Introduction

Aortic insufficiency (AI) or regurgitation (AR) secondary to valvular incompetence, remains a major health burden, causing aortic flow reversal and left ventricular dilatation1. It is the fourth most common valvular heart disease globally2 and the third most common in the United States3, with a lifetime risk of 13% in men and 8.5% in women4 and an estimated U.S. prevalence of 4.9%5,6. The clinical progression of AI is complex and often insidious. While early disease may be well-tolerated for years, chronic volume and pressure overload leads to eccentric left ventricular remodeling and enlargement that can progress to myocardial dysfunction before symptom onset7. Recent evidence suggests that adverse remodeling may occur at lower thresholds than previously recognized, with end-systolic volume index > 45 mL/m2 and ejection fraction < 60% potentially identifying early dysfunction8. This highlights the importance of timely intervention, as waiting for traditional clinical triggers may allow significant adverse remodeling to occur.

Current guidelines recommend surgical aortic valve replacement (SAVR) for symptomatic severe AI and for asymptomatic severe AI with left ventricular dysfunction or other specific indications9. However, emerging data suggest increased mortality even in patients with moderate AI10,11, who don’t meet current surgical criteria. Additionally, recent studies indicate that patients meeting Class I surgical indications may already have advanced disease progression, potentially limiting the benefits of intervention12. The lack of transcatheter options for isolated AI, unlike aortic stenosis, may further impact treatment decisions13, though promising early trials of transcatheter therapies are underway.

The purpose of this study is to analyze and compare the healthcare burden of patients with symptomatic and asymptomatic AI using Optum’s United Healthcare database. By examining real-world patterns of mortality, healthcare utilization, and expenditures, we aim to better understand the impact of delayed intervention and identify opportunities for earlier treatment to optimize outcomes.

Methods

Data source and study inclusion

This study analyzed administrative healthcare data from Optum’s integrated database, which contains United Health Care claims from Q3 2017 to Q3 2022. The database comprises longitudinally-linked medical and pharmacy claims combined with electronic health records (EHR) data. The dataset includes comprehensive patient-level information across multiple therapeutic areas, with deterministically linked medical claims and EHR data. Clinical information is extracted from both structured data fields and unstructured clinical notes using natural language processing, encompassing approximately 100 million patient histories. Available data elements include diagnostic codes, clinical signs and symptoms, biomarkers, laboratory values, and diagnostic test results, representing approximately 20,000 clinical variables from ambulatory care assessments and measurements14. All data used to perform this analysis were de-identified and accessed in compliance with the Health Insurance Portability and Accountability Act. As a retrospective analysis of a de-identified database, this research was exempt from IRB review under 45 CFR 46.101(b)(4). This study followed the guidelines set forth by the International Society of Pharmacoeconomic Outcomes Research (ISPOR) for retrospective claims based studies15. If data from this study is requested, please contact the corresponding author Dr. Vallabhajosyula at the following address: prashanth.vallabhajosyula@yale.edu.

To be included in the retrospective database analyses of Optum claims data, patients must have had a claim for International Classification of Diseases, Tenth Revision (ICD10) for AI (defined at time zero) between Q3 2017 and Q3 2022 and have 12 months of continuous enrollment (baseline period) prior to their AI diagnosis in order to capture comorbid conditions. Patients were excluded if they had a record of aortic valve replacement (AVR) during this baseline period. Since this research focuses on isolated AI, patients with ICD10 diagnosis of aortic stenosis (AS) at any time of follow-up were excluded. Please see Appendix A for all ICD10 coding detail.

Variables of interest

We investigated two main study cohorts: (1) patients with symptomatic isolated AI at time zero, and (2) patients with asymptomatic isolated AI at time zero. AI was considered symptomatic if patients had ≥ 2 visits (inpatient or outpatient) for heart failure, angina, dyspnea, or syncope at baseline, with time zero being the last date all conditions have been met. Our definition of symptomatic aortic insufficiency was used in a previous publication (See Still et al., JAHA 2024)16 and reflects a clinically meaningful threshold when working with administrative claims data where echocardiographic findings are unavailable. Multiple healthcare encounters for these symptoms suggest persistent, clinically significant disease rather than isolated incidents. Furthermore, these symptoms (heart failure, angina, dyspnea, syncope) typically manifest in patients with moderate to severe aortic insufficiency, as mild disease is often asymptomatic. The frequency of healthcare visits (≥ 2) for these symptoms suggests hemodynamically significant valve dysfunction requiring medical attention. While not a direct measure of valve function like regurgitant fraction or valve area, symptom burden correlates with disease severity and is a key factor in clinical decision-making. The 2020 ACC/AHA Guideline for the Management of Patients with Valvular Heart Disease9 recommends aortic valve replacement for symptomatic patients with severe aortic regurgitation (Class I recommendation), as symptoms indicate significant disease progression.

