Abstract
Background:
Administrative claims for aortic stenosis (AS) regurgitation (AR) may be useful but their accuracy and ability to identify individuals at risk for valve-related outcomes has not been well characterized.
Methods:
Using echocardiographic (TTE) reports linked to US Medicare claims, 2017–2018, the performance of candidate International Classification of Diseases, 10th Revision (ICD-10) claims to ascertain AS/AR was evaluated. The optimal performing algorithm was tested against outcomes at 1-year after TTE in a separate 100% sample of US Medicare claims, 2017–2019.
Results:
Of those included in the derivation (N = 5497, mean age 74.4 ± 11.0 years, 49.7% female), any AS or AR was present in 24% and 38.8%, respectively. The sensitivity and specificity of ICD-10 code I35.0 for identification of any AS was 53.1% and 94.8%, respectively. Amongst those with an I35.0 code, 40.3% had severe AS. Claims were unable to distinguish disease severity (i.e. severe vs. non-severe) or subtype (e.g. bicuspid or rheumatic AS), and were insensitive and nonspecific for AR of any severity. Among all beneficiaries who received a TTE (N = 4,033,844), adjusting for age, sex, and 27 comorbidities, those with an I35.0 code had a higher adjusted risk of all-cause mortality (adjusted hazard ratio [HR] 1.33, 95% CI 1.31–1.34), heart failure hospitalization (adjusted HR 1.37, 95% CI 1.34–1.41), and aortic valve replacement (adjusted HR 34.96, 95% CI 33.74–36.22).
Conclusions:
Among US Medicare beneficiaries receiving a TTE, ICD-10 claims, though identifying a population at significant greater risk of valve-related outcomes, failed to identify nearly half of individuals with AS and were unable to distinguish disease severity or subtype. These results argue against the widespread use of ICD-10 claims to screen for patients with AS and suggest the need for improved coding algorithms and alternative systems to extract TTE data for quality improvement and hospital benchmarking.
Keywords: Aortic stenosis, aortic regurgitation, echocardiography, claims, ICD-10
INTRODUCTION:
Aortic stenosis (AS) is estimated to be the most prevalent form of valvular heart disease in developed countries, affecting 11.0% of those ≥ 75 years1–4, with progression to severe AS being universally fatal without treatment5. Based on current projections, with aging of the population and improvements in coronary heart disease survival, the expected number of deaths from valvular heart disease without treatment is expected to double over the next 25 years, with a disproportionate impact on older aged adults4,6. At the same time, the perioperative risk for aortic valve replacement (AVR) has declined significantly, particularly since the advent of transcatheter aortic valve replacement (TAVR) in 2002, which made treatments available to many previously at a prohibitively high risk of operation.7–9
Despite these promising developments, the true prevalence of AS as well as its associated medical conditions and impact on healthcare utilization remain unknown due to absence of universal screening. Current prevalence estimates are based on large prospective cohort studies in which echocardiography was performed, and thus reflect the bias of those enrolled in these cohorts1–4. Moreover, it remains challenging to identify cohorts of individuals at future risk of valve-related adverse events who may benefit from treatment.
In this setting, administrative claims represent a promising source of information on an individual’s interactions with the healthcare system that may prove beneficial in identifying populations with aortic valve disease at risk for undertreatment. Claims capture the burden of healthcare utilization in this population and may be useful for cohort identification and characterization as well as projections regarding future treatment needs and costs. Despite the existence of specific claims for AS and aortic regurgitation (AR), the validity of these claims in identifying aortic valve disease on echocardiography remains unclear, due to the absence of echocardiographic data linked to complete claims, and represents an important gap in evaluating patterns of aortic valve disease treatment nationally. Moreover, it remains unclear if the validity of claims for identifying aortic valve disease differs by severity of disease, hemodynamics (e.g. low flow vs. normal transvalvular flow), or etiologic subtype (e.g. congenital and rheumatic disease).
As such, in this study, we evaluated the agreement between International Classification of Diseases, 10th revision (ICD-10) claims for AS and AR and echocardiographic measures of AS and AR severity in a large single-center echocardiographic report dataset linked to Medicare Fee-for-service (FFS) claims, to our knowledge, the only such health-system dataset of its type in current existence. We aimed to 1) evaluate the validity of ICD-10 claims for aortic valve disease against echocardiographic measures as the reference standard, 2) to evaluate how the performance of codes differed across disease subsets, and 3) to test whether the optimal coding algorithm identified individuals at higher risk for valve-related adverse outcomes in an external dataset of Medicare FFS beneficiaries.
METHODS
Study Population
The data supporting this study are not available as per prior data use agreements with the Centers for Medicare and Medicaid Services. We evaluated transthoracic echocardiogram (TTE) reports from the Beth Israel Deaconess Medical Center (BIDMC) and four affiliated satellite sites linked to Medicare FFS claims as part of the Echocardiography and Cardiovascular Health Outcomes (ECHO) study. As part of routine care across the BIDMC network from July 6, 2000 to September 4, 2018, TTE data were entered by attending TTE readers into an electronic dataset which stored structured quantitative and qualitative information from TTE reports. Data from Medicare beneficiaries were previously linked to 100% Medicare FFS ICD-10 claims, available October 1, 2015 to December 31, 2017, and have been used in prior research10. Medicare beneficiary summary files were used to ascertain patient vital status information.
For the current study, only individuals in the TTE dataset with linked Medicare claims from January 1, 2017 to December 31, 2018 were considered to allow for a full calendar year of historical claims information for all individuals. Only an individual’s most recent TTE in the dataset was included to avoid double counting of individuals. Individuals missing aortic valve profiling information (i.e. either peak aortic transvalvular velocity [PAV], mean aortic transvalvular gradient [MG], or aortic valve area [AVA]) or those with a prior aortic valve replacement or repair reported on TTE were excluded. Additionally, those individuals without continuous enrollment in the Medicare FFS program for the 12 months prior to the date of service for the TTE and 3 months following (or death during this period) were excluded as were those without any claims during this time period. As the majority of individuals in the dataset had only one TTE, claims within the 3 months following the TTE were included to identify corresponding claims for individuals whose initial diagnosis of aortic valve disease was on the index TTE. This study was approved by the Institutional Review Board at Beth Israel Deaconess Medical Center with a waiver of informed consent.
