Author + information
- Received April 4, 2018
- Revision received June 21, 2018
- Accepted July 3, 2018
- Published online October 17, 2018.
- Harun Kundi, MDa,
- Jordan B. Strom, MD, MSca,
- Linda R. Valsdottir, MSa,
- Sammy Elmariah, MDb,
- Jeffrey J. Popma, MDa,
- Changyu Shen, PhDa and
- Robert W. Yeh, MD, MSca,∗ ()
- aRichard A. and Susan F. Smith Center for Outcomes Research in Cardiology, Division of Cardiovascular Medicine, Beth Israel Deaconess Medical Center, Boston, Massachusetts
- bCardiology Division, Department of Medicine, Harvard Medical School, Massachusetts General Hospital, Boston, Massachusetts
- ↵∗Address for correspondence:
Dr. Robert W. Yeh, Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology, Beth Israel Deaconess Medical Center, 375 Longwood Avenue, Fourth Floor, Boston, Massachusetts 02215.
Objectives This study sought to evaluate the trends in isolated surgical aortic valve replacement (SAVR) procedures across hospitals with different transcatheter aortic valve replacement (TAVR) volumes among Medicare beneficiaries.
Background The volume of TAVR has increased in the United States since its approval, now exceeding that of isolated SAVR.
Methods Hospitalizations of adults (≥18 years) with International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes for SAVR (35.21 or 35.22) or TAVR (35.05 or 35.06) who were included in the Medicare Provider Analysis and Review database between January 1, 2011, and December 31, 2014, were included. Trends in isolated SAVR patient characteristics, procedural volumes, and outcomes by quartile (Q) of hospital-level TAVR use were assessed over the study period.
Results A total of 37,705 isolated SAVR procedures were analyzed for the study. The annual volume of isolated SAVR procedures decreased in hospitals performing the largest number of TAVR procedures (Q3: 1,557 in 2011 to 1,391 in 2014; and Q4: 2,607 in 2011 to 1,791 in 2014). Thirty-day and 1-year mortality after SAVR also declined over the study period in hospitals with the largest TAVR volume (annual change rate in mortality for Q3: −16.4%, p < 0.001; Q4: −20.8%, p < 0.001).
Conclusions The advent of TAVR was associated with a reduction in isolated SAVR volumes, a decrease in comorbidities among patients undergoing SAVR, and corresponding reductions in observed short- and long-term SAVR mortality among hospitals performing the greatest number of TAVRs.
Symptomatic severe aortic stenosis is the most common indication for aortic valve replacement (AVR) in the elderly population in the United States (1). Whereas surgical aortic valve replacement (SAVR) has been the traditional treatment option for decades (2,3), more recently transcatheter aortic valve replacement (TAVR) has become a widely used alternative to surgery based on clinical trials suggesting comparable efficacy and safety profiles among patients at intermediate or high surgical risk (4–8). The volume of TAVR has increased rapidly since its approval and now exceeds that of isolated SAVR in the United States (9). Despite this, SAVR volume has not decreased after the advent of TAVR in the United States and Europe, suggesting that TAVR has broadened eligibility for AVR rather than displacing SAVR as an alternative treatment (10,11). It is unknown, however, whether this trend is uniform across all hospitals, and whether there is an impact of hospital TAVR volume on the patient characteristics, procedural volumes, and outcomes of isolated SAVR.
We therefore sought to evaluate the trends in isolated SAVR procedures across hospitals with different TAVR volumes among Medicare beneficiaries to evaluate the impact of hospital TAVR volume on: 1) isolated SAVR volume; 2) SAVR patient characteristics; and 3) outcomes of patients undergoing isolated SAVR.
Hospitalizations of adult patients (≥18 years) with International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes for SAVR (35.21 or 35.22) or TAVR (35.05 or 35.06) were identified within the Centers for Medicare & Medicaid Services Medicare Provider Analysis and Review database between January 1, 2011, and December 31, 2014, and included in the analysis. The Medicare Provider Analysis and Review database contains inpatient billing claims for 100% of Medicare fee-for-service beneficiaries and has been used extensively to study national patterns of procedure use in the United States (12,13).
