Construction of a risk prediction model for single intrauterine fetal death in the second and third trimesters

Background Objective To explore the risk factors for single intrauterine fetal death and to examine the clinical application value of a risk prediction model for single intrauterine fetal death occurrence. Study Design From January 1, 2011, to December 31, 2023, 137 patients with single intrauterine fetal death during the second and third trimesters of pregnancy who were registered for prenatal examination were included in the study group. In a 1:5 case-control study, 591 women pregnant with twins without single intrauterine fetal death were selected as the control group according to the age of pregnant women during the same time period of documentation. Lasso regression analysis was used to screen for risk factors for single intrauterine fetal death in twins. A multivariate logistic regression model and BP artificial neural network model were established, and the prediction efficiencies of the models were evaluated. Results Single-factor logistic regression analysis showed that the risk factors for single intrauterine fetal death were irregular prenatal examination, treatment at a non-3A hospital, low educational background, history of spontaneous preterm birth, monochorionic placenta, HDCP, placenta previa, placental abruption, fetal distress, hypothyroidism, moderate-to-severe anemia, short cord, fetal growth restriction, and twin-to-twin transfusion syndrome. Lasso regression analysis showed that the risk factors for sIUFD were monochorionic placenta, HDCP, placental abruption, fetal distress, moderate-to-severe anemia, and FGR. To establish a risk prediction model for sIUFD in twins, the Hosmer and Lemeshow Test statistics of the model were χ ² =0.760, P=0.979, indicating that the model fitted the data well. The ROC curve of the prediction model was drawn, and the area under the curve was calculated to be 0.840. The accuracy of the BP artificial neural network model was 90.0%, the positive predictive value was 93.2%, the negative predictive value was 89.6%, the sensitivity value was 50.4%, and the specificity was 99.2%. Conclusion Lasso regression analysis showed that the risk factors for single intrauterine fetal death were monochorionic placenta, HDCP, placental abruption, fetal distress, moderate-to-severe anemia, and FGR. The risk prediction model of single intrauterine fetal death based on multivariate logistic regression and the BP artificial neural network had good prediction efficiency.