Measured metrics for primary outcome of interest were all-cause mortality and secondary outcomes of interest included time to home health, time to skilled nursing facility (SNF), annualized healthcare utilization (all-cause hospitalizations, cardiovascular-related hospitalizations, total hospital days, emergency department visits, outpatient visits) and healthcare costs (annualized drug [Rx] and total costs). All healthcare utilization and cost outcomes were annualized to reflect the total outcome in the post-period divided by the number of years of follow-up, to control for the fact that patients have varying levels of follow-up. All-count outcomes were rounded to the nearest whole number to employ Poisson and negative binomial count-regression methodologies. Covariates included: age, sex, race, region, insurance type (commercial or Medicare Advantage), index year, and the Elixhauser Comorbidity Index (ECI), which identifies 31 categories of comorbidities associated with mortality.

Statistical methods

Summary statistics were univariately compared using t-tests (for continuous variables) and chi-squared tests (for categorical variables) in Tables 1 and 2. The Cox proportional model was used to estimate the cause-specific hazard ratios and adjusted curves for mortality. For time to home health and SNF, the cause-speciific hazard ratios were calculated in conjunction with competing risk models used to estimate the adjusted curves. Death was considered as a competing risk for home health and SNF. The proportional hazards assumption was tested by generating initial Kaplan-Meier curves for each outcome and ensuring the curves were parallel. Both cohorts had approximately 2 years of follow-up (1.9 for AAI, 1.8 for SAI) on average, which minimizes any selection bias due to varying levels of follow-up. General linear models were estimated for annualized healthcare utilization (all-cause hospitalizations, cardiovascular-related hospitalizations, emergency department visits, hospital days, outpatient hospital visits) and annualized Rx costs and total costs.

Table 1 Patient demographics for asymptomatic and symptomatic AI at index presentation are shown.
Full size table
Table 2 Elixhauser comorbid conditions in asymptomatic and symptomatic AI patients are shown.
Full size table

All outcomes were modeled separately (using SAS 9.4), with the main independent variable being isolated asymptomatic aortic insufficiency (AAI) patients versus isolated symptomatic aortic insufficiency (SAI) patients. All models adjusted for age, sex, race, region, insurance type (commercial or Medicare advantage), index year, and the Elixhauser Comorbidity Index.

Results

Figure 1 displays the attrition diagram for this study. The total number of patients meeting all inclusion/exclusion criteria having isolated AI was 249,660, with 58.23% (145,378) having SAI at time zero. Table 1 displays patient demographics for both cohorts (AAI and SAI). Patients with incident SAI were older (average age [SD] of 74.60 [10.49] versus 70.92 [13.25]), a higher percentage were African American (12.59% versus 9.51%), and they had a higher rate of Medicare Advantage insurance (88.30% versus 79.85%). Regional distributions were similar, with the South dominant in both cohorts. Distributions of gender and index year were also similar between the two groups.

Fig. 1
Fig. 1
Full size image

Attrition diagram.

Table 2 displays the 31 chronic comorbid conditions that make up the ECI as well as the average (SD) ECI score for each cohort (AAI and SAI). Patients with SAI at time zero on average were sicker than patients with AAI (5.83 [3.46] versus 3.28 [2.35]), with higher rates for every one of the comorbidities of interest. In particular, SAI patients had higher rates of congestive heart failure (35.68% versus 3.93%), cardiac arrhythmias (49.51% versus 29.57%), valvular disease (27.13% versus 17.18%), pulmonary circulation disorders (5.53% versus 1.43%), peripheral vascular disease (33.67% versus 18.17%), COPD (32.78% versus 13.15%), renal failure (26.05% versus 12.42%), liver disease (8.40% versus 5.63%), coagulopathy (9.09% versus 4.43%), and obesity (21.03% versus 13.59%). Patients in both cohorts had high rates of hypertension and diabetes (85.52% and 33.97% [SAI] versus 70.63% and 21.88% [AAI]), and the SAI cohort had higher rates of complicated disease (see Table 2).

