eTable 1. ICD-9-CM codes used to build the study cohort.
eTable 2. Characteristics of Patients Hospitalized for Aortic Valve Surgery, 1999-2011
eTable 3. Hospitalization Rates for Aortic Valve Surgery, 1999-2011
eTable 4. Age-sex-race-adjusted annual trends in rates of aortic valve replacements, overall and by subgroups
eTable 5. Outcomes of Aortic Valve Surgery, 1999-2011
eTable 6. Age-sex-race-comorbidity-adjusted annual trends in 30-day mortality rates of aortic valve replacements, overall and by subgroups
eTable 7. Age-sex-race-comorbidity-adjusted annual trends in 1-year mortality rates of aortic valve replacements, overall and by subgroups
eTable 8. Age-sex-race-comorbidity-adjusted annual trends in 30-day readmission rates of aortic valve replacements, overall and by subgroups
eTable 9. Trends in the type of aortic valve replacement and utilization of concomitant coronary artery bypass graft surgery, 1999-2011
eFigure. Annual Change Rates in Procedure, Mortality, and Readmission
Barreto-Filho JA, Wang Y, Dodson JA, Desai MM, Sugeng L, Geirsson A, Krumholz HM. Trends in Aortic Valve Replacement for Elderly Patients in the United States, 1999-2011. JAMA. 2013;310(19):2078-2084. doi:10.1001/jama.2013.282437
Copyright 2013 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
There is a need to describe contemporary outcomes of surgical aortic valve replacement (AVR) as the population ages and transcatheter options emerge.
To assess procedure rates and outcomes of surgical AVR over time.
Design, Setting, and Participants
A serial cross-sectional cohort study of 82 755 924 Medicare fee-for-service beneficiaries undergoing AVR in the United States between 1999 and 2011.
Main Outcomes and Measures
Procedure rates for surgical AVR alone and with coronary artery bypass graft (CABG) surgery, 30-day and 1-year mortality, and 30-day readmission rates.
The AVR procedure rate increased by 19 (95% CI, 19-20) procedures per 100 000 person-years over the 12-year period (P<.001), with an age-, sex-, and race-adjusted rate increase of 1.6% (95% CI, 1.0%-1.8%) per year. Mortality decreased at 30 days (absolute decrease, 3.4%; 95% CI, 3.0%-3.8%; adjusted annual decrease, 4.1%; 95% CI, 3.7%- 4.4%) per year and at 1 year (absolute decrease, 2.6%; 95% CI, 2.1%-3.2%; adjusted annual decrease, 2.5%; 95% CI, 2.3%-2.8%). Thirty-day all-cause readmission also decreased by 1.1% (95% CI, 0.9%-1.3%) per year. Aortic valve replacement with CABG surgery decreased, women and black patients had lower procedure and higher mortality rates, and mechanical prosethetic implants decreased, but 23.9% of patients 85 years and older continued to receive a mechanical prosthesis in 2011.
Conclusions and Relevance
Between 1999 and 2011, the rate of surgical AVR for elderly patients in the United States increased and outcomes improved substantially. Medicare data preclude the identification of the causes of the findings and the trends in procedure rates and outcomes cannot be causally linked. Nevertheless, the findings may be a useful benchmark for outcomes with surgical AVR for older patients eligible for surgery considering newer transcatheter treatments.
Aortic valve disease in the United States is a major cardiovascular problem that is likely to grow as the population ages.1- 5 Aortic valve replacement is the standard treatment even for very elderly patients despite its risks in this age group.6 With transcatheter aortic valve replacement emerging as a less invasive option,7- 9 contemporary data from real-world practice are needed to provide a perspective on the outcomes that are being achieved with surgery.