Definition of Aortic Valve Disease
All TTEs were performed according to American Society of Echocardiography (ASE) guidelines11. Individuals were classified based on standard TTE criteria12,13 as having: a) no AS (MG < 10.0 mmHg or PAV < 2.0 m/s), b) mild AS (MG 10.0–19.9 mmHg or PAV 2.0–2.9 m/s and AVA > 1.0 cm2), c) moderate AS (MG 20.0–39.9 mmHg or PAV 3.0–3.9 m/s and AVA > 1.0 cm2), and d) severe AS (MG ≥ 40.0 mmHg or PAV ≥ 4.0 m/s or AVA ≤ 1.0 cm2). Additionally, consistent with published guidelines, an AVA ≤ 1.0 cm2 was used to reclassify those with severe, low-flow, low-gradient (LFLG) AS (AVA ≤ 1.0 cm2 and a MG < 40 mmHg or PAV < 4.0 m/s) and severe, high-gradient AS (MG ≥ 40.0 mmHg or PAV ≥ 4.0 m/s). AR severity was graded using an integrative approach as recommended by ASE guidelines14 as no AR (0+), mild (1+), moderate (2+), moderate-to-severe (3+), and severe (4+). Trace AR was categorized as mild for the analysis. Severity was confirmed by blinded core lab adjudication in a random subset of 20 patients by a single experienced echocardiographer (J.B.S.).
Aortic Valve Claims
The presence of any ICD-10 claim for AS (I35.0, I35.2, I35.8, I35.9, I08.X, Q23.0, I06.X, I42.1) or AR (I35.1, I35.2, I35.8, I35.9, I08.X, Q23.1, I06.X) (Supplemental Table S1) in the 12 months preceding the date of TTE or 3 months following was ascertained and used to create indicator variables for the presence or absence of a given claim during this timeframe.
Additional Covariates
Individual demographic and patient data were recorded in the TTE report at the time of image acquisition, including age, self-reported sex, systolic and diastolic blood pressures, patient reported height and weight (used to derive body mass index15 and body surface area16 using the Mosteller equation), inpatient/outpatient status, and image technical quality. Race and ethnicity were determined using the self-reported categorization included in the Medicare Beneficiary Summary File.
Echocardiographic measures included moderate or greater mitral or tricuspid regurgitation (assessed using ASE guideline-recommended integrative techniques), peak tricuspid regurgitant velocity, left ventricular ejection fraction (physician-reported), left atrial volume index, left ventricular end-diastolic and end-systolic dimension, left ventricular mass index, transmitral E/e’ ratio, transmitral E/A ratio, and stroke volume index. Additionally, the presence of rheumatic aortic valve deformity and a bicuspid aortic valve were identified using structured fields in the dataset.
Comorbidities were assigned using Medicare Chronic Conditions Warehouse indicator variables17, which use validated algorithms based on Medicare inpatient and outpatient claims within the two years prior to the date of index TTE to ascertain disease status for distinct chronic illness categories. Additionally, history of mitral or tricuspid valve interventions were identified based on appearance on the index TTE.
Statistical Analysis
Demographic, echocardiographic, and comorbidity information at baseline for the cohort are described using means and standard deviations or medians and interquartile ranges for continuous variables and counts and proportions for categorical variables. Core lab adjudicated AS/AR severity, blinded to an individual’s AS/AR status and severity, was compared to dataset values using kappa statistics. For each candidate claim, the sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and positive (+LR) and negative (−LR) likelihood ratios for identification of any AS on TTE (as the reference standard) were calculated. Unweighted kappa statistics and 95% confidence intervals (CIs) were used as measures of agreement between each claim and presence of any AS on TTE. This process was repeated considering several combinations of claims including presence of any candidate AS claim, presence of any claim except for I35.9 (which was found on preliminary analysis to be nonspecific for AS), and presence of any claim except for I35.9 and I42.1 (which may represent subaortic stenosis). Amongst individuals with any AR (≥ 1+), the above process was repeated to identify the performance of each candidate ICD-10 claim as well as the presence of any candidate AR claim. Subsequently, the above process was repeated for each candidate code and code combination across differing levels of AS and AR severity on TTE (e.g. identification of code performance for identifying severe AS/AR vs. less than severe disease on TTE).
Sensitivity Analyses
Several sensitivity analyses were performed to identify the optimal coding algorithm for AS/AR. The optimal algorithm performance was identified as the coding set with the highest kappa statistic for agreement with AS/AR definitions on TTE. To evaluate the role of code position, the above process was repeated for candidate AS/AR claims in the primary vs. all coding positions. To evaluate disease subtypes, this process was repeated for candidate codes to evaluate to identify LFLG severe AS, bicuspid aortic valve disease with any AS/AR, and rheumatic AS/AR on TTE. As those with LFLG moderate AS could have been categorized as mild AS, these individuals were reclassified as having LFLG AS and the performance of candidate codes to identify LFLG AS was reassessed. Furthermore, as the restriction to codes within 3 months after the TTE date could impact sensitivity, a 12-month window after the date of TTE was used to see if this would improve coding performance. Furthermore, change in code performance was evaluated if there was a concurrent code for heart failure (HF; ICD-10 I50.X) in any position in the 12 months preceding the TTE.
Finally, as performance of claims algorithms may be optimistic when tested in the derivation dataset, a data-driven classification tree approach using recursive partitioning was used to identify the optimal combination of codes to identify AS, accounting for model optimism using 10-fold cross validation. In building the classification trees, both Entropy and Gini index criteria were evaluated to determine each tree split, and cost complexity method was performed to prune the classification tree to avoid overfitting. First, recursive partitioning was run with all candidate ICD-10 diagnosis codes and procedural codes with greater than 5 occurrences. Second, this process was repeated including all ICD-10 codes for these individuals plus age and sex. Third, to identify if recursive partitioning could better discriminate AS severity, this process was repeated calculating the weighted kappa statistic for agreement between severity of AS ascertained by claims and severity on TTE. Based on the results of these analyses, the optimal coding algorithm to identify AS was identified for testing in the overall Medicare dataset to determine if this algorithm identified individuals at risk for valve-related adverse events.
Utility of Optimal Claims Algorithm in Medicare Data
The performance of the optimal coding algorithm to predict valve-related outcomes of interest was evaluated in 100% Medicare FFS inpatient and outpatient claims, 2017–2020. Individuals were included if they had a code for a comprehensive TTE (Current Procedural Terminology code 93306), had been covered by the Medicare FFS program during 12-months prior to and 3 months following the date of their TTE, and survived at least 3 months following their index TTE. Aortic stenosis was defined similarly using 12-months of historical claims prior to the date of TTE and up to 3 months after. We evaluated three endpoints: all-cause mortality, receipt of TAVR or SAVR, and first hospitalization for HF. Time to mortality was evaluated using Kaplan-Meier curves in those with AS vs. no AS and compared using log-rank tests (censoring on December 31, 2020 for those without an antecedent event). The date 3 months after the TTE date was used to define time zero for the survival analysis. Death information was complete for all individuals. For those with multiple AS claims or TTEs, only the most recent was considered. One-year all-cause mortality rates for those with AS were determined.
