Hospital Variation in Aortic Valve Replacement Approach Selection for Aortic Stenosis in Patients Under 65

Abstract

Objectives:

Although surgical aortic valve replacement (SAVR) has a class 1 A recommendation for treating severe aortic stenosis in patients <65 years, transcatheter aortic valve replacement (TAVR) in this population is increasing. This study evaluates the impact of hospital variation in TAVR use in patients <65 years on clinical outcomes.

Methods:

Using US 3-state data from 2013 through 2021, we assessed the hospitals’ preference for TAVR vs SAVR by generating the observed-to-expected TAVR ratio. Hospitals were ranked into tertiles based on their ratio. The risk of mortality, stroke, infective endocarditis (IE), and permanent pacemaker implantation (PPI) for patients undergoing aortic valve replacement (AVR) in each tertile was assessed at 30 days and 6 years using logistic regression and Cox-proportional hazard models.

Results:

Among 189 hospitals, 103 were in the low, 55 in the medium, and 31 in the high-tertile. Patients who underwent AVR in the high tertile had lower rates of comorbidities than patients in mid or low tertiles. Patients at high and medium ratio hospitals had higher rates of PPI at 30 days than those from low TAVR-use hospitals (17% vs 7.6% vs 5.6%, P < .001). Patients in the high vs low tertile experienced a higher 6-year risk-adjusted mortality (8.1% vs 5.3%, HR: 1.63 [1.37–1.93], P < .001), stroke (2.2% vs 1.1%, sub-distribution hazard ratio [sHR]: 2.15 [1.50–3.06], P < .001), and IE (2.9% vs 0.3%, sHR: 9.91 [5.59–17.56], P < .001).

Conclusions:

The decision on TAVR utilization in patients younger than 65 should be made carefully considering the patient’s clinical profile and life expectancy.

Graphical Abstract

INTRODUCTION

Originally, transcatheter aortic valve replacement (TAVR) was intended as an alternative to surgical aortic valve replacement (SAVR) for high-surgical-risk patients with severe aortic stenosis; its use has since expanded to patients across the entire spectrum of surgical risk.¹ However, the 2 clinical trials motivating this shift in practice included patients with a mean age of 74 years, raising concerns about the generalizability of these results to younger aortic stenosis patients.^2–6 Consequently, the current American College of Cardiology and American Heart Association (ACC/AHA) guidelines still consider SAVR a Class I recommendation in severe aortic stenosis patients less than 65 years of age.⁷ According to the European guidelines, TAVR is recommended for low-risk patients older than 70 years.⁸ However, published reports demonstrated a progressive trend in TAVR utilization in younger patients, highlighting concerns for indication creep.^9,10

A recent STS/ACC TVT and Japanese Transcatheter Valvular Therapy registry analysis described the cause of variability in TAVR outcomes between hospitals as complex, multifactorial, and unexplained by patient STS risk scores or hospital characteristics.¹¹ However, comparisons of outcomes across hospitals using unadjusted metrics are discouraged, instead, various risk adjustment approaches incorporating patient or hospital-related predictors in regression modelling have been developed.¹² An example is the observed-to-expected (O/E) ratio, which has served as a quality-of-care indicator for various measures, including adherence to clinical guidelines and was described in previous studies.^12–15 Current data regarding hospital variability in TAVR utilization in patients younger than 65 years old and its impact on patient outcomes are unknown.

We sought to assess whether the outcomes of patients younger than 65 who are treated for aortic stenosis are similar among hospitals with different utilization of TAVR as defined by the O/E TAVR ratios. The outcomes assessed include mortality, stroke, infective endocarditis (IE), and the need for permanent pacemaker implantation (PPI).

PATIENTS AND METHODS

Data source

We used 3 statewide administrative healthcare claims databases: the Department of Health Care Access and Information of California State (HCAI), New Jersey’s Discharge Data Collection System (NJDDCS), and New York’s Statewide Planning and Research Cooperative System (SPARCS). Each database was linked to the state’s vital death records to obtain death dates. This study was approved by the Committee for the Protection of Human Subjects of the State of California, the Health and Human Services Agency privacy board of the New York State Department of Health, the Department of Health of NJ, and the Institutional Review Board of Icahn School of Medicine at Mount Sinai on 7/17/2024 (STUDY-15–00807-CR005) with a waiver of informed consent.

