Abstract
Background
Patients with symptomatic severe aortic stenosis (ssAS) have a high mortality risk and compromised quality of life. Surgical/transcatheter aortic valve replacement (AVR) is a Class I recommendation, but it is unclear if this recommendation is uniformly applied. We determined the impact of managing cardiologists on the likelihood of ssAS treatment.
Methods and Results
Using natural language processing of Optum electronic health records, we identified 26 438 patients with newly diagnosed ssAS (2011–2016). Multilevel, multivariable Fine‐Gray competing risk models clustered by cardiologists were used to determine the impact of cardiologists on the likelihood of 1‐year AVR treatment. Within 1 year of diagnosis, 35.6% of patients with ssAS received an AVR; however, rates varied widely among managing cardiologists (0%, lowest quartile; 100%, highest quartile [median, 29.6%; 25th–75th percentiles, 13.3%–47.0%]). The odds of receiving AVR varied >2‐fold depending on the cardiologist (median odds ratio for AVR, 2.25; 95% CI, 2.14–2.36). Compared with patients with ssAS of cardiologists with the highest treatment rates, those treated by cardiologists with the lowest AVR rates experienced significantly higher 1‐year mortality (lowest quartile, adjusted hazard ratio, 1.22, 95% CI, 1.13–1.33).
Conclusions
Overall AVR rates for ssAS were low, highlighting a potential challenge for ssAS management in the United States. Cardiologist AVR use varied substantially; patients treated by cardiologists with lower AVR rates had higher mortality rates than those treated by cardiologists with higher AVR rates.
Keywords: aortic valve replacement, physician variability, symptomatic severe aortic stenosis
Subject Categories: Aortic Valve Replacement/Transcather Aortic Valve Implantation
Nonstandard Abbreviations and Acronyms
- AS
aortic stenosis
- AVR
aortic valve replacement
- MOR
median odds ratio
- SAVR
surgical aortic valve replacement
- ssAS
symptomatic severe aortic stenosis
- TAVR
transcatheter aortic valve replacement
Clinical Perspective
What Is New?
We identified patients with symptomatic severe aortic stenosis (ssAS) from a large database, evaluated the receipt of aortic valve replacement (AVR) within a year of ssAS diagnosis, and found that overall AVR rates for ssAS were low; the majority of patients with ssAS (64.4%) did not have an AVR within 1 year of diagnosis.
AVR rates varied widely by managing cardiologist (>2‐fold depending on cardiologist); patients with ssAS treated by cardiologists with lower AVR rates had higher mortality rates than those treated by cardiologists with higher AVR rates.
What Are the Clinical Implications?
Given the availability of effective treatments, these findings underscore the need to implement targeted initiatives to raise disease awareness, promote more objective diagnostic criteria, and reduce barriers to treatment for patients with ssAS.
Symptomatic severe aortic stenosis (ssAS) is associated with a poor prognosis if left untreated. Aortic valve replacement (AVR) improves survival among patients with ssAS and is a Class IA recommendation by the American Heart Association/American College of Cardiology and European Society of Cardiology.1 Although valve replacement was traditionally performed via open heart surgery (surgical AVR [SAVR]), the development of transcatheter AVR (TAVR) expanded options for AVR to the majority of patients with ssAS.2, 3, 4, 5 Yet despite the widespread availability of effective treatment options, barriers to treatment persist, and the contemporary penetrance of AVR remains unknown.6, 7
Clinicians have a strong influence on the likelihood of many cardiovascular therapies, including amputation8 in lower extremity peripheral arterial disease and the use of cardiac defibrillators9 in heart failure. To date, the role of the managing cardiologist in shaping ssAS treatment has not been evaluated. In this analysis, we examined contemporary rates of AVR for the treatment of ssAS in the United States using the Optum database, which has been used for previous cardiac studies.10, 11 In addition, we evaluated clinician‐level variation in the management of ssAS among cardiologists and its association with 1‐year survival.
METHODS
Data Source
This retrospective study was conducted using Optum deidentified electronic health records (EHRs),12 which is a patient‐level database that aggregates EHR systems from >2000 US hospitals and 7000 clinics, including 82 million distinct patients, into a tabular format for research purposes. This data set is available through contract with Optum. Available information includes patient‐level data from both the ambulatory and inpatient settings and provides unique identification numbers to physicians, allowing clinicians to be followed over time. The Optum database has been employed using similar methods from relevant previously published studies.10, 11
Study Population
This study included patients newly diagnosed with ssAS between 2011 and 2016 within Optum's integrated delivery network, where care and coverage are offered through the same provider reducing the risk of missing records. Because the Optum data set does not routinely contain structured data elements for echocardiographic variables other than ejection fraction (Table S1), a review of physician notes was used to identify patients with severe aortic stenosis (AS).13 Severe AS was defined by the inclusion of the terms severe or critical or a combination in the presence of the words aortic stenosis.14, 15, 16 We excluded patients with neutral or negative terms associated with their AS diagnosis such as negative, deny, not, suspect, potential, or rule out or a combination thereof. Sensitivity analyses were performed to validate reported severity—a Kaplan–Meier analysis to stratify survival by AS grade and a review of 1206 patients with data for each of 3 metrics of aortic valve stenosis to evaluate correspondence with physician reports (Figure S1, Data S1, Tables S2 through S4). To address variability in the assignment of AS severity (particularly among low flow, low gradient cases), a sensitivity analysis was conducted limiting the analysis to patients with recorded left ventricular ejection fraction (LVEF) values, stratified by LVEF (LVEF <35, 35–49, <50, and ≥50).
Patients were classified as symptomatic if there were at least 2 positive entries for cardinal symptoms (heart failure, angina, dyspnea on exertion, dyspnea, presyncope, syncope) in the 6 months before severe AS diagnosis, similar to previously described methodologies.15, 17, 18 Again, the use of negative terms was excluded from the symptomatic definition. Newly diagnosed patients with ssAS had either no documented history of severe AS within the year before their diagnosis or had severe AS, but no mention of symptoms in the 6 months before their diagnosis.
Included patients had at least 1 year of history in the EHR before ssAS diagnosis and at least 1 year of follow‐up or a record of death in the year after the date of ssAS to allow for the evaluation of patient status. A total of 10 patients with a preexisting left ventricular assist device were excluded. An additional 11 461 patients without an identifiable managing cardiologist (defined in Exposure) were also excluded. The final cohort included 26 438 patients (Figure 1).
Figure 1. Modified consort diagram.
The 26 438 patients were managed by 1627 cardiologists. EHR indicates electronic health record; IDN, integrated delivery network; LVAD, left ventricular assist device; and ssAS, symptomatic severe aortic stenosis.
Risk‐Adjustment Covariates
The risk‐adjustment set was chosen a priori based on clinical factors that could impact the likelihood of treatment. Patient history was evaluated in the year before diagnosis using both International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM) and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM) codes with characteristics listed in Table. Records related to inpatient visits were reviewed to identify hospitalizations in the year before AS diagnosis and to determine the setting of the initial ssAS diagnosis. ICD‐9‐CM/ICD‐10‐CM and Current Procedural Terminology codes were used to assess select cardiac procedures and dialysis. Multimorbidity was assessed using the Deyo modification of the Charlson Comorbidity Index.19 The full list of ICD‐9‐CM and ICD‐10‐CM codes used in this analysis is presented in Table S5. LVEF was obtained from structured extracts of echocardiography reports, as were patient age, sex, area income and education level, census region/division, insurance, and smoking status.
