Author + information
- Received January 26, 2017
- Revision received April 17, 2017
- Accepted April 19, 2017
- Published online June 26, 2017.
- John D. Carroll, MDa,∗ (, )
- Sreekanth Vemulapalli, MDb,
- Dadi Dai, PhDc,
- Roland Matsouaka, PhDd,
- Eugene Blackstone, MDe,
- Fred Edwards, MDf,
- Frederick A. Masoudi, MD, MSPHa,
- Michael Mack, MDg,
- Eric D. Peterson, MD, MPHb,
- David Holmes, MDh,
- John S. Rumsfeld, MD, PhDa,
- E. Murat Tuzcu, MDe and
- Frederick Grover, MDi
- aDivision of Cardiology, Department of Medicine, University of Colorado Denver, Aurora, Colorado
- bDivision of Cardiology, Department of Medicine, Duke University School of Medicine, Durham, North Carolina
- cOutcomes Department, Duke Clinical Research Institute, Duke University, Durham, North Carolina
- dDepartment of Biostatistics and Bioinformatics & Duke Clinical Research Institute, Duke University, Durham, North Carolina
- eDepartment of Cardiovascular Medicine, Heart and Vascular Institute, Cleveland Clinic, Cleveland, Ohio
- fDepartment of Surgery, University of Florida, Jacksonville, Florida
- gBaylor Scott & White Health, Plano, Texas
- hDepartment of Cardiovascular Diseases, Mayo Clinic, Rochester, Minnesota
- iDivision of Cardiothoracic Surgery, Department of Surgery, University of Colorado School of Medicine, Anschutz Medical Campus, Aurora, Colorado
- ↵∗Address for correspondence:
Dr. John D. Carroll, University of Colorado, Anschutz Medical Campus, Mail Stop B132, Leprino Office Building, 12401 East 17th Avenue, Room 524, Aurora, Colorado 80045.
Background Transcatheter aortic valve replacement (TAVR) has been introduced into U.S. clinical practice with efforts to optimize outcomes and minimize the learning curve.
Objectives The goal of this study was to assess the degree to which increasing experience during the introduction of this procedure, separated from other outcome determinants including patient and procedural characteristics, is associated with outcomes.
Methods The authors evaluated the association of hospital TAVR volume and patient outcomes for TAVR by using data from 42,988 commercial procedures conducted at 395 hospitals submitting to the Transcatheter Valve Therapy Registry from 2011 through 2015. Outcomes assessed included adjusted and unadjusted in-hospital major adverse events.
Results Increasing site volume was associated with lower in-hospital risk-adjusted outcomes, including mortality (p < 0.02), vascular complications (p < 0.003), and bleeding (p < 0.001) but was not associated with stroke (p = 0.14). From the first case to the 400th case in the volume–outcome model, risk-adjusted adverse outcomes declined, including mortality (3.57% to 2.15%), bleeding (9.56% to 5.08%), vascular complications (6.11% to 4.20%), and stroke (2.03% to 1.66%). Vascular and bleeding volume–outcome associations were nonlinear with a higher risk of adverse outcomes in the first 100 cases. An association of procedure volume with risk-adjusted outcomes was also seen in the subgroup having transfemoral access.
Conclusions The initial adoption of TAVR into practice in the United States showed that increasing experience was associated with better outcomes. This association, whether deemed a prolonged learning curve or a manifestation of a volume–outcome relationship, suggested that concentrating experience in higher volume heart valve centers might be a means of improving outcomes. (STS/ACC Transcatheter Valve Therapy Registry [TVT Registry]; NCT01737528)
The introduction of transcatheter aortic valve replacement (TAVR) into U.S. clinical practice has been a controlled process, with a major goal to optimize patient outcomes. Due to the technique’s novelty and complexity and lack of other good treatment options for aortic stenosis, high-risk, mostly elderly, patients have been selected for this treatment. Approaches used to achieve rational dispersion and minimize the learning curve for this new technology have included site-selection criteria linked to reimbursement by the U.S. Centers for Medicare & Medicaid Services (CMS), extensive professional labeling by the U.S. Food and Drug Administration, institutional and operator requirements recommended by professional medical societies, and device training plus comprehensive support during the procedures by medical device companies (1–3).
