///2017 Abstract Details
2017 Abstract Details2019-08-02T15:54:53-06:00

The Development and Validation of an Obstetrical Hemorrhage Risk Prediction Index

Abstract Number: F-46
Abstract Type: Original Research

Jennifer Cottral MD1 ; Brian Bateman MD, MSc2

Background: Identification of women at heightened risk for obstetrical hemorrhage may lead to improved outcomes by allowing the care team to prepare for this complication, including performing appropriate pre-transfusion testing. Currently available risk stratification tools classify women as low, medium, and high risk based on individual risk factors. (Main et al, Obstet Gynecol, 2015) A risk score that weights risk factors based on the strength of their association with hemorrhage and that captures the impact of the presence of multiple risk factors has not been previously described.

Objective: To develop and validate an obstetrical hemorrhage risk prediction index.

Methods: Data were derived from a database of delivery hospitals in New York from 1998-2007. The primary outcome was defined as the transfusion of ≥4 units of packed red blood cells (pRBCs) during the delivery hospitalization, which is the definition used by the Joint Commission to define severe hemorrhage. The data set was divided into a development (2/3 of sample) and validation cohort (1/3 of sample). For the development cohort, a logistic regression model predicting the primary outcome was created using a stepwise selection algorithm that included 23 candidate antepartum risk factors for obstetrical hemorrhage. Each of the risk factors included in the final model was weighted based on its beta coefficient, and these were summed to calculate a risk score. Using the validation cohort, the performance characteristics of the index were evaluated. The performance of the risk index was also defined for alternative thresholds for defining hemorrhage including ≥1, ≥2, ≥3, and ≥10 units of pRBCs.

Results: A total of 690,742 completed pregnancies were analyzed, of which 0.4% (n=2,764) were complicated by the primary study outcome of transfusion of ≥4 units of pRBCs. The derived index included 17 risk factors (Table). Potential total scores on the index range from 0 to 68. For each point increase in the index, the relative risk for the primary outcome was 1.34 (95% confidence interval 1.32–1.36), such that a woman with an index of 0 had a 0.2% risk of the primary outcome and with an index of ≥13 the risk was 17%. The c-statistic for the model was 0.74. Calibration was similarly robust across the alternative thresholds for defining hemorrhage.

Conclusion: Our risk index provides a simple tool that can be used in during the antepartum period to identify women at excess risk for hemorrhage.

SOAP 2017