Exposure
A managing cardiologist for each patient with ssAS was identified in the Optum records using unique provider identification numbers and the reported specialty. The managing cardiologist was defined as the majority provider cardiologist most frequently seen by a patient (as an inpatient or outpatient) in the 3 months before and after ssAS diagnosis. If multiple cardiologists met these criteria, then the cardiologist with visits closest to diagnosis was selected. We performed a sensitivity analysis focusing only on outpatient cardiologists to comment on the physician most likely responsible for long‐term care (Data S2). Furthermore, to understand if an interventional subspecialty shaped treatment likelihood, we stratified cardiologists based on whether they performed percutaneous interventions.
Outcomes
The primary outcome of this study was treatment of ssAS using AVR (SAVR or TAVR) in the year following the first report of symptoms. Specific ICD‐9‐CM/ICD‐10‐CM and Current Procedural Terminology codes are listed in Table S5. In addition, we examined all‐cause mortality at 1 year after diagnosis to evaluate the impact of provider treatment rates on patient outcomes. Date of death was captured in Optum using the Social Security Death Master File. Follow‐up was evaluated through 2017.
Statistical Analyses
We compared patient characteristics across quartiles of observed AVR rates for managing cardiologists using chi‐square tests. The quartiles were established via the PROC RANK SAS procedure, which assigns quartiles without ties, assigning the 0% treatment rate to the lowest quartile and then building the other quartiles based on patient numbers. This procedure can result in quartiles being of unequal size. For the initial assessment of the variability in treatment, we evaluated the observed and adjusted rates of AVR across quartiles of observed AVR rate using previously described methodology20, 21 and the multilevel model described in the next paragraph. Similar analysis was performed for TAVR.
The primary analysis in this study used a multivariable, multilevel, logistic model with the managing cardiologist as the random intercept to assess the contribution of unique cardiologists on the likelihood of AVR at 1 year following ssAS diagnosis. The risk‐adjustment set included variables presented in the Table. The median odds ratio (MOR)22, 23, 24 was used to express the relative association of the managing cardiologist on the likelihood of AVR. The MOR expresses the likelihood of a patient receiving a different outcome if the patient were to switch to another randomly selected cardiologist. For example, a MOR of 1 would indicate no difference in the likelihood of AVR between managing cardiologists; however, a MOR of 1.5 would indicate a 50% greater odds of a different outcome if a patient was treated by another randomly selected cardiologist. A subsequent analysis examined the relative contribution of the cardiologist on receipt of TAVR compared with SAVR using similar methods. A sensitivity analysis was performed in a claims‐linked set of 926 patients to confirm the results (Data S3) and in a patient subset with recorded ejection fraction, creatinine, and body mass index to better control for patient status (Data S4, Tables S6 and S7, and Figure S2). An additional sensitivity analysis was conducted to evaluate the impact of cardiologist caseload (which included both volume of patients with ssAS and AVR procedure volume) on the likelihood of AVR.
Table 1.
Patient Characteristics for the Overall Cohort and Stratified by Quartiles of AVR Treatment Rates by Managing Cardiologists
| Patient characteristics | Overall, n=26 438 | Quartiles of AVR treatment rates by managing cardiologists | P value | |||
|---|---|---|---|---|---|---|
| First quartile lowest AVR, n=4257 | Second quartile, n=7062 | Third quartile, n=7138 | Fourth quartile* highest AVR, n=7981 | |||
| Treatment in 1 y from first symptom report, n (%) | ||||||
| SAVR | 5894 (22.29) | 146 (3.43) | 1156 (16.37) | 1902 (26.65) | 2690 (33.70) | <0.001 |
| TAVR | 3513 (13.29) | 58 (28.43) | 412 (5.83) | 736 (10.31) | 2307 (28.91) | <0.001 |
| SAVR or TAVR | 9407 (35.58) | 204 (4.79) | 1568 (22.20) | 2638 (36.96) | 4997 (62.61) | <0.001 |
| Sex, n (%) | ||||||
| Female patient | 12 140 (45.92) | 2070 (48.63) | 3313 (46.91) | 3193 (44.73) | 3564 (44.66) | <0.001 |
| Age, y, n (%) | <0.001 | |||||
| Unknown | 48 (0.18) | 6 (0.14) | 15 (0.21) | 11 (0.15) | 16 (0.20) | |
| <65 | 3685 (13.94) | 665 (15.62) | 960 (13.59) | 1013 (14.19) | 1047 (13.12) | |
| 65–79 | 9892 (37.42) | 1431 (33.62) | 2635 (37.31) | 2707 (37.92) | 3119 (39.08) | |
| 80+ | 12 813 (48.46) | 2155 (50.62) | 3452 (48.88) | 3407 (47.73) | 3799 (47.60) | |
| Charlson Comorbidity Index, n (%) | 0.476 | |||||
| 0 | 6821 (25.80) | 1096 (25.75) | 1813 (25.67) | 1839 (25.76) | 2073 (25.97) | |
| 1 | 5657 (21.40) | 903 (21.21) | 1520 (21.52) | 1470 (20.59) | 1764 (22.10) | |
| 2 | 4114 (15.56) | 631 (14.82) | 1114 (15.77) | 1160 (16.25) | 1209 (15.15) | |
| 3 | 3287 (12.43) | 534 (12.54) | 863 (12.22) | 892 (12.50) | 998 (12.50) | |
| 4+ | 6559 (24.81) | 1093 (25.68) | 1752 (24.81) | 1777 (24.89) | 1937 (24.27) | |
| Atrial fibrillation, n (%) | 7521 (28.45) | 1318 (30.96) | 2135 (30.23) | 2135 (29.91) | 1933 (24.22) | <0.001 |
| Cancer, n (%) | 3408 (12.89) | 478 (11.23) | 870 (12.32) | 990 (13.87) | 1070 (13.41) | <0.001 |
