EP07.29: Construction of prediction model for adverse pregnancy outcome of umbilical artery thrombosis fetuses

To construct a prediction model for adverse pregnancy outcome of umbilical artery thrombosis (UAT) fetuses and evaluate the hemodynamic performance with a relatively large sample. We included singleton fetuses with least one available Doppler evaluation before the end of pregnancy at the Women and Children's Hospital of Chongqing Medical University from 2018 to 2023 as prospective study subjects. Fetuses with structural abnormalities and with no follow-up results were excluded. Basic clinical indicators, ultrasound performance were collected. All two-dimensional and Doppler examination were performed in line with the ISUOG Practice Guidelines. The least absolute shrinkage and selection operator (LASSO) regression method was used to select optimal features and analyse the predictive value of each indicator in the model. 182 singleton fetuses were included in the final statistical analysis, of whom 82 had UAT. 5 predictive features, including a history of maternal diabetes, coagulation, umbilical artery pulsatility index (UA-PI), middle cerebral artery pulsatility index (MCA-PI), ductus venosus pulsatility index (DV-PIV), were identified by LASSO regression. The nomogram prediction model constructed based on the above predictors predicted the adverse pregnancy outcome of UAT more accurately, and the area under the curve (AUC) in the validation cohort was 0.76 (95% CI: 0.66-0.85). The multi-index prediction model of UAT for adverse pregnancy outcome is feasible and has good predictive value.