The Transcatheter Valve Therapy (TVT) Registry is the approved national registry organized by the Society of Thoracic Surgeons (STS) and the American College of Cardiology (ACC); all commercial TAVR cases are required to be submitted to this registry for reimbursement. The registry is thus the comprehensive national repository of individual patient data used in the present study (4).
The purpose of this study was to examine the association between procedural experience with TAVR, measured by using cumulative hospital volume and in-hospital risk-unadjusted and risk-adjusted outcomes, in U.S. patients treated with commercially approved devices. Site volume was chosen rather than operator volume given the combined multi-operator approach to TAVR performance involving both cardiology and cardiovascular surgery specialists.
The source of patient data included all cases submitted to the TAVR module of the STS/ACC TVT registry performed from November 2011 through November 2015. This analysis focused on patients undergoing TAVR for severe aortic stenosis of a native valve according to the indication approved by the U.S. Food and Drug Administration.
The TVT registry was developed by the STS and ACC in response to the CMS National Coverage Determination of May 2012 requiring national registry participation of all U.S. TAVR centers (1). The registry’s data elements characterize the patients, the procedure, and subsequent outcomes. Details of the elements in the TVT registry data collection form and definitions can be found at the TVT registry website (5). Participating centers use standardized definitions of data elements (6) to collect clinical information on consecutive TAVR cases using commercially approved devices. Race and ethnicity data are captured by each hospital and are based on U.S. Census Bureau data standards. Data quality checks are implemented at the ACC National Cardiovascular Data Registry data warehouse and Duke Clinical Research Institute to optimize data completeness and accuracy. The Chesapeake Research Review Incorporated institutional review board and the Duke University institutional review board granted a waiver of informed consent for the TVT registry protocol.
The in-hospital outcomes studied were death, vascular complications, bleeding complications, and stroke. These outcomes were selected for their clinical importance and were defined according to standardized consensus-derived criteria, including those of the Valve Academic Research Consortium (7). In-hospital stroke was independently adjudicated at the Duke Clinical Research Institute by at least 2 cardiologists using definitions of the Valve Academic Research Consortium. This process involved review of additional clinical documentation and de-identified source documents as needed.
Case sequence approach to volume–outcomes relationship
Experience, as a potential determinant of outcomes, was assessed for the composite of all U.S. TAVR centers using the case sequence approach rather than stratifying hospitals according to their cumulative case volume. The case sequence approach has been used to assess the volume–outcome relationship for procedures. Specifically, it has been used to assess the learning curve in clinical trials of TAVR, although not in broader contemporary clinical practice (7–10). In the case sequence approach, unadjusted and risk-adjusted outcomes are assessed as a function of an increasing number of procedures performed. This approach is easily comprehensible for a single center and commonly applied in that setting. However, because each U.S. TAVR center may have its own unique relationship between increasing procedure number and outcomes, hierarchical modeling must be used to generate an average event rate for each consecutive case at an average site for a hypothetical “average” patient within the dataset. In the current TVT Registry dataset, most hospitals included in this analysis had completed <65 cases and 25% had completed <30 cases. Thus, the majority of TAVR hospitals started their programs during the period of this analysis and many had small volumes, making a hospital-based analysis of outcomes statistically challenging (11).
In this setting, the case sequence approach has 2 main advantages over a traditional analysis of hospital volume versus outcomes in that: 1) it describes both the learning curve associated with this emerging technology and any stable volume–outcome relationship; and 2) it includes low-volume centers in the analysis without penalizing them for being within their learning curve. Therefore, the case sequence approach was chosen to isolate the association of increasing experience on outcomes after adjustment for other known outcome determinants, including patient characteristics and procedure-related variables that evolved over time.