| Conduction, n (%) | 2628 (9.94) | 441 (10.36) | 749 (10.61) | 708 (9.92) | 730 (9.15) | 0.019 |
| COPD, n (%) | 2650 (10.02) | 479 (11.25) | 702 (9.94) | 672 (9.41) | 797 (9.99) | <0.001 |
| Dementia, n (%) | 607 (2.30) | 133 (3.12) | 186 (2.63) | 157 (2.20) | 131 (1.64) | <0.001 |
| Diabetes mellitus without complications, n (%) | 7401 (27.99) | 1182 (27.77) | 2021 (28.62) | 1986 (27.82) | 2212 (27.72) | 0.597 |
| Diabetes mellitus with complications, n (%) | 1622 (6.14) | 214 (5.03) | 442 (6.26) | 461 (6.46) | 505 (6.33) | 0.011 |
| Prior MI, n (%) | 3049 (11.53) | 491 (11.53) | 853 (12.08) | 797 (11.17) | 908 (11.38) | 0.362 |
| Osteoarthritis, n (%) | 3756 (14.21) | 624 (14.66) | 1034 (14.64) | 1018 (14.26) | 1080 (13.53) | 0.187 |
| Peripheral vascular disease, n (%) | 4164 (15.75) | 609 (14.31) | 1047 (14.83) | 1170 (16.39) | 1338 (16.76) | <0.001 |
| Heart failure, n (%) | 5535 (20.94) | 940 (22.08) | 1457 (20.63) | 1538 (21.55) | 1600 (20.05) | 0.027 |
| Moderate to severe renal disease, n (%) | 5652 (21.38) | 983 (23.09) | 1522 (21.55) | 1575 (22.07) | 1572 (19.7) | <0.001 |
| Current smoking, n (%) | 2917 (11.03) | 490 (11.51) | 776 (10.99) | 782 (10.96) | 869 (10.89) | 0.156 |
| Use of supplemental oxygen, n (%) | 1057 (4.00) | 166 (3.90) | 288 (4.08) | 279 (3.91) | 324 (4.06) | 0.929 |
| PCI, n (%) | 583 (2.21) | 75 (1.76) | 131 (1.85) | 134 (1.88) | 243 (3.04) | <0.001 |
| Pacemaker, n (%) | 275 (1.04) | 57 (1.34) | 81 (1.15) | 66 (0.92) | 71 (0.89) | 0.067 |
| Hemodialysis, n (%) | 379 (1.43) | 86 (2.02) | 103 (1.46) | 96 (1.34) | 94 (1.18) | 0.002 |
| Dyspnea, n (%) | 23 910 (90.44) | 3761 (88.35) | 6320 (89.49) | 6432 (90.11) | 7397 (92.68) | <0.001 |
| Dyspnea on exertion, n (%) | 3941 (14.91) | 616 (14.47) | 1006 (14.25) | 1027 (14.39) | 1292 (16.19) | 0.002 |
| Angina, n (%) | 7851 (29.70) | 1269 (29.81) | 2150 (30.44) | 2029 (28.43) | 2403 (30.11) | 0.044 |
| Syncope, n (%) | 7183 (27.17) | 1203 (28.26) | 2012 (28.49) | 1920 (26.90) | 2048 (25.66) | <0.001 |
| Ejection fraction, n (%) | <0.001 | |||||
| <35 | 2076 (7.85) | 360 (8.46) | 559 (7.92) | 569 (7.97) | 588 (7.37) | |
| 35–49 | 2784 (10.53) | 413 (9.70) | 756 (10.71) | 752 (10.54) | 863 (10.81) | |
| 50+ | 14 100 (53.33) | 1914 (44.96) | 3803 (53.85) | 3904 (54.69) | 4479 (56.12) | |
| Unknown | 7478 (28.29) | 1570 (36.88) | 1944 (27.53) | 1913 (26.80) | 2051 (25.7) | |
| Creatinine, mg/dL, n (%) | <0.001 | |||||
| <1.0 | 8571 (32.42) | 1282 (30.12) | 2265 (32.07) | 2306 (32.31) | 2718 (34.06) | |
| 1.0–1.4 | 8269 (31.28) | 1250 (29.36) | 2145 (30.37) | 2286 (32.03) | 2588 (32.43) | |
| 1.5–1.9 | 2383 (9.01) | 399 (9.37) | 638 (9.03) | 636 (8.91) | 710 (8.90) | |
| 2.0+ | 2053 (7.77) | 410 (9.63) | 567 (8.03) | 556 (7.79) | 520 (6.52) | |
| Unknown | 5162 (19.52) | 916 (21.52) | 1447 (20.49) | 1354 (18.97) | 1445 (18.11) | |
| BMI, kg/m2, n (%) | <0.001 | |||||
| <20.1 | 1248 (4.72) | 276 (6.48) | 344 (4.87) | 298 (4.17) | 330 (4.13) | |
| 20.1–25.0 | 5970 (22.58) | 1079 (25.35) | 1616 (22.88) | 1599 (22.40) | 1676 (21.00) | |
| 25.1–30.0 | 8331 (31.51) | 1270 (29.83) | 2252 (31.89) | 2230 (31.24) | 2579 (32.31) | |
| 30.1+ | 9228 (34.90) | 1358 (31.90) | 2409 (34.11) | 2590 (36.28) | 2871 (35.97) | |
| Unknown | 1661 (6.28) | 274 (6.44) | 441 (6.24) | 421 (5.90) | 525 (6.58) | |
| Diagnosed in inpatient | 10 013 (37.87) | 1881 (44.19) | 2653 (37.57) | 2720 (38.11) | 2759 (34.57) | <0.001 |
| Percent hospitalized in year prior | 12 329 (46.63) | 2214 (52.01) | 3325 (47.08) | 3351 (46.95) | 3439 (43.09) | <0.001 |
| Region, n (%) | <0.001 | |||||
| Midwest | 13 332 (50.43) | 1793 (42.12) | 3684 (52.17) | 3937 (55.16) | 3918 (49.09) | |
| Northeast | 3037 (11.49) | 928 (21.80) | 780 (11.05) | 574 (8.04) | 755 (9.46) | |
| Other/unknown | 600 (2.27) | 104 (2.44) | 183 (2.59) | 141 (1.98) | 172 (2.16) | |
| South | 6244 (23.62) | 890 (20.91) | 1571 (22.25) | 1610 (22.56) | 2173 (27.23) | |
| West | 3225 (12.20) | 542 (12.73) | 844 (11.95) | 876 (12.27) | 963 (12.07) | |
| Year of diagnosis, n (%) | <0.001 | |||||
| 2011–2012 | 5688 (21.51) | 1041 (24.45) | 1545 (21.88) | 1651 (23.13) | 1451 (18.18) | |
| 2013–2014 | 8910 (33.7) | 1520 (35.71) | 2457 (34.79) | 2352 (32.95) | 2581 (32.34) | |
| 2015–2016 | 11 840 (44.78) | 1696 (39.84) | 3060 (43.33) | 3135 (43.92) | 3949 (49.48) | |
| Insurance, n (%) | <0.001 | |||||
| Commercial | 4914 (18.59) | 769 (18.06) | 1311 (18.56) | 1282 (17.96) | 1552 (19.45) | |
| Medicaid | 673 (2.55) | 105 (2.47) | 151 (2.14) | 241 (3.38) | 176 (2.21) | |
| Medicare | 13 089 (49.51) | 2401 (56.4) | 3676 (52.05) | 3372 (47.24) | 3640 (45.61) | |
| Other or unknown | 7232 (27.35) | 906 (21.28) | 1755 (24.86) | 2128 (29.81) | 2443 (30.61) | |
| Uninsured | 530 (2.00) | 76 (1.79) | 169 (2.39) | 115 (1.61) | 170 (2.13) | |
| Income level (25th, 75th percentiles)† | $40 125 ($35 814, $46 714) | $42 046 ($35 229, $47 758) | $40 125 ($35 268, $46 955) | $39 816 ($35 981, $44 376) | $40 550 ($35 020, $46 454) | <0.001 |
| Percent college educated (25th, 75th percentiles)† | 22.00 (18.00, 27.00) | 23.00 (17.00, 29.00) | 22.00 (19.00, 27.00) | 22.00 (19.00, 27.00) | 22.00 (18.00, 27.00) | <0.001 |
AVR indicates aortic valve replacement; BMI, body mass index; COPD, chronic obstructive pulmonary disease; MI, myocardial infarction; PCI, percutaneous coronary intervention; SAVR, surgical aortic valve replacement; and TAVR, transcatheter aortic valve replacement.
