PB1110 - Implementation of International Society on Thrombosis and Haemostasis definitions of major and clinically relevant non-major bleed events in electronic health records

Background: International Society on Thrombosis and Haemostasis (ISTH) definitions for major and clinically relevant non-major (CRNM) bleeding are widely used standards for reporting bleeding outcomes in clinical trials. However, bleeding events in electronic health record (EHR) datasets are not recorded using these definitions. To enable comparison with clinical trials, methods are required to identify ISTH-defined bleeding events in RWD sources.

Aims: To create a novel method to identify ISTH-defined bleeding events within an EHR data set.

Methods: Three sets of diagnosis codes corresponding to different bleeding event types were curated using expert medical knowledge. The existence of one of these codes in the patient record was a necessary condition for the identification of an ISTH major or CRNM bleed. Additional contextual information around the time of the diagnosis event was required to differentiate major and CRNM bleeds. This included evidence of blood loss through haemoglobin measurements, transfusion codes, death information, diagnosis setting (inpatient vs outpatient), and diagnosis position. A systematic process for assessing all possible combinations of these variables was developed. All combinations were assessed by three medical doctors and determined as defining either an ISTH major or CRNM bleed.

Results: An algorithm was created and implemented in the US Optum® EHR. Calculated incidence rates for ISTH major and CRNM bleeding in patients with atrial fibrillation taking anticoagulants are consistent with those in the medical literature. This algorithm was used to replicate the results of the PACIFIC-AF (NCT04218266).

Conclusion(s): The proposed algorithm may identify bleeding events in EHR data sources equivalent to those commonly investigated in clinical trials. This algorithm could facilitate comparison between the bleeding outcome frequencies recorded in clinical trials and RWD. Validation of this code algorithm is in progress. The process of programmatic implementation of bleeding definitions has the potential to improve the quality of standardization efforts.