Patient selection

We included adults younger than 65 who underwent isolated TAVR or SAVR procedures from January 1, 2013, to December 31, 2021, in California and New York, and through December 31, 2020, in New Jersey. Procedure and outcomes were identified using the International Classification of Diseases-9th Revision-Clinical Modification (ICD-9-CM) and the ICD-10th Revision-Procedures Coding System (ICD-10-PCS) (Table S1). Patients < 18 years old and those who underwent concomitant cardiac procedures were excluded (Figure S1). The American Hospital Association (AHA) annual survey was linked to the state databases to extract the characteristics of hospitals.

Study end-points and definitions

The primary end-point was 6-year mortality by site of care. Secondary end-points were 30-day and 6-year incidence of stroke and IE, and the 30-day rate of new PPI. To identify the incidence of new PPI, patients with a pacemaker or defibrillator history were excluded.

Statistical analysis

Baseline characteristics were reported as frequencies (%) and medians (interquartile range [IQR]) for categorical and continuous variables, respectively. For group comparisons, the chi-square test or Wilcoxon non-parametric test was used depending on data distribution.

To define hospitals’ TAVR over SAVR utilization, we used the data of all patients with aortic stenosis who underwent aortic valve replacement (AVR) to calculate the observed-over-expected TAVR ratio. First, a non-parsimonious logistic regression model estimated the risk-adjusted probability of undergoing TAVR over SAVR for patients < 65 years old. The model was adjusted for age, sex, race, insurance, comorbidities, admission status, frailty score spline, and the year of index AVR. For each hospital, the predicted probabilities of having TAVR vs SAVR were summed and divided by the total number of AVR to obtain the expected TAVR rate. Observed TAVR rates were determined using the hospitals’ actual proportion of TAVR patients. The O/E TAVR ratio was calculated for each institution. Hospitals were categorized into tertiles based on their O/E ratio. Ratios diverging from 1 indicate discrepancies between expected and actual TAVR use, potentially reflecting inappropriate procedure selection.

In the main analysis, the risk of death, stroke, IE, and PPI for all patients who had AVR (either TAVR or SAVR) was assessed. We used a logistic regression model to estimate the risk of 30-day outcomes in each hospital tertile. The model was adjusted for hospital characteristics, including annual volume of AVR, bed size, and teaching status. We used the stepwise selection method, with probabilities of entry 0.15 and stay of 0.05. Six-year mortality for each tertile was estimated using Kaplan-Meier survival tables. Fine-Gray models with death as a competing risk event were used to assess the 6-year risk of secondary outcomes across hospital groups using robust sandwich estimator to account for in-hospital clustering and adjusting for hospital characteristics. The results were reported as sub-distribution hazard ratios (sHRs) with 95% CI. The proportional hazard assumption was tested using Schoenfeld residuals.

Sensitivity analyses

First, we stratified the results by procedure type and compared outcomes across hospital tertiles. The interaction between hospital rank and procedure type was used to assess the effect of hospital tertile on long-term outcomes in different procedure groups.

Second, we used a restricted cubic spline (SAS %RCT macro) to assess the relationship between the O/E ratio and the primary outcome (Figure S2). We defined 15 as the O/E ratio cutoff and compared the outcomes of patients in hospitals with O/E ≤ 15 vs >15. Third, we excluded the hospitals not performing TAVR in patients < 65 from the outcomes’ comparison. Additionally, 1:1 propensity score matching was performed between SAVR and TAVR patients within each tertile to assess the primary outcome by procedure type. Matching variables included patients’ demographics, comorbid conditions, and frailty score. Patients with missing data were excluded from analysis (0.6%). The statistical significance threshold for all tests was 0.05. All statistical analyses were done using the Statistical Analysis System, version 9.4 (SAS Institute Inc, Cary, NC).

RESULTS

We utilized state-mandated discharge databases for California, New Jersey, and New York, which represent 20% of the US population, and included long-term follow-up data. In 189 hospitals, 9655 patients <65 years old underwent isolated SAVR and 2706 underwent isolated TAVR. Patients <65 years were more likely to undergo TAVR vs SAVR if older, female, and had a history of major comorbidities such as renal failure, heart failure, coronary artery disease, and cancer. More recent years were associated with higher odds of undergoing TAVR (OR: 1.52 [1.48–1.57], P = .007), Table 1.

Table 1.