Case sequence volume was analyzed as a continuous variable for modeling, but to characterize patient and procedural features for descriptive purposes, quartiles of case sequence TAVR volume were used (<30, 31 to 71, 72 to 137, and >138 cases). These groups represented running tallies of case sequence TAVR volume such that individual sites contributed to >1 quartile and served as their own controls. This approach allowed for assessment of average change in patient selection and procedural metrics over case sequence volume curves. Categorical variables are presented as frequencies and percentages, and continuous variables are summarized as medians with interquartile ranges. Comparisons among categorical and continuous variables were performed by using the Pearson chi-square and Kruskal-Wallis tests, respectively. Observed outcome rates were reported per case sequence TAVR volume experience.
To examine the relationship between case sequence TAVR volume and outcomes, generalized linear mixed models were developed by using case sequence TAVR volume as a continuous variable. Restricted cubic splines were used to explore potential nonlinear relationships between case volume and outcomes. Relationships were plotted as curves for case sequence volume versus outcome such that any given slope along the curve represents the rate of change in outcome with increasing procedural experience. A 3-level (patients, operators, and hospitals) hierarchical structure was adopted by using random intercept models with a covariance matrix that accounts for interhospital variability, interoperator variability nested within sites, intrasite clustering of TAVR volume, and case mix. Intrasite clustering of TAVR volume was accounted for in the mixed models through an autoregressive (order 1) covariance matrix which assumes that procedures performed closer in time to each other are likely to be more similar with respect to site-level effects than those performed further apart. This mixed model also accounts for site-specific effects, such as case mix selection. As a result, the data contributed by all TAVR sites in the present analysis necessarily include each site’s commercial “learning curve.” The applied methodology allows for visual estimation of the overall learning curve as well as any stable volume–outcomes relationship.
Analyses were repeated after adjustment of outcomes for their risk factors (i.e., other determinants). (See Online Table 1 for the list of covariates included.) Variables considered for the adjusted models were derived from those initially considered for the in-hospital risk model based on expert consensus (12). For outcomes other than mortality, in which risk models have not yet been finalized, additional covariates for adjustment were chosen on the basis of each endpoint’s clinical relevance.
In addition, the procedure date relative to November 2011 (i.e., the commercial approval of TAVR in the United States) was included as an independent variable. By including a variable for procedure date in the adjusted models (thereby taking into account changing device technology, practice patterns, and procedural technique over time), we sought to isolate the effect of volume on procedural outcomes. Other evolving technology variables, such as sheath size, were also included in the adjusted models.
All the risk factors for each model were first combined into a single risk score before constructing the hierarchical model (12). The risk score was used, rather than individual variables, to simplify the model and decrease the likelihood of nonconvergence (failure of the program to find a model that fits the observed data), which is a particular problem with hierarchical models. To create this score, we performed an ordinary logistic regression model and used the predicted log odds for the outcome rather than the predicted probability because log odds are linear with respect to the model’s variables. The risk score was then added as a single independent variable in the subsequent hierarchical model. All analyses were performed at the Duke Clinical Research Institute by using SAS software version 9.3 (SAS Institute, Inc., Cary, North Carolina).
Between November 2011 and November 2015, a total of 47,270 TAVR procedures were performed by 1,927 operators at 395 sites and entered into the STS/ACC TVT registry. We excluded patients receiving TAVR for primary aortic insufficiency (n = 274), valve-in-valve procedures for failed bioprosthetics (n = 2,839), bicuspid valves (n = 862), emergent or salvage procedures (n = 66), and repeat TAVR procedures (n = 241). After applying these criteria, a total of 42,988 TAVR procedures were available for analysis (Figure 1). The hospital procedural volume distribution of these sites is shown in Figure 2. The median cumulative number of TAVRs performed per site during the study period was 80 (interquartile range: 39 to 154). The mean cumulative volume per site was 108.8 ± 96.0. New sites were opened throughout the study period.
Although all 395 hospital sites contributed cases to the first category of case sequence TAVR volume (Table 1), only 119 hospital sites contributed cases to the highest volume category. The mean date of performed TAVR was progressively later with each category.