TAVR-performing hospitals were defined as those billing for at least 1 TAVR procedure annually during the study period. Quartiles of hospital-level TAVR volume were defined according to the total volume of TAVR procedures. During the same time period, SAVR-only hospitals were defined as those performing less than 1 TAVR procedure annually while performing at least 1 isolated SAVR annually. To identify a population of patients with severe aortic valve disease and exclude those patients who might have undergone AVR for more moderate disease in conjunction with other procedures, patients who underwent concomitant valve or coronary artery bypass graft surgeries were excluded from the study (Online Table 1).
Covariates and outcomes
Possible risk factors for mortality were defined based on prior studies (14–16) and clinical knowledge, and corresponding International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes were used to identify surrogates for these variables in the dataset (Online Table 1). All covariates were ascertained from secondary diagnosis codes that were coded as “present on admission” during the index hospitalization as well as from principal and secondary diagnosis codes from all hospitalizations for up to 1 year before the hospitalization for AVR (Online Table 1). The Charlson comorbidity index as defined in Online Table 2 was also calculated for each patient to estimate 10-year survival rates (14).
The primary outcome was 30-day all-cause mortality among isolated SAVR patients, identified through linkage to the Medicare Denominator File. The rate of 1-year all-cause mortality among patients with SAVR was evaluated as a secondary endpoint.
Continuous variables are presented as mean ± SD and categorical variables as frequencies with percentages. Covariates in patients undergoing isolated SAVR were assessed for changes over time using 1-way analysis of variance for continuous variables and the Pearson chi-square test for categorical variables.
Isolated SAVR volume trends in non-TAVR and TAVR-performing hospitals by quartile of TAVR use were plotted over the study period. For each TAVR quartile, means of isolated SAVR, TAVR, and combined volume trends were plotted. Within each year, 30-day and 1-year all-cause mortality rates were compared between non-TAVR hospitals and the 4 quartiles (Q1, Q2, Q3, Q4) of TAVR-performing hospitals using chi-squared statistics. A trend test was used to evaluate for changes in unadjusted within-hospital mortality rates of patients undergoing isolated SAVR over time by quartile of TAVR volume. All statistical analyses were performed in STATA version 15.0 (STATA Corporation, College Station, Texas), using a 2-tailed p value <0.05 to define significance.
A total of 1,165 U.S. hospitals performed at least 1 isolated SAVR during the study period, of which 639 (54.8%) did not perform a single TAVR. One hundred of these SAVR-only hospitals (15.6%) were excluded from analysis because they did not perform at least 1 case annually, resulting in a final analytic sample of 539 SAVR-only hospitals. There were 526 hospitals (45.2%) that performed at least 1 TAVR during the study period. A total of 441 of these TAVR hospitals (83.8%) were excluded from analysis because they did not perform at least 1 TAVR case annually, resulting in a final analytic sample of 85 TAVR-performing hospitals (Figure 1).
Among those undergoing SAVR in non-TAVR hospitals, 7 covariates including age, chronic heart failure, coronary artery disease, coronary artery bypass surgery, peripheral vascular disease, chronic kidney disease, and chronic obstructive pulmonary disease consistently and significantly declined over the study period (Table 1). In addition to those 7 covariates, 5 more covariates including diabetes mellitus, prior myocardial infarction, prior percutaneous coronary intervention, cerebrovascular disease, and home oxygen use also consistently and significantly declined in TAVR-performing hospitals over the study period among patients undergoing SAVR (Table 2).
Trends in rates of comorbidities for isolated SAVR patients by quartile of TAVR performance are noted in Online Tables 3 to 6. With the exception of peripheral vascular disease and coronary artery disease, there was no change in rates of comorbidities over time in hospitals performing the lowest numbers of TAVR (Q1). With increasing hospital TAVR volume, a greater percentage of comorbidities in patients undergoing SAVR were observed to decline over the study period, without an observed threshold effect. Although there was no difference in estimated 10-year survival rates based on the Charlson comorbidity index in Q1 hospitals (annual change rate = 0.99%; p = 0.47), it improved most in hospitals with the highest TAVR volume over the study period (Q3: annual change rate = 6.9%, p < 0.001; Q4: annual change rate = 10.1%, p < 0.001).