Other endpoints evaluated included receipt of TAVR (defined using ICD-10 Procedural Coding System claims: 02RF37H, 02RF37Z, 02RF38H, 02RF38Z, 02RF3JH, 02RF3JZ, 02RF3KH, and 02RF3KZ) or SAVR (ICD-10-PCS claims: 02RF07Z, 02RF08Z, 02RF0KZ, 02RF0JZ), and subsequently first hospitalization for HF (defined as a principal discharge diagnosis of I50.X), using cumulative incidence functions to determine 1-year event rates for these secondary endpoints, and Fine-Gray techniques18 to account for the competing risk of interval mortality. Adjusted hazard ratios (HR), comparing those with and without AS according to the optimal coding algorithm, were computed using Cox Proportional Hazards models, for all outcomes, adjusting for age, sex, and all 27 Medicare Chronic Conditions Warehouse variables. As the goal of this evaluation was to determine the magnitude of the association with outcomes, not whether any association was independent of demographic and clinical factors, no multivariable adjustment was performed. All analyses were conducted using JMP v16.0 or SAS v9.4 (SAS Institute, Cary, NC) using a two-tailed alpha < 0.05 to define significance.
RESULTS
Overall Results
Of 9879 individuals with TTE measurements, 2017–2018 in the CMS-linked dataset, 5497 (55.6%) met inclusion criteria (Figure 1). Of these, 1319 (24.0%) had any AS and 2224 (38.8%) had any AR. The mean age of the cohort was 74.4 ± 11.0 years, 49.7% were female, and 76.3% were Caucasian (Table 1). Comorbidities were common with ischemic heart disease present in 60.6%, hypertension in 82.3%, and diabetes mellitus in 38.3%. Agreement between core lab adjudication of AS/AR severity and database designation was high (AS: kappa, 0.90, 95% CI 0.71–1.00; AR: kappa, 0.92, 95% CI 0.77–1.00) (Supplemental Table S2).
Figure 1: Flowchart of Included Individuals in the Derivation Cohort.
Shown is a flowchart illustrating those included in the derivation cohort (the Beth Israel Deaconess Medical Center TTE dataset linked to Medicare Fee-for-service claims). Of a total of 9,879 individuals in the original Medicare-linked dataset, 2017–2018, a total of 2,893 individuals (29.3%) were excluded due to lack of consecutive Medicare Fee-for-service enrollment in the 12 months prior or 3 months following the TTE or death during this time interval, 573 (5.8%) were excluded due to lack of any claims during the same time window, and 916 (9.3%) due to lack of aortic valve profiling, resulting in a final sample size of 5,497 individuals. N = number of individuals, TTE = transthoracic echocardiogram.
Table 1:
Baseline Characteristics of Included Patients
| Variable | N obs | Total Cohort (N = 5497) |
|---|---|---|
| Demographics | ||
| Age (y) | 5497 | 74.4 ± 11.0 |
| Female sex – no. (%) | 5497 | 2732 (49.7) |
| Other | 376 (6.8) | |
| Hispanic Ethnicity – no. (%) | 5176 | 89 (1.6) |
| Inpatient – no. (%) | 5497 | 2549 (46.4) |
| Suboptimal image quality – no. (%) | 5497 | 1180 (21.5) |
| Systolic blood pressure (mmHg) | 5470 | 132.2 ± 21.9 |
| Diastolic blood pressure (mmHg) | 5462 | 71.6 ± 12.6 |
| Body mass index (kg/m2) | 5493 | 28.6 ± 6.5 |
| Body surface area (m2) | 5493 | 1.92 ± 0.28 |
| Comorbidities | ||
| Dementia – no. (%) | 5497 | 1277 (23.2) |
| Acute myocardial infarction – no. (%) | 5497 | 508 (9.2) |
| Anemia – no. (%) | 5497 | 2841 (51.7) |
| Asthma – no. (%) | 5497 | 628 (11.4) |
| Atrial fibrillation – no. (%) | 5497 | 1454 (26.5) |
| Prostate | 308 (5.6) | |
| Congestive heart failure – no. (%) | 5497 | 2666 (48.5) |
| Chronic kidney disease – no. (%) | 5497 | 2848 (51.8) |
| Chronic obstructive pulmonary disease and bronchiectasis – no. (%) | 5497 | 1170 (21.3) |
| Diabetes mellitus – no. (%) | 5497 | 2103 (38.3) |
| Hyperlipidemia – no. (%) | 5497 | 3627 (66.0) |
| Hypertension – no. (%) | 5497 | 4526 (82.3) |
| Acquired hypothyroidism – no. (%) | 5497 | 1089 (19.8) |
| Ischemic heart disease – no. (%) | 5497 | 3333 (60.6) |
| Depressive disorders – no. (%) | 5497 | 1837 (33.4) |
| Rheumatoid arthritis/osteoarthritis – no. (%) | 5497 | 2087 (38.0) |
| Stroke/transient ischemic attack – no. (%) | 5497 | 656 (11.9) |
| History of mitral valve intervention– no. (%) | 5497 | 175 (3.2) |
| History of tricuspid valve intervention – no. (%) | 5947 | 28 (0.5) |
| Laboratories and Medications | ||
| NT-proBNP – median (IQR) | 382 | 1952 (523–6146) |
| Cholesterol medications – no. (%) | 1912 | 1004 (18.3) |
| Antiplatelet medications – no. (%) | 1912 | 854 (15.5) |
| Anticoagulant medications – no. (%) | 1912 | 295 (5.4) |
| Beta blockers – no. (%) | 1912 | 857 (15.6) |
| Renin-angiotensin-aldosterone inhibitors – no. (%) | 1912 | 739 (13.4) |
| Diuretics – no. (%) | 1912 | 613 (11.2) |
| Antiarrhythmic medications – no. (%) | 1912 | 792 (14.4) |
| Insulin – no. (%) | 1912 | 135 (2.5) |
| Non-insulin diabetic medications – no. (%) | 1912 | 257 (4.7) |
| Nitrates – no. (%) | 1912 | 168 (3.1) |
| Digoxin/digitalis – no. (%) | 1912 | 24 (0.4) |
| Psychiatric medications – no. (%) | 1912 | 612 (11.1) |
| Anti-inflammatory medications – no. (%) | 1912 | 901 (16.4) |
| Echocardiographic Measures | ||
| Peak tricuspid regurgitant velocity (m/s) | 4476 | 2.7 ± 0.5 |
| Moderate or greater mitral regurgitation – no. (%) | 5155 | 939 (17.1) |
| Moderate or greater tricuspid regurgitation – no. (%) | 5038 | 999 (18.2) |
| Left ventricular ejection fraction (%) | 5495 | 61.3 ± 15.6 |
| Left atrial volume index (mL/m2) | 2885 | 32.2 ± 11.7 |
| Left ventricular end-diastolic dimension (cm) | 4894 | 4.5 ± 0.8 |
| Left ventricular end-systolic dimension (cm) | 3412 | 2.9 ± 0.8 |
| Left ventricular mass index (g/m2) | 4861 | 113.2 ± 36.2 |
| Transmitral E/e’ ratio | 4069 | 11.8 ± 5.7 |
| Transmitral E/A ratio | 4540 | 1.0 ± 0.6 |
| Stroke volume index (mL/m2) | 4302 | 38.4 ± 12.1 |
| Rheumatic aortic valve deformity – no. (%) | 5497 | 26 (0.5) |
| Bicuspid aortic valve – no. (%) | 5497 | 33 (0.6) |
Displayed are baseline characteristics (at the time of performance of TTE) for individuals included in the linked Medicare and echocardiographic dataset. Continuous measures are listed as means ± standard deviations unless otherwise indicated. N = number of individuals, N obs = number of observations, no. = number, IQR = interquartile range, y = year.
