Persistently low inferior vena cava filter retrieval rates in a population-based cohort

Background: Practice patterns associated with inferior vena cava (IVC) filter placement have seen considerable variation in the last decade. We used a statewide administrative database to examine trends in IVC filter placement and retrieval in the general population. Methods: We reviewed Florida state inpatient and ambulatory surgery databases from 2004 to 2014. International Classification of Diseases, Ninth Revision diagnosis and procedure codes and Current Procedural Terminology codes were searched for patients undergoing inpatient or outpatient IVC filter placement, and each patient was longitudinally tracked to the time of inpatient or outpatient filter retrieval. For inpatient filter placements, associated diagnoses were reviewed to identify indications for placement. Univariate and multivariate logistic regression models were constructed to identify factors associated with improved retrieval rates. Results: During the 11-year period, 131,791 IVC filter placements were identified, with a 50% increase from 2004 to 2010 and a 24% decline from 2010 to 2014. Median age at filter placement was 71 years (interquartile range, 57-81 years). Mean follow-up after filter placement was 17.3 +/- 25.5 months. Only 8637 filters (6.6%) were retrieved. The annual retrieval rate trended upward, from 3.4% in 2004 to 8.5% in 2013 (P < .001). Median filter dwell time was 96.5 days (interquartile range, 44-178 days). Diagnoses associated with filter placement included venous thromboembolism (75.9%), trauma (35.0%), hemorrhage (29.9%), malignant disease (29.4%), and stroke (5.1%). Retrieval rates were highest in younger patients (34.0% in patients younger than 20 years) and lowest in Medicare patients (2.5%). In a multivariate logistic regression model, Medicare was associated with decreased retrieval rates (odds ratio, 0.33; 95% confidence interval, 0.31-0.35; P < .001) after adjusting for age and associated diagnoses. Weaker risk factors included increased age, white race, and diagnoses of deep venous thrombosis, pulmonary embolism, and malignant disease. A trauma diagnosis was associated with improved retrieval. To further investigate the Medicare effect, a propensity score-matched model was created to better account for confounding effects. In this model, Medicare persisted as a risk factor for decreased filter retrieval (odds ratio, 0.43; 95% confidence interval, 0.40-0.46; P < .001). Conclusions: IVC filter placements, after a substantial increase between 2004 and 2010, have been declining since 2010. Retrieval rates in the general population are steadily improving but continue to lag behind those described in center-specific literature. Increased age and Medicare as the primary payer are the strongest risk factors for lack of filter retrieval. Widespread improvements on a national scale are needed to improve the appropriateness of filter placements and to enhance filter retrieval rates.