Objective To investigate the occurrence and risk factors of short-term adverse hospital outcomes in pregnant women of advanced maternal age with early-onset severe preeclampsia (SPE), and to construct a quantitative nomogram prediction model to guide clinical practice.

Methods From January 2016 to January 2020, 316 pregnant women of advanced maternal age with early-onset SPE diagnosed by the Department of Obstetrics and Gynecology of Hai'an Hospital Affiliated to Nantong University were selected as training sets. They were divided into poor group(52 cases) and good group(264 cases) according to the different outcomes in the hospital. In addition, 90 pregnant women of advanced maternal age with early-onset SPE from February 2020 to May 2022 were selected as the validation set, and 16 of them had adverse outcomes. Single factor analysis was performed on the clinical data and blood biochemical indexes of the patients in the poor and good groups of the training set. The lasso algorithm (LASSO) and multifactor Logistic regression analysis were used to screen the best predictors. R software was used to establish a nomogram. The area under the curve(AUC) of the patient′s receiver operating characteristic(ROC) curve was calculated to predict the adverse outcome in the training set and the validation set.

Results The number of symptoms, systolic blood pressure, prothrombin time (PT), alanine transaminase (ALT), uric acid, lactate dehydrogenase (LDH), urea nitrogen (BUN) and creatinine, ratio of maximum systolic flow velocity to end-diastolic flow velocity (S/D) in fetal umbilical artery and resistance index (RI) in the poor group were higher or longer than those in the good group, while the blood plate count, PT activity (PTA) and albumin in the poor group were lower than those in the good group (P < 0.05). LASSO screened 6 non-collinear indicators. Logistic regression analysis showed that the number of symptoms ≥1, BUN≥5 mmol/L, PT≥10 s, LDH≥250 U/L, platelet count < 100×10⁹/L and albumin < 30 g/L were independent predictors of adverse outcomes. The internal validation of the training set showed that the AUC of the nomogram predicting the adverse outcome was 0.895, and the Hosmer-Lemeshow test showed that its goodness of fit was good(χ²=12.325, P=0.548), the calibration curve showed good consistency. External validation was performed on the validation set. The AUC of the nomogram predicting adverse outcomes was 0.846. The Hosmer-Lemeshow test showed that its goodness of fit was good(χ²=9.627, P=0.324), the calibration curve showed a good consistency.

Conclusion This study has developed a nomograph model with strong visualization and simple operation for guiding clinical early identification of adverse outcomes in hospital of advanced-aged pregnant women with early-onset SPE, which has a good predictive effect, and important potential for clinical prognosis of regional advanced-aged pregnant women with early-onset SPE in China.