Per-patient characteristics (Table 2), the overall patient population included predominantly older individuals with multiple comorbid illnesses that made them eligible for TAVR based on their high-to-prohibitive risk for surgical aortic valve replacement.
Although there were some statistically significant patient differences in baseline characteristics among the case sequence volume categories, there were no consistent trends as a function of procedural volume. If anything, patients undergoing TAVR in the highest sequential volume category had an even greater burden of high-risk features than those in the lowest volume category. For example, the fourth volume category had higher STS-predicted risk of mortality, increased frequency of New York Heart Association functional class III to IV heart failure symptoms, longer 5-meter walk times, increased frequency of atrial fibrillation, and a greater proportion of patients with severely lowered ejection fraction.
The final population of 42,988 patients included 30,566 patients who underwent TAVR using femoral artery access, with the remaining 12,422 cases having an alternative access for valve delivery system insertion. The most common alternative access was transapical, with small numbers of cases in the upper 2 volume categories and only 7 sites performing >100 transapical procedures.
TAVRs were performed electively in 91.1%, using general anesthesia in 92.1% and via a transfemoral approach in 71.1% (Table 3). Several procedural characteristics were associated with higher volume category. For example, comparing the lowest volume category with the highest, there were progressively higher rates of transfemoral artery access (68.6% to 82.6%), less use of general anesthesia (98.0% to 83.6%), more use of percutaneous valve sheath access technique (24.3% to 66.9%), and a smaller femoral sheath size in those with transfemoral access (21.6-F to 17.8-F).
Overall, unadjusted in-hospital outcomes included: mortality (4.0%), vascular complications (7.1%), bleeding (8.6%), and stroke (2.0%) (Table 4). In higher volume categories, unadjusted rates were lower for these specific outcomes.
The Central Illustration displays the case sequence volume–outcome association for both unadjusted outcomes and risk-adjusted outcomes (i.e., adjusted for patient and procedural characteristics). In the overall population, there was a significant linear association between increasing TAVR volume and in-hospital mortality both before and after adjustment for baseline patient and procedural characteristics. A significant nonlinear association was recorded between increasing TAVR volume and in-hospital vascular complications before and a linear association after adjustment for baseline patient and procedural characteristics. There was a significant nonlinear association between increasing TAVR volume and in-hospital bleeding both before and after adjustment for baseline patient and procedural characteristics. In contrast, although a significant linear association was noted between TAVR volume and in-hospital stroke, this association was lost after adjustment for patient and procedural characteristics.
The clinically important absolute differences in outcomes from the plots of case sequence volume as a continuous variable can also be quantitatively illustrated when comparing cases early and later in the development of experience. The modeled rates of mortality for the first case versus the 400th case were 3.57% (95% confidence interval [CI]: 3.14% to 4.07%) versus 2.15% (95% CI: 1.52% to 3.02%). The modeled rates for vascular complications for the first case versus the 400th case were 6.11% (95% CI: 5.40% to 6.89%) versus 4.20% (95% CI: 3.27% to 5.38%). For bleeding complications, the modeled rates for the first versus the 400th case were 9.59% (95% CI: 8.65% to 10.57%) versus 5.08% (95% CI: 4.06% to 6.35%). These rates were calculated for an “average” patient who carries the average characteristics of the overall study population.
Figure 3 displays the volume–outcome association for unadjusted and risk-adjusted outcomes for those 30,566 patients having transfemoral artery access for TAVR delivery. There were statistically significant and clinically important associations between volume and outcomes for vascular and bleeding complications, both before and after adjustments for patient and procedural characteristics. There was an association between volume and mortality that became statistically insignificant after adjustments for patient and procedural complications. The broader confidence limits in Figure 3 at high volumes reflect the fact that fewer sites have completed this number of TAVR cases and the low absolute rate of complications. Volume and stroke (plot not shown) were not associated, neither before nor after adjustments for patient and procedural characteristics in those with transfemoral access.