Changes in practice during the past decade are partly due to growing recognition that the benefits of aortic valve replacement extend to extremes in patient age.10,11 There is also a recommendation for consideration of earlier, preemptive aortic valve replacement in selected asymptomatic patients12 and strong guidance that bioprostheses rather than mechanical valves be used for patients 65 years or older.13,14 There is also uncertainty about the benefit of performing coronary artery bypass graft (CABG) surgery during aortic valve replacement in patients with stable coronary disease.8,15,16 Previous studies have indicated that rates of aortic valve replacement are increasing and outcomes are improving17- 20 but do not provide population-based rates and long-term follow-up.17- 20
We therefore studied aortic valve replacement among Medicare fee-for-service beneficiaries between 1999 and 2011. We calculated trends in rates of surgical aortic valve replacement and outcomes defined as mortality, readmission, and length of stay. In addition, we compared the outcomes of patients who had CABG surgery and aortic valve replacement with those who had replacement alone and assessed rates of use of mechanical prostheses, which generally are not recommended for patients 65 years or older in the absence of another reason for anticoagulation.21,22 Because variation in patterns of care and outcomes across subgroups may indicate opportunities for quality improvement, we also calculated these trends by age, sex, and race.
Institutional review board approval for this study was obtained through the Yale University Human Investigation Committee. We identified all Medicare fee-for-service patients from January 1, 1999, to December 31, 2011, using the inpatient administrative data from the Centers for Medicare & Medicaid Services (CMS). We identified patients who underwent aortic valve replacement based on the principal discharge diagnosis (International Classification of Diseases, Ninth Revision, Clinical Modification codes 35.21 [aortic valve surgery replacement with bioprosthesis], and 35.22 [aortic valve surgery replacement with mechanical prosthesis]; eTable 1 in the Supplement). We excluded 566 patients who underwent aortic valve repair, 37 412 who underwent aortic valve replacement with concomitant mitral valve surgery, and 4007 who underwent tricuspid valve surgery from the years 1999 to 2011, and 2961 who had endocarditis as their principal diagnosis. For a small group of patients who had more than 1 aortic valve replacement during any of the study years—a proportion that decreased over time (n=389, or 1.5% in 1999 and 44 or 0.1% in 2011)—we selected the first procedure. We linked aortic valve replacement data with Medicare denominator files to obtain mortality information and to determine the eligibility of the beneficiaries and their length of time in the fee-for-service program.
We collected data on patients’ age, sex, race, and common comorbidities. Race was determined from the Medicare denominator files, which used patient-reported data from the Social Security Administration.23 We selected 21 comorbidities in categories of cardiovascular disease (hypertension, diabetes mellitus, atherosclerotic disease, unstable angina, prior myocardial infarction, prior heart failure, peripheral vascular disease, stroke, non–stroke cerebrovascular disease), geriatric conditions (dementia, functional disability, malnutrition), and miscellaneous (renal failure, chronic obstructive pulmonary disease, pneumonia, respiratory failure, liver disease, cancer, major psychiatric disorders, depression, and trauma) based on the method used by CMS for profiling hospitals for acute myocardial infarction and heart failure.24,25 We identified comorbidities from diagnosis codes of all patient hospitalizations for any cause, primary or secondary, up to 1 year before the initial hospitalization for aortic valve replacement. Comorbidity data from 1998 were used for patients who underwent an aortic valve replacement in 1999.
We calculated person-years for each beneficiary to account for new enrollment, disenrollment, or death during an index year. For each year, we linked person-year data with aortic valve replacement hospitalization data to obtain rates of aortic valve replacement by dividing the total number of aortic valve replacements by the corresponding person-years of beneficiaries.26 Using CABG codes (ICD-9, 36.1x), we determined the proportion of aortic valve replacement performed with and without CABG surgery (eTable 1 in the Supplement). We also determined the type of aortic valve replacement performed using codes for replacement with bioprosthesis and replacement with mechanical prosthesis, as described above.