The fourth quartile represents the highest treating clinicians.
Area‐level variable (zip 3).
Kaplan–Meier curves were used to evaluate the survival associated with each quartile of AVR rates for managing cardiologists. Multilevel, multivariable Fine‐Gray competing risk models clustered by cardiologists were used to determine the impact of cardiologists on the likelihood of 1‐year AVR treatment. We adjusted for patient demographics and comorbidities listed in the Table. Subdistribution hazard ratios (HRs) were used to describe the impact of fixed covariates, and the MOR was used to describe the relative impact of the individual cardiologists on the likelihood of receiving the 1‐year AVR treatment. A cubic spline was used to assess the proportional hazards assumption.25 To model the relationship between cardiologist AVR treatment rate and 1‐year all‐cause mortality, the AVR rate was modeled as a restricted cubic spline with 4 degrees of freedom. A sensitivity analysis was performed to limit the impact of immortal time bias (Data S5). To evaluate the relative impact of cardiologists with high AVR rates on 1‐year all‐cause mortality, we conducted a sensitivity analysis by restricting the cohort to patients treated by cardiologists with AVR rates ≤70% (Data S6).
Imputation of missing variables with <10% of missing data was accomplished via multivariate imputation by chained equations using the version 2.9 package.26 Details are found in Data S7. All analyses were conducted using SAS 9.4 and R version 3.5.2 with P≤0.05 considered significant.
Ethical/Institutional Review Board
We did not obtain ethical/institutional review board approval because this was a retrospective study of deidentified EHR data.
RESULTS
The study cohort of 26 438 patients with ssAS included 45.9% women with a median age of 79 years (25th–75th percentiles, 70–84 years) and median Charlson Comorbidity Index of 2 (25th–75th percentiles, 0–3). Median follow‐up was 701 days (25th–75th percentiles, 395–1179 days). Dyspnea was the predominant symptom, affecting 90.4% of patients, with angina affecting 29.7% of patients. ssAS was primarily diagnosed in the outpatient setting (62.1%). Complete characteristics are presented in the Table.
Clinician Variability in AVR Use
The average general cardiologist saw a median of 10 newly diagnosed patients with ssAS during the study interval (25th–75th percentiles, 5–20). Within a year of the first report of symptoms, 35.6% of patients with ssAS underwent AVR (13.3% TAVR, 22.3% SAVR), and treatment rates increased from 26.8% in 2011 to 2012 to 40.9% for patients diagnosed in 2015 to 2016 (Figure 2).
Figure 2. Treatment rates stratified by TAVR/SAVR over time.
Overall, 35.6% patients with symptomatic severe aortic stenosis (ssAS) had aortic valve replacement (n=9407) in the year after date of first ssAS diagnosis, whereas 37.3% of patients who had aortic valve replacement underwent TAVR (n=3513). SAVR indicates surgical aortic valve replacement; and TAVR, transcatheter aortic valve replacement.
Rates of AVR within a year of the first symptom report varied substantially by cardiologist from 0% in the lowest quartile to 100% in the highest (median, 29.6%; 25th–75th percentiles, 13.3%–47.0%). The distribution of AVR rates is provided in Figure S3 and by quartiles in Figure S4. Ranges for AVR rates by quartile were the following: quartile 1 (lowest treatment rates), 0 to 0.130; quartile 2, 0.133 to 0.294; quartile 3, 0.296 to 0.469; and quartile 4 (highest treatment rates), 0.470 to 1.000. Adjusted median AVR rates ranged from 13.6% in the lowest quartile to 57.2% in the highest quartile. Likewise, among patients who underwent AVR, substantial variation was observed in the use of TAVR, with 0% in the lowest quartile to 100% in the highest (median, 25%; 25th–75th percentiles, 0%–50%; adjusted median TAVR rates [lowest, highest], 32.7, 46.3%).
Factors Associated With AVR
Characteristics most strongly associated with no AVR within 1 year included age ≥80 years (adjusted HR for AVR, 0.56; 95% CI, 0.51–0.63) and dementia (adjusted HR, 0.32; 95% CI, 0.24–0.42). Among those treated, patient characteristics most strongly associated with TAVR (versus SAVR) included age ≥80 years (adjusted HR for TAVR, 20.8; 95% CI, 16.3–26.7) and prior percutaneous coronary intervention (adjusted HR, 2.27; 95% CI, 1.57–3.30). The full models are presented in Table S8. There was a significant correlation between the managing cardiologist's AVR rate and their rate of TAVR, with a Pearson correlation coefficient of 0.63 (P<0.001; Figure S5).
The managing cardiologist was among the strongest determinants of treatment for ssAS when compared with other covariates. Using the MOR, patients had a 125% and 141% increased chance of receiving a different treatment strategy if they had randomly selected a different managing cardiologist for AVR (MOR, 2.25; 95% CI, 2.14–2.36) and TAVR (MOR, 2.41; 95% CI, 2.21–2.61), respectively. Similarly, the median HR27 was 1.265 (95% CI, 1.209–1.315; P<0.001), which demonstrates that the median increase in the hazard of mortality in the year after AVR was 26% when comparing a patient treated by a cardiologist with a lower referral rate with a patient treated by a cardiologist with a higher referral rate.
The strength of this association persisted over time and did not vary substantively by region (Table S9) or whether the managing cardiologist performed interventional procedures (adjusted MOR, 2.10; 95% CI, 1.89–2.31; non‐cardiologist‐adjusted MOR, 2.16; 95% CI, 2.04–2.28). By adjusting the model for volume of patients with ssAS and AVR volume, sensitivity analysis also showed the association persisted (Data S8). Treatment rates did vary substantively by the volume of patients with ssAS managed by the cardiologist (lowest tertile, highest tertile: 29.1%, 44.5%). Managing cardiologists with the highest volume of patients with ssAS by tertile had a slightly greater variation in AVR likelihood (adjusted MOR, 2.75; 95% CI, 2.37–3.12) compared with those with the lowest volume of patients with ssAS (adjusted MOR, 2.04; 95% CI, 1.89–2.19). A sensitivity analysis among a subset of patients with ejection fraction data showed that for patients with LVEF <50 (which is a Class I indication for AVR treatment), the MOR of AVR was 2.08 (95% CI, 1.87–2.28), similar to the MOR for the full model (2.25; 95% CI, 2.14–2.36). Further stratification showed similar results; the MOR of the likelihood of AVR was 2.03 (95% CI, 1.68–2.36) for LVEF <35, 2.01 (95% CI, 1.74–2.26) for LVEF 35–49, 2.18 (95% CI, 2.05–2.31) for LVEF ≥50, and 2.49 (95% CI, 2.30–2.69) for patients with LVEF “unknown,” indicating that similar results are observed when the analysis is stratified by LVEF (MOR 95% CIs overlap for groups LVEF <35 [most severe] to LVEF ≥50 [less severe]; patients with unknown LVEF were most likely to undergo AVR).