Analysis of the volume–outcome association for those having alternative access was deemed to be invalid because of the few sites performing >100 cases and there not being a robust representation of outcomes at high case sequences.
Patient outcomes from transcatheter treatment of severe aortic stenosis improved with experience as quantified by using case sequence volume. Using data from the STS/ACC TVT registry on 42,988 patients from 395 U.S. hospital sites, we found that with increasing TAVR procedural experience, there was a statistically significant and clinically important decline in the risk for major adverse outcomes for patients treated in U.S. clinical practice. After adjustment for patient factors, date of procedure, and specific procedural characteristics (including device iterations), the inverse associations persisted between increasing case volume and lower in-hospital mortality, vascular complications, and bleeding. This association was most pronounced during the first 100 cases, indicating the effect of an early learning curve for TAVR. Beyond these initial 100 cases, procedural risk continued to decline but at a more gradual rate.
The absolute differences in these risk-adjusted key outcomes associated with experience were of a clinically important magnitude. For example, the mortality rate calculated for an average patient who carried the average characteristics of the overall study population declined from 3.57% as the first case to 2.15% as the 400th case. The modeled rate of bleeding, vascular complications, and stroke exhibited similar trends, although stroke was not statistically significantly different.
The results reported here were for the composite of all sites performing TAVR in the United States and represent averages by using the case sequence approach. The composite learning curves represent that of the entire community. Classifying cumulative volume on a hospital or individual operator level had the disadvantage that most TAVR sites in the United States had low volumes, with the median cumulative hospital volume being only 80 cases, and many sites began performing TAVR during the study period. As pointed out in an editorial addressing the challenges of assessing hospital-specific outcomes, estimates of hospital performance are increasingly imprecise with lower procedural volumes (8).
The observation that higher procedural experience for TAVR was associated with better in-hospital outcomes did not prove causality but strongly implied that there was a substantial initial and ongoing learning process to optimally provide patients this novel treatment for severe aortic stenosis. This outcome was expected because this treatment is a new paradigm in several dimensions. TAVR requires the use of a complex first-in-class medical device technology and requires interdisciplinary collaboration that might not have existed before the institution of valve programs. Furthermore, the patient population selected for TAVR is typically older, frail, burdened with other medical conditions, and at high (if not prohibitive) risk for surgical valve replacement.
Volume–outcome associations and the uniqueness of TAVR
This study of the association between volume and outcome in TAVR differed in important ways from the existing literature of the volume–outcome relationship for well-established procedures and operations with years of experience in optimizing techniques and mature technology (13–15). Such investigations likely occurred after completion of the early learning period and focused on whether there is a stable volume–outcome relationship. TAVR is a new, rapidly evolving technology with numerous versions of the delivery systems and valves that are manufactured for use in the United States by 2 companies. Our investigation used data from this period of rapid technology evolution as well as during a period of rapid site expansion and training. Therefore, it is unrealistic to separate the initial learning curve and the volume–outcome relationship in TAVR’s introduction to U.S. clinical practice. Furthermore, there is value in understanding and quantifying the entire learning process with the introduction of novel technologies into the field of transcatheter treatment of valvular heart disease.
There have been numerous efforts in the United States to attenuate the impact of learning at sites and achieve results typical of an established experienced center. Three major efforts were a standardized site selection process, professional educational conferences, and extensive external clinical support for TAVR sites from valve manufacturers. Site selection requirements are outlined in both the National Coverage Determination from the CMS as well as professional society recommendations (1–3).
One previous report of U.S. hospital volume–outcome relationships in TAVR used National Inpatient Sample administrative data from 7,635 patients during a 1-year period (16). This smaller study did not show a relationship of annual hospital volume to TAVR mortality and morbidity. However, risk adjustment was performed by using a multivariate regression analysis based on International Classification of Diseases-Ninth Revision-Clinical Modification codes because patient-level data of key variables were not available. Furthermore, this study classified annual volume at the hospital level and did not consider the effect of a learning curve that might have masked any stable volume–outcome relationship. The present analysis showed that even after accounting for the site initial learning curve, there seemed to be a stable volume–outcome relationship.
