(600) Multivariable prediction model for red blood cell transfusion associated with cesarean delivery

Secondary analysis of a multicenter trial of tranexamic acid (TXA) versus placebo for the prevention of red blood cell transfusion after cesarean delivery. The primary outcome for this analysis was transfusion of packed red blood cells by hospital discharge or 7 days postpartum, whichever came first. Maternal characteristics and risk factors related to labor were compared between patients who received a transfusion and those who did not. Any of the characteristics that differed with a p-value < 0.05 were considered in a multivariable model. Model selection and internal validation were performed using k-fold cross-validation. Using backward elimination, variables that remained significant at p < 0.05 were retained in each model. The final model was chosen based on the highest area under the receiver operating characteristic curve.


Results: 433 of 10,993 patients (3.9%) received a red blood cell transfusion. Several factors were associated with transfusion and retained in the final model (Table). The final multivariate model had strong predictive ability with an area under the curve (AUC) of 0.80 (95% CI 0.77, 0.82). Exclusion of suspected placenta accreta cases (N=32) from the multivariable model yielded a similar AUC (0.79, 95% CI 0.76, 0.81). Using this model, a low-risk non-laboring patient at term with a singleton pregnancy and a normal hemoglobin level (12.5 g/dL) has an 8.0 per 1,000 risk of transfusion. Conversely, a laboring patient with oxytocin duration for 36 hours, second stage of labor, chorioamnionitis, and hemoglobin level 9.3 g/dL has a 65% risk of transfusion.

Conclusion: We identified several factors associated with risk of red blood cell transfusion in patients who underwent cesarean delivery and a model with strong predictive ability. Validation of the model in an independent population is needed, and a calculator to risk-stratify patients for preparation for transfusion is being developed.