Association Between the Managing Cardiologist's ssAS Treatment Variability and Patient Survival
The managing cardiologists' rates of AVR were directly associated with the likelihood of 1‐year survival for their patients with ssAS (Figures 3 and 4). Patients managed by cardiologists in the highest quartile of treatment rates had a 1‐year survival rate of 81% compared with 73% for the lowest quartile. After adjusting for differences in patient characteristics, patients with ssAS cared for by cardiologists in the lowest quartile of AVR rates experienced a higher associated risk of mortality than those treated by managing cardiologists in the highest quartile of AVR rates (adjusted HR, 1.22; 95% CI, 1.13–1.33). The following 2 sensitivity analyses were conducted to evaluate the impact on mortality: (1) an analysis to limit the impact of immortal time bias by limiting the window for AVR treatment to 3 months showed similar results (results found in Data S5 and Figures S6 and S7), and (2) an evaluation of the impact of cardiologists when removing physicians with >70% AVR treatment rates also revealed similar findings (see Data S6 for analysis results).
Figure 3. Survival stratified by managing cardiologist treatment rate.
Kaplan–Meier curves for survival when stratified by managing cardiologist AVR treatment rate with 1 representing the lowest quartile of AVR rates at the 1‐year AVR rate and 4 the highest. Patients treated by cardiologists with higher AVR rates have a significantly higher survival at 1 year. The colored bands around each survival cure represent the 95% CI. The number of patients at risk at each 60‐day interval for each quartile are displayed below the survival curves. AVR indicates aortic valve replacement.
Figure 4. Association between the managing cardiologists' AVR treatment rate and 1‐year all‐cause mortality.
Association between managing cardiologists' 1‐year AVR treatment rate and 1‐year all‐cause mortality was modeled as a restricted cubic spline with 4 degrees of freedom. The hazard presented was adjusted for patient factors and demographics and demonstrates that a higher clinician 1‐year treatment rate is associated with a significantly reduced 1‐year mortality risk. The distribution of clinicians by 1‐year AVR rate is shown below the curve with each strike representing an individual clinician. The light blue band around the line represents the 95% CI. AVR indicates aortic valve replacement.
DISCUSSION
A substantial body of evidence documents the ability of AVR to prolong survival and alleviate suffering in patients with ssAS, resulting in the highest recommendation (Class 1A) for treating patients with ssAS with AVR in both US and European guidelines.1, 28 Despite widespread availability of this therapy, the majority of patients with severe AS in our study did not undergo AVR within a year of symptom development. The current study is unique in that we identified patients with ssAS from a large database, evaluated the receipt of AVR within a year of ssAS diagnosis, and were able to describe an association between patient outcomes and the AVR rate of the managing cardiologist.
The percentage of patients treated with AVR in our study is similar to a previously published study by Lancellotti et al, which showed AVR treatment rates of 30% for patients with moderate AS and 45.1% for patients with severe AS upon study entry.29 Nevertheless, these rates are not directly comparable with our results because the study by Lancellotti et al had longer follow‐up time, excluded symptomatic patients with AS, and included non‐US‐based clinics.29
Across managing cardiologists, there was marked variability in a patient's likelihood for AVR, with the managing cardiologist among the strongest determinants of AVR. This variation was clinically important; patients treated by cardiologists with low treatment rates had significantly lower 1‐year survival compared with those treated by cardiologists with high treatment rates. Although more research is needed to determine what exactly drives the apparent variability in use of AVR by cardiologists, our findings show that there might potentially be issues in the current management of patients diagnosed with ssAS and that reducing variability in AVR treatment rates may present an opportunity for meaningful improvement.
The decision not to treat a patient with ssAS is complex, including anatomic considerations, comorbidities that affect the likelihood of therapeutic benefit, and a patient's individual values and preferences. A recent analysis of 407 patients with ssAS evaluated at 9 US heart valve treatment centers from 2013 to 2014 (including 17% who were ultimately medically managed) demonstrated a 31% rate of patient refusal.30 Yet nearly one‐third of these patients also expressed uncertainty with their ultimate treatment strategy; this uncertainty raises concerns as to whether patients were adequately educated as well as whether they were encouraged to be participants in a shared clinician–patient decision‐making process. Nonetheless, considerations other than these affect the treatment options offered, including a patient's race and sex, as well as the local availability of treatment options.31 As a result, patients who would be candidates for treatment may not be given the option.
Our analysis revealed that although there are some predictable factors (such as increased patient age and dementia) that were negatively associated with AVR (versus no AVR; age >80 years HR, 0.52 [95% CI, 0.45–0.61]; dementia HR, 0.33 [95% CI, 0.23–0.46]; Table S8), other factors, such as education and the Charlson Comorbidity Index (up through 3) were not found to be associated with AVR versus no AVR (education in 80th quantile HR, 0.90 [95% CI, 0.76–1.07]; Charlson Comorbidity Index 3 HR, 0.94 [95% CI, 0.79–1.11]; Table S8). Our study is consistent with other research that has shown a gradual decline in number of SAVR cases and a dramatic increase in number of TAVR cases in recent years.32, 33 Because TAVR is recommended for patients with comorbidities that would preclude them from referral for SAVR,34 this may explain why our findings (our study used a definition of AVR that includes both TAVR and SAVR) showed no association between Charlson Comorbidity Index of 0 to 3 and AVR versus non‐AVR. Of note, increasing Charlson comorbidity scores (2, 3, or 4+) were associated with an increased likelihood of TAVR versus SAVR (HR, 1.65–2.43, respectively; Table S8), corresponding with recommendations for the use of TAVR in higher risk individuals. The lack of association between income or education and propensity for AVR versus non‐AVR may in part be attributed to insufficient granularity of the Optum database to accurately assess patient socioeconomic status because it provides geographic information only at the zip‐code level.
The use of a subjective surrogate (symptom report) for the recognition of clinically significant left ventricular strain may limit a more standardized application of AVR therapy. Of the individuals in our study, >90% presented with dyspnea (a nonspecific symptom that is notoriously difficult to elicit) as their primary clinical complaint. This finding is similar to other series35, 36, 37 and raises concerns regarding the use of subjective clinical symptoms as the primary trigger for the treatment of this fatal disease. The low treatment rates of ssAS observed in this US cohort may speak to a broader challenge in symptom attribution for patients with ssAS. In an often debilitated, comorbid, and elderly population, symptoms of dyspnea can be difficult to elicit in a limited clinical visit.7, 38 When a more intensive search for symptoms has been undertaken, 3 separate studies have demonstrated that the vast majority (>80%) of patients with severe AS experience valve‐associated symptoms.18, 39, 40 Likewise, objective markers such as pro–B‐type natriuretic peptide levels and 6‐minute walk distance41 have demonstrated left ventricular strain in large segments of patients who were otherwise asymptomatic. Clinical management guidelines have started a process of incorporating the use of these tests into their suggested treatment algorithms, and a more widespread use of objective metrics to trigger valve replacement for patients with ssAS may simplify clinical care, reduce treatment variability, and improve overall treatment rates in this terminal disease state.
