A Regression Model for Prediction of Cesarean-Associated Blood Transfusion.

Published

Journal Article

OBJECTIVE: To develop a model to predict cesarean-associated red blood cell transfusion. STUDY DESIGN: Secondary analysis of all cesarean deliveries in the Maternal-Fetal Medicine Units Network Cesarean Registry. Using a split-sample technique, the derivation group was used to identify associated factors and build predictive models, and the validation group was used to estimate classification errors and determine test characteristics. Using factors available at the time of cesarean, we developed a multivariable logistic regression prediction model. RESULTS: A total of 59,468 women were split evenly and randomly into the derivation and validation groups. The overall rate of transfusion was 2.7%. The area under the receiver operating characteristic curve for the derivation and validation groups were 0.82 (95% confidence interval [CI]: 0.80-0.84) and 0.84 (95% CI: 0.82-0.85), respectively (p = 0.16). The strongest predictors of transfusion were placenta previa (odds ratio [OR]: 7.06, 95% CI: 5.19-9.61) and eclampsia/Hemolysis Elevated Liver Enzymes Low Platelets syndrome (OR: 5.67, 95% CI: 3.77-8.51). In the validation group, the model had a sensitivity, specificity, positive, and negative predictive values of 55.8, 91.5, 16.2, and 98.6%, respectively. Overall, 90.5% of patients were correctly classified. CONCLUSION: A regression model incorporating variables available at the time of cesarean accurately predicts the need for intra- or postoperative transfusion.

Full Text

Duke Authors

Cited Authors

  • Albright, CM; Spillane, TE; Hughes, BL; Rouse, DJ

Published Date

  • July 2019

Published In

Volume / Issue

  • 36 / 9

Start / End Page

  • 879 - 885

PubMed ID

  • 30743270

Pubmed Central ID

  • 30743270

Electronic International Standard Serial Number (EISSN)

  • 1098-8785

Digital Object Identifier (DOI)

  • 10.1055/s-0039-1678604

Language

  • eng

Conference Location

  • United States