In-hospital mortality was defined as deaths occurring during the index aortic valve replacement-specific hospitalization. To standardize the follow-up period, we used 30-day and 1-year mortality rates, defined as the rate of deaths due to all causes that occurred within 30 days or 1 year from the first procedure date during the index hospitalization for aortic valve replacement. We divided length of stay into 2 phases: preprocedure length of stay, defined as the difference in days between the procedure and admission dates, and postprocedure length of stay, defined as the difference in days between the discharge and procedure dates. We defined 30-day readmission as rehospitalizations for all causes occurring within 30 days from the date of discharge, using November 30, 2011, as the final date of discharge for complete follow-up.26,27
We express the rate of aortic valve replacement per 100 000 person-years, the rates of mortality and readmission as percentages, and length of stay as median (interquartile range) days. We used the Mantel-Haenszel χ2 test to determine the statistical significance of temporal changes in observed outcomes and patient characteristics. To evaluate changes in rates of aortic valve replacement, we fit a mixed-effects model with a Poisson link function and state-specific random intercepts, adjusting for age, sex, and race. To estimate changes in the rates of mortality (30-day and 1-year), we used the same mixed model with a logit-link function and hospital-specific random intercepts, adjusting for patient age, sex, race, and comorbidities. To assess change in rates of 30-day readmission, we constructed a Cox proportional hazards model with death as a censoring event and adjusting for age, sex, race and comorbidity. We used the method developed by Lin et al28 to check the adequacy of the Cox regression model and found the proportional hazards assumption was satisfied.
All models included an ordinal time variable ranged from 0 to 12, corresponding to the years 1999 to 2011, to represent the annual changes in outcomes. The incidence rate ratio (RR) of the time variable was used to represent the age-, sex-, and race-adjusted annual changes in the aortic valve replacement rate, and the odds ratio (OR) and hazard ratio (HR) of the time variable were used to represent the age-, sex-, and race-comorbidity–adjusted annual changes in mortality and readmission rates, respectively. We repeated models for age, sex, and race subgroups.
Dr Wang performed all analyses using SAS version 9.2 (SAS Institute Inc). Statistical tests were 2-sided at a significance level of .05.
The final sample consisted of 409 591 889 records, representing 82 755 924 beneficiaries aged 65 years or older with at least 1 month of enrollment in Medicare fee-for-service during the study period (2 542 827 477 person-years). Patient age, sex, and race remained mostly unchanged over time but several comorbidities were more commonly coded, including hypertension (51.8% in 1999, 65.7% in 2011), diabetes mellitus (21.0% in 1999, 27.7% in 2011), and renal failure (2.4% in 1999, 10.4% in 2011). Atherosclerotic disease (60.7% in 1999, 57.2% in 2011) and history of heart failure (18.4% in 1999, 17.6% in 2011) were reported less frequently (P <.001 for trend) (Table 1 and eTable 2, in the Supplement).
Rates of aortic valve replacement increased by 19 (95% CI, 19-20) procedures per 100 000 person-years between 1999 and 2011 (93 in 1999, 112 in 2011; P < .001 for trend); the rate of aortic valve replacement without CABG surgery increased (40 in 1999, 64 in 2011; increase, 24 [95% CI, 23-24] procedures per 100 000 person-years; P < .001 for trend) and the rate of aortic valve replacement with CABG surgery decreased (53 in 1999, 48 in 2011; decrease, 5 [95% CI, 4-5] procedures per 100 000 person-years; P < .001 for trend). Procedure rates increased in all age, sex, and race strata, most notably in patients 75 years or older (Table 2 and eTable 3, in the Supplement). The findings did not change substantially after accounting for age, sex, race, and state.
The age-, sex-, and race-adjusted change was 1.6% (95% CI, 1.0%-1.8%; incidence RR, 1.016 [95% CI, 1.01-1.018]) increase per year for all aortic valve replacement procedures, 4.1% (95% CI, 3.9%-4.2%; incidence RR, 1.041 [95% CI, 1.039-1.042]) increase per year for aortic valve replacement without CABG surgery, and 0.5% (95% CI, 0.2%-0.6%; incidence RR, 0.995 [95% CI, 0.994-0.998]) decrease per year for aortic valve replacement with CABG surgery. The direction of change was similar by age, sex, and race strata except for a small increase in rates of aortic valve replacement with CABG surgery among patients 75 years or older (eFigure, eTable 4 in the Supplement).