Use of Claims for Identification of Any AS/AR
Use of the ICD-10 claim I35.0 in any position had the optimal performance to identify any AS with a sensitivity, specificity, NPV, PPV, and kappa for agreement of 48.1% (95% CI, 45.4–50.9%), 96.5% (95% CI, 95.9%−97.0%), 85.5% (95% CI, 84.8%−86.1%), 81.3% (95% CI, 78.6%−83.7%), kappa = 0.52 (95% CI, 0.49–0.55) (Table 2). While use of any candidate claim improved sensitivity to 71.9% (95% CI, 69.4%−74.3%), specificity decreased to 72.5% (95% CI, 71.1%−73.8%). While specificity improved when considering only the primary diagnosis code, sensitivity was markedly lower (Supplemental Table S3).
Table 2:
Performance of Candidate ICD-10 Claims in Any Diagnostic Position to Identify Aortic Stenosis on Paired Echocardiogram
| Claim(s) | Number with AS on TTE | Number with AS in Claims | Number Matched | Number Not Matched | Sensitivity (%) | Specificity (%) | NPV (%) | PPV (%) | +LR | −LR | Kappa Statistic |
|---|---|---|---|---|---|---|---|---|---|---|---|
| All candidate claims | 1319 | 2099 | 948 | 3027 | 71.9% [69.4%, 74.3%] | 72.5% [71.1%, 73.8%] | 89.1% [88.2%, 89.9%] | 45.2% [43.7%, 46.6%] | 2.6 [2.5, 2.8] | 0.4 [0.4, 0.4] | 0.37 [0.34, 0.39] |
| All claims except I35.9 | 1319 | 1135 | 757 | 3800 | 57.4% [54.7%, 60.0%] | 91.0% [90.0%, 91.8%] | 87.1% [86.1%, 88.1%] | 66.7% [63.9%, 69.4%] | 6.3 [5.7, 7.1] | 0.5 [0.4,0.5] | 0.51 [0.48,0.53] |
| All claims except 135.9 and I42.1 | 1319 | 1098 | 354 | 575 | 56.4% [53.7%, 59.1%] | 91.5% [90.6%, 92.3%] | 86.9% [85.9%, 87.9%] | 67.8% [64.9%, 70.5%] | 6.7 [6.0, 7.4] | 0.5 [0.4, 0.5] | 0.51 [0.48,0.54] |
| I35.0 | 1319 | 781 | 635 | 4032 | 48.1% [45.4%, 50.9%] | 96.5% [95.9%, 97.0%] | 85.5% [84.8%, 86.1%] | 81.3% [78.6%, 83.7%] | 13.8 [11.6, 16.3] | 0.5 [0.5, 0.6] | 0.52 [0.49, 0.55] |
| I35.2 | 1319 | 47 | 37 | 4168 | 2.8% [2.0%, 3.8%] | 99.8% [99.6%, 99.9%] | 76.5% [76.3%, 76.6%] | 78.7% [64.9%, 88.1%] | 11.7 [5.8, 23.5] | 1.0 [1.0, 1.0] | 0.04 [0.02, 0.05] |
| I35.8 | 1319 | 39 | 27 | 4166 | 2.0% [1.4%, 3.0%] | 99.7% [99.5%, 99.9%] | 76.3% [76.2%, 76.5%] | 69.2% [53.3%, 81.6%] | 7.1 [3.6, 14.0] | 1.0 [1.0, 1.0] | 0.03 [0.01, 0.04] |
| I35.9 | 1319 | 1426 | 538 | 3290 | 40.8% [38.1%, 43.5%] | 78.7% [77.5%, 80.0%] | 80.8% [80.1%, 81.5%] | 37.7% [35.7%, 39.8%] | 1.9 [1.8, 2.1] | 0.8 [0.7, 0.8] | 0.19 [0.16, 0.22] |
| Q23.0 | 1319 | < 11 | < 11 | 4178 | 0.5% [0.2%, 1.1%] | 100.0% [99.9%, 100.0%] | 76.1% [76.0%, 76.2%] | 100.00% | N/A | 1.0 [1.0, 1.0] | 0.01 [0.00, 0.01] |
| I42.1 | 1319 | 49 | 21 | 4150 | 1.6% [1.0%, 2.4%] | 99.3% [99.0%, 99.6%] | 76.2% [76.0%, 76.3%] | 42.9% [29.9%, 56.8%] | 2.4 [1.4, 4.2] | 1.0 [1.0, 1.0] | 0.01 [0.00, 0.02] |
| I08.X | 1319 | 481 | 244 | 3941 | 18.5% [16.4%, 20.7%] | 94.3% [93.6%, 95.0%] | 78.6% [78.1%, 79.0%] | 50.7% [46.5%, 54.9%] | 3.3 [2.8, 3.9] | 0.9 [0.8, 0.9] | 0.16 [0.14, 0.19] |
| I06.X | 1319 | 32 | 21 | 4167 | 1.6% [1.0%, 2.4%] | 99.7% [99.5%, 99.9%] | 76.2% [76.1%, 76.4%] | 65.6% [48.0%, 79.8%] | 6.0 [2.9, 12.5] | 1.0 [1.0, 1.0] | 0.02 [0.01, 0.03] |
Displayed is the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (+LR) and negative likelihood ratio (−LR) for each candidate aortic stenosis (AS) claim or set of claims and corresponding 95% confidence interval to identify any AS on transthoracic echocardiogram (TTE). Additionally, kappa statistics for agreement between claims and TTE ascertainment of AS are provided. Number with AS on TTE refers to the number of individuals with any AS on TTE. Number with AS in claims refers to the number of individuals assigned as having AS using the candidate AS claim(s). Number matched refers to the number of individuals with AS by both claims and TTE definitions. Number not matched refers to the number of individuals without AS by both claims and TTE definitions. Claims with an X in the first decimal place indicate that any numeric value was considered in place of the X. Cell values < 11 are suppressed per Centers for Medicare and Medicaid Services policy. N/A = not applicable.
