Facility-Level Variation Underlying Low Inferior Vena Cava Filter Retrieval in the United States

OBJECTIVES: Timely retrieval of inferior vena cava (IVC) filters is recommended to reduce complications and optimize outcomes. This study aims to quantify facility-level variation in risk-adjusted IVC filter retrieval across US hospitals and to identify patient- and hospital-level factors associated with nonretrieval. METHODS: Medicare beneficiaries undergoing IVC filter implantation were identified in the 100% claims files for years 2016 to 2020. Facility-level variation in device retrieval was quantified using Bayesian hospital profiling. Patient- and hospital-level factors associated with nonretrieval were assessed using logistic regression, adjusting for diagnostic indication, comorbidities, and implantation year. RESULTS: IVC filters were implanted in 119,613 Medicare beneficiaries across 2,485 facilities. Retrieval rates were low: median 6.2% within 3 months and 14.8% within 1 year. Excluding deaths within 3 months (30.2%), retrieval ranged from 0% to 100% across facilities. Among high-volume hospitals (top 25th percentile, implanting ≥13 filters per year), 1-year risk-adjusted retrieval ranged from 0% to 74.5%, mean 20% ± 14.2% (positive skew 0.95). Patient factors associated with IVC filter nonretrieval included age > 80 years (odds ratio 2.98, 95% confidence interval [2.73-3.24]), Black race (1.62, [1.51-1.72]), and Hispanic ethnicity (1.45, [1.16-1.80]). Among hospital factors, nonteaching (1.45 [1.37-1.53]), small bed size (1.37 [1.24-1.50]), and safety-net (1.42 [1.34-1.50]) facilities were strongly associated with IVC filter nonretrieval. DISCUSSION: High mortality within 3 months of IVC filter implantation suggests opportunity to improve patient selection and, potentially, device type choice. There is large facility-level variance underlying low aggregate IVC filter retrieval nationally; a focus on standardizing device surveillance and identifying best practices from high-performing facilities is warranted.