Between 1999 and 2011, 30-day mortality rates decreased an absolute 3.4% (95% CI, 3.0%-3.8%) for all aortic valve replacement, an absolute 3.1% (95% CI, 2.6%-3.7%) for aortic valve replacement without CABG surgery, and an absolute 3.2% (95% CI, 2.6%-3.8%) for aortic valve replacement with CABG surgery. One-year mortality rates also decreased by an absolute 2.6% (95% CI, 2.1%-3.2%) for all aortic valve replacements, 2.2% (95% CI 1.5% to 3.0%) for aortic valve replacement without CABG surgery, and 2.4% (95% CI 1.6% to 3.2%) for aortic valve replacement with CABG surgery. The decreases were similar among all age, sex, and race strata (Table 3 and eTable 5, in the Supplement) and did not change substantially after accounting for patient characteristics and hospital. Age-, sex-, and race-comorbidity–adjusted decreases in 30-day mortality rates were 4.1% (95% CI, 3.7%-4.4%) per year (OR, 0.959; 95% CI, 0.956-0.963) for overall aortic valve replacement, 4.7% (95% CI, 4.2%-5.3%) per year (OR, 0.953, 95% CI, 0.947-0.958) for aortic valve replacement without CABG, and 3.4% (95% CI, 2.9%-3.8%) per year (OR, 0.966; 95% CI, 0.962-0.971) for aortic valve replacement with CABG surgery. Similarly, the age-, sex-, and race-comorbidity–adjusted decreases in 1-year mortality rates were 2.5% (95% CI, 2.3%-2.8%) per year (OR, 0.975; 95% CI, 0.972-0.977) for all aortic valve replacement, 2.8% (95% CI, 2.3%-3.2%) per year (OR 0.972, 95% CI, 0.968-0.977) for aortic valve replacement without CABG surgery, and 2.1% (95% CI, 1.7%-2.5%) per year (OR, 0.979; 95% CI, 0.975-0.983) for aortic valve replacement with CABG surgery. The decreases were seen among all age, sex, and race strata with the exception of a nonstatistically significant estimate of no change in 1-year mortality among nonwhite nonblack women undergoing aortic valve surgery with CABG surgery (eFigure, eTables 6 and 7 in the Supplement). Although decreases in mortality were comparable within sex and race strata, women nevertheless had higher 30-day and 1-year mortality than men, and black patients had higher mortality than white patients in every study year.
Median length of stay was unchanged (Table 3 and eTable 5, in the Supplement). Thirty-day readmission rates decreased slightly for aortic valve replacement procedures overall and with and without CABG surgery (Table 3). Age-, sex-, and race-comorbidity–adjusted decreases in 30-day readmission rates were 1.1% (95% CI, 0.9%-1.3) per year (HR, 0.989; 95% CI, 0.987-0.991) for overall aortic valve replacement, 1.3% (95% CI, 1.1%-1.6%) per year (HR, 0.987, 95% CI, 0.984-0.989) for aortic valve replacement without CABG surgery and 0.8% (95% CI, 0.5%-1.1%) per year (HR, 0.992; 95% CI, 0.989-0.995) for aortic valve replacement with CABG surgery (eFigure, eTable 8 in the Supplement).
Use of bioprostheses in aortic valve replacement increased from 44.1% in 1999 to 72.7% in 2011 (P < .001). Conversely, mechanical prosethetic implants decreased from 55.9% to 27.3% (P<.001). Among age, sex, and race strata in 2011, use of bioprostheses was highest for patients who were 85 years or older, men, and white (eTable 9 in the Supplement). In 1999, 46.2% of patients 85 years or older received a mechanical prosthesis, decreasing to 23.9% in 2011.
Our study describes a national increase in the rates of aortic valve replacement and a reduction in mortality, readmission, and length of stay for Medicare beneficiaries undergoing the procedure from 1999 through 2011. Rates of aortic valve replacement without CABG surgery increased while rates of aortic valve replacement with it decreased, but adjusted annual decreases in 30-day mortality were comparable for either procedure. Mechanical aortic valve prostheses continued to be used in about a quarter of elderly patients in 2011.
