Abstract Body (Do not enter title and authors here): Background: Inferior vena cava (IVC) filters are commonly implanted in patients with venous thromboembolism (VTE) who are unable to receive anticoagulation, to protect against clot migration to the heart. With prolonged implantation, IVC filters are associated with complications: device fracture, migration, penetration into adjacent organs and worsened VTE. Two Federal Drug Administration advisories and multi-society guidelines have emphasized the importance of timely retrieval but national retrieval rates remain low (<30%).

Aims: To quantify hospital-level variation in IVC filter retrieval, describe time-based retrieval rates and examine factors associated with retrieval.

Methods: IVC Filter implantation for all patients >= 18 years was identified in the 2016-2020 inpatient and outpatient 100% limited data set Medicare files using claims codes. Hospital information for the implanting facility was cross-linked from the American Hospital Association and Healthcare Cost Report Information System files, including identifying information, teaching status, census location, operating margin and % uncompensated care. IVC filter retrieval rates and time to retrieval from implantation was calculated. Bayesian hospital profiling methods were used to quantify 1-year retrieval rate for all U.S. facilities, adjusting for patient factors – demographics, diagnostic indications and comorbidities.

Results: Among Medicare beneficiaries, there were 140,481 IVC Filter implantations across 2,850 facilities. Excluding patients who died within 90 days of implantation (25.7%), retrieval rates at 3 months, 1 year and anytime were 7.9%, 18.7% and 20.0% respectively. IVC Filter retrieval within 1 year varied significantly at the facility-level, from 0-100%. Focusing on facilities with at least 13 IVC filter implantations each year (top 25%ile volume), 1-year retrieval ranged from 0 to 74.5%. Higher 1-year retrieval was seen among higher implantation volume (12.4% bottom quartile, 20.5% top quartile), teaching (21.1% teaching vs. 16.9% nonteaching) and non-safety net (21.1% low uncompensated care, 15.0% high uncompensated care) hospitals. Retrieval rates did not vary significantly by hospital operating margin or rurality.

Conclusion(s): There is low overall IVC filter retrieval in the United States with large underlying facility-level variation. Focused examination of high-performing facilities could yield insights on how to improve device retrieval nationally.