Implications and future trends
U.S. health care policy must consider the major priorities of quality, access, and costs. TAVR represents an important advancement in the treatment of aortic stenosis, a condition associated with aging. TAVR volume is expected to increase with the major influx of the baby boomer generation with its advancing age and will further strain the health care budget of many countries (17). This study focused on answering the question of whether an association existed between the quality of a procedure’s outcome and the experience of the teams and institutions performing TAVR. These results will help inform both professional societies currently updating recommendations of institutional and operator requirements for TAVR as well as the national coverage decision for TAVR.
The number of sites needed in the United States to balance access and quality of TAVR outcomes cannot be definitively determined from the present study. There is an outcome “cost” associated with lower volumes as institutions and personnel learn how to deliver this novel treatment. The impact of additional sites in already saturated regions of the country has other potential detrimental effects. Historically, the proliferation of hospitals performing bypass surgery results in fewer high-volume sites (18). Urban hospital “clusters” in the United States have recently been characterized by consolidation and designation of specialized hospital sites for high-complexity services such as trauma and stroke intervention (19). Regionalization of specialized procedures has been recommended by policy groups and supported by industry purchasers of medical insurance, but key questions need to be addressed, including which procedures, what are the drawbacks (including potential barriers to access not related to geography), and how to define a center of excellence (20).
Patients, families, and referring clinicians need to make informed decisions regarding where they seek their medical care, especially when it involves procedures with high complexity and major risks. Recent attention to the consumer as well as the patient’s perspective of hospital and operator transparency of procedure volumes and outcomes further solidify the need for objective analyses of outcomes and their determinants, including volume (21,22).
It was not possible in the present study to extrapolate whether a clinically and statistically significant volume–outcome relationship will persist in the future after sites and physicians have achieved more experience. Technology and technique improvements may further increase procedural safety, but the majority of patients undergoing TAVR will continue to be elderly, with multiple comorbid conditions, due to the chief etiology of aortic stenosis being degenerative with onset at an advanced age (17,23,24).
This was a retrospective, observational study. Clinical outcomes in TAVR have multiple determinants, and adjustment for patient and procedural factors can only account for measured factors. The TVT registry captures a multitude of patient and procedural characteristics as well as the type of delivery system and valve technologies. Many variables were used in the risk adjustment models, but the potential exists for unmeasured confounding factors, such as social factors and factors not included in first iterations of models (e.g., frailty) (25,26).
Investigations of volume–outcome associations have multiple challenges, including the complexity of the analytic approaches, the selection of outcomes, and the boundaries of significance between any of the specific endpoints. The TVT registry is well suited for this type of study with the ability to capture patient-level relevant data elements from tens of thousands of patients with standardized definitions for clinical end points.
In-hospital outcomes were assessed and analyzed because of the large number of patients submitted to the TVT registry and the completeness of the data elements. Outcomes at 30 days and 1 year are also important, may have other determinants, and need to be separately analyzed.
This analysis does not include the diversity of other potentially complex applications of TAVR technology, such as repeat procedures, valve-in-valve, urgent procedures, use in other variations of aortic valve disease, and off-label use for mitral valve disease. Higher volume hospital sites are more likely to have encountered these cases with the consequence of broadening a site’s ability to treat all patients rather than selecting more straightforward cases.
This analysis is limited to cases performed with the commercially approved technology, and some hospital sites have additional case volume from investigative studies. Sites participating in early clinical trials were more likely to have completed the early learning curve by the time TAVR was introduced commercially.
Using the case sequence approach, we presented evidence showing a statistically significant and clinically important inverse association between increasing TAVR volume and reductions in TAVR procedural mortality and morbidity. These results can inform decisions regarding optimizing this transformational, expensive, and rapidly growing treatment for aortic valve disease in the U.S. health care system. Sustaining the specialized, experienced center model for heart valve therapy is supported by these data.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Operator experience has a substantial impact on clinical outcomes in patients undergoing transcatheter aortic valve replacement. There is typically a steep learning curve over the course of approximately 100 cases. With higher procedure volumes, outcomes continue to improve and the volume-outcome relationship becomes more stable, as reflected in lower rates of major in-hospital adverse events including mortality, bleeding, and vascular complications.