Predictors of TAVR vs SAVR Utilization in Patients Younger Than 65 Years Based on the Results of Multivariable Logistic Regression

	Odds ratio	95% Confidence interval	P-value
Age (1-year increments)	1.08	1.07–1.09	<.001
Year of surgery	1.52	1.48–1.57	.01
Sex
Male	Ref.
Female	1.38	1.22–1.57	<.001
Insurance status
Private insurance	Ref.
Medicare	2.09	1.76–2.48	<.001
Medicaid	1.54	1.30–1.81	<.001
Uninsured	1.40	0.79–2.49	.14
Other insurance	1.18	0.73–1.90	.47
Race
White	Ref.
Non-White	0.84	0.73–0.97	<.001
Type of admission
Elective	Ref.
Non-elective	0.77	0.66–0.89	<.001
Comorbidities
Immobility	1.30	0.96–1.76	.12
Cognitive disorder	0.97	0.61–1.53	.95
Oxygen dependence	1.83	1.17–2.86	.01
Bicuspid aortic valve	0.48	0.38–0.60	<.001
Drug abuse	0.62	0.49–0.77	<.001
Alcoholism	0.95	0.77–1.17	.60
CKD/dialysis	2.94	2.33–3.70	<.001
CKD no dialysis	1.50	1.26–1.77	<.001
Cancer	2.28	1.94–2.68	<.001
Heart failure	2.85	2.51–3.25	<.001
Liver disease	1.83	1.56–2.16	<.001
COPD	1.27	1.11–1.45	<.001
Stroke	0.94	0.61–1.45	.89
CVD	1.12	0.91–1.37	.41
Coagulopathy	1.10	0.93–1.31	.28
CAD/PCI	2.60	1.82–3.72	<.001
CAD/CABG	4.93	3.88–6.27	<.001
CAD	1.46	1.28–1.66	<.001
Atrial fibrillation	0.92	0.80–1.07	.18
Tobacco use	0.88	0.77–1.00	.02
PVD	1.48	1.24–1.77	<.001
DM	1.16	1.02–1.32	.01
Permanent pacemakers	1.07	0.80–1.43	.65
Atherosclerosis of aorta	0.78	0.60–1.02	.04

Abbreviations: CKD, Chronic Kidney Disease; COPD, Chronic Obstractive Pulmonary Disease; CVD, Cerebrovascular Disease; CAD, Coronary Artery Disease; CABG, CoronaryArtery Bypass Graft; DM, Diabetis Mellitus; PCI, Percutaneous Coronary Inrtervention; PVD, Peripheral Vascular Disease.

Among 189 hospitals performing AVR in patients < 65, 147 (77.8%) hospitals performed TAVR. The hospital O/E ratios ranged from 0 to 1.3 for low, 1.4 to 16.8 for medium, and 16.9.1 to 75.1 for high tertile (Figure 1). Most of the hospitals were classified in the low tertile (54.5%), followed by the medium (29.1%) and high (16.4%) tertile. The characteristics of hospitals are shown in Table S2. Hospitals in the highest tertile had the highest number of beds and were more likely to be teaching hospitals than the medium and low tertiles. TAVR was the procedure of choice in 9.7%, 29.1%, and 26.7% of the patients who required AVR in the low, mid, and high tertiles, respectively.

Figure 1.

The Distribution of Observed-to-Expected TAVR Ratios for Hospitals in Different Utilization Tertiles.

Patient characteristics

Baseline characteristics for patients by hospital tertiles are reported in Table S3. There was no difference in median patient age across the 3 hospital groups. Patients were mostly males and of the white race. High tertile hospitals had more patients without major comorbidities (24.5%) than medium (18.9%) and low (20.8%). The median follow-up time was 4.1 years (IQR: 2.1–6.4 years).

Clinical outcomes

After adjusting for patients’ demographics, comorbidities, and hospital characteristics, there was no difference in 30-day risk of mortality, stroke, or IE across tertiles. Having the procedure performed in hospitals with high utilization of TAVR was associated with a higher risk of 30-day PPI compared to low-utilization hospitals (OR: 3.07 [2.64–3.57], P < .001), Table 2.

Table 2.

The Outcomes of Patients Undergoing Aortic Valve Replacement by Hospital TAVR Utilization Tertiles

30-day outcomes	Low tertile (n = 4079)		Medium tertile (n = 4182)			High tertile (n = 4100)
	Incidence	OR [95% CI]	Incidence	OR [95% CI]	P-value	Incidence	OR [95% CI]	P-value
Mortality	36 (0.9)	Ref.	42 (1.0)	1.14 [0.73–1.78]	.87	44 (1.1)	1.22 [0.78–1.80]	.49
Stroke	22 (0.5)	Ref.	21 (0.5)	0.93 [0.51–1.70]	.74	16 (0.4)	0.72 [0.38–1.38]	.32
Infective endocarditis	<10	Ref.	<10	0.49 [0.09–2.66]	.24	<10	1.49 [0.42–5.30]	.20
Permanent pacemaker implantationa	215 (5.6)	Ref.	292 (7.6)	1.33 [1.12–2.57]	<.001	676 (17)	3.07 [2.64–3.57]	<.001
6-year outcomes	Incidence	HR [95% CI]	Incidence	HR [95% CI]	P-value	Incidence	HR [95% CI]	P-value
Mortality	215 (5.3)	Ref.	233 (5.6)	1.10 [0.91–1.32]	.33	333 (8.1)	1.63 [1.37–1.93]	<.001
Stroke	45 (1.1)	Ref.	37 (0.9)	0.82 [0.53–1.27]	.38	91 (2.2)	2.15 [1.50–3.06]	<.001
Infective endocarditis	13 (0.3)	Ref.	11 (0.3)	0.86 [0.39–1.93]	.72	119 (2.9)	9.91 [5.59–17.56]	<.001

