Accuracy of blood transfusion in postpartum hemorrhage to assess maternal morbidity

Objective: To measure the accuracy of blood transfusion (timing and number of blood units) in postpartum hemorrhage (PPH) in a perinatal network. Study design: (1) The ANONYMAT software system was used for anonymization and linkage of two large stand-alone databases, the Burgundy Perinatal Network (BPN) and the National Blood Centre (EFS) databases, which contain, respectively, clinical data from hospital discharges and information concerning any blood transfusion in France (considered as the gold standard database for identifying any transfusion). (2) Identification of prescriptions of at least one red blood cell (RBC) unit at the day of delivery (>= 22 weeks) and up to 42 days, with manual reviewing of medical records in case of discordant recording. (3) Assessing the sensitivity and positive predictive value of data from the BPN database. Results: Among the 9736 women receiving at least one blood product dispensed between 01/01/2006 and 12/31/2007 and the 35,779 women who delivered, 233 women (0.65% of deliveries) received at least one RBC unit for post partum hemorrhage. In the BPN database according to the type of hospital stay in our perinatal network (delivery stay only, delivery and post-delivery stays), sensitivity and positive predictive value for RBC transfusion ranged from 61.4% (55.1-67.6) to 67.8% (61.8-73.8) and 82.2% (76.5-87.9) to 83.2% (77.8-88.5), respectively. Linkage of both BPN and EFS databases allowed accurate recording of all but one RBC transfusion. Conclusion: Our approach allowed 100% electronic recording of PPH requiring blood transfusion, making it an important sentinel event of maternal morbidity to assess the perinatal network. (C) 2012 Elsevier Ireland Ltd. All rights reserved.