Assessing the Accuracy of Reporting of Maternal Red Blood Cell Transfusion at Birth Reported in Routinely Collected Hospital Data.

Abstract

Introduction Hospital administrative data collections have been used to describe transfusion practice, particularly in relation to the maternity population. Knowledge of the accuracy of this data is important in order to interpret the results of such studies. The aim of this study was to compare the accuracy of reporting of red cell transfusion around childbirth within hospital data with data submitted by hospital blood banks. Methods Linked hospital and birth data from New South Wales, Australia, between June 2006 and December 2010 were used to identify blood transfusions occurring at delivery. This reporting was compared with the gold standard of blood pack level information submitted by hospital blood banks, and sensitivity, specificity, and positive and negative predictive values calculated. Reporting related to quantity and timing of transfusion were also considered. Results Data were available for 235,796 births, with blood bank data identifying that 2.0% of received a blood transfusion. Overall the sensitivity of hospital data for identifying transfusion was 84.8% (95% CI 83.7%, 85.8%) with specificity 99.9% (99.9%, 99.9%). Sensitivity was better for births involving a postpartum haemorrhage [Sn 90.9% (89.9%, 91.9%)], and poorer for births in regional hospitals [Sn 78.8% (76.0%, 81.5%)]. Almost all (96%) transfusions of 10 or more units were identified in hospital data, and there was no difference in reporting depending on whether the transfusion was on the baby's date of birth or not. Discussion The reliability of hospital reporting of transfusion in maternity patients is high, however with some underreporting of cases.