Abstract: Objective To establish a multivariate prediction model for platelet transfusion in patient with severe postpartum hemorrhage (PPH) to provide a basis for the clinical formulation of platelet transfusion. Methods A total of 319 maternities with severe PPH successfully treated in Women's Hospital of Nanjing Medical University between January 2008 and May 2021 were retrospectively analyzed. The patients were divided into two groups: platelet transfusion group (64) and no platelet transfusion group (255) based on platelet transfusion. Predictors included age, body mass index (BMI), mode of delivery, pregnancy comorbidities and complications including amniotic fluid embolism, abruptio placentae, stillbirth, preeclampsia, HELLP syndrome, moderate to severe liver function abnormalities (≥3-fold increase in transaminases), amount of postpartum hemorrhage, presence of occult hemorrhage (including intra-abdominal hemorrhage, subperitoneal hematoma, and vaginal wall hematoma) that could not be counted accurately, amount of transfused red blood cell suspensions, platelet count before delivery, and whether hemorrhagic shock or disseminated intravascular coagulation occurred during resuscitation. Risk factors for platelet transfusion were identified and R software was used to create a nomogram prediction model. The performance of nomogram was evaluated by using a bootstrapped-concordance index and calibration plots. Receiver operating characteristic (ROC) analysis was used to compare the diagnostic efficacy of the nomogram model and other independent predictors. Results Binary logistic regression analysis showed that amount of postpartum hemorrhage, amount of red blood cells transfused, platelet count before delivery, presence of occult hemorrhage, placental abruption, preeclampsia and moderate to severe liver function abnormalities were independent risk factors for predicting severe reduction in platelets. The nomogram model constructed based on the amount of postpartum hemorrhage, amount of transfused red blood cell suspensions, platelet count before delivery, presence of occult hemorrhage, placental abruption, preeclampsia and moderate to severe liver function abnormalities has the highest AUC value of 0.789. The AUCs for the independent influencing factors of platelet transfusion in patients with severe postpartum hemorrhage applied alone for the amount of postpartum hemorrhage, the amount of transfused red blood cell suspensions, platelet count before delivery, presence of occult hemorrhage, placental abruption, platelet count, preeclampsia and moderate to severe liver function abnormalities before delivery were 0.631, 0.689, 0.518, 0.611, 0.641, 0.531 and 0.529, respectively. Conclusion The multivariate prediction model of platelet transfusion in patients with severe PPH based on Nomogram has high predictive efficacy. It can guide clinical foresight to prepare platelets in advance and provide a basis for timely control of platelet transfusion timing and formulation of platelet transfusion plan in severe PPH.

Key words: Severe postpartum hemorrhage, Platelet transfusion, Multivariate prediction model, Nomogram