For the AAI cohort, Fig. 2 displays Kaplan Meier curve for time to symptomatic disease (SAI) for a 5-year time horizon. Of the 104,282 patients without a record of symptomatic AI at time zero, an estimated 47.34% will become symptomatic within 5 years of their incident AI diagnosis. Given that symptomatic AI meets guidelines for surgical intervention, and that 47% of asymptomatic patients become symptomatic over 5-year follow-up, we assessed for surgical AVR rates in the asymptomatic cohort. Similarly, as symptomatic patients at time zero already meet the surgical trigger by guidelines, we also assessed for AVR rates in this group over 5-year follow-up. Importantly, regardless of whether an isolated AI patient is symptomatic or not over the 5-year horizon of this study, only a very small percentage of patients (AAI 4.18% and SAI 0.58%) will have an aortic valve replacement (see Fig. 3). Since AVR is the only curative treatment for this condition, the very low number of patients to receive it is surprising. For those without symptoms, a usage rate of only 4.18% across the horizon of this study signals underutilization, since 47% will become symptomatic over the 5 years, at which point they would meet guideline recommendations for AVR. For those with symptoms, the numbers are even smaller: only 0.58%, or < 1%, of those patients receive the potentially curative AVR over the 5-year follow-up, even though they meet guideline recommendations at the time of index diagnosis.

Fig. 2
Fig. 2
Full size image

Kaplan Meier curve for time to onset of symptoms in the asymptomatic AI cohort is shown. Of 104,282 asymptomatic patients identified at index diagnosis, 47.34% will become symptomatic within 5 years of follow-up.

Fig. 3
Fig. 3
Full size image

Surgical AVR is the only potential curative treatment for severe AI and is recommended in patients with symptomatic AI. Kaplan Meier estimates for time to AVR was assessed in the two study cohorts, asymptomatic (AAI) and symptomatic (SAI) patients, over 5-year follow-up. Although 47% of AAI patients will develop symptoms, only 4.18% of these patients will undergo AVR. Further, > 99% of patients with symptomatic AI did not undergo potentially curative AVR intervention during the 5-year follow-up.

Given the remarkably poor penetration of AVR therapy in both study groups, we then assessed the risk of death, reliance on home healthcare, and admission to SNF over the follow-up period in both the SAI and AAI groups (Fig. 4). Patients with SAI were more likely to die (HR 1.22; CI: 1.19 versus 1.25, P < .0001), utilize home healthcare (HR 1.26 CI: 1.24 versus 1.27, P < .0001), and be admitted to a SNF (HR 1.21, CI: 1.19 versus 1.24, P < .0001) than patients with AAI over the 5-year study time horizion. At 5 years, estimated rates of death, home healthcare usage, and SNF admission for SAI versus AAI patients were 27% versus 23%, 64% versus 47%, and 16% versus 7%, respectively.

Fig. 4
Fig. 4
Full size image

(A) Multivariable time to death in AAI and SAI patient cohorts. Adjusted time to death demonstrates significantly higher risk in SAI patients (p < .0001). (B) Multivariable time to home health demonstrated significantly higher risk in SAI patients (p < .0001). (C). Multivariable time to SNF demonstrated significantly higher risk in SAI patients (p < .0001).

Noting the high mortality and health care needs in both of these patient populations, we estimated the annualized healthcare utlization and cost burden for the SAI and AAI cohorts (Fig. 5). General linear models were constructed for estimations of annualized all-cause inpatient hospitalizations, cardiovascular-related inpatient hospitalizations, ER visits, hospital days, outpatient hospital visits (Fig. 5A), and annualized prescription medication costs and total costs (Fig. 5B). Regardless of the hospital visits (all-cause and cardiovascular-related), ER visits, outpatient visits, or number of annualized days spent in the hospital, the SAI cohort had significantly higher healthcare utlization than the AAI cohort (P < .0001): IP visits 0.49 versus 0.40, CV IP visits 0.15 versus 0.10, ER visits 0.62 versus 0.46, hospital days 3.8 versus 3.0, and OP visits 6.5 versus 6.2.

Fig. 5
Fig. 5
Full size image

General linear models for adjusted annualized healthcare utilization and costs is shown. (A) Annualized in-patient hospitalizations for all-cause and cardiovascular (CV) related, emergency room (ER) visits, hospital days and outpatient (OP) visits are shown for AAI and SAI cohorts. (B) Annualized prescription (Rx) medicine and total health care costs for AAI and SAI cohorts are shown.