Use of the ICD-10 claim I35.9 in any position was optimal to identify any AR, though sensitivity and specificity remained low with a sensitivity, specificity, NPV, PPV, and kappa for agreement of 41.9% (95% CI, 39.8%−43.9%), 85.4% (95% CI, 83.7%−86.9%), 56.2% (95% CI, 55.2%−57.2%), 76.6% (95% CI, 74.4%−78.7%), kappa = 0.26 (95% CI, 0.24–0.29) (Table 3). While use of any candidate claim improved sensitivity to 55.0% (95% CI, 52.9%−57.1%), specificity decreased to 79.4% (95% CI, 77.5%−81.1%). While specificity improved when considering only the primary diagnosis code, sensitivity was markedly lower (Supplemental Table S4).
Table 3:
Performance of Candidate ICD-10 Claims in Any Diagnostic Position to Identify Aortic Regurgitation on Paired Echocardiogram
| Claim(s) | Number with AS on TTE | Number with AS in Claims | NumberMatched | Number Not Matched | Sensitivity (%) | Specificity (%) | NPV (%) | PPV (%) | +LR | −LR | Kappa Statistic |
|---|---|---|---|---|---|---|---|---|---|---|---|
| All candidate claims | 2224 | 1625 | 1224 | 1541 | 55.0% [52.9%, 57.1%] | 79.4% [77.5%, 81.1%] | 60.6% [59.4%, 61.9%] | 75.3% [73.5%, 77.0%] | 2.7 [2.4, 2.9] | 0.6 [0.5, 0.6] | 0.34 [0.31, 0.36] |
| I35.1 | 2224 | 251 | 222 | 1913 | 10.0% [8.8%, 11.3%] | 98.5% [97.9%, 99.0%] | 48.9% [48.5%, 49.2%] | 88.4% [83.9%, 91.8%] | 6.7 [4.6, 9.8] | 0.9 [0.9, 0.9] | 0.08 [0.07, 0.09] |
| I35.2 | 2224 | 44 | 29 | 1927 | 1.3% [0.9%, 1.9%] | 99.2% [98.7%, 99.6%] | 46.7% [46.6%, 46.9%] | 65.9% [51.0%, 78.2%] | 1.7 [0.9, 3.1] | 1.0 [1.0, 1.0] | 0.00 [−0.00, 0.01] |
| I35.8 | 2224 | 38 | 29 | 1933 | 1.3% [0.9%, 1.9%] | 99.5% [99.1%, 99.8%] | 46.8% [46.7%, 47.0%] | 76.3% [60.5%, 87.2%] | 2.8 [1.3, 5.9] | 1.0 [1.0, 1.0] | 0.01 [0.00, 0.01] |
| I35.9 | 2224 | 1215 | 931 | 1658 | 41.9% [39.8%, 43.9%] | 85.4% [83.7%, 86.9%] | 56.2% [55.2%, 57.2%] | 76.6% [74.4%, 78.7%] | 2.9 [2.5, 3.2] | 0.7 [0.7, 0.7] | 0.26 [0.24, 0.29] |
| I08.X | 2224 | 413 | 297 | 1826 | 13.4% [12.0%, 14.8%] | 94.0% [92.9%, 95.0%] | 48.7% [48.2%, 49.1%] | 71.9% [67.6%, 75.9%] | 2.2 [1.8, 2.7] | 0.9 [0.9, 0.9] | 0.07 [0.05, 0.09] |
| Q23.1 | 2224 | 52 | 44 | 1934 | 2.0% [1.4%, 2.6%] | 99.6% [99.2%, 99.8%] | 47.0% [46.8%, 47.2%] | 84.6% [72.2%, 92.1%] | 4.8 [2.3, 10.2] | 1.0 [1.0, 1.0] | 0.01 [0.01, 0.02] |
| I06.X | 2224 | 29 | 21 | 1934 | 0.9% [0.6%, 1.4%] | 99.6% [99.2%, 99.8%] | 46.7% [46.6%, 46.9%] | 72.4% [53.8%, 85.5%] | 2.3 [1.0, 5.2] | 1.0 [1.0, 1.0] | 0.00 [0.00, 0.01] |
Displayed is the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (+LR) and negative likelihood ratio (−LR) for each candidate aortic regurgitation (AR) claim or set of claims and corresponding 95% confidence interval to identify any AR on transthoracic echocardiogram (TTE). Additionally, kappa statistics for agreement between claims and TTE ascertainment of AR are provided. Number with AR on TTE refers to the number of individuals with any AR on TTE. Number with AR in claims refers to the number of individuals assigned as having AR using the candidate AR claim(s). Number matched refers to the number of individuals with AR by both claims and TTE definitions. Number not matched refers to the number of individuals without AR by both claims and TTE definitions. Claims with an X in the first decimal place indicate that any numeric value was considered in place of the X. Cell values < 11 are suppressed per Centers for Medicare and Medicaid Services policy. N/A = not applicable.
Use in Claims in Clinical Subsets
A total of 724 individuals (13.2%) had severe AS on TTE. Of individual codes, while I35.0 in any position was optimal to identify severe AS with a sensitivity, specificity, NPV, PPV, and kappa for agreement of 2.3% (95% CI, 1.4%−3.7%), 99.5% (95% CI, 99.3%−99.7%), 87.0% (95% CI, 86.9%−87.2%), 43.6% (95% CI, 29.2%−59.2%), kappa = 0.03 (95% CI, 0.01–0.05) respectively, sensitivity and overall agreement between claims and TTE data remained low for severe AS (Supplemental Table S5). Sensitivity was lower and specificity was marginally higher when considering codes in the primary position (Supplemental Table S6). Fewer than ten patients had severe AR on TTE, the majority of whom (60%) were identified using I35.9.
A total of 23 individuals (0.4%) had a bicuspid aortic valve and AS on TTE. Amongst candidate claims, while the PPV was highest with Q23.0 (14.3%, 95% CI 2.6%−51.3%), sensitivity was poor (4.3%, 95% CI 0.8–2.1%) and highest for I35.0 (Supplemental Table S7). Amongst this limited subgroup, no patients had a bicuspid aortic valve and AR on TTE. A total of 11 individuals (0.2%) had rheumatic aortic valve deformity and AS on TTE. Amongst candidate claims, while the PPV was highest with I06.X (12.5%, 95% CI 5.0%−28.1%), sensitivity was modest (36.4%, 15.2%−64.6%) and highest for I08.X (81.8%, 95% CI 52.3–94.9%) (Supplemental Table S8). No patients had a rheumatic deformity and AR on TTE. A total of 488 individuals (8.9% of total, 67.4% of those with severe AS) had LFLG severe AS on TTE. Consistent with overall findings, agreement was highest with I35.0 (Supplemental Table S9). Results were substantially unchanged when reclassifying those with mild AS and a stroke volume index ≤ 35 mL/m2 as LFLG (Supplemental Table S10). Use of a 12-month window after the date of the TTE modestly improved sensitivity with decreased specificity (Supplemental Table S11). Use of a concurrent HF code to ascertain AS (Supplemental Table S12) or AR (Supplemental Table S13) did not substantially change algorithm performance.