Several studies have investigated this topic with less contemporary cohorts, but direct comparison between our results and those of other studies is difficult because other studies looked at smaller cohorts and few provided a comprehensive national perspective or reported standardized short- and long-term outcomes.8,18,19,29,30 Goodney et al,31 using Medicare data from 1994 to 1999, reported that in-hospital mortality for aortic valve replacement (54% performed with CABG surgery) was 8.8%, ranging from 6% to 13%. Lee et al,19 using the Society of Thoracic Surgeons database, showed that operative mortality for aortic valve replacement decreased from 5.6% during the years 1993-1997 to 4.4% during the years 2003-2007. Brown et al18 focused on aortic valve replacement without CABG surgery in the Society of Thoracic Surgeons database and reported higher survival rates, with age-, sex-, and race-comorbidity–adjusted operative mortality decreasing from 3.5% to 2.4% between 1997 and 2006. Our data set extended to 2011, and we found higher mortality for patients undergoing aortic valve replacement without CABG surgery (6.6% in 1999 and 3.5% in 2011), which we believe is likely due to the older age of our population and our use of 30-day (rather than in-hospital) mortality as an outcome measure. In addition to providing more contemporary estimates, our study provides longer-term (1-year) mortality outcomes, data on readmission, and outcomes from centers that may not have been captured in Society of Thoracic Surgeons data.
Across all years of observation, we found lower procedure rates and higher mortality in black patients than in white patients. Racial differences in aortic valve disease or black patients being referred at a later stage of illness and having less access to the procedure may be factors.32- 36 Similarly for women, procedure rates were lower and mortality was higher. A report from the Society of Thoracic Surgeons database of patients undergoing aortic valve replacement without CABG surgery described a similar finding: fewer patients undergoing the procedure were women (43% women vs 57% men) and mortality was higher among women (2006 mortality: 3.2% in women, 2.1% in men).19 We could not determine reasons for any differences from the Medicare data set, but they point to a worthwhile direction for future investigation. Encouragingly, mortality rates declined among all subgroups from 1999 to 2011, although residual differences persisted.
There are several potential explanations for the increase in the use of aortic valve replacement among older adults, including better access to specialized health care centers and enhanced awareness of the therapeutic benefit of aortic valve replacement independent of patient age among health care professionals.11 The increase in age and comorbidities among patients undergoing aortic valve replacement suggest that the recommendation to perform preemptive aortic valve surgery earlier in the disease course had not taken hold during the study period. Although the largest increase in rates of aortic valve replacement occurred in those aged 85 years or older, a group in which the rate of 1-year mortality was 17.5% by 2011, these patients still underwent surgery less commonly than those aged 75 through 84 years, despite the known increasing prevalence of aortic stenosis with advancing age.1 This finding suggests that there is a significant residual population of very elderly patients with aortic stenosis for whom transcatheter aortic valve replacement may emerge as a common treatment modality. In that context, our findings could provide a useful benchmark of outcomes with surgical aortic valve replacement among eligible patients who may also be considering transcatheter aortic valve replacement.
Approximately 50% of aortic valve replacement procedures were performed with CABG surgery, and we could not determine from our data whether the decrease in aortic valve replacement with CABG surgery over time reflected changing coronary anatomy among the study population (eg, lower prevalence of lesions amenable to surgical revascularization), an increasing use of preprocedural or postprocedural percutaneous coronary intervention, or some other factor. The treatment of chronic coronary artery disease in the setting of aortic valve replacement is uncertain and evolving.15 CABG surgery continues to be performed during aortic valve replacement in cases of anatomical coronary obstruction. Future studies are needed to investigate whether the current practice of aortic valve replacement plus CABG surgery for patients with stable coronary artery disease is preferable to other strategies such as aortic valve replacement plus watchful waiting, aortic valve replacement plus percutaneous coronary intervention (hybrid procedure), or transcatheter aortic valve replacement plus percutaneous coronary intervention.8
Previous studies reported a progressive shift toward the use of bioprostheses in the aortic position, but rates of use of mechanical prostheses are still quite high.17,18,37 The debate surrounding the optimal prosthesis for elderly patients remains unresolved.30,38,39 In similar instances of clinical uncertainty, some experts recommend shared decision making.40 The 1998 guidelines for managing valvular heart disease from the American Heart Association and the American College of Cardiology recommended the use of bioprosthetic valves in the aortic position for patients 65 years or older.13,38,39 Our administrative data, however, lacked detailed information that could explain the high rate of use of mechanical prostheses. Small native aortic valve annulus or preexisting clinical conditions for anticoagulation may represent some appropriate uses of mechanical prostheses in the elderly. The high rates of mechanical prostheses that we observed in 2011 raise concern about whether this practice is in the best interest of patients.