For healthcare costs, after multivariable modeling, there was no difference between annualized prescription medication costs (P = .5226) for SAI and AAI patients ($11,463 versus $11,413); however, there was a statistically significant difference (P < .0001) of $6,394 in total adjusted annualized costs between the two cohorts (SAI $54,982 versus AAI $48,588). Overall, the healthcare cost burden of isolated AI in both symptomatic and asymptomatic patients was high.

Discussion

In this retrospective, real-world analysis of 249,660 patients with isolated aortic valve insufficiency, 58.23% (145,378) had symptomatic disease at the time of their incident diagnosis (time zero); of the remaining 41.86% (104,282) of patients, who were asymptomatic at the time of their incident AI diagnosis, almost half (47.34%) became symptomatic within 5 years. This rapid progression aligns with recent evidence showing that adverse left ventricular remodeling may occur at lower thresholds than previously recognized, with end-systolic volume index > 45 mL/m2 and ejection fraction < 60% potentially identifying early dysfunction8. This study highlights the substantial healthcare burden associated with AI in general, and especially for those with symptomatic AI. Patients with symptomatic AI at the time of their diagnosis were older and sicker at baseline than asymptomatic AI patients. After adjusting for the differences in baseline demographics and chronic comorbid conditions, we found the healthcare burden for the symptomatic cohort to be consistently higher than that of the asymptomatic cohort in the following measures: rates of death (27% versus 23%), home healthcare utilization (64% versus 47%), SNF admissions (16% versus 7%), IP visits (0.49 versus 0.40), CV IP visits (0.15 versus 0.10), ER visits (0.62 versus 0.46), hospital days (3.8 versus 3.0), and OP visits (6.5 versus 6.2). Overall, we found a statistically significant difference (of $6,394 in total adjusted annualized costs (SAI $54,982 versus AAI $48,588), emphasizing the cost burden incurred to society from this disease. The high rates of home healthcare utilization (64% versus 47%) and SNF admissions (16% versus 7%) in the symptomatic cohort particularly highlight the downstream functional decline associated with untreated AI. These outcomes likely reflect the higher comorbidity burden and cardiovascular complications in symptomatic patients, suggesting more rapid progression to functional dependency when definitive treatment is delayed or withheld.

Notably, regardless of whether an AI patient is symptomatic or not over the 5-year time horizion of this study, only a very small percentage of patients (AAI 4.18% and SAI 0.58%) will have an aortic valve replacement. Particularly for patients with symptoms, undergoing AVR, which is per current clinical guidelines, is being severely underutilized by our cardiovascular community17. Given that AI presents a heavy healthcare burden and timely intervention with AVR is an excellent treatment modality, understanding the causes for such low rates of surgical intervention in both the study cohorts will be important to address this huge clinical gap.

Our findings dovetail with results of another recent study, which showed that timely intervention with SAVR provided a significant survival benefit and decreased overall healthcare utilization and cost burden in patients with symptomatic aortic regurgitation (sAR). According to that analysis, patients who received a SAVR within 12 months of their sAR diagnosis had a significant mortality benefit compared with those who did not; and those who received the surgery within 5 years of diagnosis were associated with decreased mortality (sAR without SAVR reached a 39% mortality rate over 5 years versus sAR with SAVR, at 15%, p < .0001). The analysis also demonstrated lower healthcare utilization and costs over 12 months between sAR patients with and without SAVR ($41,399 < sAR without SAVR versus $33,744 < sAR with SAVR, for a difference of $7,655, p < .0001). Additionally, patients who received SAVR demonstrated an 18% reduction in estimated annualized costs through decreased emergency department visits and average hospital days16.

