A Framework for Improving Characterization of Obstetric Hemorrhage Using Informatics Data

Abstract

Objective: To characterize postpartum hemorrhage trends and outcomes using bioinformatics and electronic health record data.

Methods: This retrospective analysis included all women who delivered in a four-hospital system from July 2014 to July 2017 during implementation of a postpartum hemorrhage bundle. Data on billing codes, uterotonics, transfusion, intrauterine tamponade device placement, and hysterectomy were analyzed. A framework of four postpartum hemorrhage levels based on hemorrhage interventions was created using this informatics data. Levels were analyzed in relation to hematocrit drop from the highest predelivery to the lowest postpartum level. Changes in treatment patterns were assessed with risk-adjusted regression models with adjusted odds ratios (aOR) and 95% CI as the measures of effect. Postpartum hemorrhage-associated severe maternal morbidities were analyzed with adjusted models.

Results: The cohort included 43,657 deliveries. Four mutually exclusive postpartum hemorrhage levels were created based on informatics and billing criteria. Level 1: receipt of uterotonic other than oxytocin (3.7% of patients); level 2: billing diagnosis code for postpartum hemorrhage (3.0% of patients); level 3: placement of the intrauterine tamponade device, transfusion of 1-3 units red blood cells (RBCs), or both (1.8% of patients); and Level 4: hysterectomy, 4 or more units RBCs, or both (0.6% of patients). Higher postpartum hemorrhage levels were associated with higher hematocrit drops. In postpartum hemorrhage levels 1 through 4, 1.6%, 5.6%, 30.2%, and 30.7% of women had hematocrit drops greater than 40%, compared with 0.4% of women without postpartum hemorrhage. Over the course of the study, hematocrit drops within a given level did not change. Postpartum hemorrhage interventions such as uterotonics increased significantly (aOR 1.16, 95% CI 1.11-1.21, with aOR denoting change in outcome across 1 year). Although severe maternal morbidity did not change significantly, risk of hysterectomy decreased significantly (aOR 0.52, 95% CI 0.40-0.68).

Conclusion: Postpartum hemorrhage can be characterized in a granular fashion with informatics data. Informatics data are becoming increasingly available and can provide detailed assessment of postpartum hemorrhage incidence, management, and outcomes to facilitate surveillance and quality improvement.