The volume of isolated SAVR procedures decreased only in hospitals performing the largest number of TAVR procedures (Q3: 1,557 to 1,391 SAVR hospitalizations per year; Q4: 2,607 to 1,791 SAVR hospitalizations per year) (Figure 2). Individual hospital trends according to quartiles of TAVR volume demonstrated wide interhospital variation in the volume of SAVR procedures performed annually (Online Figure 1). The mean TAVR volume exceeded that of SAVR in all hospitals except those in the lowest quartile of TAVR performance, with the crossover point occurring as early as 2012 in the highest-volume TAVR-performing hospitals (Online Figure 2).
There were no significant changes in 30-day all-cause mortality of isolated SAVR patients over the study period for hospitals in the lowest quartiles of TAVR volume (non-TAVR: −4.6%, p = 0.06; Q1: 10.7%, p = 0.08; Q2: −7.2%, p = 0.33) (Table 3). By contrast, 30-day all-cause mortality significantly declined for hospitals in the highest TAVR volume quartiles (Q3: −16.4%, p < 0.001; Q4: −20.8, p < 0.001) with the largest annual rate of decline seen in Q4 (Figure 3). With the exception of Q1, the 1-year all-cause mortality rates for SAVR patients significantly declined in all quartiles over the study period (p < 0.001; Q2: −16.4%; Q3: −15.7%; Q4: −16.3%).
Because we excluded 83.8% of TAVR-performing hospitals in our primary analysis, we performed a sensitivity analysis without excluding hospitals that performed less than 1 TAVR procedure annually, but that performed at least 1 such procedure during the study period. In this analysis, there were 488 TAVR-performing hospitals that performed a total of 54,603 SAVR procedures. As shown in Online Tables 7 to 9 and Online Figures 3 and 4, the results of this analysis were consistent with our primary findings.
Although TAVR has now eclipsed SAVR worldwide as the predominant treatment for severe symptomatic aortic stenosis, knowledge about how the establishment and growth of TAVR programs have influenced the volumes and outcomes of patients undergoing SAVR at hospitals is limited. The current study demonstrates that SAVR volume declined from 2011 to 2014 only in hospitals performing the highest volume of TAVR procedures. Additionally, rates of comorbidities and 30-day and 1-year all-cause mortality all declined during the study period in patients undergoing SAVR, with the largest decline seen in hospitals performing the highest volume of TAVR procedures.
Prior studies from the United States and Europe have suggested that isolated SAVR volume was not affected by the advent of TAVR (11,12). By contrast, the current study indicates that this conclusion may not apply to the highest volume TAVR-performing hospitals. In fact, we demonstrate that hospitals performing the highest number of TAVR procedures had the greatest observed decline in SAVR volume, with TAVR volume exceeding that of SAVR nationally in the highest volume TAVR-performing hospitals earlier than previously reported (9). These data suggest that the degree of displacement of SAVR by TAVR correlates directly with the number of TAVR procedures performed at a given hospital.
Additionally, the patient case mix changed over time, with patients undergoing SAVR exhibiting the greatest reduction in rates of comorbidities in high TAVR volume hospitals. This finding may suggest a shift from SAVR to TAVR use in the highest surgical risk patients. Accordingly, there was a reduction over time in 30-day and 1-year all-cause mortality rates for SAVR patients, with the greatest declines observed in high-volume TAVR-performing hospitals. In 2014, we found that 30-day and 1-year mortality rates in non-TAVR hospitals were 3.4% and 8.2%, respectively. In comparison, 30-day and 1-year mortality rates in high-volume TAVR hospitals were lower, at 1.3% and 5.5%, respectively. The estimated 10-year survival rates also differed between hospitals, with the rate of change in 10-year survival progressively increasing in TAVR-performing hospitals but not in non-TAVR hospitals. These findings suggest that TAVR availability may have led to the selective use of SAVR in lower risk patients. It may also partially reflect secular trends in mortality for AVR patients, because SAVR mortality declined even in hospitals not performing TAVR. We also found that the prevalence of chronic heart failure was higher in TAVR hospitals than non-TAVR hospitals. This may be caused by differences in case mix of AVR patients at TAVR-performing versus nonperforming hospitals. However, it may also be possible that these differences are caused by discrepancies in coding practices.