Results of Recursive Partitioning
Using tree-based methods, I35.0 was determined to be optimal for identifying any AS with a sensitivity, specificity, and c-statistic for the tree-based method on cross-validation of 53.1%, 94.8%, and 0.77 respectively after considering all diagnosis codes, age, and sex (Supplemental Figure S1). Amongst those with a code for I35.0 (N = 781), 635 (81.3%) had AS on TTE, including 217 (27.8%) with mild AS, 84 (10.7%) with moderate AS, and 315 (40.3%) with severe AS, and 165 (21.1%) had no AS. While the code of next greatest importance for identifying any AS was Z95.2 (presence of an aortic valve replacement), the sensitivity of this claim to identify AS amongst those without a claim for I35.0 was only 8.9% on cross-validation (Supplemental Figure S2). When considering all diagnoses, age, and sex to identify severity of AS, the tree-based model failed to identify an optimal algorithm.
Evaluation of Optimal Claims Algorithm in Medicare Data
After applying exclusions, a total of 4,033,844 individuals were included in the final sample (Figure 2). Of those with AS (identified via a claim for I35.0), 13.6% had the primary endpoint of all-cause mortality at 1-year vs. 7.3% without AS (adjusted HR 1.33, 95% CI 1.31–1.34) (Table 4). Amongst secondary endpoints, 1,8% of those with AS by this definition had a HF hospitalization vs. 0.8% without AS (adjusted HR 1.37, 95% CI 1.34–1.41) and 4.1% with AS underwent AVR vs. 0.1% without AS (adjusted HR 34.96, 95% CI 33.74–36.22). Of these, 3.0% with AS and 0.1% without underwent TAVR (adjusted HR 45.09, 95% CI 42.97–47.32) and 1.1% with AS and 0.1% without underwent SAVR (adjusted HR 23.64, 95% CI 22.35–25.00). Use of a 12-month window after the date of TTE to define AS did not substantially change the association of AS and valve-related outcomes of interest (Supplemental Table S14).
Figure 2: Flowchart of Included Individuals in the Overall Medicare Cohort.
Shown is a flowchart illustrating those included in the overall Medicare dataset (Medicare Fee-for-service beneficiaries with a TTE performed between January 1, 2017 and December 31, 2019). Of a total of 5,010,524 individuals initially considered for inclusion, 117,046 individuals (2.3%) were excluded due to death prior to the first day of follow-up (3 months following the date of the TTE), 586,717 individuals (11.7%) were excluded due to being under age 65 at the time of their index TTE, and 272,917 individuals (5.4%) were excluded due to lacking continuous enrollment in the Medicare Fee-for-service program for 12 months prior and 3 months after the date of the index TTE. A total of 4,033,844 (80.5%) were included the final cohort. N = number of individuals, TTE = transthoracic echocardiogram.
Table 4:
Rates and Risks of Outcomes at 1-Year Amongst those with an ICD-10 Claim of I35.0 in the Overall Medicare Dataset
| Outcome | I35.0 Present (N = 402,559) | I35.0 Absent (N = 3,631,285) | Adjusted Hazard Ratio (95% CI) |
|---|---|---|---|
| All-cause mortality | 13.6% (13.5%, 13.7%) | 7.3% (7.3%, 7.3%) | 1.33 (1.31, 1.34) |
| Heart failure hospitalization | 1.8% (1.8%, 1.9%) | 0.8% (0.8%, 0.8%) | 1.37 (1.34, 1.41) |
| Aortic valve replacement | 4.1% (4.1%, 4.2%) | 0.1% (0.1%, 0.1%) | 34.96 (33.74, 36.22) |
| Transcatheter aortic valve replacement | 3.0% (3.0%, 3.1%) | 0.1% (0.1%, 0.1%) | 45.09 (42.97, 47.32) |
| Surgical aortic valve replacement | 1.1% (1.1%, 1.1%) | 0.1% (0.1%, 0.1%) | 23.64 (22.35, 25.00) |
Shown are the unadjusted rates (determined using Kaplan-Meier estimates) and 95% confidence intervals (in parentheses) of relevant outcomes amongst those with and without an ICD-10 claim of 135.0 in any position in the overall Medicare dataset (a complete sample of Medicare inpatient and outpatient claims, 2017–2019) at 1-year following baseline (3 months after the most recent echocardiogram). Additionally shown are the hazard ratios and 95% confidence intervals (in parentheses) for the occurrence of each outcome using the group without an I35.0 claim as reference, adjusting for age, sex, and all 27 Medicare Chronic Condition Warehouse variables. Aortic valve replacement refers to the receipt of either a transcatheter or surgical aortic valve replacement. Non-death outcomes are adjusted for the competing risk of death with event rates taken from cumulative incidence functions.
DISCUSSION
In this large echocardiographic study, the ICD-10 code I35.0 (i.e. nonrheumatic aortic stenosis) in any position, though optimal compared to other coding algorithms and identifying a population at a significantly greater risk of all-cause mortality, HF hospitalization, and receipt of AVR in Medicare FFS claims, was nevertheless insensitive for classification of AS in Medicare FFS claims, failing to identify nearly half of individuals with AS. While severe AS was found in 40.3% of those with this claim, claims were unable to distinguish disease severity or subtypes. Moreover, ICD-10 claims were insensitive and nonspecific for the diagnosis of AR of any severity. These results argue against the widespread use of claims to screen for patients with AS and suggest the need for improved coding algorithms and alternative systems to extract TTE data to be included in quality efforts.
To date, there has been limited validation of administrative claims data against echocardiographic criteria for assessment of AS status. As part of a multicenter epidemiologic study of genetic single nucleotide polymorphisms and presence of aortic valvular calcification within the Cohorts for Heart and Aging Research in Genome Epidemiology (CHARGE) consortium19, three ICD-9 and 2 ICD-10 codes (I35.0 and I35.2) were validated against TTE data in 100 randomly selected patients with prevalent or incident AS. Of these, only 9 patients did not have TTE evidence of AS, suggesting high accuracy of codes, and most patients with AS had moderate to severe disease. While this study did not assess the optimal coding algorithm or the performance of claims for subtypes of AS, it nevertheless aligns with results of the present study which suggest that the 81.3% of individuals with an I35.0 claim have at least mild AS with the majority (51%) of individuals having moderate to severe AS and 40.3% having severe AS. Thus, though claims can feasibly identify a population enriched with individuals with severe AS who ultimately receive therapy, they are not specific to AS severity.