There are numerous barriers to surgery—all of which warrant further study—that may account for such low penetration of AVR intervention in this patient population. First, the increased comorbid burden in the symptomatic AI group may lead physicians/healthcare providers to consider the cause of patient symptoms secondary to the comorbid conditions rather than to AI. Second, there may be some resistance in the medical community to refer patients with symptomatic AI to surgeons, as the disease can have a relatively indolent course. This resistance for surgical referral may also be partially driven by patients themselves, who often express anxiety over open-heart operations. Therefore, continuing and escalating medical therapy may seem preferable to surgery by both physician/provider and patient. While our symptomatic cohort had higher comorbidity burden (ECI 5.83 vs. 3.28) and cardiovascular complications, these differences alone don’t fully explain the markedly low AVR rates. A recent analysis of 4,608 symptomatic AI patients found that 25.7% underwent SAVR within one year of diagnosis, with 9% mortality at one year versus 24% if left untreated. This suggests that a larger proportion of symptomatic patients can safely undergo and benefit from surgery than what we observed in our cohort (< 1%). Further research is needed to understand whether specific comorbidity combinations or other clinical factors are driving surgical decision-making in these patients. Recent advances in imaging have improved our ability to detect early myocardial dysfunction7, yet this hasn’t translated to earlier intervention. This is particularly concerning as recent studies indicate that patients meeting Class I surgical indications may already have advanced disease progression that limits intervention benefits12.

Third, the lack of a transcatheter valve option for treating AI—unlike aortic stenosis, which can be treated with either surgical AVR or transcatheter AVR—may contribute to the delay in surgical referral. Interestingly, from a technical standpoint, surgical AVR for AI can be less complex than that for aortic stenosis, which carries a higher chance of increased calcification of all elements of the aortic root complex in addition to the aortic valve— including the coronary arteries and the sinuses of Valsalva, which also tend to be smaller. Therefore, from a surgical AVR standpoint, symptomatic AI patients possibly have the options for a surgical therapy that may be less technically complex than surgery for aortic stenosis. A better understanding of surgical outcomes in patients with symptomatic AI will help understand the relative survival benefit of timely intervention.

In a recent analysis of 4,608 symptomatic AI patients, 25.7% had surgical AVR within one year. Within that 25.7%, mortality was 9% at one year versus 24% if left untreated17. Another study investigated guideline-based surgical outcomes for symptomatic AI and found that class I indication triggers for surgical intervention carried higher long-term risks than class IIa or class IIb indications18. This suggests that given the high mortality seen with medical management even for asymptomatic severe AI, patients meeting class I indications by current guidelines may be too far progressed in the disease process to maximize the benefit of surgical AVR. This supports the premise that earlier intervention in this patient population may render improved survival benefit. Collectively, these studies emphasize the importance of early recognition of asymptomatic and symptomatic AI in patients, and the importance of timely referral for consideration for surgical therapy. Furthermore, identification of hurdles leading to such a large clinical gap in patient care will be critical for improving the care of patients with isolated AI.

Limitations of this study include the fact that AI was ascertained through sources of automated data that rely on coding. This could have been biased in terms of over- or under-coding. For example, we do not have AR severity through echocardiographic data, as such, our study relied on diagnosis codes and health claims data. This limitation may have resulted in inclusion of patients with less severe disease in our cohort, contributing to the lower observed SAVR rates. We could only control for known confounders—largely of generalized illness—and lacked the ability to control for unknown confounders, ideally by mechanisms such as randomization, stratification, and matching. We used statistical modeling to control for the potential confounding effect of known variables with between-group differences. While statistical models controlled for several factors, models could not control for some variables that are not included in an administrative database, such as echocardiography results. A strength of the present study, however, is that the data reflect real-world patient characteristics and outcomes across the country as compared to evidence from controlled clinical trials.

Conclusions

In this retrospective, claims-based, real-world analysis involving 249,660 isolated AI patients, our findings revealed that 58.23% (145,378) presented with symptomatic disease at the time of their initial diagnosis, while approximately 50% of the remaining patients were projected to develop symptoms within the 5-year study duration. Remarkably, only a very small proportion (less than 1%) of symptomatic patients had any documented record of receiving a guideline-recommended intervention, such as aortic valve replacement.

This study demonstrates that patients with symptomatic AI have higher rates of mortality, reliance on home healthcare and SNF, and increased utilization of inpatient care, emergency room visits, and outpatient consultations, leading to an adjusted annualized cost difference of $6,394 between the two cohorts (SAI $54,982 versus AAI $48,588). These findings emphasize the urgent need to address this large clinical gap in optimal, timely care for patients with AI. It also emphasizes the need to develop other interventions that can effectively slow the progression of isolated AI. By addressing the disease progression, we can potentially alleviate the healthcare burden experienced by patients and healthcare systems alike. Implementing strategies aimed at early detection, appropriate management, and timely interventions, such as AVR, could significantly limit the adverse impact of symptomatic AI and its associated costs.