Although TAVR is a novel and expanding treatment choice for severe aortic stenosis, most U.S. hospitals perform a relatively small number of TAVR cases annually. Our study suggests that in low TAVR volume facilities, the number of transcatheter procedures was not large enough to displace surgical volumes. As TAVR use continues to expand worldwide, these results suggest that the volume of SAVR may continue to decline. Our sensitivity analysis for TAVR-performing hospitals found consistent results. The results have significant implications for training and experience of SAVR centers. We found that the average number of isolated SAVRs performed per hospital was 60.4 over the study period. Prior studies have suggested a negative relationship between volume and outcomes after cardiac surgery (15). However, the continued rapid growth and dissemination of TAVR promote a shift from SAVR to TAVR. Whether this shift toward less complex surgical patients may mitigate the adverse effects of declining surgeon volumes for SAVR is unknown.
First, the study is retrospective and based on administrative data, and is therefore subject to residual confounding because of unmeasured variables and inaccuracies in coding. Second, the patterns observed may in part be caused by changes in secular trends over the study period. Third, because claims represent billing diagnoses, they may not accurately reflect the presence and severity of clinical conditions. Fourth, because of limited granularity in the administrative dataset, traditional surgical risk scores, such as the Society for Thoracic Surgery Predicted Risk of Mortality (16,17) or logistic EuroSCORE (18), could not be determined for each patient. The data also do not allow us to see how many patients were transferred from non-TAVR hospitals to have the procedure elsewhere. Because TAVR patients during this time period were restricted to those at extreme or high surgical risk, we purposefully did not compare outcomes between TAVR and SAVR patients due to the potential for unmeasured confounding. Finally, the observed patterns may not reflect overall trends in countries outside the United States and in non-Medicare beneficiaries.
In Medicare beneficiaries from 2011 to 2014, the advent of TAVR was associated with a reduction in SAVR volumes, the selection of lower risk patients for SAVR, and corresponding declines in short- and long-term mortality in hospitals performing the most TAVR procedures.
WHAT IS KNOWN? Isolated SAVR volume has not decreased after the advent of TAVR worldwide.
WHAT IS NEW? The current study demonstrates that the degree of displacement of SAVR by TAVR correlates directly with the number of TAVR procedures performed at a given hospital. In addition, SAVR mortality rates declined the most at hospitals with the highest TAVR volumes, likely representing declining risk of the SAVR case mix.
WHAT IS NEXT? The continued rapid growth and dissemination of TAVR may promote a shift from SAVR to TAVR. Whether this shift toward less complex surgical patients may mitigate the adverse effects of declining surgeon volumes for SAVR is unknown.
Members of the study team are supported by funding from the National Heart, Lung, and Blood Institute (1F32HL1407-11 [J.B.S.], R01HS024520-01 [C.S.], and 1R01HL136708-01 [R.W.Y.]). Dr. Elmariah is funded by the American Heart Association (14 FTF20440012). The funding organizations had no role in the final edits or submission of this manuscript. Dr. Elmariah has received institutional research support from Siemens and Boehringer Ingelheim Pharmaceuticals, Inc.; and consulting fees from Medtronic and Edwards Lifesciences. Dr. Popma has received grants from Boston Scientific, Medtronic, Abbott Vascular, Edwards Lifesciences, and Direct Flow Medical; is on the Advisory Board of Edwards Lifesciences; and received personal fees from Boston Scientific, Cordis, and Direct Flow Medical, outside the submitted work. Dr. Yeh has received investigator-initiated grant funding from Abiomed; grant support from Boston Scientific; and consulting fees from Abbott, Medtronic, Boston Scientific, and Teleflex, outside the submitted work. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic valve replacement
- surgical aortic valve replacement
- transcatheter aortic valve replacement
- Received April 4, 2018.
- Revision received June 21, 2018.
- Accepted July 3, 2018.
- 2018 American College of Cardiology Foundation
- Leon M.B.,
- Smith C.R.,
- Mack M.,
- et al.
- Reardon M.J.,
- Van Mieghem N.M.,
- Popma J.J.,
- et al.
- Baron S.J.,
- Arnold S.V.,
- Wang K.,
- et al.
- Grover F.L.,
- Vemulapalli S.,
- Carroll J.D.,
- et al.
- Culler S.D.,
- Cohen D.J.,
- Brown P.P.,
- et al.
- Edwards F.H.,
- Cohen D.J.,
- O’Brien S.M.,
- et al.