The current study builds upon these preliminary data by suggesting that the optimal claim for assessment of AS was I35.0 which had a sensitivity and specificity on cross-validation of only 53.1% and 94.8% respectively for presence of any AS. This sensitivity did not substantially improve when expanding the window for evaluation of claims after the TTE to 12 months from 3 months. Furthermore, our results suggest that amongst those lacking a claim of I35.0, there is no single code or combination of codes, clinical diagnoses, and demographic variables that reliably identifies those with AS. Amongst those with an I35.0 claim, 18.7% had no AS on TTE. While the majority (52.7%) of those without AS on TTE had AR and this finding may reflect local coding practices, this observation nevertheless underlines the challenges of using administrative claims for ascertainment of AS at present. Furthermore, despite the existence of specific codes under the ICD-10 framework for the presence of bicuspid AS (i.e. Q23.0 – congenital stenosis of the aortic valve) or rheumatic AS (i.e. I06.X – rheumatic aortic stenosis), I35.0 nevertheless had the best overall performance in identifying AS in these disease subsets. Presence of I35.0 in the primary coding position vs. any coding position resulted in increased specificity but markedly lower sensitivity for identification of AS. Moreover, I35.0 was optimal for identification of low-flow, low-gradient AS though was not specific for this hemodynamic state. When considering the severity of AS, as mentioned previously, though the majority of patients with I35.0 had moderate to severe disease, claims were unable to identify those individuals that only had severe AS present. Lastly, while the use of claim I35.9 (i.e. nonrheumatic aortic valve disorder, unspecified) was optimal to identify any AR, the sensitivity and specificity for detection of any AR were both modest (41.9% and 85.4% respectively), suggesting that claims overall cannot reliably distinguish those with AR on TTE.
Use of I35.0 to ascertain AS cases in the overall Medicare cohort nonetheless identified a cohort at increased risk of valve-related adverse outcomes and need for future treatment. Individuals identified in the Medicare dataset as having AS had a 33% higher risk of all-cause mortality and a 37% higher risk of heart failure hospitalization at 1-year following baseline, suggesting that this claim identified a high-risk cohort. Consistent with this finding, 4.1% of those with a code for I35.0 underwent an AVR in the following year vs. only 0.1% without this code, representing a 35-fold increased risk for requiring a valve replacement. Thus, this claim identified those who ultimately went on to receive therapy for AS, and though insensitive, could be used to define a denominator of individuals at risk for clinical worsening without treatment.
Thus, in aggregate, these findings suggest that claims should be used with caution in identifying a cohort with AS. Though the use of ICD-10 code I35.0 is specific for the diagnosis of any AS, it is not specific to any particular severity or subtype of AS and furthermore misses roughly half of AS cases. Furthermore, claims were insensitive and nonspecific for the diagnosis of AR of any severity. Despite the availability of claims for other subtypes of aortic valve disease such as bicuspid or rheumatic AS, these claims were not consistently used to code for AS in these settings. Thus, these data suggest against using claims in screening for AS/AR. Though use of I35.0 identified a population who ultimately went on to receive AVR and thus could be used to define a denominator of individuals at risk for clinical worsening, the modest sensitivity limits use of claims for screening.
Nevertheless, algorithms to identify cohorts with severe aortic valve disease are important for several reasons. Such algorithms could be used for hospital benchmarking and process improvement (e.g. by intensifying resources for a population in order to prevent gaps in referral and treatment), evaluating healthcare utilization and treatment needs of individuals with AS/AR, improving knowledge about AS and its associated biological underpinnings,19 and disaggregating diagnosis from treatment in the study of racial, ethnic, and sociodemographic disparities that exist in the receipt of TAVR.10,20 Though limited to a single medical center, the current study indicates that coding for AS/AR remains inconsistent. Improved guidance for coding of AS/AR (particularly for subtypes of AS/AR such as bicuspid and rheumatic disease where specific codes exist already) and creation of new codes specific to the severity of AS/AR could potentially improve the utility of claims to evaluate AS/AR. Furthermore, the standardization of TTE reporting and harmonization of data elements across reporting vendors could improve the ability to extract relevant valvular disease data from TTE reports for the purposes of quality improvement and benchmarking.
While large and multisite, our study nevertheless has several recognized limitations. First, as individuals received TTEs for clinical indications, there is referral bias impacting the prevalence of VHD. Thus, the true prevalence of AS in the general population may be lower. Second, though TTE is the standard for evaluation of AS, multimodality assessment of AS severity was not used. Third, derivation of the claims algorithm may be influenced by practices at Beth Israel Deaconess Medical Center and it is possible that certain claims may perform differently in disease subgroups at other sites and these algorithms should be tested externally. Fourth, the prevalence of AS/AR increases with age1 and thus prevalence and algorithm performance may differ amongst individuals < 65 years old at entry who were not included. Fifth, though AS/AR severity closely approximated assessment by a single blinded and experienced reader, it is possible that inaccuracies in the estimation of AS/AR severity may influence the observed results. Sixth, as only Medicare FFS beneficiaries were included, results may not generalize to those Medicare Advantage or other insurance types.
CONCLUSION
In a large study of Medicare FFS beneficiaries who received TTEs, the ICD-10 code I35.0 for nonrheumatic aortic stenosis, despite missing nearly 50% of those with AS, identified a population at significantly greater risk of all-cause mortality, HF hospitalization, and receipt of AVR. While the majority of individuals with this claim had moderate to severe AS on paired TTE, claims were unable to distinguish a group with solely severe AS or disease subtypes (e.g. bicuspid or rheumatic AS). Moreover, ICD-10 claims had only modest accuracy in the diagnosis of AR. Overall, these results argue against the widespread use of claims to screen for patients with AS and suggest the need for improved coding algorithms and alternative systems to extract TTE data for quality improvement and health system performance evaluation.
Supplementary Material
WHAT IS KNOWN / WHAT IS ADDED.
What Is Known:
Administrative claims may be useful for identifying patients with aortic valve disease and estimating disease prevalence, but have not been extensively validated.
What Is Added:
In this large echocardiographic cohort, though strongly predictive of future receipt of aortic valve replacement, claims were insensitive for the diagnosis of aortic valve disease and were unable to distinguish amongst disease severity and subtypes.
What Is Added:
The results argue against use of AS/AR claims for screening of aortic valve disease and suggest the need for improved coding systems and methods to extract echocardiographic data for quality improvement and benchmarking.
SOURCES OF FUNDING:
The project was funded by a grant from the National, Heart, Lung, and Blood Institute (1K23HL144907 - Strom).
DISCLOSURES:
Dr. Strom additionally reports grant funding from Edwards Lifesciences, Anumana, Ultromics, and HeartSciences, consulting for Bracco Diagnostics, and speaker fees from Northwest Imaging Forums, unrelated to the submitted work. Dr. Yeh reports grant funding and consulting fees from Abbott Vacsular, Boston Scientific and Medtronic. Other authors report no relevant disclosures.