TRANSLATIONAL OUTLOOK: Further studies of the association of volume with outcomes will be needed as TAVR instrumentation matures, new operators are trained, processes of care evolve, and dispersion of the technology expands across an aging population with greater frailty and a heavier burden of comorbidities.
For a supplemental table, please see the online version of this article.
Dr. Carroll is a local site investigator in transcatheter aortic valve replacement studies sponsored by Edwards Lifesciences and Medtronic. Dr. Vemulapalli has received grants from the American College of Cardiology, the Society of Thoracic Surgeons, the Patient-Centered Outcomes Research Institute, Abbott Vascular, and the Agency for Healthcare Quality and Research; and has served as a consultant for Novella. Dr. Blackstone has received a grant through MedStar to independently analyze PARTNER trial data provided by Edwards Lifesciences for investigator-initiated research as one of the branches of the PARTNER Publication Office. Dr. Masoudi served as chief science officer of the National Cardiovascular Data Registry programs. Dr. Mack was a trial co–principal investigator for Edwards Lifesciences and Abbott Vascular; and was on the trial steering committee for Medtronic. Dr. Rumsfeld was chief science officer for the National Cardiovascular Data Registry during the conduct of this study. Drs. Carroll, Mack, Holmes, Tuzcu, and Grover are members of the Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry Steering Committee. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- American College of Cardiology
- Centers for Medicare & Medicaid Services
- Society of Thoracic Surgeons
- transcatheter aortic valve replacement
- transcatheter valve therapy
- Received January 26, 2017.
- Revision received April 17, 2017.
- Accepted April 19, 2017.
- 2017 American College of Cardiology Foundation
- ↵National Coverage Determination (NCD) for Transcatheter Aortic Valve Replacement (TAVR) (20.32). Available at: https://www.cms.gov/medicare-coverage-database/details/ncd-details.aspx. Accessed January 17, 2017.
- Tommaso C.L.,
- Bolman R.M.,
- Feldman T.,
- et al.
- Holmes D.R. Jr..,
- Mack M.J.,
- Kaul S.,
- et al.
- Carroll J.D.,
- Shuren J.,
- Jensen T.S.,
- et al.
- ↵STS-ACC TVT Registry data collection forms and coder’s data dictionary. Available at: https://www.ncdr.com/webncdr/tvt/home/data-collection. Accessed January 17, 2017.
- Kappetein A.P.,
- Head S.J.,
- Genereux P.,
- et al.
- Hess C.N.,
- Peterson E.D.,
- Neely M.L.,
- et al.
- Alli O.,
- Rihal C.S.,
- Suri R.M.,
- et al.
- Suri R.M.,
- Minha S.,
- Alli O.,
- et al.
- Desai N.R.,
- Joynt K.E.
- Strom J.B.,
- Wimmer N.J.,
- Wasfy J.H.,
- Kennedy K.,
- Yeh R.W.
- Freeman J.V.,
- Wang Y.,
- Curtis J.P.,
- Heidedreich P.A.,
- Hlatky M.A.
- de Biasi A.R.,
- Paul S.,
- Nasar A.,
- Girardi L.N.
- Carroll J.D.
- Wilson C.T.,
- Fisher E.S.,
- Welch H.G.,
- Slewers A.E.,
- Lucas F.L.
- Like R.D.,
- Luke T.,
- Muller N.
- Chhabra K.R.,
- Dimick J.B.
- Frankel N.Z.
- Osnabrugge R.L.J.,
- Mylotte D.,
- Head S.J.,
- et al.
- Havranek E.P.,
- Mujahid M.S.,
- Barr D.A.,
- et al.
- O’Brien S.M.,
- Cohen D.J.,
- Rumsfeld J.S.,
- et al.