Can we use routinely collected outcome data for a nationwide trial on venous thromboembolism prophylaxis following primary joint replacement? A feasibility study

Background
UK hospitals nationally report venous thromboembolism (VTE) within 90-days of hospital admission, with Hospital Acquired Thrombosis (HAT) registers at each centre used for this. We assessed the accuracy of our HAT register in identifying VTE following primary THR and TKR.
Methods
We assessed 982 elective admissions for primary THR and TKR at a large tertiary centre during 2018. The primary outcome was any VTE (DVT and/or PE) within 90days post-surgery. VTEs were identified by systematically searching hospital databases (including discharge and outpatient letters, readmissions, emergency department visits, and imaging) for every patient. VTEs were also collected using the HAT database at our centre, which is maintained regularly by a specialist nursing team and used to report VTEs nationally. Diagnostic test characteristics were assessed for HAT in identifying VTEs compared to the gold standard (i.e. VTE’s from the hospital databases).
Results
The prevalence of VTE was 2.7% (n=27), with 20 VTEs identified by HAT. The accuracy of HAT in identifying VTEs were: sensitivity=74.1% (95% CI=53.7-88.9%), specificity=100% (CI=99.6-100%), positive predictive value=100% (CI=83.2-100%), negative predictive value=99.3% (CI=98.5-99.7%).
Conclusions
One-quarter of VTEs occurring after THR and TKR were not identified by the HAT register. These cases would be missing when our hospitals data is sent for national VTE reporting, and therefore would have substantial implications if HAT were primarily used to identify VTEs in a trial. Further work is needed to improve the accuracy of HAT VTE reporting before this could be relied upon in this setting.