Recognition of the scope of this possible performance gap is the first step to addressing the issue. Ensuring that AVR, a potentially life‐saving treatment, is offered to all appropriate patients with ssAS in the United States will involve collaborative efforts throughout the valvular heart disease ecosystem. An educated patient is more likely to choose treatment, and patient education cannot rely solely on clinician‐initiated programs. Although shared decision‐making is critical in this preference‐sensitive field, direct‐to‐patient educational programs from professional societies and patient advocacy groups may be helpful. Furthermore, both clinicians and health systems should examine their own practices to identify untreated patients and to implement systems that foster early recognition, active surveillance, and timely treatment of this terminal illness. With the availability of EHR analytics, new opportunities are emerging to take a more holistic, data‐driven view of patient status. Finally, policy makers should continue working to reduce barriers to care by ensuring local access to treatment42 and addressing the existing profit‐margin differentials that have the potential to influence treatment decisions between TAVR and SAVR.
Limitations
This study has several notable limitations. First, the identification of severe AS for cohort development was based on a review of clinician notes rather than echocardiography reports. This approach may be subject to error, but it was subsequently validated through both evaluation of stratified survival curves and comparison to echocardiography reports in a subset of the cohort. Second, we were not able to validate symptom status in the Optum database, and although attempts were made to validate symptoms by evaluating treatment rates in the claims set, there were too few patients to conduct this analysis. Third, we used multivariate imputation by chained equations to impute missing variables with <10% of missing data and created an “unknown” category for variables in which data >10% were missing, which may have served a potential source of bias. Fourth, 90.4% of patients in our study had dyspnea, which is a subjective symptom. Therefore, we acknowledge that there is a possibility that patients not referred for AVR were dyspneic for reasons other than severe AS or that the dyspnea was mild in the judgment of the clinician. Fifth, the treatment rates observed here are among patients with an existing diagnosis of severe AS. Although the scope of the problem has not been well described, it is expected that a sizable cohort of patients never reach diagnosis, including among certain vulnerable populations.43 As a result, the treatment rates observed here are likely an overestimate of actual treatment rates for this disease. Sixth, we identified managing cardiologists based on the frequency of interactions. This definition may not have accurately determined the “cardiology home” for some patients, particularly those with frequent hospitalizations; however, our results were consistent when applied to cardiologists in the outpatient setting. Seventh, we recognize that the decision to receive AVR involves a complex interaction between patient and clinician, as well as including patient values and preferences, neither of which could be evaluated in this study. Rates of patient refusal of AVR procedures are not recorded in the database and present a limitation; refusal rates are reported to be 20% to 33% (depending on race and other patient‐related characteristics).37 Eighth, in the survival benefit analysis, there is a risk of the immortal time bias since patients must survive long enough to undergo an AVR. A sensitivity analysis conducted using a shorter time window (3 months) from diagnosis to treatment revealed results similar to our primary analysis (1‐year window). Ninth, we acknowledge the possibility that cardiologists with higher AVR treatment rates (referring more often) also worked with surgeons with good outcomes and that this may have contributed to potential confounding. Tenth, our analyses were based on the assumption of random distribution of unmeasured confounding; however, similar to any observational analysis, this assumption may not hold and should be considered when interpreting the results. Finally, we acknowledge that the current study was conducted before the expansion of TAVR as an option for patients and that our results will need to be verified in future studies with more recent data.
CONCLUSIONS
In conclusion, a majority of patients in the United States with severe AS do not undergo AVR within a year of symptom development. We have identified substantial variation at the level of the managing cardiologist in the use of this therapy and an association between AVR treatment rates and 1‐year survival of patients with ssAS. Given the availability of effective treatments, there may be value in implementing targeted initiatives to raise disease awareness, promote more objective diagnostic criteria, and reduce barriers to treatment.
Sources of Funding
This study was funded by Edwards Lifesciences, Irvine, CA, which was not involved in the study design, analysis and interpretation of data, or the writing of the report.
Disclosures
Dr Brennan reports consulting and speaking funds from Edwards Lifesciences and AtriCure. Dr Boero reports consulting for Edwards Lifesciences. Dr Thourani reports research and advising for Edwards Lifesciences. Dr Vemulapalli reports grants/contracts from the American College of Cardiology, Society of Thoracic Surgeons, Abbott Vascular, Boston Scientific, National Institutes of Health (R01 and Small Business Innovation Research grants), Food and Drug Administration National Evaluation System for health Technology Coordinating Center (FDA NESTcc), and Cytokinetics and advisory board/consulting/honoraria with Boston Scientific, American College of Physicians, Janssen, Edwards Lifesciences, and HeartFlow. Dr Wang reports research grants to the Duke Clinical Research Institute from Abbott, AstraZeneca, Bristol Myers Squibb, Boston Scientific, Cryolife, Chiesi, Merck, Portola, and Regeneron and consulting honoraria from AstraZeneca, Bristol Myers Squibb, Cryolife, and Novartis. Mr Liska, Mr Gander, and Mr Jager report consulting for Edwards Lifesciences. Dr Peterson reports being a coinvestigator on the American College of Cardiology Society of Thoracic Surgeons Transcatheter Valve Therapy TAVR Registry. The remaining authors have no disclosures to report.
Supporting information
Data S1–S8
Tables S1–S9
Figures S1–S7
Acknowledgments
The authors thank Erin Campbell, MS, for her editorial contributions to this manuscript and Boye Gricar, PhD, Emily Farrar, PhD, and Sibyl Munson, PhD, of Boston Strategic Partners, Inc., for assistance with manuscript preparation.
(J Am Heart Assoc. 2021;10:e020490. DOI: 10.1161/JAHA.120.020490.)
Supplementary Material for this article is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.120.020490
For Sources of Funding and Disclosures, see page 11.
