6%. The Hosmer-Lemeshow test showed that except for the PREP model (which had a small validation population and was not evaluated for calibration), the P-values of all the other 5 models were less than 0.05.

  Conclusions  The six prediction models have certain application value in the prediction of adverse outcomes of HDP patients in the eastern and western regions of China, but the fitting is poor. The predictors involved in some models are not routine inspection indicators, and the feasibility of large-scale model application is still open to question. It is still necessary to establish a better prognostic model suitable for local areas based on Chinese characteristics.