^aThe rate of new PPI is calculated from the number of patients without a history of PPI, which is 3846 in low, 3862 in medium, and 3975 in high tertile of hospitals.

Abbreviation: HR: hazard ratio

Patients from high ratio hospitals had a higher risk of mortality at 6 years compared to patients from low ratio hospitals (HR: 1.63 [1.37–1.93], P < .001), Table 2. Long-term mortality was similar in low vs medium TAVR-use hospitals (Figure 2). Moreover, patients in high tertile hospitals were more likely to experience a stroke (sHR: 2.15 [1.50–3.06], P < .001) or IE at 6 years (sHR: 9.91 [5.59–17.56], P < .001).

Figure 2.

Six-year Survival for Patients Undergoing Aortic Valve Replacement in Low, Medium, and High Utilization-of-TAVR Tertiles. Abbreviation: HR: hazard ratio

Sensitivity analysis

The stratified analysis showed that TAVR patients in high tertile hospitals had lower rates of major comorbidities (Table S4). Similarly, SAVR patients in high tertile hospitals had lower rates of comorbidities than those in other tertiles but higher rates of atrial fibrillation. The rates of 30-day mortality, IE, and stroke were similar among the 3 hospital groups regardless of procedure (Table S5). TAVR and SAVR patients in high TAVR utilization hospitals had higher rates of PPI implantation within 30 days post-op.

At 6 years, the risk of mortality and IE for TAVR patients in high-tertile hospitals was higher than in low-tertile hospitals. The 6-year risk of stroke was higher in high vs low-tertile hospitals for TAVR patients. SAVR patients in different hospital groups had comparable long-term stroke (Table S6).

In the second sensitivity analysis, the spline curve demonstrated that around the O/E ratio of 15, the hazard of death became significantly higher. We compared the outcomes of patients who underwent AVR in hospitals with O/E ≤ 15 vs O/E > 15 in Table S7. There was no difference in 30-day outcomes of patients who underwent AVR in high vs low ratio hospitals. The risk of 6-year mortality, stroke, and IE was higher in hospitals with an O/E ratio > 15.

In the third sensitivity analysis, we excluded hospitals that do not perform TAVR in patients <65. One hundred and forty-seven hospitals with 11 849 isolated AVRs (TAVR+SAVR) remained in the cohort. There was no difference in 30-day risk of death, stroke, and IE between the 3 hospital tertiles (Table S8). The risk of PPI at 30 days was significantly higher among patients who underwent procedures in medium and high TAVR-use hospitals compared to the low TAVR-use hospitals.

At 6 years post-procedure, having procedures in high-tertile hospitals was associated with a higher risk of mortality, stroke, and IE. The risk of mortality, stroke, and IE was similar among patients in the medium and low TAVR utilization hospitals.

After propensity score matching, TAVR patients consistently showed higher mortality at 6 years in all hospital tertiles (Table S9).

DISCUSSION

In this study, we assess the outcomes of patients with aortic stenosis younger than 65 years old who underwent AVR in hospitals with different TAVR vs SAVR utilization. Patients who underwent SAVR or TAVR for aortic stenosis in hospitals with high TAVR use had lower rates of comorbidities compared to patients from hospitals with medium or low use. Yet, they experienced a higher risk of late mortality. In addition, having procedures at hospitals in the medium- and high-tertiles was associated with a higher risk of PPI at 30 days.

The classification of hospitals is not based on crude rates of TAVR but instead on the difference between the observed TAVR rates and expected rates. We reported the outcomes for all patients undergoing intervention to account for those inappropriately treated with SAVR, inappropriately treated with TAVR, and those treated appropriately. Most hospitals in the high tertile were in New York. This may be related to the fact that the Northeast region of the United States had the highest rate of TAVR use compared to other regions.¹⁶ According to our analysis, the variation in survival between hospitals was independent of their volume or teaching status but depended on the hospital’s use of TAVR vs SAVR. For instance, patients who underwent AVR in hospitals with high TAVR use despite having fewer comorbidities had worse survival and a higher risk of postoperative complications compared to patients receiving care in low or medium-tertile hospitals.