This study has several limitations. First, our findings are observational and multiple factors may have accounted for the observed trends (eg, patient selection, health care system changes, secular changes). We cannot determine the underlying causes due to the limitations of administrative data. Second, our mortality adjustment models used comorbidity information based on administrative data and did not include some relevant clinical information that could have improved the modeling process, such as patient functional status, left ventricular function, the pathological process associated with aortic valve dysfunction, previous cardiac surgery, or the distinction between elective or urgent aortic valve replacement.41 Third, we used Medicare claims data to infer trends in the type of surgery performed (bioprosthetic implant or mechanical prostheses). Because we used codes for the primary diagnosis, aortic valve disease may have been misclassified as the primary diagnosis in cases of aortic dissection or aneurysm with concomitant aortic valve disease. However, this would represent a very small number of patients. Fourth, our cohort was limited to Medicare fee-for-service beneficiaries, and thus we cannot comment on trends among patients enrolled in Medicare managed care programs or among patients younger than 65 years. Because more patients have enrolled in Medicare managed care programs over time,42 the Medicare fee-for-service population may have changed and therefore affected the observed trends. However, we are reporting rates that take into account any change in the denominator over the years. Finally, the use of inpatient claims limited our ability to include comorbidity information in our procedure rate model, so we were therefore unable to estimate the influence of changes in patient comorbidity on patient selection for an aortic valve replacement.
We found an increase in rates of aortic valve replacement and an improvement in mortality and other outcomes among Medicare beneficiaries over a 12-year period. Rates of aortic valve replacement with CABG surgery decreased, older patients are still receiving mechanical prostheses despite recommendations to the contrary, and women and black patients experienced higher mortality than men and white patients. These findings may provide a useful benchmark for outcomes of aortic valve replacement surgery for older patients eligible for surgery considering newer transcatheter treatments.
Corresponding Author: Harlan M. Krumholz, MD, SM, Yale University School of Medicine, 1 Church St, Ste 200, New Haven, CT 06510 (firstname.lastname@example.org).
Author Contributions: Drs Barreto-Filho and Wang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Barreto-Filho, Wang, Krumholz
Acquisition of data: Krumholz
Analysis and interpretation of data: Barreto-Filho, Wang, Dodson, Desai, Sugeng, Geirsson, Krumholz
Drafting of the manuscript: Barreto-Filho, Krumholz
Critical revision of the manuscript for important intellectual content: Barreto-Filho, Wang, Dodson, Sugeng, Geirsson, Krumholz
Statistical analysis: Wang
Obtained funding: Krumholz
Administrative, technical, or material support: Krumholz
Study supervision: Krumholz
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Krumholz reported that he works under contract with the Centers for Medicare & Medicaid Services to develop and maintain performance measures. He is chair of a cardiac scientific advisory board for UnitedHealth and the recipient of a research grant from Medtronic, through Yale University, to develop methods of clinical trial data sharing. No other conflicts were reported.
Funding/Support: This work was supported by grant U01HL105270-04 (Center for Cardiovascular Outcomes Research at Yale University) from the National Heart, Lung, and Blood Institute. During the time that this work was conducted, Dr Barreto-Filho was a postdoctoral fellow at Yale University supported by grant 3436-10-1 from CAPES (Coordenação de Aperfeicoamento de Pessoal de Nível Superior, Ministry of Education, Brazil).
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; in the preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Correction: This artcle was corrected on November 19, 2013, to amend an author affiliation.