REFERENCES
- 1.Nishimura RA, Otto CM, Bonow RO, Carabello BA, Erwin JP III, Fleisher LA, Jneid H, Mack MJ, McLeod CJ, O’Gara PT, et al. 2017 AHA/ACC focused update of the 2014 AHA/ACC guideline for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2017;135:e1159–e1195. DOI: 10.1161/CIR.0000000000000503. [DOI] [PubMed] [Google Scholar]
- 2.Mack MJ, Leon MB, Smith CR, Miller DC, Moses JW, Tuzcu EM, Webb JG, Douglas PS, Anderson WN, Blackstone EH, et al.; PARTNER 1 Trial Investigators . 5‐year outcomes of transcatheter aortic valve replacement or surgical aortic valve replacement for high surgical risk patients with aortic stenosis (PARTNER 1): a randomised controlled trial. Lancet. 2015;385:2477–2484. DOI: 10.1016/S0140-6736(15)60308-7. [DOI] [PubMed] [Google Scholar]
- 3.Baron SJ, Arnold SV, Wang K, Magnuson EA, Chinnakondepali K, Makkar R, Herrmann HC, Kodali S, Thourani VH, Kapadia S, et al.; PARTNER 2 Investigators . Health status benefits of transcatheter vs surgical aortic valve replacement in patients with severe aortic stenosis at intermediate surgical risk: results from the PARTNER 2 randomized clinical trial. JAMA Cardiol. 2017;2:837–845. DOI: 10.1001/jamacardio.2017.2039. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.ClinicalTrials.gov . PARTNER 3 trial: the safety and effectiveness of the SAPIEN 3 transcatheter heart valve in low risk patients with aortic stenosis (P3). ClinicalTrials.gov web site. Updated April 8, 2021. Available at: https://clinicaltrials.gov/ct2/show/NCT02675114. Accessed November 1, 2019.
- 5.ClinicalTrials.gov Medtronic evolut transcatheter aortic valve replacement in low risk patients. ClinicalTrials.gov web site. Available at: https://clinicaltrials.gov/ct2/show/NCT02701283. Updated October 30, 2019. Accessed November 1, 2019.
- 6.Dua A, Dang P, Shaker R, Varadarajan P, Pai RG. Barriers to surgery in severe aortic stenosis patients with Class I indications for aortic valve replacement. J Heart Valve Dis. 2011;20:396–400. [PubMed] [Google Scholar]
- 7.Bonow RO, Leon MB, Doshi D, Moat N. Management strategies and future challenges for aortic valve disease. Lancet. 2016;387:1312–1323. DOI: 10.1016/S0140-6736(16)00586-9. [DOI] [PubMed] [Google Scholar]
- 8.Jones WS, Patel MR, Dai D, Subherwal S, Stafford J, Calhoun S, Peterson ED. Temporal trends and geographic variation of lower‐extremity amputation in patients with peripheral artery disease: results from U.S. Medicare 2000–2008. J Am Coll Cardiol. 2012;60:2230–2236. DOI: 10.1016/j.jacc.2012.08.983. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Marzec LN, Peterson PN, Bao H, Curtis JP, Masoudi FA, Varosy PD, Bradley SM. Use of cardiac resynchronization therapy among eligible patients receiving an implantable cardioverter defibrillator: insights from the National Cardiovascular Data Registry Implantable Cardioverter Defibrillator Registry. JAMA Cardiol. 2017;2:561–565. DOI: 10.1001/jamacardio.2016.5388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Messika‐Zeitoun D, Verta P, Gregson J, Pocock SJ, Boero I, Feldman TE, Abraham WT, Lindenfeld J, Bax J, Leon M, et al. Impact of tricuspid regurgitation on survival in patients with heart failure: a large electronic health record patient‐level database analysis. Eur J Heart Fail. 2020;22:1803–1813. DOI: 10.1002/ejhf.1830. [DOI] [PubMed] [Google Scholar]
- 11.Matthew Brennan J, Leon MB, Sheridan P, Boero IJ, Chen Q, Lowenstern A, Thourani V, Vemulapalli S, Thomas K, Wang TY, et al. Racial differences in the use of aortic valve replacement for treatment of symptomatic severe aortic valve stenosis in the transcatheter aortic valve replacement era. J Am Heart Assoc. 2020;9:e015879. DOI: 10.1161/JAHA.119.015879. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wallace PJ, Shah ND, Dennen T, Bleicher PA, Crown WH. Optum Labs: building a novel node in the learning health care system. Health Aff (Millwood). 2014;33:1187–1194. DOI: 10.1377/hlthaff.2014.0038. [DOI] [PubMed] [Google Scholar]
- 13.Halpern R, Seare J, Tong J, Hartry A, Olaoye A, Aigbogun MS. Using electronic health records to estimate the prevalence of agitation in Alzheimer disease/dementia. Int J Geriatr Psychiatry. 2019;34:420–431. DOI: 10.1002/gps.5030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Nunes AP, Iglay K, Radican L, Engel SS, Yang J, Doherty MC, Dore DD. Hypoglycaemia seriousness and weight gain as determinants of cardiovascular disease outcomes among sulfonylurea users. Diabetes Obes Metab. 2017;19:1425–1435. DOI: 10.1111/dom.13000. [DOI] [PubMed] [Google Scholar]
- 15.Nunes AP, Liang C, Gradishar WJ, Dalvi T, Lewis J, Jones N, Green E, Doherty M, Seeger JD. U.S. prevalence of endocrine therapy‐naïve locally advanced or metastatic breast cancer. Curr Oncol. 2019;26:e180–e187. DOI: 10.3747/co.26.4163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Loughlin AM, Lin N, Abler V, Carroll B. Tardive dyskinesia among patients using antipsychotic medications in customary clinical care in the United States. PLoS One. 2019;14:e0216044. DOI: 10.1371/journal.pone.0216044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bach DS, Siao D, Girard SE, Duvernoy C, McCallister BD, Gualano SK. Evaluation of patients with severe symptomatic aortic stenosis who do not undergo aortic valve replacement: the potential role of subjectively overestimated operative risk. Circ Cardiovasc Qual Outcomes. 2009;2:533–539. DOI: 10.1161/CIRCOUTCOMES.109.848259. [DOI] [PubMed] [Google Scholar]
- 18.Campo J, Tsoris A, Kruse J, Karim A, Andrei AC, Liu M, Bonow RO, McCarthy P, Malaisrie SC. Prognosis of severe asymptomatic aortic stenosis with and without surgery. Ann Thorac Surg. 2019;108:74–79. DOI: 10.1016/j.athoracsur.2019.01.031. [DOI] [PubMed] [Google Scholar]
- 19.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD‐9‐CM administrative databases. J Clin Epidemiol. 1992;45:613–619. DOI: 10.1016/0895-4356(92)90133-8. [DOI] [PubMed] [Google Scholar]