NONSTANDARD ABBREVIATIONS AND ACRONYMS:
- AR
aortic regurgitation
- AS
aortic stenosis
- AVA
aortic valve area
- AVR
aortic valve replacement
- BIDMC
Beth Israel Deaconess Medical Center
- CHARGE
Cohorts for Heart and Aging Research in Genome Epidemiology
- CI
confidence interval
- ECHO
Echocardiography and Cardiovascular Health Outcomes
- FFS
Fee-for-service
- HF
heart failure
- HR
hazard ratio
- ICD-10
International Classification of Diseases, 10th revision
- LFLG
low-flow, low-gradient
- +LR
positive likelihood ratio
- −LR
negative likelihood ratio
- MG
mean aortic transvalvular gradient
- NPV
positive predictive value
- PAV
peak aortic transvalvular velocity
- PPV
positive predictive value
- TAVR
transcatheter aortic valve replacement
- TTE
transthoracic echocardiogram
Footnotes
REFERENCES
- 1.Nkomo VT, Gardin JM, Skelton TN, Gottdiener JS, Scott CG, Enriquez-Sarano M. Burden of valvular heart diseases: a population-based study. Lancet. 2006;368:1005–1011. [DOI] [PubMed] [Google Scholar]
- 2.Iung B, Baron G, Butchart EG, Delahaye F, Gohlke-Bärwolf C, Levang OW, Tornos P, Vanoverschelde JL, Vermeer F, Boersma E, et al. A prospective survey of patients with valvular heart disease in Europe: The Euro Heart Survey on Valvular Heart Disease. Eur Heart J. 2003;24:1231–1243. [DOI] [PubMed] [Google Scholar]
- 3.Eveborn GW, Schirmer H, Heggelund G, Lunde P, Rasmussen K. The evolving epidemiology of valvular aortic stenosis. the Tromso study. Heart. 2013;99:396–400. d [DOI] [PubMed] [Google Scholar]
- 4.Danielsen R, Aspelund T, Harris TB, Gudnason V. The prevalence of aortic stenosis in the elderly in Iceland and predictions for the coming decades: the AGES-Reykjavík study. Int Jn Cardiol. 2014;176:916–922. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ross J Jr., Braunwald E. Aortic stenosis. Circulation. 1968;38:61–67. [DOI] [PubMed] [Google Scholar]
- 6.Coffey S, Cox B, Williams MJ. Lack of progress in valvular heart disease in the pre-transcatheter aortic valve replacement era: increasing deaths and minimal change in mortality rate over the past three decades. Am Heart J. 2014;167:562–567. [DOI] [PubMed] [Google Scholar]
- 7.Kundi H, Cohen DJ, Strom JB, Valsdottir L, Shen C, Choi E, Popma JJ, Yeh RW. Trends in isolated aortic valve replacement in the United States in the early phase of expansion of TAVR. Int J Cardiol. 2019;292:68–72. [DOI] [PubMed] [Google Scholar]
- 8.Kundi H, Popma JJ, Khabbaz KR, Chu LM, Strom JB, Valsdottir LR, Shen C, Yeh RW. Association of Hospital Surgical Aortic Valve Replacement Quality With 30-Day and 1-Year Mortality After Transcatheter Aortic Valve Replacement. JAMA Cardiol. 2019;4:16–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kundi H, Strom JB, Valsdottir LR, Elmariah S, Popma JJ, Shen C, Yeh RW. Trends in Isolated Surgical Aortic Valve Replacement According to Hospital-Based Transcatheter Aortic Valve Replacement Volumes. JACC Cardiovasc Interv. 2018;11:2148–2156. [DOI] [PubMed] [Google Scholar]
- 10.Hyland PM XJ, Shen C, Markson LK, Manning WJ, Strom JB.. Comorbidities and the Associated Long-Term Utilization of Transthoracic Echocardiography among Medicare Beneficiaries at a Large Academic Center. 2022;108:956–963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Mitchell C, Rahko PS, Blauwet LA, Canaday B, Finstuen JA, Foster MC, Horton K, Ogunyankin KO, Palma RA, Velazquez EJ. Guidelines for Performing a Comprehensive Transthoracic Echocardiographic Examination in Adults: Recommendations from the American Society of Echocardiography. J Am Soc Echocardiogr. 2019;32:1–64. [DOI] [PubMed] [Google Scholar]
- 12.Otto CM, Nishimura RA, Bonow RO, Carabello BA, Erwin JP, Gentile F, Jneid H, Krieger EV, Mack M, McLeod C, et al. 2020 ACC/AHA Guideline for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation. 2021;143:e72–e227. [DOI] [PubMed] [Google Scholar]
- 13.Vahanian A, Beyersdorf F, Praz F, Milojevic M, Baldus S, Bauersachs J, Capodanno D, Conradi L, De Bonis M, De Paulis R, et al. 2021 ESC/EACTS Guidelines for the management of valvular heart disease: Developed by the Task Force for the management of valvular heart disease of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Eur Heart J. 2021;43:561–632. [Google Scholar]
- 14.Zoghbi WA, Adams D, Bonow RO, Enriquez-Sarano M, Foster E, Grayburn PA, Hahn RT, Han Y, Hung J, Lang RM, et al. Recommendations for Noninvasive Evaluation of Native Valvular Regurgitation: A Report from the American Society of Echocardiography Developed in Collaboration with the Society for Cardiovascular Magnetic Resonance. J Am Soc Echocardiogr.. 2017;30:303–371. [DOI] [PubMed] [Google Scholar]
- 15.Nuttall FQ. Body Mass Index: Obesity, BMI, and Health: A Critical Review. Nutr Today. 2015;50:117–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Mosteller RD. Simplified calculation of body-surface area. N Engl J Med. 1987;317:1098. [DOI] [PubMed] [Google Scholar]
- 17.Gorina Y, Kramarow EA. Identifying chronic conditions in Medicare claims data: evaluating the Chronic Condition Data Warehouse algorithm. Health Serv Res. 2011;46:1610–1627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Fine JP, Gray RJ. A Proportional Hazards Model for the Subdistribution of a Competing Risk. J Am Stat Assoc. 1999;94:496–509. [Google Scholar]
- 19.Thanassoulis G, Campbell CY, Owens DS, Smith JG, Smith AV, Peloso GM, Kerr KF, Pechlivanis S, Budoff MJ, Harris TB, et al. Genetic associations with valvular calcification and aortic stenosis. N Engl J Med. 2013;368:503–512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nathan AS, Yang L, Yang N, Eberly LA, Khatana SAM, Dayoub EJ, Vemulapalli S, Julien H, Cohen DJ, Nallamothu BK, et al. Racial, Ethnic, and Socioeconomic Disparities in Access to Transcatheter Aortic Valve Replacement Within Major Metropolitan Areas. JAMA Cardiol. 2022;7:150–157. [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.