Establishment And Validation Of Clinical Prediction Model For Postpartum Hemorrhage During Cesarean Delivery

Objective:To establish a clinical prediction model of postpartum hemorrhage risk after cesarean delivery based on the risk factors,and verify this model,so as to supplement the problems existing in the current clinical prediction model of postpartum hemorrhage,further optimize the research on clinical prediction model of postpartum hemorrhage and provide a theoretical basis.Methods:In this study,parturient women hospitalized in our hospital from January 1,2016 to December 31,2020 and receiving cesarean delivery were selected as the subjects.By a retrospective analysis,the case data of the subjects were extracted,including age,birth history,gestational age at delivery,history of magnesium sulfate use during pregnancy,pregnancy complications,history of prenatal bleeding,laboratory examination results,auxiliary examination results,etc.According to the inclusion and exclusion criteria,the case database was constructed.The modeling cohort and verifying cohort were constructed through data sampling.The risk factors were screened using univariate analysis and multivariate logistic regression analysis,to establish a clinical prediction model of postpartum hemorrhage risk after cesarean delivery.The clinical prediction model was presented in the form of a nomogram,and evaluated with differentiation,calibration and clinical effectiveness.Results:The comparison of the population baseline revealed statistically significant differences（P<0.05）in the frequency of previous cesarean deliveries,thrombocytopenia,and placenta previa between the modeling cohort and the validation cohort.In univariate analysis,13 factors such as age,the frequency of previous induced abortions,the frequency of previous cesarean deliveries,prenatal bleeding,polycyesis,gestational age at termination of pregnancy,hypertensive disorder complicating pregnancy,pregnancy with anemia,thrombocytopenia,placental location,placenta previa,placental implantation,and fundal height were included in the risk factors for postpartum hemorrhage after cesarean deliveries.By univariate analysis and multivariate logistic regression analysis,age,the frequency of induced abortions,the frequency of previous cesarean deliveries,prenatal bleeding,polycyesis,hypertensive disorder complicating pregnancy,placenta previa,placental implantation,anemia and thrombocytopenia were included in the establishment of the clinical prediction model of postpartum hemorrhage risk after cesarean delivery.The established logistic regression model was statistically significant（χ2=29.914,P<0.001）.This model could accurately classify 83.6%of subjects.The sensitivity of the prediction model was 83.3%,the specificity was 88.7%,the negative predictive value was 86.2%,and the positive predictive value was 86.2%.The clinical prediction model was presented in the form of a nomogram.The receiver operating characteristic（ROC）curve of the clinical prediction model was drawn in the modeling cohort and verifying cohort,and the area under the curve（VUC）were 0.899 and 0.909 respectively,suggesting that this clinical prediction model can well distinguish postpartum hemorrhage patients and non-postpartum hemorrhage patients.In the modeling cohort and verifying cohort,the calibration line and standard line of the calibration curve fitted well,and the mean absolute errors were 0.022 and 0.008 respectively,indicating that the risk of postpartum hemorrhage predicted by this clinical prediction model is consistent with the actual risk.In the analysis of the clinical decision-making curve,the threshold probability of the curve was 0-1 and 0.25-0.8 respectively,which were higher than the two polar lines,suggesting that the model has certain clinical effectiveness.Conclusion:The clinical prediction model comprehensively includes the independent risk factors of postpartum hemorrhage after cesarean delivery as the predicting factors,and is visualized,which makes the clinical prediction model easy to understand and operate.In the analysis of discrimination,calibration and clinical decision-making curve,the clinical prediction model performs well and can be applied to clinical practice.In the future,we need to further expand the sample size and improve the prediction model.