- 20.Merchant RM, Berg RA, Yang L, Becker LB, Groeneveld PW, Chan PS; American Heart Association’s Get With the Guidelines‐Resuscitation Investigators . Hospital variation in survival after in‐hospital cardiac arrest. J Am Heart Assoc. 2014;3:e000400. DOI: 10.1161/JAHA.113.000400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Chan PS, Berg RA, Spertus JA, Schwamm LH, Bhatt DL, Fonarow GC, Heidenreich PA, Nallamothu BK, Tang F, Merchant RM; AHA GWTG‐Resuscitation Investigators . Risk‐standardizing survival for in‐hospital cardiac arrest to facilitate hospital comparisons. J Am Coll Cardiol. 2013;62:601–609. DOI: 10.1016/j.jacc.2013.05.051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Merlo J, Chaix B, Ohlsson H, Beckman A, Johnell K, Hjerpe P, Råstam L, Larsen K. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60:290–297. DOI: 10.1136/jech.2004.029454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol. 2005;161:81–88. DOI: 10.1093/aje/kwi017. [DOI] [PubMed] [Google Scholar]
- 24.Larsen K, Petersen JH, Budtz‐Jørgensen E, Endahl L. Interpreting parameters in the logistic regression model with random effects. Biometrics. 2004;56:909–914. DOI: 10.1111/j.0006-341X.2000.00909.x. [DOI] [PubMed] [Google Scholar]
- 25.Hess KR. Assessing time‐by‐covariate interactions in proportional hazards regression models using cubic spline functions. Stat Med. 1994;13:1045–1062. DOI: 10.1002/sim.4780131007. [DOI] [PubMed] [Google Scholar]
- 26.van Buuren S, Groothuis‐Oudshoom K. mice: multivariate imputation by chained equations in R. J Stat Softw. 2011;45:1–67. [Google Scholar]
- 27.Austin PC, Wagner P, Merlo J. The median hazard ratio: a useful measure of variance and general contextual effects in multilevel survival analysis. Stat Med. 2017;36:928–938. DOI: 10.1002/sim.7188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Nishimura RA, Otto CM, Bonow RO, Carabellow BA, Erwin JP III, Guyton RA, O’Gara PT, Ruiz CE, Skubas NJ, Sorajja P, et al.; American College of Cardiology/American Heart Association Task Force on Practice Guidelines . 2014 AHA/ACC guideline for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. 2014;63:e57–e185. DOI: 10.1016/j.jacc.2014.02.536. [DOI] [PubMed] [Google Scholar]
- 29.Lancellotti P, Magne J, Dulgheru R, Clavel MA, Donal E, Vannan MA, Chambers J, Rosenhek R, Habib G, Lloyd G, et al. Outcomes of patients with asymptomatic aortic stenosis followed up in heart valve clinics. JAMA Cardiol. 2018;3:1060–1068. DOI: 10.1001/jamacardio.2018.3152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Dharmarajan K, Foster J, Coylewright M, Green P, Vavalle JP, Faheem O, Huang P‐H, Krishnaswamy A, Thourani VH, McCoy LA, et al. The medically managed patient with severe symptomatic aortic stenosis in the TAVR era: patient characteristics, reasons for medical management, and quality of shared decision making at heart valve treatment centers. PLoS One. 2017;12:e0175926. DOI: 10.1371/journal.pone.0175926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Brennan JM, Bryant A, Boero I, Wang T, Vemulapalli S, Thourani V, Leon M, Peterson E. Race and sex‐based disparities persist in the treatment of patients with severe, symptomatic aortic valve stenosis (Abstract number 1207). J Am Coll Cardiol. 2019;73(Suppl. 1):1207. DOI: 10.1016/S0735-1097(19)31814-5.30871704 [DOI] [Google Scholar]
- 32.Kim KM, Shannon F, Paone G, Lall S, Batra S, Boeve T, DeLucia A, Patel HJ, Theurer PF, He C, et al. Evolving trends in aortic valve replacement: a statewide experience. J Card Surg. 2018;33:424–430. DOI: 10.1111/jocs.13740. [DOI] [PubMed] [Google Scholar]
- 33.Badhwar V, Rankin JS, Thourani VH, D'Agostino RS, Habib RH, Shahian DM, Jacobs JP. The Society of Thoracic Surgeons adult cardiac surgery database: 2018 update on research: outcomes analysis, quality improvement, and patient safety. Ann Thorac Surg. 2018;106:8–13. DOI: 10.1016/j.athoracsur.2018.04.052. [DOI] [PubMed] [Google Scholar]
- 34.Finn M, Green P. Transcatheter aortic valve implantation in the elderly: who to refer? Prog Cardiovasc Dis. 2014;57:215–225. DOI: 10.1016/j.pcad.2014.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Martínez‐Sellés M, Gómez Doblas JJ, Carro Hevia A, García de la Villa B, Ferreira‐González I, Alonso Tello A, Andión Ogando R, Ripoll Vera T, Arribas Jiménez A, Carrillo P, et al.; PEGASO Registry Group . Prospective registry of symptomatic severe aortic stenosis in octogenarians: a need for intervention. J Intern Med. 2014;275:608–620. DOI: 10.1111/joim.12174. [DOI] [PubMed] [Google Scholar]
- 36.Hartzell M, Malhotra R, Yared K, Rosenfield HR, Walker JD, Wood MJ. Effect of gender on treatment and outcomes in severe aortic stenosis. Am J Cardiol. 2011;107:1681–1686. DOI: 10.1016/j.amjcard.2011.01.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Yeung M, Kerrigan J, Sodhi S, Huang PH, Novak E, Maniar H, Zajarias A. Racial differences in rates of aortic valve replacement in patients with severe aortic stenosis. Am J Cardiol. 2013;112:991–995. DOI: 10.1016/j.amjcard.2013.05.030. [DOI] [PubMed] [Google Scholar]
- 38.Taniguchi T, Morimoto T, Shiomi H, Ando K, Kanamori N, Murata K, Kitai T, Kawase Y, Izumi C, Miyake M, et al.; CURRENT AS Registry Investigators . Initial surgical versus conservative strategies in patients with asymptomatic severe aortic stenosis. J Am Coll Cardiol. 2015;66:2827–2838. DOI: 10.1016/j.jacc.2015.10.001. [DOI] [PubMed] [Google Scholar]
- 39.Attias D, Macron L, Dreyfus J, Monin JL, Brochet E, Lepage L, Hekimian G, Lung B, Vahanian A, Messika‐Zeitoun D. Relationship between longitudinal strain and symptomatic status in aortic stenosis. J Am Soc Echocardiogr. 2013;26:868–874. DOI: 10.1016/j.echo.2013.05.004. [DOI] [PubMed] [Google Scholar]
- 40.Goliasch G, Kammerlander AA, Nitsche C, Dona C, Schachner L, Öztürk B, Binder C, Duca F, Aschauer S, Laufer G, et al. Syncope: the underestimated threat in severe aortic stenosis. JACC Cardiovasc Imaging. 2019;12:225–232. DOI: 10.1016/j.jcmg.2018.09.020. [DOI] [PubMed] [Google Scholar]
- 41.Redfors B, Pibarot P, Gillam LD, Burkhoff D, Bax JJ, Lindman BR, Bonow RO, O'Gara PT, Leon MB, Généreux P. Stress testing in asymptomatic aortic stenosis. Circulation. 2017;135:1956–1976. DOI: 10.1161/CIRCULATIONAHA.116.025457. [DOI] [PubMed] [Google Scholar]
- 42.Vora AN, Vemulapalli S. Geographic dispersion of TAVR services: ensuring availability while maintaining quality. Am Heart J. 2016;177:160–162. DOI: 10.1016/j.ahj.2016.04.005. [DOI] [PubMed] [Google Scholar]
- 43.Beydoun HA, Beydoun MA, Liang H, Dore GA, Shaked D, Zonderman AB, Eid SM. Sex, race, and socioeconomic disparities in patients with aortic stenosis (from a nationwide inpatient sample). Am J Cardiol. 2016;118:860–865. DOI: 10.1016/j.amjcard.2016.06.039. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1–S8
Tables S1–S9
Figures S1–S7