Addressing variation in treatment allocation in lower-risk younger patients is important for improving clinical outcomes. The variation in procedure selection among hospitals may be attributed to the inconsistencies in implementing guidelines within clinical practice. Although the ACC/AHA guidelines recommend SAVR for patients < 65, they acknowledged that age is used as a surrogate for life expectancy, and the final decision should be individualized based on longevity and quality of life, leaving space for clinicians’ judgement in making the ultimate procedure choice.⁷ Other possible reasons contributing to the disagreement between guidelines and clinical practice include patients’ preference, which was described as the most prominent driver for TAVR choice in younger patients.¹⁷ Accredited to its minimally invasive nature, TAVR is considered an attractive option for patients. This comes despite data demonstrating superior outcomes of aortic regurgitation patients treated according to guideline recommendations compared to those who were not treated per recommendations, further validating the beneficial influence of adherence to guidelines.¹⁸

On the other hand, indication creep in TAVR patients has been previously described and remains a pressing concern.^10,19 A single-centre study reported that “off-label” use of TAVR was documented in 67% of patients between 2005 and 2008.²⁰ “Off-label use” was defined according to a set of patient-related and anatomically-related criteria consistent with previous clinical trial inclusion/exclusion criteria and the product labelling of CoreValve device used in the study. One of the “off-label” criteria was age < 65 years without severe comorbidities. This further confirms the lack of adherence to ACC/AHA guidelines recommendations related to cardiovascular procedures, which was well-documented in previous studies, rooting for a more complex cause of disunity.^15,21–23

For general cardiologists, valvular heart disease remains a relatively small area of their clinical practice, and as such, guideline unfamiliarity may contribute to the lack of adherence.²⁴ In the revascularization space, Leape and colleagues²⁵ found that familiarity with ACC/AHA guidelines for CABG and PCI was insufficient to ensure compliance with recommendations. However, physicians were more likely to follow recommendations when supported by evidence from randomized clinical trials. This highlights the need for clinical trials assessing TAVR vs SAVR utilization criteria in patients younger than 65.

Well-validated clinical trials are also important in addressing the rising concerns regarding the durability of bioprosthetic valves and TAVR complications in young patients. For instance, the increased risk of new PPI post-TAVR is associated with reduced left ventricular function. Although previous studies showed it did not affect the 2-year survival of elderly patients, this may not be the case in younger patients with longer life expectancy.²⁶

Long-term durability data for TAVR valves are limited, primarily because most clinical trials enrolled octogenarian patients. Although short-term outcomes in slightly younger low-risk patients are promising up to 2 years, the applicability of these findings to younger adults with longer life expectancy remains uncertain.²⁷ Previous reports suggest that structural valve deterioration after TAVR is underestimated in low-risk patients.^28,29 As the use of TAVR expands to younger patients, the increased risk of aortic valve reintervention has to be taken into consideration when choosing the first AVR approach.

The substantial increase in the number of patients requiring surgical reintervention after TAVR in recent years and the associated risk of perioperative mortality necessitate the careful assessment of the patient’s clinical profile and life expectancy before deciding on TAVR as the first procedure.³⁰ Until further evidence from clinical trials is generated, adhering to the current guidelines’ recommendations remains prudent. In addition, implementing a heart team approach to decision-making, consistent with the US Centres for Medicare and Medicaid Services (CMS) reimbursement requirements, is essential to assess the suitability of the treatment modality.³¹

This study has limitations. First, the administrative databases lack clinical variables, such as echocardiogram findings, the cause of death, type of device, access site, and other unmeasured confounders, including physician/surgeon skills. Second, we have no information about referral patterns, which may have impacted the procedure choice. Also, we lack data about postoperative care that can impact patient outcomes. Finally, events that patients experienced while out of state were not reported, possibly leading to an underestimation of outcome incidence. However, we have no reason to believe that the rates of patients moving out of state differ between comparison groups.

CONCLUSION

Patients younger than 65 undergoing aortic valve intervention at hospitals with high use of TAVR experienced worse early and late postoperative mortality and a higher risk of pacemaker implantation. Randomized clinical trials are needed to identify the best procedure for treating aortic stenosis in patients <65 years.

Supplementary Material

Supplementary material is available at EJCTS online.

DATA AVAILABILITY

The data for this study is protected under data use agreements with states and has not been shared with individuals not included in these agreements.